abstract_text
stringlengths 56
26.7k
| article_id
stringlengths 9
15
| article_text
stringlengths 27.2k
74.7k
|
|---|---|---|
in the third part of the series presenting the optical gravitational lensing experiment ( ogle ) microlensing studies of the dark matter halo compact objects ( machos ) we describe results of the ogle - iii monitoring of the large magellanic cloud ( lmc ) . this unprecedented data set contains almost continuous photometric coverage over 8 years of about 35 million objects spread over 40 square degrees . we report a detection of two candidate microlensing events found with the automated pipeline and an additional two , less probable , candidate events found manually . the optical depth derived for the two main candidates was calculated following a detailed blending examination and detection efficiency determination and was found to be . if the microlensing signal we observe originates from machos it means their masses are around and they compose only per cent of the mass of the galactic halo . however , the more likely explanation of our detections does not involve dark matter compact objects at all and rely on natural effect of self - lensing of lmc stars by lmc lenses . in such a scenario we can almost completely rule out machos in the sub - solar mass range with an upper limit at per cent reaching its minimum of per cent at . for masses around the constraints on the machos are more lenient with per cent . owing to limitations of the survey there is no reasonable limit found for heavier masses , leaving only a tiny window of mass spectrum still available for dark matter compact objects . [ firstpage ] cosmology : dark matter , gravitational lensing , galaxy : structure , halo , galaxies : large magellanic cloud | 1012.1154 | the milky way s halo is probably one of the least known parts of our galaxy . numerous recent detections of tidal debris leftover after close encounters between smaller dwarf galaxies and our giant spiral confirm the presence of cold dark matter ( cdm ) substructures in the halo ( e.g.,@xcite ) . however , the following question still remains unanswered : what actually is the dark matter ( dm ) or is not . compact dark matter objects ( machos ) would have been the most convenient explanation and in the last two decades this theory has been tested using various methods sensitive to different ranges of masses . in the high mass regime ( @xmath8 ) wide halo binary objects were studied , but no signature of disturbance due to machos was detected ( see @xcite and @xcite ) . for detecting stellar - mass compact dm objects the technique of gravitational microlensing was suggested by @xcite . it employs a unique feature of gravitational lensing , namely its sensitivity to unseen objects when they bend and amplify the light of a distant source . the idea was simple : let us observe some distant background rich in stars ( e.g.,magellanic clouds ) and wait for their temporal brightening due to passage of a massive object located along the line - of - sight between us and the source . several observing campaigns started after paczynski s proposal : macho @xcite , ogle @xcite , eros @xcite and moa @xcite , which for many years observed the large and small magellanic clouds ( lmc and smc , respectively ) . the macho collaboration was first to publish their results and claimed 20 per cent of mass of the galactic halo is composed of machos with an average mass of 0.4 @xmath9 ( @xcite , @xcite ) . based on detection of 10 candidates for microlensing events found in the central 15 sq . deg of the lmc observed over 5.7 years they derived the optical depth towards the lmc of @xmath10 @xcite . in 2007 the eros group analysed their data and published their conclusions , which contradicted the macho result quite severely . in their data comprising of 6.7 years of continuous observations of 84 deg@xmath11 they found no candidates for microlensing events among their bright sample of stars @xcite . this led them to the upper limit for the optical depth of @xmath12 , which translated to @xmath13/@xmath14 per cent for machos with masses @xmath15 . the ogle project monitored the lmc during its second ( 1996 - 2000 ) and third ( 2001 - 2009 ) phases ( hereafter ogle ii and ogle iii , respectively ) , and the observations are still carried on in the current ogle our study of the ogle ii data ( @xcite , herafter paper i ) led to the detection of two candidates for microlensing events , ogle - lmc-01 and ogle - lmc-02 , and @xmath16 , which is closer to the limit of eros than the value of macho . moreover , all detected microlensing signal can be attributed to self - lensing ( i.e. , when foreground lmc stars microlens backgound lmc stars ) , leaving no room for lensing due to dm halo objects in the sub - solar and solar mass range . however , the ogle ii phase lasted only 4 years and covered only parts of the central bar of the lmc ( about 4.7 deg@xmath11 ) , therefore its result is naturally limited and is subject to some uncertainty . due to small number statistics , the upper limit on the macho presence in the milky way s halo was estimated only at @xmath17 per cent . in this paper we present yet another voice in this turbulent story of microlensing towards the magellanic clouds and report the results of the search for microlensing events in the ogle iii data gathered towards the lmc . ogle iii overwhelms most previous studies in terms of its duration ( 8 years ) and coverage ( 40 deg@xmath11 ) but more importantly , it uniformly covers the entire lmc bar region and much of lmc outskirts . the paper is organised as follows . in section [ sec : data ] , the ogle iii lmc dataset is described . in the following section , the search algorithm is presented . section [ sec : results ] contains the results of the search and detailed description of all detected candidates for microlensing events . next , the blending study is described and the optical depth is derived . the paper finishes with discussion of the results and conclusions . cccrrc|cccrrc field & @xmath18 & @xmath19 & & @xmath20 & field & @xmath18 & @xmath19 & & @xmath20 + & & & tpl & real & & & & & tpl & real & + lmc100 & 5:19:02.2 & -69:15:07 & 712.2 & 973.9 & 5.13 & lmc158 & 4:30:59.9 & -70:26:01 & 20.0 & 20.8 & 3.63 + lmc101 & 5:19:03.1 & -68:39:19 & 279.4 & 323.2 & 4.84 & lmc159 & 5:25:11.4 & -68:03:58 & 129.1 & 137.6 & 4.47 + lmc102 & 5:19:03.4 & -68:03:48 & 125.9 & 134.7 & 4.52 & lmc160 & 5:25:20.9 & -68:39:24 & 264.4 & 296.0 & 4.74 + lmc103 & 5:19:02.9 & -69:50:26 & 540.6 & 687.4 & 5.02 & lmc161 & 5:25:32.5 & -69:14:59 & 453.6 & 561.8 & 4.99 + lmc104 & 5:19:02.4 & -70:26:03 & 272.1 & 313.2 & 4.84 & lmc162 & 5:25:43.3 & -69:50:24 & 817.7 & 1159.9 & 5.17 + lmc105 & 5:19:01.6 & -71:01:31 & 205.0 & 224.2 & 4.65 & lmc163 & 5:25:52.2 & -70:25:55 & 524.6 & 669.5 & 5.00 + lmc106 & 5:19:01.0 & -71:36:57 & 135.5 & 144.6 & 4.47 & lmc164 & 5:26:08.4 & -71:01:23 & 197.5 & 217.1 & 4.68 + lmc107 & 5:13:01.5 & -66:52:57 & 93.2 & 99.3 & 4.45 & lmc165 & 5:26:20.9 & -71:37:01 & 140.6 & 152.3 & 4.61 + lmc108 & 5:13:01.9 & -67:28:40 & 105.8 & 113.2 & 4.51 & lmc166 & 5:31:20.1 & -68:03:51 & 127.8 & 138.5 & 4.59 + lmc109 & 5:12:53.3 & -68:04:06 & 138.8 & 149.4 & 4.57 & lmc167 & 5:31:39.6 & -68:39:32 & 180.9 & 200.1 & 4.69 + lmc110 & 5:12:43.6 & -68:39:42 & 293.4 & 335.6 & 4.82 & lmc168 & 5:32:01.4 & -69:15:00 & 316.2 & 374.0 & 4.90 + lmc111 & 5:12:32.7 & -69:15:02 & 441.2 & 526.7 & 4.94 & lmc169 & 5:32:22.8 & -69:50:26 & 757.6 & 1054.8 & 5.14 + lmc112 & 5:12:21.5 & -69:50:21 & 481.7 & 594.0 & 4.97 & lmc170 & 5:32:48.1 & -70:25:53 & 588.6 & 769.3 & 5.04 + lmc113 & 5:12:10.9 & -70:25:48 & 289.7 & 334.2 & 4.85 & lmc171 & 5:33:10.6 & -71:01:30 & 235.8 & 268.3 & 4.81 + lmc114 & 5:11:58.9 & -71:01:22 & 109.4 & 116.2 & 4.44 & lmc172 & 5:33:34.4 & -71:36:54 & 185.7 & 205.9 & 4.72 + lmc115 & 5:07:09.7 & -66:52:59 & 133.8 & 143.0 & 4.47 & lmc173 & 5:37:29.3 & -68:03:50 & 126.9 & 135.2 & 4.44 + lmc116 & 5:07:00.9 & -67:28:29 & 117.3 & 124.4 & 4.41 & lmc174 & 5:37:59.8 & -68:39:26 & 155.5 & 169.2 & 4.63 + lmc117 & 5:06:55.3 & -68:03:58 & 260.4 & 298.2 & 4.82 & lmc175 & 5:38:32.3 & -69:15:01 & 268.7 & 305.3 & 4.79 + lmc118 & 5:06:25.4 & -68:39:25 & 383.6 & 463.1 & 4.94 & lmc176 & 5:39:01.6 & -69:50:30 & 357.0 & 414.9 & 4.86 + lmc119 & 5:06:02.5 & -69:15:02 & 578.7 & 723.4 & 5.01 & lmc177 & 5:39:38.0 & -70:25:49 & 459.3 & 577.8 & 5.01 + lmc120 & 5:05:39.8 & -69:50:28 & 339.4 & 399.4 & 4.89 & lmc178 & 5:40:14.1 & -71:01:27 & 260.6 & 292.4 & 4.77 + lmc121 & 5:05:14.4 & -70:25:59 & 200.6 & 223.5 & 4.74 & lmc179 & 5:40:52.3 & -71:36:58 & 167.3 & 180.7 & 4.57 + lmc122 & 5:04:52.9 & -71:01:25 & 139.3 & 150.6 & 4.56 & lmc180 & 5:40:51.5 & -72:12:28 & 111.6 & 120.0 & 4.51 + lmc123 & 5:01:18.0 & -66:53:00 & 113.5 & 121.3 & 4.48 & lmc181 & 5:43:35.7 & -68:03:58 & 97.2 & 103.1 & 4.40 + lmc124 & 5:01:00.3 & -67:28:27 & 136.9 & 147.6 & 4.56 & lmc182 & 5:44:16.0 & -68:39:32 & 141.0 & 152.9 & 4.59 + lmc125 & 5:00:36.1 & -68:03:54 & 162.5 & 177.3 & 4.65 & lmc183 & 5:45:02.8 & -69:14:59 & 173.3 & 189.9 & 4.66 + lmc126 & 5:00:02.4 & -68:39:31 & 270.9 & 309.7 & 4.82 & lmc184 & 5:45:43.2 & -69:50:33 & 243.2 & 277.0 & 4.78 + lmc127 & 4:59:33.6 & -69:14:54 & 296.1 & 340.8 & 4.84 & lmc185 & 5:46:30.8 & -70:25:51 & 350.9 & 413.7 & 4.90 + lmc128 & 4:59:03.6 & -69:50:24 & 184.2 & 203.4 & 4.71 & lmc186 & 5:47:21.2 & -71:01:24 & 205.6 & 225.9 & 4.67 + lmc129 & 4:58:24.6 & -70:26:07 & 151.9 & 164.6 & 4.58 & lmc187 & 5:48:12.6 & -71:36:52 & 141.6 & 154.2 & 4.57 + lmc130 & 4:57:50.8 & -71:01:20 & 118.1 & 126.2 & 4.46 & lmc188 & 5:48:26.6 & -72:12:27 & 68.0 & 71.9 & 4.28 + lmc131 & 4:55:28.6 & -66:52:46 & 133.3 & 142.8 & 4.49 & lmc189 & 5:50:37.9 & -68:39:26 & 81.7 & 86.1 & 4.28 + lmc132 & 4:55:00.6 & -67:28:36 & 107.1 & 114.0 & 4.44 & lmc190 & 5:51:33.2 & -69:14:55 & 107.7 & 114.4 & 4.40 + lmc133 & 4:54:29.2 & -68:03:47 & 186.6 & 204.2 & 4.64 & lmc191 & 5:52:20.1 & -69:50:24 & 137.7 & 147.9 & 4.48 + lmc134 & 4:53:49.2 & -68:39:18 & 158.5 & 171.4 & 4.58 & lmc192 & 5:53:24.1 & -70:25:51 & 143.3 & 154.0 & 4.46 + lmc135 & 4:53:05.2 & -69:14:51 & 140.0 & 150.6 & 4.55 & lmc193 & 5:54:21.7 & -71:01:34 & 83.4 & 88.1 & 4.24 + lmc136 & 4:52:23.7 & -69:50:25 & 117.3 & 125.1 & 4.49 & lmc194 & 5:55:29.7 & -71:36:59 & 44.8 & 46.8 & 3.99 + lmc137 & 4:51:30.2 & -70:26:01 & 87.0 & 92.3 & 4.40 & lmc195 & 5:56:00.0 & -72:12:25 & 24.9 & 25.8 & 3.73 + lmc138 & 4:49:34.7 & -66:53:07 & 52.2 & 54.8 & 4.17 & lmc196 & 5:56:54.7 & -68:39:29 & 51.7 & 54.2 & 4.16 + lmc139 & 4:49:05.2 & -67:28:30 & 52.8 & 55.5 & 4.21 & lmc197 & 5:58:02.7 & -69:15:06 & 50.2 & 52.5 & 4.06 + lmc140 & 4:48:18.2 & -68:04:05 & 81.9 & 86.8 & 4.37 & lmc198 & 5:59:02.5 & -69:50:35 & 43.0 & 44.8 & 3.95 + lmc141 & 4:47:26.7 & -68:39:36 & 85.6 & 90.7 & 4.38 & lmc199 & 6:00:14.7 & -70:26:00 & 39.2 & 40.8 & 3.90 + lmc142 & 4:46:31.9 & -69:15:08 & 112.7 & 120.4 & 4.45 & lmc200 & 6:01:27.5 & -71:01:36 & 35.1 & 36.6 & 3.87 + lmc143 & 4:45:43.1 & -69:50:19 & 78.7 & 82.9 & 4.29 & lmc201 & 6:02:45.9 & -71:37:04 & 46.9 & 49.2 & 4.06 + lmc144 & 4:44:40.2 & -70:26:01 & 55.8 & 58.5 & 4.16 & lmc202 & 6:03:28.3 & -72:12:34 & 42.9 & 44.9 & 4.00 + lmc145 & 4:43:47.5 & -66:52:43 & 30.1 & 31.3 & 3.91 & lmc203 & 6:03:29.9 & -72:48:04 & 40.9 & 42.8 & 3.95 + lmc146 & 4:43:03.0 & -67:28:17 & 39.4 & 41.2 & 4.03 & lmc204 & 6:03:14.6 & -68:39:25 & 55.3 & 58.1 & 4.13 + lmc147 & 4:42:07.8 & -68:03:55 & 46.9 & 49.1 & 4.11 & lmc205 & 6:04:32.9 & -69:15:04 & 36.7 & 38.4 & 4.01 + lmc148 & 4:41:06.8 & -68:39:27 & 49.1 & 51.4 & 4.14 & lmc206 & 6:05:40.3 & -69:50:27 & 38.3 & 40.0 & 3.99 + lmc149 & 4:40:05.1 & -69:14:57 & 52.3 & 54.8 & 4.16 & lmc207 & 6:07:04.2 & -70:25:55 & 35.9 & 37.4 & 3.95 + lmc150 & 4:39:05.3 & -69:50:16 & 44.8 & 46.8 & 4.08 & lmc208 & 6:08:30.4 & -71:01:27 & 44.1 & 46.2 & 4.04 + lmc151 & 4:37:51.6 & -70:25:45 & 38.1 & 39.8 & 4.05 & lmc209 & 6:10:07.0 & -71:37:00 & 37.5 & 39.1 & 3.91 + lmc152 & 4:37:54.1 & -66:52:52 & 25.3 & 26.5 & 3.76 & lmc210 & 6:10:55.7 & -72:12:37 & 34.5 & 36.1 & 3.94 + lmc153 & 4:37:01.7 & -67:28:30 & 27.3 & 28.5 & 3.85 & lmc211 & 6:11:22.0 & -72:48:04 & 31.3 & 32.7 & 3.88 + lmc154 & 4:35:59.1 & -68:04:02 & 26.6 & 27.8 & 3.91 & lmc212 & 6:11:04.0 & -69:14:50 & 39.9 & 41.7 & 4.01 + lmc155 & 4:34:49.4 & -68:39:32 & 33.0 & 34.5 & 3.95 & lmc213 & 6:12:17.9 & -69:50:37 & 32.7 & 34.1 & 3.81 + lmc156 & 4:33:32.7 & -69:15:00 & 34.3 & 35.9 & 3.96 & lmc214 & 6:13:58.2 & -70:26:08 & 31.6 & 32.9 & 3.88 + lmc157 & 4:32:23.8 & -69:50:26 & 25.6 & 26.6 & 3.74 & lmc215 & 6:15:36.4 & -71:01:28 & 32.0 & 33.3 & 3.88 + & 19,424.4 & 22,740.0 & + _ note : _ coordinates point to the centre of the field ( centre of the mosaic ) , each being @xmath21 . number of `` good '' objects in the template is provided ( @xmath22 and @xmath23 mag ) together with the estimated number of real monitored stars ( see section [ sec : blending ] ) . mean number of all objects detected on a single ccd used for calculating the density of a field is given in the last column . the photometric observations presented in this work were carried out during the third phase of the ogle project ( 2001 - 2009 ) with the 1.3-m warsaw telescope located at las campanas observatory , chile , operated by the carnegie institution of washington . the `` second generation '' camera comprised of eight site @xmath24 ccd detectors with 15 @xmath25 m pixels resulting in 0.26 arcsec / pixel scale and @xmath26 arcmins field of view . the details on the instrumentation setup can be found in @xcite . there were 116 fields in the lmc covering a total of 40 square degrees . their distribution on the sky is shown in fig . [ fig : fields ] . the fields are listed in table [ tab : fields ] along with the coordinates of their centers , the number of `` good '' template objects in the @xmath27-band , the blending - corrected number of stars ( see section [ sec : blending ] ) and the mean number of all objects visible on a single ccd ( of 8) used for deriving the blending density level . by `` good '' we mean those template objects which have at least 80 good data points ( excluding measurements with very large error - bars ) and mean magnitude brighter than @xmath28 mag . the limiting magnitude was selected based on the mean observed luminosity function . the very first observations of the lmc within the ogle iii phase were taken in july 2001 ( jd=@xmath29 ) , however , regular monitoring started in september 2001 and continued until may 2009 ( jd=@xmath30 ) . the vast majority of observations were done through the cousins @xmath27-band filter with exposure time of 180 s. between 385 and 637 measurements were taken in each field with an average sampling varying from 3.0 to 4.6 days between subsequent frames ( excluding the gaps between the seasons ) . in addition to that , between 30 and 137 observations per field were obtained in johnson @xmath31-band and integration time of 225 s. average sampling frequency in the @xmath31-band was between 4.2 and 13.6 days . the images were reduced with the pipeline based on the difference image analysis ( dia ; @xcite , @xcite ) . the photometry used in this work comes from the final reduction calibrated to the standard system . full description of the reduction techniques , photometric calibration and astrometric transformations can be found in @xcite . photometric errors produced by the dia package were corrected following the method similar to that described in paper i. for each field the constant stars were selected based on their low photometric dispersion . these stars were then used for deriving the magnitude dependent correction factor for the error - bars which minimized the difference between observed scatter of the data points and the error - bar . for each field and filters @xmath32 we derived parameters @xmath33 and @xmath34 , which correct the original error - bar with the formula : @xmath35 where @xmath36 is the original error - bar returned by the photometry pipeline . for all the lmc fields the mean values of the error - correction parameters were : @xmath37 , @xmath38 , @xmath39 , @xmath40 . as a side product of the error correction study we obtained also a formula for calculating error - bars of synthetic @xmath27-band magnitudes used in the light curves simulations : @xmath41 where @xmath42 is the simulated magnitude for which the error bar ( @xmath43 ) is required , @xmath44 and @xmath45 are the magnitude and the error bar of the reference star at a given epoch . such calculated error bars still need to be corrected with eq . ( [ eq : errors ] ) . error corrections for @xmath27 and @xmath31 passbands for the first couple of fields are gathered in table [ tab : errorcor ] . the full table is available on - line on the ogle website . .error correction coefficients for each ccd chip of the first four ogle iii lmc fields for @xmath27- and @xmath46 bands . the full table is available on - line from the ogle website . [ cols="^,^,^,^,^",options="header " , ] lrrrrrr + parameter & & & + @xmath47 & 4136.8 & @xmath48 & 4137.0 & @xmath48 & 4136.8 & @xmath48 + & & & & & & + @xmath49 & 34.97 & @xmath50 & 24.65 & @xmath51 & 36.16 & @xmath52 + & & & & & & + @xmath53 & 0.08972 & @xmath54 & 0.22681 & @xmath55 & 0.08124 & @xmath56 + & & & & & & + @xmath57 & 18.423 & @xmath58 & 18.373 & @xmath58 & 18.423 & @xmath58 + & & & & & & + @xmath59 & 0.514 & @xmath60 & 1.0 & & 0.486 & @xmath61 + & & & & & & + @xmath62 & & & & & 19.517 & @xmath63 + & & & & & & + @xmath64 & & & & & 0.520 & @xmath65 + & & & & & & + @xmath66 & 1114.1 & & 1123.0 & & 1309.3 & + & & & & & & + @xmath67 & 2.52 & & 2.53 & & 2.71 & + + parameter & & & + @xmath47 & 2227.9 & @xmath68 & 2227.9 & @xmath68 & & + & & & & & & + @xmath49 & 32.76 & @xmath69 & 29.26 & @xmath70 & & + & & & & & & + @xmath53 & 0.87763 & @xmath71 & 1.0440 & @xmath72 & & + & & & & & & + @xmath57 & 17.238 & @xmath73 & 17.238 & @xmath73 & & + & & & & & & + @xmath59 & 0.70253 & @xmath74 & 1.0 & & & + & & & & & & + @xmath62 & & & & & & + & & & & & & + @xmath64 & & & & & & + & & & & & & + @xmath66 & 911.4 & & 911.5 & & & + & & & & & & + @xmath67 & 1.56 & & 1.56 & & & + + parameter & & & + @xmath47 & 3106.0 & @xmath75 & 3108.6 & @xmath76 & 3105.6 & @xmath773.8 + & & & & & & + @xmath49 & 224.1 & @xmath78 & 135.6 & @xmath79 & 346.9 & @xmath80 + & & & & & & + @xmath53 & 0.12014 & @xmath81 & 0.24846 & @xmath82 & 0.070525 & @xmath83 + & & & & & & + @xmath57 & 21.25 & @xmath84 & 21.24 & @xmath84 & 21.27 & @xmath770.03 + & & & & & & + @xmath59 & 0.43664 & @xmath85 & 1.0 & & 0.25126 & @xmath86 + & & & & & & + @xmath62 & & & & & 22.31 & @xmath770.06 + & & & & & & + @xmath64 & & & & & 0.46176 & @xmath87 + & & & & & & + @xmath66 & 274.1 & & 275.2 & & 330.46 & + & & & & & & + @xmath67 & 0.673 & & 0.674 & & 0.749 & + + light curves of the events are shown in fig . [ fig : events ] , along with the best microlensing models obtained for the available @xmath27- and @xmath31-band ogle data . parameters of the standard paczyski model fit for applicable events are gathered in table [ tab : models ] . models with 7 parameters were performed on @xmath27 and @xmath31 data , where available . 5- and 4-parameters models were fit to the @xmath27-band data only , with blending parameter as a free parameter and fixed to unity , respectively . [ fig : cmd ] shows derived positions of the sources of all ogle iii candidates for microlensing events on the colour - magnitude diagram ( cmd ) , with the locus of lmc stars based on the ogle and hst data . also plotted are the positions of two ogle ii events from paper i and events reported by macho , colour- and shape - coded to differentiate between self lensing candidates according to @xcite , the binary event macho - lmc-9 , the thick disk lens candidates macho - lmc-5 @xcite and macho - lmc-20 @xcite and the remaining candidates . [ fig : cmdlenses ] shows the same cmd , but with positions of the baselines of ogle iii events and estimated locations of the lenses in these events . the lens locations were derived based on an assumption that the entire blending light belongs to the lens . fig . [ fig : mapevents ] again presents all events ever detected towards the lmc from macho , ogle ii and ogle iii , over - plotted on top of the density map of the red clump stars from the lmc . below we discuss each of the ogle iii events . + this event was the only event alerted by the ogle s early warning system towards the lmc and was designated as 2007-lmc-01 . the baseline of this candidate exhibits some irregular variations . these are the most likely caused by the presence of the very bright ( 6 mag , hd 33923 ) star ( saturated on ogle frames ) less than one arc minute away . otherwise , the baseline does not show any additional event - like phenomenon , supported by the macho and eros observations since 1993 . this makes a very strong case for the microlensing origin of this event . a single lens microlensing model fit to the data was performed using @xmath27- and @xmath31-band ogle data and its parameters are gathered in table [ tab : events ] . the event s light curve along with the best fit model is shown in fig . [ fig : events ] . from the multi - band fit we were able to derive the source magnitude and colour of @xmath88 mag and @xmath89 mag . this places the source on the red giant branch of the cmd ( fig . [ fig : cmd ] ) , indicating the source intrinsically belongs to the lmc . field lmc122 , in which ogle - lmc-03 is located , is one of the sparsest according to its stellar density ( see table [ tab : fields ] ) . at that density and magnitude level , there is about 9 per cent of stars being blended with another one ( see section [ sec : blending ] ) , therefore we can safely assume the entire remaining light in the blend constituting the baseline of the event comes from a single star - the lens . another constraint comes from the astrometric measurements of the dia residuals around the peak of the event . they clearly indicate lack of any shift of the centroid during the microlensing magnification , which suggests the additional flux sits exactly on top of the source star . the blending parameters in both pass - bands obtained in the light curve fitting indicate the lens is very similar to the source in terms of the brightness and colour . hence , the lens may also be an lmc red giant ( see fig.[fig : cmdlenses ] ) . this relatively bright red star ( 17.3 mag ) shows an event of small amplitude ( @xmath900.5 mag ) just at the beginning of the ogle iii observations in nov 2001 ( the peak occurred on @xmath91 ) . its light curve is shown in fig . [ fig : events ] . there was no @xmath31-band ogle data available during the course of the event , but 37 @xmath31-band data points after the event allowed a derivation of the colour of the blend . a standard microlensing model fit to the @xmath27-band data shown in fig . [ fig : events ] ( solid line ) is best described with the time - scale of @xmath92 days , a maximum amplification of @xmath93 and the blending parameter @xmath94 . blending less than 1 indicates there is an additional light present in the overall flux of the object . the source magnitude can be calculated from the model as @xmath95 mag , but the colour can be only assumed to be equal to the blend s , i.e.,@xmath96 . even with these values and their uncertainties , the cmd location of the source can be linked to the lmc stars locus at the red giant branch , just above the red clump ( see fig . [ fig : cmd ] ) . if all additional light within a blend comes from the lens ( also hinted by zero - shift of the light centroid ) its brightness can be calculated , however because there were no @xmath31-band data during the event we could not derive its colour . the estimated location of the lens in this scenario is shown on fig . [ fig : cmdlenses ] . we note the large positive error on the blending parameter , which might indicate that the standard microlensing model is in fact not favoured for this event . its relatively bright baseline and small amplification may also suggest the brightness bump is caused by some kind of a variable star , for example a `` blue bumper '' blended with a red star to move its colour towards the red part of the cmd . moreover , the light curve indicates some small asymmetry around the peak , which can be reproduced with a microlensed binary source ( dashed line in fig.[fig : events ] ) with the goodness of fit @xmath97 . however , that model has very little constraints and the `` wiggle '' around the peak could be as well caused by instrumental inaccuracies . further photometric observations of this star are necessary as `` blue bumpers '' usually have secondary brightenings after a decade or so . this event was not detected in our regular search procedure described above , because its baseline magnitude ( 21.22 mag ) was well below our threshold ( 20.4 mag ) . it was found when the search for events was conducted with magnitude cut removed ( see table [ tab : conditions ] ) , which means it would pass through the pipeline if the magnitude cut was different . the microlensing model fit to its @xmath27- and @xmath31-band light curves ( shown in fig . [ fig : events ] ) yielded the time - scale of @xmath98 days and @xmath99 with blending parameters of @xmath100 and @xmath101 . the goodness of fit to multi - band data was @xmath102 , whereas when only @xmath27-band data were used it was @xmath103 . the single passband model returned also a relatively long time - scale of @xmath104 days with @xmath105 and @xmath106 . from the blended multi - colour models we were able to estimate the magnitude and colour of the source as @xmath107 mag and @xmath108 , which locates the source on the blue end of the main sequence of the lmc ( see fig . [ fig : cmd ] ) . severe blending indicates there is a lot of additional light within the seeing disk of the object with the event . judging from the sparse stellar density of the field lmc159 containing the event , we could assume all the remaining light belongs to the lens . astrometry of the centroid from the dia indicates no shift compared to the baseline template position . this also supports the assumption that there is no additional light taking part in the event except the source and the lens . this places the lens at @xmath109 mag and @xmath110 , which is far from the main lmc locus ( see fig . [ fig : cmdlenses ] ) . its cmd location in the vicinity of the two macho events caused by thick - disk lenses suggests this could be another example of a galactic lens towards the lmc . this , however , is relatively difficult to confirm with a follow - up using currently available instruments given the very faint magnitudes of both the source and the lens and large uncertainties of event s parameters . this candidate was also not detected by the regular search pipeline , but was visually found in the data when inspecting large signal to noise ratio light curves . its light curve exhibits asymmetric bump with amplitude of about 3 mag , but the peak is covered very sparsely by the ogle observations . however , single lens model does not reproduce the light curve . the simplest static binary source model is doing a much better job ( see fig . [ fig : events ] ) , however due to insufficient number of datapoints we can not exclude numerous possible binary lens models here . the binary source model gave time - scale of about 27 days and amplifications of 3 and 137 on each of the components with @xmath111 , compared to @xmath112 for a standard single source single lens model . the lack of a sensible model for the event prevents a derivation of the position of the lensed source on the colour - magnitude diagram . the colour and magnitude of the blend ( source and lens , plus any additional blending light ) , marked on fig . [ fig : cmd ] places it about 0.4 mag to the right of the red giant branch of the lmc . if this was the position of the source , it would suggest it does not belong to the lmc and is either in the foreground ( in the milky way halo ) or in the background behind the lmc . this candidate was already flagged by the eros group @xcite as anomalous . their data also indicate a slight reddening during the brightening of the event . this could hint towards some kind of nova eruption , however the light curve in its rising part is too smooth for a typical outburst - like object . taking all this into account we can not firmly conclude on the nature of this event . showing estimated positions of the lenses of the candidates for microlensing events found in the ogle - iii data towards the lmc . open circles mark the positions of the blended objects as seen at the baseline of each event . the error - bars for @xmath113 of the lens shown with dashed lines indicate that the colour of the lens could not be determined and was taken from the blend / baseline . , width=302 ] for completeness , we checked the previously known microlensing event candidates if their ogle - iii counterparts were exhibiting additional variation atypical for microlensing events . it was not necessary for the ogle - ii events as their light curves collected during the ogle - iii were already investigated in paper i , where it was confirmed they had constant baselines in years 2001 - 2009 . macho group @xcite have reported 17 candidate events and all of them were located within the ogle - iii fields . we confirm the constant baseline behaviour during the ogle - iii phase in most of them , except macho - lmc-7 and macho - lmc-23 . the secondary peak in event macho - lmc-23 we see in our data occurred at jd@xmath114 and was already reported by @xcite . this object was on our list of objects which were subject to cuts applied above , but its light curve was clearly asymmetrical and it was removed in the cut 8 with its blended model not converging and with large negative blending parameter . event macho - lmc-7 originally passed through improved selection criteria for macho events and was included in the optical depth determination of @xcite . the ogle - iii light curve of this object reveals three small amplitude bumps separated by @xmath902 years , all resembling microlensing curve with time - scales between 20 and 40 days and amplitudes around 0.5 mag . it clearly indicates this is not a genuine microlensing event but some sort of repeating outbursting variable star . it contributed about 10 per cent to the optical depth derived in @xcite , therefore their value should be now decreased to about @xmath115 . among the events reported by eros group @xcite we successfully cross - matched with the ogle - iii data only two candidates : eros - lmc-1 and eros - lmc-11 . the latter showed constant baseline over the course of the third phase of the ogle project , however the former exhibited a nice symmetric microlensing - like bump at jd@xmath116 and , as mentioned above , actually passed through our search pipeline . eros has already informed about the second peak they detected in that event occurring after nearly 5 years after the first one . the bump we detected happened after another 7 years , showing the underlying contaminant population of mysterious bumpers towards the lmc can exhibit a microlensing - like repeating episodes of brightening with random periodicity . the only possibility of ruling them out is to rely on as long time - baseline of observations as possible . the earth s atmosphere blurs stellar images . astronomical images taken with medium- and large - sized telescopes are almost never diffraction limited , unless active and/or adaptive optics is used . the earth s atmosphere prevents us from seeing much sharper and clearer images than we would otherwise see in its absence . our atmosphere affects crowded stellar fields the most , such as the magellanic clouds , as many stars merge together into composite objects ( `` blends '' ) that often can be no longer resolved by psf fitting . since the gravitational microlensing is equally likely to happen on any star in a blend , no matter if it is bright or faint , we would like to know how many stars contribute to one composite object , because in general the number of observed objects is not equal , and usually less , than the real number of monitored stars . this is of particular importance for the optical depth estimator , since at least three of its ingredients are directly affected by blending . the most obvious dependent quantity is the number of monitored stars . without correction for blending our estimate of the optical depth would be overestimated . another directly dependent quantity from the optical depth estimator is the detection efficiency of microlensing events . detecting a microlensing event on a bright source blended with fainter stars would yield almost 100 per cent detection efficiency . on the other hand , in case where a faint microlensed star is blended with a much brighter star the detection efficiency for such an event will be very small . also the time - scale of an event is affected by blending and appears shorter with blending ( see e.g. , @xcite ; @xcite ) . it is critically important to understand the blending and its effects on the optical depth estimator . the amount of blending can be estimated by comparing the ground - based images with much more detailed archival _ hst _ images and then grouping fields according to their stellar density and applying a fixed correction to the number of monitored stars for each density level . such a method was applied in paper i and paper ii for the ogle - ii data . however , the area covered in a single _ hst _ image is tiny , hence such comparison results in very low number statistics of cross - matched stars . moreover , ogle iii fields cover a very wide spectrum of stellar densities , which is difficult to cover with enough number of archival hst images . therefore to address the issue of blending here , we performed simulations of the ogle - iii images based on combined stellar luminosity functions derived from the _ hst _ and ogle images . based on known properties of the ogle - iii images , such as the point spread function s ( psf ) size and shape as well as the shape of the underlying stellar luminosity function , we are able to simulate ogle - iii images with different stellar densities and luminosity functions . we used the lmc photometry from _ the hst local group stellar photometry _ archive @xcite . three selected hst fields , lmc_u4b115 , lmc_u65008 , and lmc_u65007 , were each observed in f555v ( f814w ) filter with the wfpc2 for 2000 sec ( 2000 sec ) , 2560 sec ( 2460 sec ) , and 1560 sec ( 1560 sec ) seconds . these fields were calibrated to the standard @xmath27-band filter and were located in ogle s fields lmc162.6 ( dense ) , lmc119.2 ( medium density ) , and lmc136.1 ( sparse ) . then , the ogle and _ hst _ luminosity functions were combined with a stitching point around the red clump @xmath27-band magnitude , i.e.,@xmath117 mag . due to low number statistics for stars with @xmath118 mag we approximated the luminosity function with @xmath119 . each of the luminosity functions had a slightly different shape shown in fig . [ fig : lf ] , but , as show later , this affects the correction factor for the number of monitored stars by less than 5 per cent ( fig . [ fig : cf2 ] ) . , through panels b ) with @xmath120 , and c ) with @xmath121 , up to panel d ) with @xmath122.,width=321 ] before each simulation the basic parameters of the ogle - iii template image were measured . this included the psf shapes and sizes , and the background level . first , we created a mock image ( @xmath123 pixels ) with the background light and poisson noise that matched the original ogle images . next , we chose an _ hst _ stellar density as @xmath124 stars / arcmin@xmath11 , where @xmath125 is the simulation number , and @xmath126 to 15 . then , we injected stars that were drawn from the combined luminosity function ( top panel in fig . [ fig : lf ] ) and were brighter than @xmath127 mag . the faint end ( @xmath128 mag ) of the luminosity function had a minimal impact on our study ( see fig . [ fig : cf2 ] ) as our limiting search for microlensing events is @xmath129 mag ; the faint end simply serves as a background here . here , @xmath130 mag is the magnitude threshold for stars(blends ) we allow in our search for microlensing events , hence stars fainter by @xmath131 mag are the faintest which can contribute to a combined blended objects as sources for events . [ fig : simogle ] shows four examples of simulated ogle images with different stellar densities . once the image was created we obtained the stellar photometry in the way identical to the ogle - iii treatment of the template images using dophot photometry package @xcite . finally , we matched the input ( simulated _ hst _ ) and output ( simulated ogle ) catalogues of stars . after some experimentation , we find the matching radius of @xmath132 ogle - iii pixels for stars with @xmath133 mag and @xmath134 pixels , where @xmath135 is a size of the gaussian function at the level of @xmath130 mag , @xmath136 , for stars @xmath137 mag and psf size @xmath138 . the output catalogue was calibrated in such a way , that magnitudes of recovered single ( not blended ) stars were nearly identical to the injected ones , but also the number of flux counts on a simulated image for a given magnitude was nearly identical to real images . [ fig : cal ] shows a difference between recovered and injected magnitudes as a function of the injected magnitude . for each ogle star of magnitude @xmath139 within its radius @xmath140 we count all _ hst _ stars with brightness of @xmath141 mag . then , we constructed distributions of blending , where blending for each star is calculated as @xmath142 . the blending distributions for selected simulated density levels are shown in fig . [ fig : blending ] . for very dense , dense , medium and sparse , respectively . , width=321 ] -band magnitude as a function of injected magnitude for a dense field with @xmath143 . the two dashed lines show 2@xmath138 ogle - iii photometric uncertainties.,width=321 ] by comparing number of input _ hst _ stars and resulting number of ogle objects on simulated frames we were able to obtain a `` correction factor '' ( cf ) . the correction factor converts from the number of observed stars in an ogle - iii ccd chip to the number of real stars hidden ( unresolvable ) due to blending , which may act as a source in a microlensing event . figs . [ fig : cf ] and [ fig : cf2 ] show the cf as a function of the number of observed stars in a single ogle - iii ccd chip . plot in fig . [ fig : cf2 ] indicates that the correction factor only weakly changes ( with 5 per cent ) when different shapes of the luminosity functions are used ( see fig . [ fig : lf ] ) . we derived the number of stars correction for three magnitude bins , corresponding to the bins in which the blending distributions were derived : bright @xmath144 mag ( b ) , middle @xmath145 mag ( m ) , and faint @xmath146 mag ( f ) . then , for each ogle template object brighter than @xmath28 mag we calculated a number of corresponding unresolved stars , using the cf for the given field s density , encoded in a number of all objects found on a template image ( see table [ tab : fields ] ) . the resulting estimated number of real stars in each field ( combined for all 8 ccd chips ) is provided in table [ tab : fields ] . in total there were around 19.4 million objects on the ogle template images . these were estimated to be composed of about 22.7 million real stars , all of which could potentially be microlensed . in the bright star sample there were 5.8 million objects on the templates and 6.3 million estimated real stars . mag is described by function b ( bright ) , @xmath145 mag by function m ( middle ) , and the @xmath146 mag by function f ( faint ) . the function based on the entire magnitude range , @xmath147 is marked as a ( all).,width=321 ] above we reported a detection of 4 candidates for microlensing events . however , in order to calculate the optical depth towards the lmc we could only use events which were returned by our automated search pipeline , i.e.,events which satisfied well specified criteria and were possible to describe with a point - source - point - lens microlensing model . these were ogle - lmc-03 and ogle - lmc-04 . the remaining events were found manually and were either below the magnitude threshold ( ogle - lmc-05 ) or clearly not due to a single lens ( ogle - lmc-06 ) . the microlensing model of @xcite gives an amplification @xmath148 described with eq.[eq : a ] . the time - scale ( einstein radius cross time ) is the only parameter of the model which is related to some physical values , but is a degenerate function of the lens mass , distance , and relative velocity . for a number of microlensing events with time - scales @xmath149 and detection efficiency @xmath150 , the optical depth towards the lmc from the ogle - iii data is calculated from the formula : @xmath151 where @xmath152 is the time - span of all observations , @xmath153 is the total number of monitored stars , @xmath154 is the total number of events . the detection efficiency was derived separately for each field containing a microlensing event in the monte carlo simulations of the events covering the range of time - scales between 1 and 1000 days . figure [ fig : eff ] shows exemplary curves of the efficiency for all and bright stars samples and for dense and sparse fields . in these simulations all microlensing model s parameters except @xmath49 were randomly drawn from appropriate ranges : @xmath47 from @xmath155 to @xmath156 days ( the entire range of the ogle iii lmc coverage ) , @xmath53 from 0 to 1 , @xmath57 from the luminosity function of a given simulated field and @xmath157 from blending distribution derived for that field ( see section [ sec : blending ] and fig.[fig : blending ] ) . the optical depth was then calculated for @xmath158 events and their time - scales , with @xmath159 and @xmath160 days . following @xcite and papers i and ii we applied a correction for binary events for which our automated search pipeline is not sensitive to . it meant all the efficiencies were reduced down by 10 per cent . table [ tab : tau ] gathers the values of time - scales and detection efficiencies used for deriving the optical depth . for both events we found the optical depth of @xmath161 and @xmath0 for binary - corrected efficiency . cccc event & @xmath162 & @xmath163 & @xmath164 + + & & & + ogle - lmc-03 & 34.97@xmath50 & 0.202489 & 0.04 + & & & + ogle - lmc-04 & 32.76@xmath69 & 0.075691 & 0.11 + total @xmath165 & & & @xmath166 + + & & & + ogle - lmc-03 & 34.97@xmath50 & 0.182240 & 0.05 + & & & + ogle - lmc-04 & 32.76@xmath69 & 0.068122 & 0.11 + total @xmath165 & & & @xmath167 + if we allowed for non - single events contributing to the optical depth , hence included the candidate ogle - lmc-06 with its time - scale of around @xmath168 days and estimated efficiency of @xmath169 , it would add around @xmath170 to the total value of @xmath171 . in the other scenario , we could also speculate that ogle - lmc-04 is actually a `` blue bumper '' and has to be excluded from the @xmath171 measurements . that would reduce the value of the optical depth to @xmath172 or @xmath173 if ogle - lmc-06 was included as above . in the bright star sample there were the same two events as in the all stars sample . the efficiencies of their detection were about 1.4 times larger than efficiencies derived in the all stars sample . on the other hand , the number of monitored stars was smaller by a factor of around 3 . these two factors do not cancel out and the resulting optical depth for the bright star sample is @xmath174 , which is around 2 times larger than for all stars . however , given very low number of events and resulting large error bars on @xmath171 , comparison between these two values can not lead to any meaningful conclusions . the number of microlensing event candidates detected towards the lmc during the 8 years of the ogle - iii survey is remarkably small . the overall optical depth calculated using the two most reliable events is @xmath175 and is smaller than obtained with two events found in the ogle ii data ( @xmath176 , paper i ) . on the other hand , large systematic errors in both of these measurements only prove we are dealing with a very low statistics in terms of the numbers of events . this is in clear contradiction to a significantly larger number of events claimed by the macho collaboration ( @xcite , @xcite ) , but is in agreement with results obtained with the eros group data @xcite . can ogle events still be caused by the hypothetical machos ? according to the halo model of @xcite ( model `` s '' ) , if the galaxy halo was entirely filled with dark matter compact objects it would produce the optical depth of @xmath177 . because microlensing optical depth depends on the total mass of the lensing objects , the fraction of measured @xmath171 in @xmath178 gives directly a fraction of halo mass in compact objects . based on our measurement of the optical depth , if both microlensing events were due to machos this fraction would be @xmath179 per cent , with a typical mass around 0.2 @xmath9 . this mass can be estimated from the mean time - scale of the events found using this formula : @xmath180 ( @xcite , @xcite ) . the derived mass fraction is considerably smaller than @xmath181 per cent claimed by the macho group . because of the complexity of the microlensing events in general we are not able to definitively tell where the lenses and sources come from and what they are this is only possible in case of events exhibiting additional effects , e.g.,parallax or when the lens is a binary , or , in some special cases , after detailed high resolution imaging and spectroscopic follow - up . therefore , in case of our events we can only speculate about the nature of the events ; there are , however , certain hints towards a non - dark matter scenario . due to some line - of - sight depth of the lmc it is possible that stars of the cloud get microlensed by other lmc s stars located in front . the effect of `` self - lensing '' ( sl ) was estimated based on the internal structure studies of the lmc by @xcite and is the strongest for the lmc bar ( see fig . [ fig : mapevents ] ) . its contribution to the total optical depth is around @xmath182 in the central parts of the lmc and around @xmath183 when averaged over all ogle - iii fields . the lenses should be dark in case of macho lensing , whereas in self - lensing the lens can be any star from the lmc . such a luminous lens can be indirectly `` detected '' when microlensing model indicates some amount of light blended in with the source . for crowded fields such as those in the magellanic clouds observed with ground - based seeing , blending is a relatively common phenomenon . in section [ sec : blending ] we tackled this issue with scrutiny and concluded that a single ogle object is composed of about 1.2 real stars on average over the entire ogle iii lmc coverage . for the fields containing two our event candidates and for their respective magnitude range ( fig.[fig : cf ] ) these factors are even smaller than average with 1.09 and 1.16 for lmc122.1 and lmc163.6 , respectively . it means that there is less than about 15 per cent chance the blending detected in the microlensing model fit comes from random line - of - sight alignment of stars within a single seeing disk . there is yet another tool we have at hand to confirm this last statement further . photometry processing using the dia method provides also a precise position of the lensed flux with respect to the centroid of light measured on the template . the accuracy of such displacement measurement strongly depends on the magnitude at the peak and in the case of our events is at a level of 100 mas ( lensed images of the source are usually separated by less than 1 mas hence only centroid shift can be measured with such accuracy ) . because stars randomly blended together are usually displaced slightly from each other within the seeing disk , astrometric detection of a shift of the light barycentre provides an additional proof for `` natural '' random blending . in the case of the two events presented here there was clearly no shift visible in astrometry of the residuals on the dia subtracted images . unless we are very `` lucky '' and the blended star happen to be sitting exactly on the source of the event , this can be attributed to the fact that all the blending light comes from the lensing object . hence , the lenses must be luminous . their locations on the cmd , calculated from the blended microlensing model , indicate they belong to the lmc star locus . this clearly hints towards the self - lensing nature of the detected events . the two events found manually are also difficult to attribute to macho lensing . ogle - lmc-05 is likely to be caused by a very red lens , probably located in the galaxy disk . ogle - lmc-06 is generally very puzzling , with its light curve resembling a binary lens / source event and its very red colour , suggesting rather a variable star than a microlensing event . it means we have no strong candidates for microlensing events caused by lenses from the halo . we can , however , put an upper limit on the macho presence in the halo , similarly as done in @xcite , who detected no events in their bright star sample . in our case , however , we have two events we associated with origin other than macho lensing . this imply we can not apply straightforward zero - detection poisson statistics , but should follow the suggestion of @xcite , also applied in paper ii and treat our sl candidate events as an expected background . this is an obvious estimate , as more detailed studies involving lmc modelling are necessary in order to obtain exact amount of expected self - lensing events over the entire ogle iii lmc sky coverage . such analysis is planned to be performed in a way similar to the ogle ii lmc study in @xcite . based on a mean detection efficiency over all fields we were able to estimate the number of expected events due to machos considering model `` s '' of @xcite . this number was calculated for a wide range of macho masses from @xmath184 to @xmath185 and was translated to a fraction of halo mass using zero - statistics of @xcite . it is shown in fig . [ fig : taulimit ] . for masses around @xmath186 we expected @xmath187 events for a halo full of machos , it translates to an upper limit of @xmath3 per cent at 95 per cent confidence and @xmath188 per cent at 90 per cent . the limit reaches its minimum at around @xmath5 with @xmath4 per cent and is less rigid on masses higher than @xmath189 reaching around 20 per cent at @xmath6 and more at higher masses . our result is in agreement not only with previous lmc microlensing findings @xcite , but also with studies of the microlensing effects of compact objects in distant galaxies observed in the lensed quasars ( e.g. @xcite ) , which ruled out machos in mass range between 0.1 and 10 @xmath9 . for the mass window @xmath190 there is still no reasonable constraint . there is actually some hint for heavy mass ( around 10@xmath9 ) compact objects in the halo - @xcite studied the ogle-2005-smc-001 microlensing event and concluded it was caused by a binary black hole most likely located in the halo . however , it still remains a mystery why we do nt see such events towards the lmc , therefore it is too early to conclude on dark matter compact objects existing in that mass window . iii smc data , including that unique event , will be presented and studied in the forthcoming paper ( wyrzykowski et al . in prep . ) . in this study we analysed almost 8 years of observations of the large magellanic cloud by ogle iii . the data set with its volume , coverage and quality supersedes all previous determinations of the microlensing optical depth , including the one based on the ogle ii data ( paper i ) . we detected two sound candidates for microlensing events and further two possible candidates . neither of them , however , is likely to be caused by dark matter compact lenses from the halo of our galaxy . the two best candidates can be explained as an expected signal from the self - lensing within the lmc . of the remaining two , one is either a binary event or a some kind of chromatic outburst , whereas the other is a candidate for galactic disk lens . such null detection for macho lensing led to estimating the upper limit on their contribution to the mass of the halo of the galaxy . the upper limit set at a level of 6 - 7 per cent at @xmath186 leaves very little room for dark matter compact objects . still , at the moment we can not exclude more heavy dark matter lenses , like black holes . our survey puts a 20 per cent halo mass fraction limit on compact objects with masses of @xmath6 and actually no limit on higher masses . this heavy mass end window should be now explored with more attention . as a side product of our analysis we also discovered that event macho - lmc-7 , reported by the macho group and used in their final optical depth determination , exhibited couple of additional brightening episodes in the ogle - iii data , a feature which excludes it as a genuine microlensing event . with the ogle project continuing now in its fourth phase we hope the sensitivity to extremely long events will improve significantly within next years . it should result in the increase in the statistics of potential black - hole lenses or allow us to rule out heavy dark matter compact objects as well and close that topic definitively . we would like to thank for their help at various stages of this work to drs nicholas rattenbury , vasily belokurov and patrick tisserand . we also thank the anonymous referee for their invaluable comments and remarks . this work was partially supported by ec fr7 grant perg04-ga-2008 - 234784 to w . js acknowledges support through the polish mnisw grant no . n20300832/0709 and space exploration research fund of the ohio state university . the ogle project acknowledges funding received from the european research council under the european community s seventh framework programme ( fp7/2007 - 2013 ) , erc grant agreement no . 246678 . n. , nguyen h. t. , werner m. w. , alcock c. , patten b. m. , stern d. , 2004 , in bulletin of the american astronomical society vol . 36 of bulletin of the american astronomical society , spitzer space telescope observations of the aftermath of microlensing event macho - lmc-5 . |
the paper analyzes and compares some spectral filtering methods as truncated singular / eigen - value decompositions and tikhonov / re - blurring regularizations in the case of the recently proposed reflective and anti - reflective boundary conditions . we give numerical evidence to the fact that spectral decompositions ( sds ) provide a good image restoration quality and this is true in particular for the anti - reflective sd , despite the loss of orthogonality in the associated transform . the related computational cost is comparable with previously known spectral decompositions , and results substantially lower than the singular value decomposition . the model extension to the cross - channel blurring phenomenon of color images is also considered and the related spectral filtering methods are suitably adapted . filtering methods , spectral decompositions , boundary conditions . 65f10 , 65f15 , 65y20 | 0710.5272 | in this paper we deal with the classical image restoration problem of blurred and noisy images in the case of a space invariant blurring . under such assumption the image formation process is modelled according to the following integral equation with space invariant kernel @xmath0 where @xmath1 denotes the true physical object to be restored , @xmath2 is the recorded blurred and noisy image , @xmath3 takes into account unknown errors in the collected data , e.g. measurement errors and noise.as customary , we consider the discretization of ( [ eq : modello_continuo ] ) by means of a standard @xmath4 generalization of the rectangle quadrature formula on an equispaced grid , ordered row - wise from the top - left corner to the bottom - right one . hence , we obtain the relations @xmath5 in which an infinite and a shift - invariant matrix @xmath6_{(i , j)=((i_1,i_2),(j_1,j_2))}$ ] , i.e. , a two - level toeplitz matrix , is involved.in principle , ( [ eq : modello_discreto_base ] ) presents an infinite summation since the true image scene does not have a finite boundary . nevertheless , the data @xmath7 are clearly collected only at a finite number of values , so representing only a finite region of such an infinite scene . in addition , the blurring operator typically shows a finite support , so that it is completely described by a point spread function ( psf ) mask such as @xmath8_{i_1=-q_1,\ldots , q_1,i_2=-q_2,\ldots , q_2}\ ] ] where @xmath9 for any @xmath10 and @xmath11 , @xmath12 , @xmath13 ( normalization according to a suitable conservation law).therefore , relations ( [ eq : modello_discreto_base ] ) imply @xmath14 where the range of collected data defines the so called field of view ( fov).once again , we are assuming that all the involved data in ( [ eq : modello_discreto_sistema ] ) , similarly to ( [ eq : modello_discreto_base ] ) , are reshaped in a row - wise ordering . in such a way we obtain the linear system @xmath15 where @xmath16 is a finite principal sub - matrix of @xmath17 , with main diagonal containing @xmath18 , @xmath19 , @xmath20 and with @xmath21 , for any two - index @xmath22 . such a reshape is considered just to perform the theoretical analysis , since all the deblurring / denoising methods are able to deal directly with data in matrix form . for instance , it is evident that the blurring process in ( [ eq : modello_discreto ] ) consists in a discrete convolution between the psf mask , after a rotation of @xmath23 , and the proper true image data in @xmath24_{i_1=-q_1 + 1,\ldots , n_1+q_1,i_2=-q_2 + 1,\ldots , n_2+q_2}.\ ] ] hereafter , with a two - index notation , we denote by @xmath25_{i_1=1,\ldots , n_1,i_2=1,\ldots , n_2}$ ] the true image inside the fov and by @xmath26_{i_1=1,\ldots , n_1,i_2=1,\ldots , n_2}$ ] the recorded image . thus , assuming the knowledge of psf mask in ( [ eq : coefficienti_psf ] ) and of some statistical properties of @xmath3 , the deblurring problem is defined as to restore , as best as possible , the true image @xmath27 on the basis of the recorded image @xmath28 . as evident from ( [ eq : modello_discreto ] ) , the problem is undetermined since the number of unknowns involved in the convolution exceeds the number of recorded data . boundary conditions ( bcs ) are introduced to artificially describe the scene outside the fov : the values of unknowns outside the fov are fixed or are defined as linear combinations of the unknowns inside the fov , the target being to reduce ( [ eq : modello_discreto_sistema ] ) into a square linear system @xmath29 with @xmath30 , @xmath31 , @xmath32 and @xmath33.the choice of the bcs does not affect the global spectral behavior of the matrix . however , it may have a valuable impact both with respect to the accuracy of the restored image and to the computational costs for recovering @xmath1 from the blurred datum , with or without noise . notice also that , typically , the matrix @xmath34 is very ill - conditioned and there is a significant intersection between the subspace related to small eigen / singular values and the high frequency subspace . such a feature requires the use of suitable regularization methods that allow to properly restore the image @xmath27 with controlled noise levels @xcite , among which we can cite truncated svd , tikhonov , and total variation @xcite.hereafter , we focus our attention on special case of psfs satisfying a strong symmetry property , i.e. , such that @xmath35 this assumption is fulfilled in the majority of models in real optical applications . for instance , in most 2d astronomical imaging with optical lens @xcite the model of the psf is circularly symmetric , and hence , strongly symmetric ; in the multi - image deconvolution of some recent interferometric telescopes , the psf is strongly symmetric too @xcite . moreover , in real applications when the psf is obtained by measurements ( like a guide star in astronomy ) , the influence of noise leads to a numerically nonsymmetric psf , also when the kernel of the psf is strongly ( or centro ) symmetric . in such a case , by employing a symmetrized version of the measured psf , comparable restorations are observed @xcite . the paper is organized as follows . in section [ sez : bc ] we focus on two recently proposed bcs , i.e. , the reflective @xcite and anti - reflective bcs @xcite and their relevant properties . section [ sez : filtering_methods ] summarizes some classical filtering techniques as the truncated singular / eigen - values decomposition and the tikhonov method . the re - blurring method @xcite is considered in the case of anti - reflective bcs and its re - interpretation in the framework of the classical tikhonov regularization is given . in section [ sez : cross_channel_blurring ] the model is generalized for taking into account the cross - channel blurring phenomenon and the previous filtering methods are suitable adapted . lastly , section [ sez : numerical_tests ] deals with some computational issues and reports several numerical tests , the aim being to compare the quoted filtering methods and the two type of bcs , both in the case of gray - scale and color images . in section [ sez : conclusioni ] some conclusions and remarks end the paper . in this section we summarize the relevant properties of two recently proposed type of bcs , i.e. , the reflective @xcite and anti - reflective bcs @xcite . special attention is given to the structural and spectral properties of the arising matrices . in fact , though the choice of the bcs does not affect the global spectral behavior of the matrix @xmath34 , it can have a valuable impact with respect both to the accuracy of the restoration ( especially close to the boundaries where ringing effects can appear ) , and the computational costs for recovering the image from the blurred one , with or without noise . moreover , tanking into account the scale of the problem , the regularization methods analysis can be greatly simplified whenever a spectral ( or singular value ) decomposition of @xmath34 is easily available . this means that the target is to obtain the best possible approximation properties , keeping unaltered the fact that the arising matrix shows an exploitable structure . for instance , the use of periodic bcs enforces a circulant structure , so that the spectral decomposition can be computed efficiently with the fast fourier transform ( fft ) @xcite . despite these computational facilities , they give rise to significant ringing effects when a significant discontinuity is introduced into the image . hereafter , we focus on two recently proposed boundary conditions , that more carefully describe the scene outside the fov.clearly , several other methods deal with this topic in the image processing literature , e.g. local mean value @xcite or extrapolation techniques ( see @xcite and references therein ) . nevertheless , the penalty of their good approximation properties could lie in a linear algebra problem more difficult to cope with . in @xcite ng _ et al . _ analyze the use of reflective bcs , both from model and linear algebra point of view . the improvement with respect to periodic bcs is due to the preservation of the continuity of the image . in fact , the scene outside the fov is assumed to be a reflection of the scene inside the fov . for example , with a boundary at @xmath36 and @xmath37 the reflective condition is given by @xmath38.more precisely , along the borders , the bcs impose @xmath39 and , at the corners , the bcs impose for any @xmath40 , @xmath41 @xmath42 i.e. , a double reflection , first with respect to one axis and after with respect to the other , no matter about the order.as a consequence the rectangular matrix @xmath43 is reduced to a square toeplitz - plus - hankel block matrix with toeplitz - plus - hankel blocks , i.e. , @xmath44 shows the two - level toeplitz - plus - hankel structure . moreover , if the blurring operator satisfies the strong symmetry condition ( [ eq : strong_simmetry_condition ] ) then the matrix @xmath44 belongs to dct - iii matrix algebra . therefore , its spectral decomposition can be computed very efficiently using the fast discrete cosine transform ( dct - iii ) @xcite.more in detail , let @xmath45 be the two - level dct - iii matrix algebra , i.e. , the algebra of matrices that are simultaneously diagonalized by the orthogonal transform @xmath46_{s , t=1}^m,\ ] ] with @xmath47 denoting the kronecker symbol.thus , the explicit structure of the matrix is @xmath48 , with @xmath49 $ ] and where each @xmath50 , @xmath51 is the unilevel dct - iii matrix associated to the @xmath52 row of the psf mask , i.e. , @xmath53 , with @xmath54 $ ] . here , we denote by @xmath55 the shift operator such that @xmath56 $ ] and by @xmath57 the usual flip matrix ; at the block level the same operations are intended in block - wise sense.beside this structural characterization , the spectral description is completely known . in fact , let @xmath1 be the bivariate generating function associated to the psf mask ( [ eq : coefficienti_psf ] ) , that is @xmath58 then the eigenvalues of the corresponding matrix @xmath59 are given by @xmath60},x_{s_2}^{[n_2]}\right),\ s=(s_1,s_2 ) , \quad x_r^{[m]}=\frac{(r-1)\pi}{m},\ ] ] where @xmath61 , @xmath62 , and where the two - index notation highlights the tensorial structure of the corresponding eigenvectors.lastly , notice that standard operations like matrix - vector products , resolution of linear systems and eigenvalues evaluations can be performed by means of fct - iii @xcite within @xmath63 arithmetic operations ( ops ) . for example , by multiplying by @xmath64^t$ ] both the sides of @xmath65 , it holds that @xmath66_{(i_1,i_2)}= [ r_n^t ( a_n e_1)]_{(i_1,i_2 ) } / [ r_n^t e_1]_{(i_1,i_2 ) } , \quad i_1=1,\ldots , n_1 , i_2=1,\ldots , n_2,\ ] ] i.e. , it is enough to consider an inverse fct - iii applied to the first column of @xmath44 , with a computational cost of @xmath67 ops . more recently , anti - reflective boundary conditions ( ar - bcs ) have been proposed in @xcite and studied @xcite . the improvement is due to the fact that not only the continuity of the image , but also of the normal derivative , are guaranteed at the boundary . this regularity , which is not shared with dirichlet or periodic bcs , and only partially shared with reflective bcs , significantly reduces typical ringing artifacts.the key idea is simply to assume that the scene outside the fov is the anti - reflection of the scene inside the fov . for example , with a boundary at @xmath36 the anti - reflexive condition impose @xmath68 where @xmath69 is the center of the one - dimensional anti - reflection , i.e. , @xmath70 in order to preserve a tensorial structure , at the corners , a double anti - reflection , first with respect to one axis and after with respect to the other , is considered , so that the bcs impose @xmath71 where @xmath72 is the center of the two - dimensional anti-reflection.more precisely , by choosing as center of the anti - reflection the first available data , along the borders , the bcs impose @xmath73 at the corners , the bcs impose for any @xmath40 and @xmath41 , @xmath74 as a consequence the rectangular matrix @xmath75 is reduced to a square toeplitz - plus - hankel block matrix with toeplitz - plus - hankel blocks , plus an additional structured low rank matrix . moreover , under the assumption of strong symmetry of the psf and of a mild finite support condition ( more precisely @xmath76 if @xmath77 , for some @xmath78 ) , the resulting linear system @xmath79 is such that @xmath44 belongs to the @xmath80 commutative matrix algebra @xcite . this new algebra shares some properties with the @xmath81 ( or dst - i ) algebra @xcite.going inside the definition , a matrix @xmath82 has the following block structure @xmath83 } & 0^t & 0 \\ \hline d_{1}+z^{[2 ] } & & 0 \\ \vdots & & \vdots \\ d_{q_1 - 1}+z^{[q_1 ] } & & 0 \\ d_{q_1 } & \ \tau(d_0,\ldots , d_{q_1 } ) \ & d_{q_1}\\ 0 & & d_{q_1 - 1}+z^{[q_1 ] } \\ \vdots & & \vdots \\ 0 & & d_{1}+z^{[2]}\\ \hline 0 & 0^t & d_{0}+z^{[1 ] } \\ \end{array } \right],\ ] ] where @xmath84 is a block @xmath81 matrix with respect to the @xmath85 blocks @xmath86 , @xmath51 and @xmath87}= 2\sum_{t = k}^{q_1 } d_{t}$ ] for @xmath88 . in particular , the @xmath85 block @xmath86 is associated to @xmath52 row of the psf , i.e. , @xmath89}=[h_{i_1,i_2}]_{i_2=-q_2,\ldots , q_2}$ ] and it is defined as @xmath90 } & 0^t & 0 \\ \hline h_{i_1,1}+z_{i_1}^{[2 ] } & & 0 \\ \vdots & & \vdots \\ h_{i_1,q_2 - 1}+z_{i_1}^{[q_2 ] } & & 0 \\ h_{i_1,q_2 } & \ \tau(h_{i_1,0},\ldots , h_{i_1,q_2})\ & h_{i_1,q_2 } \\ 0 & & h_{i_1,q_2 - 1}+z_{i_1}^{[q_2 ] } \\ \vdots & & \vdots\\ 0 & & h_{i_1,1}+z_{i_1}^{[2]}\\ \hline 0 & 0^t & h_{i_1,0}+z_{i_1}^{[1 ] } \\ \end{array } \right],\ ] ] where @xmath91}= 2\sum_{t = k}^{q_2 } h_{i_1,t}$ ] for @xmath92 and @xmath93 is the unilevel @xmath81 matrix associated to the one - dimensional psf @xmath89}$ ] previously defined.notice that the rank-1 correction given by the elements @xmath91}$ ] pertains to the contribution of the anti - reflection centers with respect to the vertical borders , while the low rank correction given by the matrices @xmath94}$ ] pertains to the contribution of the anti - reflection centers with respect to the horizontal borders.it is evident from the above matrix structure that favorable computational properties are guaranteed also by virtue of the @xmath81 structure . therefore , firstly we recall the relevant properties of the two - level @xmath81 algebra @xcite.let @xmath95 be the two - level @xmath81 matrix algebra , i.e. , the algebra of matrices that are simultaneously diagonalized by the symmetric orthogonal transform @xmath96_{s , t=1}^m.\ ] ] with the same notation as the dct - iii algebra case , the explicit structure of the matrix is two level toeplitz - plus - hankel . more precisely , @xmath97 with @xmath49 $ ] , where each @xmath50 , @xmath51 is a the unilevel @xmath81 matrix associated to the @xmath52 row of the psf mask , i.e. , @xmath98 with @xmath54 $ ] . here , we denote by @xmath99 the double shift operator such that @xmath100 $ ] ; at the block level the same operations are intended in block - wise sense.once more , the spectral characterization is completely known since for any @xmath101 the related eigenvalues are given by @xmath60},x_{s_2}^{[n_2]}\right ) , s=(s_1,s_2 ) , \quad x_r^{[m]}=\frac{r \pi}{m+1},\ ] ] where @xmath61 , @xmath62 , and @xmath1 is the bivariate generating function associated to the @xmath102 defined in ( [ eq : funzione_generatrice ] ) . as in the dct - iii case , standard operations like matrix - vector products , resolution of linear systems and eigenvalues evaluations can be performed by means of fst - i within @xmath103 ( ops ) . for instance , it is enough to consider a fst - i applied to the first column of @xmath44 to obtain the eigenvalues @xmath66_{(i_1,i_2)}= [ q_n ( a_n e_1)]_{(i_1,i_2 ) } / [ q_n e_1]_{(i_1,i_2 ) } , \quad i_1=1,\ldots , n_1 , i_2=1,\ldots , n_2.\ ] ] now , with respect to the @xmath80 matrix algebra , a complete spectral characterization is given in @xcite . a really useful fact is the existence of a transform @xmath104 that simultaneously diagonalizes all the matrices belonging to @xmath80 , although the orthogonality property is partially lost . @xcite any matrix @xmath82 , @xmath31 , can be diagonalized by @xmath104 , i.e. , @xmath105 where @xmath106 , @xmath107 , with @xmath108 \quad and \quad \widetilde{t}_m=\left [ \begin{array}{ccc } \alpha_m & 0^t & 0\\ \ \\ -q_{m-2}p & q_{m-2 } & -q_{m-2}jp \\ \ \\ 0 & 0^t & \alpha_m \end{array } \right ] \ ] ] the entries of the vector @xmath109 are defined as @xmath110 , @xmath111 , @xmath112 is the flip matrix , and @xmath113 is a normalizing factor chosen such that the euclidean norm of the first and last column of @xmath114 will be equal to @xmath115 . @xcite [ teo : teo_autovalori ] let @xmath82 , @xmath31 , the matrix related to the psf @xmath116_{i_1=-q_1,\ldots , q_1 , i_2=-q_2,\ldots , q_2}$ ] . then , the eigenvalues of @xmath44 are given by * @xmath115 with algebraic multiplicity @xmath117 , * the @xmath118 eigenvalues of the unilevel @xmath81 matrix related to the one - dimensional psf @xmath119 $ ] , each one with algebraic multiplicity @xmath120 , * the @xmath121 eigenvalues of the unilevel @xmath81 matrix related to the one - dimensional psf @xmath122 $ ] , each one with algebraic multiplicity @xmath120 , * the @xmath123 eigenvalues of the two - level @xmath81 matrix related to the two - dimensional psf @xmath124 . notice that the three sets of multiple eigenvalues are exactly related to the type of low rank correction imposed by the bcs through the centers of the anti - reflections . more in detail , the eigenvalues of @xmath125 and of @xmath126 take into account the condensed psf information considered along the horizontal and vertical borders respectively , while the eigenvalue equal to @xmath115 takes into account the condensed information of the whole psf at the four corners.in addition , it is worth noticing that the spectral characterization can be completely described in terms of the generating function associated to the @xmath102 defined in ( [ eq : funzione_generatrice ] ) , simply by extending to @xmath127 the standard @xmath81 evaluation grid , i.e. , it holds @xmath128},x_{s_2}^{[n_2]}\right ) , s=(s_1,s_2 ) , s_j=0,\ldots , n_j , \quad x_r^{[m]}=\frac{r \pi}{m+1},\ ] ] where the @xmath129index refers to the first / last columns of the matrix @xmath114 @xcite.see @xcite for some algorithms related to standard operations like matrix - vector products , resolution of linear systems and eigenvalues evaluations with a computational cost of @xmath130 ops.it is worthwhile stressing that the computational cost of the inverse transform is comparable with that of the direct transform and , at least at first sight , the very true penalty is the loss of orthogonality due to the first / last column of the matrix @xmath114 . owing to the ill - conditioning , the standard solution @xmath131 is not physically meaningful since it is completely corrupted by the noise propagation from data to solution , i.e. , by the so called inverted noise . for this reason , restoration methods look for an approximate solution with controlled noise levels : widely considered regularization methods are obtained through spectral filtering @xcite . hereafter , we consider the truncated singular values decompositions ( svds ) ( or spectral decompositions ( sds ) ) and the tikhonov ( or re - blurring ) regularization method . the singular values decomposition ( svd ) highlights a standard perspective for dealing with the inverted noise . more precisely , if @xmath132 is the svd of @xmath44 , i.e. , @xmath133 and @xmath134 are orthogonal matrices and @xmath135 is a diagonal matrix with entries @xmath136 , then the solution of the linear system @xmath137 can be written as @xmath138 where @xmath139 and @xmath140 denote the @xmath141 column of the matrix @xmath133 and @xmath134 , respectively.with regard to the image restoration problem , the idea is to consider a sharp filter , i.e. , to take in the summation only the terms corresponding to singular values greater than a certain threshold value @xmath142 , so damping the effects caused by division by the small singular values . therefore , by setting the filter factors as @xmath143 the filtered solution is defined as @xmath144 due to scale of the problem , the svd of the matrix @xmath44 is in general an expensive computational task ( and not negligible also in the case of a separable psf ) . thus , an `` a priori '' known spectral decomposition , whenever available , can give rise to a valuable simplification . more precisely , let @xmath145 be a spectral decomposition of @xmath44 , then the filtered solution is defined as @xmath146 where @xmath140 and @xmath147 denote the @xmath141 column of @xmath134 and the @xmath141 row of @xmath148 , respectively , and where @xmath149 if @xmath150 , @xmath127 otherwise . in the classical tikhonov regularization method , the image filtering is obtained by looking for the solution of the following minimization problem @xmath151 where @xmath152 is the regularization parameter and @xmath153 is a carefully chosen matrix ( typically @xmath154 or represents the discretization of a differential operator , properly adapted with respect to the chosen bcs).the target is to minimize the euclidean norm of the residual @xmath155 without explosions with respect to the quantity @xmath156 . as well know , ( [ eq : min_tikhnonov ] ) is equivalent to the solution to the damped least square problem @xmath157 in addition , the regularization tikhonov method can be reinterpreted in the framework of classical spectral filtering method . for instance , in the case of @xmath154 , by making use of the svd of @xmath158 , the solution of ( [ eq : dampednes ] ) can be rewritten as @xmath159 where @xmath160 with @xmath161 , @xmath162.a severe drawback in adopting the tikhonov regularization approach in the case of @xmath82 is due to the fact that @xmath163 , so that all the favorable computational properties are substantially spoiled . an alternative approach , named re - blurring , has been proposed in @xcite : the proposal is to replace @xmath164 by @xmath165 in ( [ eq : dampednes ] ) , where @xmath165 is the blurring matrix related to the current bcs with a psf rotated by @xmath166 . this approach is completely equivalent to ( [ eq : dampednes ] ) in the case of dirichlet and periodic bcs , while the novelty concerns both reflective bcs and anti - reflective bcs , where in general @xmath167 . the authors show that the re - blurring with anti - reflective bcs is computationally convenient and leads to a larger reduction of the ringing effects arising in classical deblurring schemes . from the modelling point of view , the authors motivation relies upon the fact that re - blurring smoothes the noise in the right hand side of the system , in the same manner as this happens in the case of dirichlet , periodic and reflective bcs . hereafter , we consider an explanation of the observed approximation results . as previously claimed , we focus our attention on the case of a strongly symmetric psf , so that the matrix @xmath165 equals the matrix @xmath44 . moreover , also in this case it is evident that the linear system @xmath168 is not equivalent to a minimization problem , again because the matrix @xmath169 is not symmetric . nevertheless , the symmetrization of ( [ eq : damped_reblurring ] ) can be performed by diagonalization , so obtaining @xmath170 where @xmath171 and @xmath172 . in such a way ( [ eq : damped_reblurring_diag ] ) is again equivalent to the minimization problem @xmath173 or equivalently , again by making use of the diagonalization result , to @xmath174 clearly , the last formulation in ( [ eq : min_reblurring_2 ] ) is the most natural and it allows to claim that the re - blurring method can be interpreted as a standard tikhonov regularization method in the space transformed by means of @xmath175 . recalling that @xmath175 is not an orthogonal transformation , the goal becomes to compare @xmath176 and @xmath177 , that is to bound @xmath178 , being @xmath179 . a quite sharp estimate of such a norm can be found by exploiting the structure of the unilevel matrix @xmath180 . let @xmath181 $ ] , it holds that @xmath182 being @xmath183 . since , by definition , @xmath184 , we have @xmath185 notice that the bound given in ( [ eq : noma2_ttilde ] ) is quite sharp , since for instance @xmath186 equals @xmath187 . hereafter , we extend the analysis of the deblurring problem to the case of color images digitalized , for instance , according to the standard rgb system . several techniques can be used for recording color images , but the main problem concerns the fact that light from one color channel can end up on a pixel assigned to another color . the consequence of this phenomenon is called cross - channel blurring among the three channels of the image and it sums up to the previously analyzed blurring of each one of the three colors , named within - channel blurring.by assuming that the cross - channel blurring takes place after the within - channel blurring of the image , that it is spatially invariant and by assuming that the same within - channel blurring occurs in all the three color channels , the problem can be modelled @xcite as @xmath188 with @xmath30 , @xmath31 , @xmath32 , and @xmath189.\ ] ] the row - entries denote the amount of within - channel blurring pertaining to each color channel ; a normalized conservation law prescribes that @xmath190 , @xmath191^t$ ] . lastly , the vectors @xmath192 are assumed to collect the three color channels in the rgb order.clearly , if @xmath193 , i.e. , the blurring is only of within - channel type , the problem is simply decoupled into three independent gray - scale deblurring problems.in the general case , taking into account the tensorial structure of the whole blurring matrix @xmath194 is evident that the truncated svds and sds can be formulated as the natural extension of those considered in the within - blurring case . notice that in the case of sds , we will consider a svd for the matrix @xmath195 , since it naturally assures an orthogonal decomposition , no matter about the specific matrix , while its computational cost is negligible with respect to the scale of the problem . in addition , we tune the filtering strategy with respect the spectral information given only by the matrix @xmath44 , i.e. , for any fixed @xmath196 ( or @xmath197 ) we simultaneously sum , or discard , the three contribution on @xmath1 related to the three singular values of @xmath195.with respect to the tikhonov regularization method , the approach is a bit more involved . under the assumption @xmath198 , the damped least square problem @xmath199 f = ( a_{\mathrm{color}}\otimes a_n)^t g\ ] ] can be rewritten as @xmath200 f = ( a_{\mathrm{color } } \otimes v_n \lambda_n \widetilde v_n)^t g. \label{eq : damped_color}\ ] ] thus , by setting @xmath201 , @xmath202 , @xmath203 , ( [ eq : damped_color ] ) can be transformed in @xmath204 s_{3n}^{-1}\hat f = s_{3n } ( a_{\mathrm{color}}\otimes v_n \lambda_n \widetilde v_n)^t s_{3n}^{-1}\hat g , \ ] ] so obtaining the linear system @xmath205 \hat f = ( a_{\mathrm{color}}^t \otimes \lambda_n ) \hat g , \ ] ] that can easily be decoupled into @xmath206 linear systems of dimension @xmath207.clearly , in the case of any matrix @xmath59 , all these manipulations can be performed by means of an orthogonal transformation @xmath208 . notice also that the computational cost is always @xmath209 ops.with respect to @xmath210 , we can consider the same strategy by referring to the re - blurring regularization method . more precisely , the linear system @xmath211 f = ( a_{\mathrm{color}}^t \otimes a_n ) g \ ] ] can be transformed in @xmath205 \hat f = ( a_{\mathrm{color}}^t \otimes \lambda_n ) \hat g. \ ] ] though the transformation @xmath212 is not orthogonal as in the reflective case , the obtained restored image are fully comparable with the previous ones and the computational cost is still @xmath213 ops . before analyzing the image restoration results , we discuss how the methods can work without reshaping the involved data . in fact , the tensorial structure of the matrices , obtained by considering reflective and anti - reflective bcs , can be exploited in depth , so that the algorithms can deal directly , and more naturally , with the data collected in matrix form . hereafter , we consider a two - index notation in the sense of the previously adopted row - wise ordering.in the sd case considered in section [ sez : filtering_methods_tsvd_tsd ] , since @xmath214 } \otimes \tilde v_{k_2}^{[n_2]}$ ] is represented in matrix form as @xmath215})^t \tilde v_{k_2}^{[n_2]}$ ] , the required scalar product can be computed as @xmath216}\right)^t \tilde v_{k_2}^{[n_2]}\right ] \odot g,\ ] ] where @xmath217 denotes the summation of all the involved terms after a element - wise product . clearly , @xmath218 } \otimes v_{k_2}^{[n_2]}$ ] is represented in matrix form as @xmath219 } ( \tilde v_{k_2}^{[n_2]})^t$ ] . in a similar manner , in the case of the svd of @xmath44 with separable psf @xmath220 , we can represent @xmath218 } \otimes v_{k_2}^{[n_2]}$ ] in matrix form as @xmath219 } ( v_{k_2}^{[n_2]})^t$ ] and @xmath221 } \otimes u_{k_2}^{[n_2]})^t$ ] as @xmath222 } ( u_{k_2}^{[n_2]})^t$].the eigenvalues required for the sd can be stored into a matrix @xmath223 . in the case of @xmath224 this matrix can be evaluated as @xmath225 where @xmath226 denotes the first column of @xmath44 and @xmath227 the first canonical basis vector , reshaped as matrices in column - wise order . in addition , the two - level direct and inverse transform @xmath228 and @xmath229 can be directly evaluated on a matrix data as @xmath230 by referring to the corresponding unilevel transforms.in the same way , the eigenvalues required in the case of @xmath231 can be suitably stored as @xmath232 \in \mathbb{r}^{n_1\times n_2},\ ] ] with reference to the notations of theorem [ teo : teo_autovalori ] , where the eigenvalues of the unilevel and two - level @xmath81 matrices are evaluated as outlined in section [ sez : bc - ar].lastly , the linear systems obtained , for any fixed @xmath233 , in the case of tikhonov and re - blurring regularization methods can be solved with reference to the matrix @xmath234 of the corresponding filter factors by applying the reflective and anti - reflective transforms with a computational cost @xmath235 ops . in this section we compare the effectiveness of truncated spectral decompositions ( sds ) with respect to the standard truncated svds both in the case of reflective and anti - reflective bcs . due to scale of the problem , the svd of the matrix @xmath44 is in general an expensive computational task ( and not negligible also in the case of a separable psf ) . thus , a spectral decomposition , whenever available as in these cases , leads to a valuable simplification.firstly , we consider the case of the separable psf caused by atmospheric turbulence @xmath236 where @xmath237 and @xmath238 determine the width of the psf itself . since the gaussian function decays exponentially away from its center , it is customary to truncate the values in the psf mask after an assigned decay @xmath239 . it is evident from the quoted definition that the gaussian psf satisfies the strong symmetry condition ( [ eq : strong_simmetry_condition ] ) . another example of strongly symmetric psf is given by the psf representing the out - of - focus blur @xmath240 where @xmath241 is the radius of the psf.in the reported numerical tests , the blurred image @xmath2 has been perturbed by adding a gaussian noise contribution @xmath242 with @xmath243 fixed noise vector , @xmath244 , and @xmath245 assigned value . in such a way the signal noise ratio ( snr ) @xcite is given by @xmath246 and @xmath247 , and blurred image with the out - of - focus psf with support @xmath248 . ] in figure [ fig : cameraman_psfgauss_d15s2 ] we report the template true image ( the fov is delimited by a white frame ) , together with the blurred image with the gaussian psf with support @xmath248 and @xmath247 and the reference perturbation @xmath243 , reshaped in matrix form.we consider the optimal image restoration with respect to the relative restoration error ( rre ) , i.e. , @xmath249 , where @xmath250 is the computed approximation of the true image @xmath251 by considering spectral filtering . more in detail , the rre is analyzed by progressively adding a new basis element at a time , according to the non - decreasing order of the singular / eigen - values ( the eigenvalues are ordered with respect to their absolute value).in the case of sds ( or svds related to a separable psf ) this can be done as described in section [ sez : computational_issues ] and , beside the preliminary cost related to the decomposition computation , the addition of a new term has a computational cost equal to @xmath252 ops . the algorithm proposed in @xcite , that makes use of the anti - reflective direct and inverse transforms , is less expensive in the case of tests with few threshold values.hereafter , the aim is to compare the truncated svd with the truncated sd restorations both in the case of reflective and anti - reflective bcs . periodic bcs are not analyzed here , since reflective and anti - reflective bcs give better performances with respect to the approximation of the image at the boundary.in table [ tab : sigma2_cameraman_psfgauss ] and [ tab : sigma5_cameraman_psfgauss ] we report the results obtained by varying the dimension of the psf support , the parameter @xmath245 related to the amount of the noise perturbation and the variance of the considered gaussian blur . as expected the optimal rre worses as the parameter @xmath245 increases and the anti - reflective bcs show better performances in the case of low noise levels . in fact , for low @xmath245 values , the reduction of ringing artifacts is significant , while the quality of the restoration for higher @xmath245 values is essentially driven by the goal of noise filtering . therefore , in such a case , the choice of the bcs becomes more an more meaningless since it is not able to influence the image restoration quality . some examples of restored images are reported in figure [ fig : cameraman_psfgauss_d15s2ratio001 - 005 ] . psf & & & & + + svd & 0.059164 & 0.087402 & 0.090742 & 0.093856 + sd & 0.043754 & 0.087400 & 0.090746 & 0.093867 + + svd & 0.060278 & 0.091964 & 0.094468 & 0.097034 + sd & 0.060278 & 0.091964 & 0.094476 & 0.097034 + + svd & 0.091151 & 0.11214 & 0.11307 & 0.11495 + sd & 0.091152 & 0.11214 & 0.11307 & 0.11495 + + svd & 0.11635 & 0.13356 & 0.13508 & 0.13739 + sd & 0.11635 & 0.13356 & 0.13510 & 0.13739 + + svd & 0.13024 & 0.14607 & 0.14746 & 0.15047 + sd & 0.13024 & 0.14607 & 0.14746 & 0.15047 + psf & & & & + + svd & 0.039165 & 0.064081 & 0.086621 & 0.087237 + sd & 0.038316 & 0.063114 & 0.083043 & 0.083521 + + svd & 0.062182 & 0.094237 & 0.098897 & 0.10042 + sd & 0.059617 & 0.089105 & 0.092814 & 0.094343 + + svd & 0.096049 & 0.12231 & 0.12403 & 0.12536 + sd & 0.091383 & 0.11230 & 0.11343 & 0.11495 + + svd & 0.12791 & 0.15070 & 0.15188 & 0.15492 + sd & 0.11666 & 0.13414 & 0.13570 & 0.13816 + + svd & 0.14399 & 0.16756 & 0.16964 & 0.17225 + sd & 0.13083 & 0.14709 & 0.14852 & 0.15162 + psf & & & & + + svd & 0.063387 & 0.081274 & 0.097351 & 0.14634 + sd & 0.045365 & 0.081274 & 0.096387 & 0.14634 + + svd & 0.063915 & 0.096243 & 0.11449 & 0.15217 + sd & 0.063915 & 0.096274 & 0.11449 & 0.15217 + + svd & 0.089032 & 0.13343 & 0.14947 & 0.17397 + sd & 0.089032 & 0.13343 & 0.14946 & 0.17397 + + svd & 0.12203 & 0.16002 & 0.17339 & 0.18335 + sd & 0.12203 & 0.16002 & 0.17339 & 0.18335 + + svd & 0.13412 & 0.16793 & 0.17963 & 0.19057 + sd & 0.13412 & 0.16793 & 0.17963 & 0.19057 + psf & & & & + + svd & 0.040214 & 0.079543 & 0.088224 & 0.13686 + sd & 0.039437 & 0.078970 & 0.088832 & 0.13129 + + svd & 0.068197 & 0.095808 & 0.11522 & 0.15767 + sd & 0.063575 & 0.093247 & 0.1127 & 0.14893 + + svd & 0.09412 & 0.14482 & 0.16825 & 0.21148 + sd & 0.089038 & 0.13611 & 0.15270 & 0.17446 + + svd & 0.13553 & 0.18563 & 0.21006 & 0.22962 + sd & 0.12253 & 0.16269 & 0.17439 & 0.18414 + + svd & 0.15010 & 0.20164 & 0.22256 & 0.23960 + sd & 0.13487 & 0.16916 & 0.18088 & 0.19218 + @xmath253 + ( gaussian blur @xmath247 ) . , title="fig : " ] @xmath254 + ( gaussian blur @xmath247 ) . , title="fig : " ] more impressive is the fact that sds give better , or equal , results with respect to those obtained by considering svds . this numerical evidence is really interesting in the case of anti - reflective bcs : despite the loss of the orthogonality property in the spectral decomposition , the restoration results are better than those obtained by considering svd . moreover , the observed trend with respect to the reflective bcs is also conserved.a further analysis refers to the so - called picard plots ( see figure [ fig : cameraman_psfgauss_d15s2ratio001 - 005_picard_plot ] ) , where the coefficients @xmath255 , or @xmath256 , ( black dots ) are compared with the singular values @xmath196 , or the absolute values of the eigenvalues @xmath257 , ( red line ) . as expected , initially these coefficients decrease faster than @xmath196 , or @xmath257 , while afterwards they level off at a plateau determined by the level of the noise in the image.the threshold of this change of behavior is in good agreement with the optimal @xmath258 value obtained in the numerical test by monitoring the rre.moreover , notice that the picard plots related to the sds are quite in agreement with those corresponding to svds . in the case of the anti - reflective sd we observe an increasing data dispersion with respect to the plateau , but the correspondence between the threshold and the chosen optimal @xmath258 is still preserved.the computational relevance of this result is due to the significant lower computational cost required by the anti - reflective sds with respect to the corresponding svds . @xmath253 + ( gaussian blur @xmath247 ) . , title="fig:",height=226 ] + @xmath254 + ( gaussian blur @xmath247 ) . , title="fig:",height=226 ] .optimal rres of truncated sds with reference to the true image in figure [ fig : cameraman_psfgauss_d15s2 ] ( out - of - focus blur ) . [ cols= " < , < , < , < , < , < , < , < , < " , ] lastly , table [ tab : cameraman_psfff ] reports the spectral filtering results obtained in the case of out - of - focus blur by varying the dimension of the psf support and the parameter @xmath245 related to the noise perturbation . the rre follows the same trend observed in the case of gaussian blur . other image restoration tests with different gray - scale images have been considered in @xcite.a more interesting remark again pertains the computational cost . since the out - of - focus psf is not separable , but the transforms are , the use of sds related to reflective or anti - reflective bcs allows to exploit the tensorial nature of the corresponding transforms , both with respect to the computation of the eigenvalues and of the eigenvectors ( or of the reflective and anti - reflective transforms ) . here , we analyze some restoration tests in the case the template color image reported in figure [ fig : clown_psfgauss_d15s2 ] , by assuming the presence of a cross - channel blurring phenomenon modelled according to ( [ eq : sistema_quadrato_color ] ) . the entity of this mixing effect is chosen according to the matrix @xmath259.\ ] ] and @xmath247 and matrix @xmath195 in ( [ eq : def_acolor ] ) . ] in figure [ fig : clown_psfgauss_d15s2 ] is also reported the cross - channel blurred image with gaussian psf with support @xmath248 and @xmath247 . notice that the entity of the cross - channel blurring is not negligible , since the whole image results to be darkened and the color intensities of the additive rgb system are substantially altered.table [ tab : clown_psfgauss_sigma2 ] reports the optimal rres of truncated svds and sds obtained by varying the dimension of the gaussian psf support and the parameter @xmath245 related to the amount of the noise perturbation . it is worth stressing that we tune the filtering strategy with respect the spectral information given just by the matrix @xmath44 , i.e. , for any fixed @xmath196 ( or @xmath197 ) we simultaneously sum , or discard , the three contribution on @xmath1 related to the three singular values of @xmath195 . in fact , the magnitude of singular values of the considered matrix @xmath195 does not differ enough to dramatically change the filtering information given just by @xmath44 . nevertheless , also the comparison with the restoration results obtained by considering a global ordering justifies this approach.the color case behaves as the gray - scale one : as expected the optimal rre becomes worse as the parameter @xmath245 increases and the anti - reflective sd shows better performances in the case of low noise levels . @xmath253 + ( cross - channel and gaussian blur @xmath247 ) . , title="fig : " ] @xmath254 + ( cross - channel and gaussian blur @xmath247 ) . , title="fig : " ] in addition , by referring to figure [ fig : clown_psfgauss_d15s2ratio001 - 005 ] , we note that the truncated svd in the case of anti - reflective bcs shows a little more freckles than the corresponding truncated svd in the case of reflective bcs . nevertheless , for low noise levels , is just the anti - reflective sd that exhibits less freckles than the reflective sd . psf & & & & + + svd & 0.078276 & 0.12114 & 0.11654 & 0.1178 + sd & 0.078276 & 0.12114 & 0.11654 & 0.1178 + + svd & 0.078992 & 0.1212 & 0.11663 & 0.11792 + sd & 0.078992 & 0.12119 & 0.11663 & 0.11792 + + svd & 0.10152 & 0.12396 & 0.12088 & 0.12198 + sd & 0.10152 & 0.12396 & 0.12088 & 0.12198 + + svd & 0.12102 & 0.13853 & 0.13743 & 0.13844 + sd & 0.12102 & 0.13853 & 0.13743 & 0.13844 + + svd & 0.13437 & 0.14898 & 0.14854 & 0.14947 + sd & 0.13437 & 0.14898 & 0.14854 & 0.14947 + psf & & & & + + svd & 0.076646 & 0.1006 & 0.1098 & 0.10646 + sd & 0.074953 & 0.098474 & 0.10508 & 0.10216 + + svd & 0.077394 & 0.10639 & 0.1111 & 0.11002 + sd & 0.075727 & 0.10233 & 0.10612 & 0.10443 + + svd & 0.10431 & 0.12695 & 0.12624 & 0.12779 + sd & 0.10087 & 0.11737 & 0.11805 & 0.118 + + svd & 0.13017 & 0.15075 & 0.15063 & 0.15166 + sd & 0.12127 & 0.13699 & 0.13756 & 0.13795 + + svd & 0.1456 & 0.16516 & 0.16626 & 0.16647 + sd & 0.13507 & 0.14796 & 0.14955 & 0.15018 + + r & 0.041015 & 4.1e-005 & 0.079044 & 9e-006 & 0.086386 & 1.1e-005 & 0.089556 & 1.6e-005 + ar & 0.034237 & 1.1e-005 & 0.059465 & 1e-006 & 0.078963 & 1e-006 & 0.079805 & 1e-006 + + r & 0.050155 & 0.000188 & 0.087482 & 5.7e-005 & 0.090825 & 4.3e-005 & 0.093071 & 4.9e-005 + ar & 0.048556 & 0.000163 & 0.085279 & 4.6e-005 & 0.089388 & 3.3e-005 & 0.090821 & 3.3e-005 + + r & 0.083456 & 0.005555 & 0.10748 & 0.001786 & 0.10863 & 0.001678 & 0.11023 & 0.001573 + ar & 0.083436 & 0.005536 & 0.10744 & 0.001792 & 0.10868 & 0.001691 & 0.11019 & 0.001575 + + r & 0.12024 & 0.038152 & 0.12982 & 0.01929 & 0.13071 & 0.018417 & 0.13307 & 0.017892 + ar & 0.12049 & 0.038379 & 0.13006 & 0.01957 & 0.13096 & 0.018669 & 0.1333 & 0.018105 + + r & 0.14767 & 0.06587 & 0.14721 & 0.039231 & 0.14822 & 0.038181 & 0.15097 & 0.037893 + ar & 0.14813 & 0.066251 & 0.14766 & 0.039707 & 0.14866 & 0.038644 & 0.15144 & 0.038296 + + r & 0.031422 & 0.000172 & 0.05346 & 6.9e-005 & 0.060954 & 3.5e-005 & 0.074785 & 2.7e-005 + ar & 0.036213 & 0.000302 & 0.051236 & 6.8e-005 & 0.06683 & 5.7e-005 & 0.084482 & 5.8e-005 + + r & 0.034441 & 0.000271 & 0.061465 & 0.000145 & 0.073751 & 0.000101 & 0.09074 & 7.9e-005 + ar & 0.038313 & 0.000402 & 0.059957 & 0.000138 & 0.076695 & 0.000126 & 0.095274 & 0.000106 + + r & 0.069647 & 0.008493 & 0.11361 & 0.004117 & 0.12881 & 0.003037 & 0.14914 & 0.001873 + ar & 0.070384 & 0.008923 & 0.11404 & 0.00422 & 0.12982 & 0.003139 & 0.15061 & 0.001969 + + r & 0.12204 & 0.053687 & 0.1532 & 0.030719 & 0.16614 & 0.022121 & 0.18769 & 0.01346 + ar & 0.12256 & 0.05423 & 0.15402 & 0.031574 & 0.16739 & 0.023213 & 0.18933 & 0.014472 + + r & 0.16366 & 0.092379 & 0.17357 & 0.055919 & 0.1829 & 0.042944 & 0.20323 & 0.028803 + ar & 0.16433 & 0.093069 & 0.17485 & 0.057326 & 0.18457 & 0.044901 & 0.20511 & 0.031011 + by considering a gaussian blurring of the true image reported in figure [ fig : cameraman_psfgauss_d15s2 ] , table [ tab : cameraman_psfgauss_tikhnov_sigma2 ] compares the optimal rre obtained in the case of the tikhonov method for reflective bcs and of the re - blurring method for anti - reflective bcs . in addition , in table [ tab : cameraman_psfff_tikhnov ] , the same comparison refers to the case of the out - of - focus psf.as expected , the rre deteriorates as the dimension of the noise level or the dimension of the psf support increases . notice also that the gap between the reflective and anti - reflective bcs is reduced also for low noise levels . further numerical tests can be found in @xcite.lastly , we focus our attention on the case of the color image in figure [ fig : clown_psfgauss_d15s2 ] . the image restorations have been obtained by considering the transformation procedure outlined at the end of section [ sez : cross_channel_blurring ] . despite the rres in table [ tab : cameraman_psfgausssigma2_tikhnov_cross_channel ] are bigger than in the gray - scale case , the perception of the image restoration quality is very satisfying and a little less freckles than in the corresponding sds and svds are observed ( see figure [ fig : clown_psfgauss_tik_d15s2ratio001 - 005 ] ) . notice , also that the lack of orthogonality in the @xmath208 transform related to the anti - reflective bcs does not deteriorate the performances of the restoration . + r & 0.069148 & 0.000203 & 0.11508 & 0.001204 & 0.1123 & 0.000717 & 0.11335 & 0.000726 + ar & 0.062854 & 0.000102 & 0.091232 & 7e-006 & 0.1014 & 4.4e-005 & 0.098266 & 1.5e-005 + + r & 0.071259 & 0.000312 & 0.11515 & 0.001228 & 0.11239 & 0.000744 & 0.11347 & 0.000755 + ar & 0.066734 & 0.000209 & 0.098658 & 5.8e-005 & 0.10276 & 7.7e-005 & 0.10111 & 4.5e-005 + + r & 0.094871 & 0.004975 & 0.11896 & 0.002919 & 0.11712 & 0.002421 & 0.1182 & 0.002459 + ar & 0.094458 & 0.004841 & 0.1144 & 0.001884 & 0.11481 & 0.00184 & 0.11507 & 0.001755 + + r & 0.13209 & 0.029798 & 0.13662 & 0.015305 & 0.13599 & 0.014896 & 0.13669 & 0.014824 + ar & 0.13239 & 0.029944 & 0.13561 & 0.014992 & 0.13593 & 0.014772 & 0.13611 & 0.014595 + + r & 0.16281 & 0.051315 & 0.15543 & 0.029068 & 0.15547 & 0.028822 & 0.15586 & 0.02868 + ar & 0.16341 & 0.051659 & 0.15526 & 0.029213 & 0.15602 & 0.029029 & 0.15588 & 0.02872 + @xmath253 + -1.5 cm ( cross - channel and gaussian blur @xmath247 - @xmath254).,title="fig : " ] -1.5 cm @xmath254 + -1.5 cm ( cross - channel and gaussian blur @xmath247 - @xmath254).,title="fig : " ] -1 cm in this paper we have analyzed and compared sd and svd filtering methods in the case both of reflective and anti - reflective bcs . numerical evidence is given of the good performances achievable through sds and with a substantially lower computational cost with respect to svds . in addition , the tensorial structure of the reflective and anti - reflective sds can be exploited in depth also in the case of not separable psfs . a special mention has to be done to the fact that the loss of orthogonality of the anti - reflective transform does not seems to have any consequence on the trend of the image restoration results . the analysis in the case of cross - channel blurring in color images allows to confirm the quoted considerations . finally , the re - blurring regularizing method has been re - interpreted as a standard tikhonov regularization method in the space transformed by means of @xmath175 . some numerical tests highlight the image restoration performances , also in the case of cross - channel blurring . future works will concern the analysis of effective strategies allowing to properly choose the optimal regularizing parameters in the anti - reflective bcs case . 99 b. anconelli , m. bertero , p. boccacci , m. carbillet , and h. lanteri , _ reduction of boundary effects in multiple image deconvolution with an application to lbt linc - nirvana _ , astron . astrophys . , 448 ( 2006 ) , pp . 12171224 . a. aric , m. donatelli , and s. serra - capizzano , _ the antireflective algebra : structural and computational analyses with application to image deblurring and denoising _ , calcolo , to appear . a. aric , m. donatelli , and s. serra capizzano , _ spectral analysis of the anti - reflective algebra _ , linear algebra appl . , to appear . a. aric , m. donatelli , j. nagy , and s. serra - capizzano , _ the anti - reflective transform and regularization by filtering _ , submitted , 2006 . m. bertero and p. boccacci , _ introduction to inverse problems in imaging _ , inst . of physics publ . london , uk , 1998 . m. bertero and p. boccacci , _ image restoration for large binocular telescope ( lbt ) _ , astron . ser . , 147 ( 2000 ) , pp . d. bini and m. capovani , _ spectral and computational properties of band symmetric toeplitz matrices _ , linear algebra appl . , 52/53 ( 1983 ) , pp . p. j. davis , _ circulant matrices _ , wiley , new york , 1979 . m. donatelli , c. estatico , a. martinelli , and s. serra capizzano , _ improved image deblurring with anti - reflective boundary conditions and re - blurring _ , inverse problems , 22 ( 2006 ) , pp . m. donatelli , c. estatico , j. nagy , l. perrone , and s. serra capizzano , _ anti - reflective boundary conditions and fast 2d deblurring models _ , proceeding to spie s 48th annual meeting , san diego , ca usa , f. luk ed , 5205 ( 2003 ) , pp . 380389 . m. donatelli and s. serra capizzano , _ anti - reflective boundary conditions and re - blurring _ , inverse problems , 21 ( 2005 ) , pp . 169182 . h. engl , m. hanke , and a. neubauer , _ regularization of inverse problems _ , kluwer academic publishers , dordrecht , the netherlands , 2000 . c. w. groetsch , _ the theory of tikhonov regularization for fredholm integral equations of the first kind _ , pitman , boston , 1984 . p. c. hansen , _ rank - deficient and discrete ill - posed problems _ , siam , philadelphia , pa , 1997 . m. hanke and j. nagy , _ restoration of atmospherically blurred images by symmetric indefinite conjugate gradient technique _ , inverse problems , 12 ( 1996 ) , pp . 157173 . p. c. hansen , j. nagy , and d. p. oleary , _ deblurring images matrices , spectra and filtering _ , siam publications , philadelphia , 2006 . r. l. lagendijk and j. biemond , _ iterative identification and restoration of images _ , springer - verlag new york , inc . m. k. ng , r. h. chan , and w. c. tang , _ a fast algorithm for deblurring models with neumann boundary conditions _ , siam j. sci . comput . , 21 ( 1999 ) , no . . l. perrone , _ kronecker product approximations for image restoration with anti - reflective boundary conditions _ , numer . linear algebra appl . , 131 ( 2006 ) , pp . _ tecniche di filtraggio nella ricostruzione di immagini con condizioni al contorno antiriflettenti _ , ( in italian ) , basic degree thesis , university of milano - bicocca , milano , 2006 . s. serra capizzano , _ a note on anti - reflective boundary conditions and fast deblurring models _ , siam j. sci . , 253 ( 2003 ) , pp . 13071325 . y. shi and q. chang , _ acceleration methods for image restoration problem with different boundary conditions _ , appl . math . , in press . |
we briefly recall the previous publications plb1996,2013,2014,2015,npb1997,2000 and prd2000,2001,2016 on the effective field theory of massive ( tev ) composite particles and their sm gauge couplings in the domain of uv - stable fixed point of relevant four - fermion operators , and on the resonant and nonresonant phenomena for experimental searches , in particular the final states of multijets . we then focus the discussions on the decay and annihilation channels of composite particles into the final states of the sm gauge bosons , leptons and quarks in connection with the on - going searches of the atlas and cms experiments at energy tev . we calculate the rates of composite bosons ( fermions ) produced by lhc-collision decaying ( annihilating ) into two gauge bosons , , , and give the definite ratios of decay ( annihilation ) rates of different channels . it is shown that a composite fermion decays into an elementary fermion and a composite boson , the latter being an intermediate state decays into two gauge bosons , leading to a peculiar kinematics of final states of a jet ( or a lepton ) and two gauge bosons . these provide the justification for the verification or falsification of such a theoretical framework beyond the sm . we also give some speculative discussions on the channels of composite fermions decaying to , and two boson - tagged jets with quark jets , or to four - quark jets . moreover , at the same tev energy scale , the composite particles can be formed by high - energy sterile neutrino ( dark matter ) collisions , then they decay , leading to an excess of cosmic ray particles in space , producing signals of sm particles in underground . | 1601.06845 | 0.1 cm the parity - violating ( chiral ) gauge symmetries and spontaneous / explicit breaking of these symmetries for the hierarchy of fermion masses have been at the center of a conceptual elaboration that has played a major role in donating to mankind the beauty of the sm for particle physics . the nambu - jona - lasinio ( njl ) model @xcite of dimension-6 four - fermion operators at high energies and its effective counterpart , the phenomenological model @xcite of the elementary higgs boson and its yukawa - coupling to fermions at low energies , provide an elegant and simple description for the electroweak symmetry breaking and intermediate gauge boson masses . after a great experimental effort for many years , using data recorded at @xmath9 tev experiments of @xmath1-collision at the large hadron collider ( lhc ) , the atlas @xcite and cms @xcite collaborations have shown the first observations of a 125 gev scalar particle in the search for the sm higgs boson . this far - reaching result begins to shed light on this most elusive and fascinating arena of fundamental particle physics . recently , in the run-2 of the upgraded lhc , the preliminary results on @xmath0 tev @xmath1 collision data are prepared by atlas @xcite and cms @xcite to search for new ( beyond the sm ) resonant and/or nonresonant phenomena that might manifest themselves in the high - energy regime with final states of diboson , dilepton , dijet and so forth . the diphoton channel shows the trace of a new resonance at the diphoton invariant mass @xmath10 gev , which however needs to be confirmed by increasing the integrated luminosity of @xmath1 collisions in time . in refs . @xcite , we have already presented discussions on the possible decay channels of composite bosons and fermions into final states of jets and leptons in the framework of effective four - fermion operators in high energies . in this article , we focus on the discussions on the possible channels of composite bosons and fermions decaying into the final states of two sm gauge bosons ( diboson channels ) , and/or jets , in connection with the atlas amd cms on going @xmath0 tev experiments . it is shown that the diphoton channel of composite bosons decay should have the largest branching ratio , however other diboson channels should contribute to the invariant mass @xmath11 of possible composite - boson resonance . instead composite fermions decay into a quark(jet ) or a lepton and a composite boson , the latter is an intermediate state , which then decays into two bosons , i.e. , the peculiar jet - diboson or lepton - diboson channel . the kinematic relation of composite boson and fermion decays is thus definitely fixed , so that the theoretical scenario of composite bosons and fermions can be verified or excluded , provided that the diboson channel of composite boson decay and the jet - diboson or ( lepton - diboson ) channel of composite fermion decay can be experimentally verified . in secs . [ eff4o ] and [ smeff4o ] , we present the sm gauge - symmetric operators of four - fermion interactions and the discussion of their possible origin . some clarifications of the low - energy domain for the sm and the high - energy ( tev ) domain for an effective field theory of composite particles are given in secs . [ ird ] , [ uvd ] and riro . after showing the composite particle spectrum and effective lagrangian in secs . [ composite ] and [ efflag ] , we discuss the composite boson and fermion decay in secs . ( [ cbd ] ) and ( [ cfd ] ) , as well as their annihilation in sec . ( [ acf ] ) . in the final section , we present some discussions and speculations on the possible channels of composite particles that could be relevant to some experiments at energy scale tev . 0.1 cm the dynamics of new physics at high energies may be represented by an effective theory of high - dimensional operators of fermion fields , e.g. , dimension-6 four - fermion operators , preserving at least the sm gauge symmetries . the strong technicolor dynamics of extended gauge theories at the tev scale was invoked @xcite to have a natural scheme incorporating the relevant four - fermion operator @xmath12 of the @xmath13-condensate model @xcite . on the other hand , these relevant operators can be constructed on the basis of phenomenology of the sm at low - energies . in 1989 , several authors @xcite suggested that the sm symmetry breakdown could be a dynamical mechanism of the njl type that intimately involves the top quark at the high - energy scale @xmath14 . since then , many models based on this idea have been studied @xcite . the low - energy sm physics was supposed to be achieved by the renormalization group ( rg ) equations in the ir - domain of the infrared stable ( ir ) stable fixed point with the electroweak scale @xmath15 gev @xcite . on the other hand , a well - defined quantum field theory for the sm lagrangian requires a natural regularization ( cutoff @xmath14 ) fully preserving the sm chiral - gauge symmetry . the quantum gravity provides a such regularization of discrete space - time with the minimal length @xmath16 @xcite , where the planck length @xmath17 cm . however , the no - go theorem @xcite tells us that there is no any consistent way to regularize the sm bilinear fermion lagrangian to exactly preserve the sm chiral - gauge symmetry . this implies that the natural quantum - gravity regularization for the sm leads us to consider at least four - fermion operators . actually , in the torsion - free einstein - cartan ( ec ) theory of gravitational tetrad field @xmath18 and spin - connection field @xmath19 , four - fermion operators are naturally obtained by integrating over `` static '' torsion fields , @xmath20 where the axial current @xmath21 of massless fermion fields . if intermediate torsion fields propagate short distances @xmath22 characterized by their large masses @xmath23 , this implies the four - fermion coupling @xmath24 and there are high - dimensional fermion operators for @xmath25 . the relation @xmath24 should be determined by the wilson - kadanoff renormalization group approach at a nontrivial fixed point . here we adopt the effective four - fermion operators ( [ ec0 ] ) in the context of a well - defined quantum field theory at the high - energy scale @xmath14 . in the context of the sm fermion content and gauge symmetries , considering massless left- and right - handed dirac fermions @xmath26 and @xmath27 carrying the sm quantum numbers , as well as right - handed dirac sterile neutrinos @xmath28 and their majorana counterparts @xmath29 , we obtain @xcite @xmath30 where the axial currents @xmath31 and @xmath32 , and the four - fermion operators can be written as @xmath33 by using the fierz theorem . equations ( [ art0 ] ) and ( [ art1 ] ) represent repulsive and attractive operators , respectively . 0.1 cm in the quark sector ( [ art1 ] ) , the four - fermion operators read @xmath34+\cdot\cdot\cdot , \label{bhlx}\end{aligned}\ ] ] where @xmath35 are the color indexes of the top and bottom quarks , the @xmath36 doublet @xmath37 and singlet @xmath38 are the eigenstates of electroweak interactions . the first and second terms in eq . ( [ bhlx ] ) are respectively the four - fermion operators of top - quark channel @xcite and bottom - quark channel , whereas @xmath39 " stands for the first and second quark families that can be obtained by substituting @xmath40 and @xmath41 . in the lepton sector ( [ art1 ] ) , the four - fermion operators in terms of gauge eigenstates read , @xmath42 , \label{bhlxl}\end{aligned}\ ] ] preserving all sm gauge symmetries , where the lepton @xmath36 doublets @xmath43 , singlets @xmath44 and the conjugate fields of sterile neutrinos @xmath45 . coming from the second term in eq . ( [ art1 ] ) , the last term in eq . ( [ bhlxl ] ) preserves the symmetry @xmath46 for the lepton - number conservation , although @xmath47 violates the lepton number of family `` @xmath48 '' by two units . similarly , there are following four - fermion operators @xmath49 , \label{bhlbv}\end{aligned}\ ] ] where quark fields @xmath50 and @xmath51 . there are the sm gauge - symmetric four - fermion operators of quark - lepton interactions @xcite , @xmath52 + \cdot\cdot\cdot , \label{bhlql}\end{aligned}\ ] ] where @xmath53 and @xmath54 for the first family . the `` @xmath39 '' represents for the second and third families with substitutions : @xmath55 , @xmath56 , and @xmath57 and @xmath58 . these operators are not present in the lagrangian ( [ art1 ] ) . it is necessary to study how four - fermion operators are generated in the framework of the @xmath59 or @xmath60 unification theory , see refs . @xcite . due to the unique four - fermion coupling @xmath61 and the global fermion - family @xmath62 chiral symmetry of eq . ( [ art1 ] ) for the massless sm fermions , we perform chiral transformations @xmath63 and @xmath64 so that the four - fermion operator ( [ art1 ] ) is only for each quark or lepton family , for more details see the recent article @xcite . beside we do not consider the fermion - family mixing of the sm . therefore , there are no additional four - fermion operators contributing to the flavor - changing - neutral - current ( fcnc ) processes . in the following , our notations will be only for the first sm family , however , they are the same for the second and thrid families . apart from what is possible new physics at the scale @xmath14 explaining the origin of these effective four - fermion operators , it is essential and necessary to study : ( i ) the phase diagram in the space of these effective operator couplings ; ( ii ) which dynamics of these operators undergo in terms of their couplings as functions of running energy scale @xmath65 ; ( iii ) associating to these dynamics where infrared ( ir ) and/or ultraviolet ( uv ) stable fixed point of physical couplings locates ; ( iv ) in the ir and/or uv domains ( scaling regions ) of these stable fixed points , which operators become physically relevant and renormalizable following rg equations ( scaling laws ) , and other irrelevant operators are suppressed by the cutoff at least @xmath66 . 0.1 cm in the symmetric breaking phase of weak coupling @xmath61 , upon the energetically minimized ground state of spontaneous symmetry breaking ( ssb ) @xcite , _ only _ the top - quark channel of eq . ( [ art1 ] ) is relevant , and becomes the full induced effective lagrangian in the scaling region ( ir - domain ) of the ir - stable fixed point @xcite @xmath67 where the unconventional wave - renormalization ( form factor ) @xmath68 for the composite higgs boson @xmath69 is adopted , and all renormalized quantities received fermion - loop contributions are defined with respect to the low - energy scale @xmath65 . note that @xmath70 the weak critical coupling of the njl dynamics for the ssb , only the top - quark channel in eq . ( [ art1 ] ) undergoes the njl dynamics and form the composite massive higgs boson @xmath71 , massless goldstone bosons @xmath72 and @xmath73 which are the longitudinal modes of massive intermediate gauge bosons @xcite . in refs . @xcite , we solve the full one - loop rg equations for running couplings @xmath74 and @xmath75 with the top - quark and higgs - boson mass - shell conditions @xmath76 @xmath77 and @xmath78 . as a result , we uniquely determine the solution @xmath79 and @xmath80 , as well as the energy scale @xmath81 and finite `` form factor '' @xmath82 , @xmath83 at which the quartic coupling changes its sign @xmath84 , indicating a phase transition from the symmetry breaking phase to the symmetric phase . the finite @xmath85 ( [ thvari ] ) indicates the composite higgs boson behaves as if an elementary higgs boson . from the one fermion - loop contribution ( tadpole diagram ) to the gap equation for top - quark mass @xmath86 , we obtain and the @xmath87-function @xmath88>0 , \label{gbeta}\end{aligned}\ ] ] for @xmath89 and @xmath90 , where @xmath65 is the running energy scale and @xmath91 is the number of colors . the positive @xmath87-function of eq . ( [ gbeta ] ) indicates that an ir - stable fixed point @xmath92 and the ir - domain @xmath93 as @xmath94 , as shown the part `` i '' of the @xmath95-function in fig . as the energy scale @xmath65 decreases , the rg flow is attracted to the ir - stable fixed point , and the effective sm of particle physics in low energies is realized in this ir - domain . -function in terms of the four - fermion coupling @xmath61 . we indicate the quasi ir - stable fixed point @xmath92 and a possible uv - stable fixed point @xmath96 , the latter separates the symmetry - breaking phase ( positive @xmath95-function ) from the symmetric phase ( negative @xmath95-function ) . we also indicate the positive parts `` i '' ( increasing ) and `` ii '' ( decreasing ) , as well as the negative part `` iii '' of the @xmath95-function.,title="fig:",height=120 ] ( -155,90)@xmath95 ( 5,33)@xmath61 ( -130,00)symmetry breaking phase ( 5,00)symmetric phase ( -95,60)i ( -15,60)ii ( 1,15)iii ( -155,23)@xmath92 ( -155,40)@xmath97 ( -32,23)@xmath96 ( -32,40)@xmath98 0.1 cm in the symmetric phase of strong coupling @xmath61 , we use the strong coupling expansion to calculate two - point functions of composite boson and fermion fields , and the vertex functions of their couplings to the sm gauge bosons @xcite . in the lowest non - trivial order , we obtained the propagator of the massive composite dirac fermions : @xmath36-doublet @xmath99 and @xmath36-singlet @xmath100 , as well as the massive composite bosons : @xmath36-doublet @xmath101 , and their couplings to the sm gauge bosons . these massive spectra fully respect the sm chiral gauge symmetry . these massive composite particles are formed by the such strong - coupling dynamics in the symmetric phase , when @xmath102 the strong critical coupling and @xmath103 . they are completely different from the composite particles like massive higgs boson and massless goldstone bosons formed by the weak - coupling njl dynamics in the spontaneously symmetry - breaking phase @xcite . in the lowest non - trivial order of the strong coupling expansion , i.e. , the two - fermion - loop contribution ( sun - set diagrams ) , we obtained the negative @xmath87-function @xcite @xmath104 where @xmath105 and the dimensionless lorentz scalar @xmath106 is positive and finite , monotonically decreases as the energy scale @xmath107 increases . on the basis of the @xmath95-function being positive ( [ gbeta ] ) and negative ( [ ren ] ) , as sketched as `` i '' and `` iii '' in fig . [ fixp ] , we infer there must be at least one zero - point of the @xmath95-function , i.e. , @xmath108 and @xmath109 . this zero - point @xmath110 is a uv - stable fixed point , coinciding with the strong critical coupling @xmath96 for the second - order phase transition from the symmetry breaking phase to the symmetric phase . analogously to the electroweak scale @xmath111 in the ir - domain of the fixed point @xmath92 , the correlation length @xmath112 sets the physical scale @xmath113 in the uv - domain ( scaling region ) of the fixed point ( critical point ) @xmath96 . the @xmath114 leads to the scaling invariance , i.e. , the rg invariance . we define @xcite an effective theory of composite particles , whose propagators have poles and residues that respectively represent their masses and form - factors , and their mass - shell conditions read @xmath115 where the masses @xmath116 . the scale @xmath117 tev of eq . ( [ thvari ] ) is estimated by extrapolating the solution of the rg equation to the high - energy regime . as shown in fig . [ fixp ] , the positive `` ii '' and negative `` iii '' of @xmath95-function indicate the uv - stable fixed point @xmath96 . as the running energy scale @xmath65 increases , @xmath118 , the rg flow is attracted to this uv - stable fixed point , and the effective lagrangian of composite particle physics in high energies is realized in this uv - domain @xmath119 . the effective lagrangian for these composite particles and their couplings to the sm gauge bosons will be shown later . instead , as the running energy scale @xmath65 decreases , the rg flows go away from the uv - fixed point @xmath96 and approach to the ir - fixed point @xmath92 , associating with the phase transition and composite particle decays . 0.1 cm in the weak - coupling ir - domain , as an energetically favorable solution for the ssb ground state @xcite , among four - fermion operators in eq . ( [ art1 ] ) the top - quark channel operator @xmath12 @xcite is the only physically relevant and renormalizable operators of effective dimension-4 , due to the njl - dynamics for the ssb . namely , it becomes the effective sm lagrangian with _ bilinear _ top - quark mass term and yukawa - coupling to the composite higgs boson @xmath69 , which obeys the rg equation approaching to the low - energy sm physics characterized by the energy scale @xmath15 gev . other four - fermion operators in eq . ( [ art1 ] ) , as well as repulsive four - fermion operators ( [ art0 ] ) , do not undergo the ssb and are irrelevant dimension-6 operators , whose tree - level amplitudes of four - fermion scatterings are suppressed @xmath120 , thus their deviations from the sm could not be experimentally accessible today . in the strong - coupling uv - domain , all attractive four - fermion operators ( [ art1 ] ) are physically relevant operators associating to the strong - coupling dynamics and formation of composite bosons and fermions . due to the unique four - fermion coupling @xmath61 , this strong - coupling dynamics of forming composite particles occurs in all channels of eq . ( [ art1 ] ) for each sm fermion family . the effective field theory in this uv - domain is expected to contains the massive composite particle spectrum preserving the sm chiral gauge symmetries , and relevant operators of effective dimension-4 following the rg equations . compared with the sm in the ir - domain , the effective theory of composite particles in the uv - domain has the same chiral gauge symmetries ( quantum numbers ) and couplings to gauge bosons ( @xmath121 and gluon ) , but the different vector - like spectra and coupling vertexes , apart from massive composite particles being comprised by sm elementary particles . the repulsive four - fermion operators ( [ art0 ] ) are irrelevant operators of dimension-6 , and thus suppressed @xmath120 for the reasons that they are neither associated with the njl dynamics of spontaneous symmetry breaking in the ir - domain , nor associated with the strong - coupling dynamics of forming composite particles in the uv - domain . these discussions are reminiscent of the asymptotic safety @xcite that quantum field theories regularized at uv cutoff @xmath14 might have a non - trivial uv - stable fixed point , rg flows are attracted into the uv - stable fixed point with a finite number of physically renormalizable operators . the weak and strong four - fermion coupling @xmath61 brings us into two distinct domains . this lets us also recall the qcd dynamics : asymptotically free quark states in the domain of a uv - stable fixed point and bound hadron states in the domain of a possible ir - stable fixed point . 0.1 cm we turn to discuss the composite particle spectra and interacting vertexes in the uv - domain . for the @xmath122-quark channel , the massive composite boson is an @xmath36-doublet @xmath123^{-1/2}(\bar u_{ra } \psi^{ia}_l ) \label{boundb}\ ] ] of the hypercharge @xmath124 . this composite boson combines with an elementary two - component weyl quark to form the composite two - component weyl - fermion states , @xmath125 and the massive composite four - component dirac - fermion states : the @xmath36 doublet @xmath126 and the @xmath36 singlet @xmath127 , @xmath128 the form - factors @xmath129^{1/2}$ ] ( [ boundb ] ) and @xmath130 ( [ bound ] ) are generalized wave - function renormalization of the composite boson and fermion operators . the four - component composite dirac fermions are vector - like spectra , fully preserving the parity - symmetry . the composite @xmath36 doublets @xmath126 carry the same @xmath131-hypercharge as the sm elementary and left - handed @xmath36 doublets @xmath132 . the composite @xmath36 singlets @xmath127 carry the same @xmath131-hypercharge as the sm elementary and right - handed @xmath36 singlets @xmath133 . these composite particles have the sm - gauge symmetric masses @xmath134 for the @xmath135-quark channel , the composite particles are represented by eqs . ( [ boundb]-[boundd ] ) with the replacement @xmath136 , carrying the different quantum numbers of the @xmath131 gauge group . this represents the first family of the composite particles composed by the sm elementary particles ( quarks ) in the first family . the composite boson and fermion states ( [ boundb]-[boundd ] ) are the eigenstates of both the sm gauge interactions and mass operators ( [ boundm ] ) , due to the sm gauge symmetries are preserved by the massive spectra of composite particles . the same discussions also apply for the second and third quark families by substituting the @xmath36 doublet @xmath137 into @xmath138 or @xmath139 and singlet @xmath140 into @xmath141 or @xmath142 , as well as singlet @xmath143 into @xmath144 or @xmath145 . in the lepton sector , the composite boson and weyl - fermion states formed by the first term ( @xmath146-channel ) of eq . ( [ bhlxl ] ) are : @xmath147^{1/2}(\bar\ell_{r } \ell^{i}_l ) ; \qquad { \bf\psi}^{i}_r = z^{^s}_r{\mathcal a}^i\ell_{r}\ , , \quad { \bf\psi}_{l}=z^{^s}_l{\mathcal a}^{i\dagger}\ell_{il } , \label{boundl}\end{aligned}\ ] ] and the massive composite dirac fermions @xmath148 and @xmath149 . the composite boson doublets @xmath150^{1/2}(\bar\ell_{r } \ell^{i}_l)$ ] carry the hypercharge @xmath151 , the doublets @xmath148 carry the @xmath131 hypercharge of the sm field @xmath152 and the singlets @xmath149 carry the @xmath131 hypercharge of the sm field @xmath146 . compared with their counterparts ( [ boundb]-[boundd ] ) in the quark sector , these massive composite particles have different quantum numbers of the sm gauge symmetries . the composite particles formed by the second term ( @xmath153-channel ) of eq . ( [ bhlxl ] ) are obtained by eq . ( [ boundl ] ) with the replacement @xmath154 , and they carry different charge of the @xmath131 gauge group . analogously , we present for the @xmath155-channel of quark - lepton interactions ( [ bhlql ] ) , the massive composite dirac fermions : @xmath156 doublet @xmath148 and singlet @xmath157 , where the renormalized composite boson and composite weyl - fermion states are : @xmath158^{-1/2}(\bar d^{\,a}_{r}\psi^{i}_{la});\quad { \bf\psi}^{i}_r=(z^{^s}_r)^{-1}{\mathcal a}^i e_{r}\ , , \quad { \bf\psi}_{l}=(z^{^s}_l)^{-1}{\mathcal a}^{i\dagger}\ell_{il}. \label{boundem}\ ] ] for the @xmath159-channel , the composite particles are represented by eq . ( [ boundem ] ) with the replacements @xmath160 and @xmath161 , carrying different quantum numbers of the @xmath131 gauge group . the composite particles from the second and third lepton families can be obtained by substitutions : @xmath55 , @xmath56 , and @xmath57 and @xmath58 . in general , the @xmath131-hypercharge of the composite particle is the sum of its constituents hypercharges , and electric charge @xmath162 in units of @xmath163 , where @xmath164 is the diagonal third component of @xmath36-isospin , @xmath165 for the neutrino and up quarks and @xmath166 for the electron and down quarks . the massive composite boson @xmath101 ( [ boundb ] ) carries an electric charge @xmath167 . the massive composite dirac fermion @xmath168 ( [ boundd ] ) carries an electric charge @xmath167 . these first - family composite particles are composed only by the sm elementary particles in the first family , and there are no extra elementary gauge bosons and fermions , except the right - handed sterile neutrino @xmath169 . the second - family ( third - family ) composite particles are composed only by the sm elementary particles in the second ( third ) family . the three families of composite particles are replicated corresponding to the three families of elementary fermions in the sm . thought composite particles are massive , they carry the quantum numbers of the sm chiral gauge symmetries , which are the sum of quantum numbers of their constituents of the sm elementary particles . the propagators of these composite particles have poles and residues that respectively represent their masses and form - factors @xcite . as long as their form - factors are finite , these composite particles behave as elementary particles . it should be mentioned that the gauge symmetric masses @xmath170 and form factors @xmath171 of composite particles , e.g. , ( [ boundb]-[boundm ] ) can be different from one to another , due to some other effects that we do not study here . to end this section , we present some discussions on the third - family composite bosons ( [ boundb ] ) by the top and bottom quarks , compared with the composite higgs and goldstone bosons . the former comprises of the massive bound states @xmath172 and @xmath173 formed by the strong - coupling dynamics in the uv - domain of the symmetric phase . they are different from the composite massive higgs boson @xmath71 and massless goldstone bosons @xmath72 and @xmath73 @xcite formed by the njl dynamics in the ir - domain of the spontaneously symmetry - breaking phase . nevertheless , at the critical point of phase transition that may occur around tev , see eq . ( [ thvari ] ) , we expect some relationships of the binding energy and form factor between two types of composite bosons from the point view of the different ground states of the symmetric phase and the symmetry - breaking phase . we will study this issue in future . 0.1 cm in the ir - domain of the symmetry breaking phase , the sm effective lagrangian of elementary fermions in the first family reads @xmath174 for the both quark and lepton sector in one sm family , @xmath175 is the weak isospin raising and lowering operators @xmath176 and @xmath177 . the weak angle @xmath178 and the electric charge @xmath179 . these massive fermions couple to the massless photon @xmath180 , massive intermediate gauge bosons @xmath181 and @xmath182 . the gauge and mass eigenstates of the sm elementary fermions are different , leading to the fermion - family mixing , when three fermion families are considered . in the uv - domain of the symmetric phase , massive composite bosons and fermions ( see sec . [ composite ] ) have the quantum numbers of sm chiral gauge symmetries @xmath183 , couplings @xmath184 to sm gauge bosons @xmath121 and gluon . the gauge and mass eigenstates of composite bosons and fermions are the same . the effective lagrangian of these massive composite bosons and fermions is the sm chiral gauge symmetrically invariant . based on the discussion in the previous section [ composite ] , we obtain the following effective lagrangian for the first - family composite particles ( [ boundb],[boundd],[boundl],[boundem ] ) . for the massive composite dirac fermions @xmath185 and @xmath186 , it reads @xmath187z_\mu . \label{comf}\end{aligned}\ ] ] for the massive composite bosons @xmath101 , it reads @xmath188 \label{comc}\end{aligned}\ ] ] and @xmath189 where @xmath190 is the pauli matrix and the isospin index @xmath191 stands for the component of definite hypercharge @xmath192 . the effective lagrangian ( [ comf ] ) and ( [ dcomc ] ) for the both quark and lepton sectors . these massive composite particles couple to the gauge bosons @xmath193 , @xmath194 and @xmath195 . these lagrangians show that the massive spectrum and interacting vertex are vector - like fully preserving the parity - symmetry . the lagrangian ( [ comc ] ) and ( [ dcomc ] ) are for the first - family of composite particles , and they can be replicated for the second and third families of composite particles . the sm fermion - family mixing does not present in the interacting vertexes ( [ comf ] ) of composite fermions with the sm gauge bosons , since the composite particles have the same eigenstates of gauge interactions and mass operators . to the leading order of the expansion in the powers of the perturbative sm gauge couplings ( the tree - level ) , the interacting vertexes of composite bosons and fermions coupling to sm gauge bosons ( @xmath196 ) are represented by the tree - level feynman diagrams , see fig . [ vertex ] , which show gauge bosons interacting with one of elementary particles inside composite particles . therefore , this implies that at the tree level , the calculations of feynman diagrams are similar to those in the sm , however , the fermion spectra and coupling vertexes are vectorlike , see eqs . ( [ smf ] ) and ( [ comf ] ) . the final results are characterized by the mass scale @xmath197 tev of the uv domain and functions of the sm gauge couplings @xmath198 at this characteristic scale , rather than the electroweak scale @xmath15 gev in the ir domain . moreover , the fcnc process is prohibited at the tree - level and receives the possible one - loop contribution due to the fermion - family mixing of massive elementary fermions in the sm . this is exactly the same as what happens in the sm . in the absence of fermion - family mixing of composite particles , there is no any additional interacting vertexes in the four - fermion operators ( [ art1 ] ) and the effective lagrangian ( [ comf ] ) that contribute to the fcnc process . as the running energy scale @xmath65 decreasing , and composite particle s form - factor and binding - energy vanishing , the composite particles becomes unstable and dissolve ( decay ) to their constitutes of sm elementary particles as final states . in the following , we discuss the decay and annihilation channels of the composite particles into the sm elementary particles , in particular two gauge bosons , as final states . 0.1 cm for the sake of simplicity , we adopt the first quark family @xmath199 and @xmath200 for illustrations , and the discussions can be generalized to other families including leptons . the @xmath36-charged composite boson @xmath201 or @xmath202 and its sm - gauge invariant mass term @xmath203 ( [ boundm ] ) . the complex scalar field @xmath101 has four components @xmath204 , @xmath205 and @xmath206 for @xmath207 . the second and third families of composite bosons are simply replicated by the replacements @xmath57 and @xmath58 . there is no mixing among three families of composite bosons , and the gauge interactions at the leading order ( tree - level ) of gauge couplings do not introduce the mixing of composite bosons ( [ 4comp ] ) with the counterparts in the second or third family . in eq . ( [ 4comp ] ) , the pseudo - scalars form composite meson states @xmath208 and @xmath209 carry the same quantum numbers of qcd pions @xmath210 and @xmath211 in the sm , however different masses @xmath212 . apart from the different kinematic threshold , these composite bosons @xmath213 undergo all decay channels of qcd pions . for the reasons that the properties of composite particles , e.g. , form - factors and masses ( or the bing - energy depth ) due to the strong - coupling dynamics , are unknown , we have not been able to calculate the total rate @xmath214 and width of the composite boson that decays to the final states of the sm elementary particles including the channel of two gauge bosons or two fermions . the total decay rate and width of the composite boson is very important for the collider phenomenology of the composite boson . nevertheless , at the leading order ( tree - level ) of gauge interactions , we are able to calculate the rates @xmath215 of the composite boson decaying to two gauge bosons as diboson final states , @xmath216 and their charged conjugated channels . these diboson channels ( [ pidecay ] ) and ( [ pidecay+ ] ) are expected to be the most energetically favorable and largest branching ratio , @xmath217 moreover , in the lhc experiments , the diboson final state and its kinematics might be more easily identified than the final state of two fermions ( quarks or two jets ) due to the background of the qcd dynamics . the decay rates and branching ratios ( [ bratio1 ] ) of the diboson channels ( [ pidecay ] ) and ( [ pidecay+ ] ) depend on the composite - boson mass @xmath218 and decay constant @xmath219 determined by the strong - coupling dynamics , analogously to the pion mass @xmath220 and decay constant @xmath221 determined by the qcd dynamics . however the following ratios of the branching ratios ( [ bratio1 ] ) for different decay channels , @xmath222 depend only on the sm gauge couplings at the energy scale @xmath223 tev , attributing to the peculiar gauge - interaction structure of the effective lagrangian ( [ comf ] ) and ( [ comc ] ) , as will be seen soon . these ratios ( [ bratio ] ) possibly provide the useful information for the on - going atlas and cms experiments in the lhc @xmath1 collisions . to further proceed it , it needs some more clarifications on the ratios ( [ bratio ] ) . both the composite pseudo - scalar boson @xmath208 and the composite scalar higgs boson @xmath69 decay to the two - photon state , however these two channels are not mixed . they have different parities , masses ( invariant masses ) and decay constants . the former composed by the strong - coupling dynamics in the uv - domain decay to two photons via the pseudo - scalar coupling ( see fig . [ triangle ] ) to the loop of @xmath122 and @xmath135 quarks , as described below . whereas the latter composed by the njl - dynamics in the ir - domain via the scalar coupling @xmath224 ( [ smf ] ) to a top - quark loop . in the lhc @xmath1 collision , the first family of composite particles made of @xmath122 and @xmath135 quarks are more probably produced , compared with other families of composite particles that are produced by gauge interacting vertexes . in the feynman diagram representation at the tree - level , see fig . [ triangle ] , these are the triangle diagram of quark loop with three interaction vertexes , one associates to the axial current representing the composite boson @xmath208 and other two are electroweak interactions associate to the sm gauge bosons . these feynman diagrams are the most relevant leading order contribution to the composite boson decay rate . in the rest frame of composite meson @xmath225 , the final states of two body decay follow the energy - momentum conservations @xmath226 and have the invariant mass @xmath227 . due to the electric coulomb energy , @xmath228 . their decay rates are expected to be the same as the pions decay rates , replacing the mass @xmath229 and the decay constant @xmath230 . in the neutral channel , we have @xmath231 where @xmath232 is the fine - structure ( fermi ) constant . the dipboson channel ( [ 0pidecayr ] ) is expected to have the largest branching ratio , analogously to the qcd pion decay . suppose that this channel ( [ 0pidecayr ] ) might be a candidate for the possible resonance of the diphoton channel of the invariant mass @xmath10 gev that the atlas and cms experiments are searching for . it implies that the composite meson mass @xmath233 gev , since there are other decay channels into two gauge bosons of @xmath3 , @xmath4 and @xmath5 . the channel @xmath3 corresponds to the final states of a photon and two leptons or a `` fat '' jet of two quarks . the channels @xmath4 and @xmath5 correspond to the final states of ( i ) two `` fat '' jets of four quarks for @xmath234 and @xmath235 hadronic decays or ( ii ) four leptons for @xmath234 and @xmath235 leptonic decay , or a `` fat '' jet and two leptons in the combination of cases ( i ) and ( ii ) . due to @xmath236 , the @xmath234 and @xmath235 bosons are approximately `` massless '' , and the rates of decay channels ( [ pidecay ] ) are approximately given by eq . ( [ 0pidecayr ] ) . these rates depend on the gauge couplings , see the effective lagrangian ( [ comf ] ) , which are actually the sm gauge couplings in eq . ( [ smf ] ) , since the sm gauge bosons couple to the quark or lepton inside the composite particle . using the @xmath237-coupling @xmath238 and the @xmath239-coupling @xmath240 to fermion fields , we can approximately estimate the decay - rate ratios @xmath241 where @xmath242 is the weinberg angle and @xmath243 is approximately adopted to have numbers . these relations provide the verification or falsification of the composite particle scenario , if the resonance of diphoton channel is confirmed and other diboson channels are also observed and measured . it is also possible that the composite boson @xmath208 decays into the dijet final state of two quarks which was already discussed in ref . suppose that the neutral composite meson @xmath208 is produced by @xmath1 collisions at the lhc , its resonance location @xmath244 , and the rate of each decay channel depends on the values of the mass @xmath245 , decay constant @xmath246 and sm gauge couplings . in the charged channels of final states of leptons or quarks , we have @xmath247 the charged quark channel ( [ pidecayr ] ) is favorable because of heavy quark mass . suppose that the charged composite meson @xmath209 is produced by @xmath1 collisions at the lhc , its resonance location at the invariant mass @xmath248 of the dijet final state or other possible final states produced by two quarks . whereas the charged lepton channel ( [ pidecayr1 ] ) implies the final state of @xmath249 lepton of energy @xmath250 and missing energy carried away by @xmath251-neutrino . the decay channels of dijet and dilepton were discussed in refs . @xcite . in eq . ( [ 4comp ] ) , the scalars form composite quarkonium states @xmath252 and @xmath253 carry the same quantum numbers of qcd quarknium states , however have larger masses @xmath254 . apart from different kinematic threshold , these composite quarkonium states in principle undergo all decay channels of qcd quarkonium states . the discussions are similar to the composite meson states @xmath213 , and will be presented in a lengthy article . recall that the direct decays of composite boson and quarkonium states into dijets or dilepton without an intermediate state were discussed @xcite . on the basis of the sm chiral gauge symmetries and vector - like composite fermion content , there are also other possible final decay channels , analogously to those studied in the effective theory of qcd at low energies . which channel is more relevant for detections , depending not only on its theoretical branching ratio in principle , but also on its experimental measurement in practice . the same analysis and discussion can be generalized to the decay rates of neutral and charged composite bosons made of charged lepton and/or neutrino pair ( [ boundl ] ) in the lepton channel . though the decay constant @xmath219 and mass @xmath218 should be approximately at the same scale , the rate should be @xmath255-time smaller than that of the quark channel ( [ 0pidecayr ] ) . however , the ratios of branching ratios for different diboson channels remain the same as ( [ br ] ) . beside , in the lhc @xmath1 collision , the production probability of leptonic composite bosons is smaller than hadronic one , due to the leptonic production rate is proportional to the small fine - structure constant @xmath256 . 0.1 cm for the sake of simplicity , we adopt the first quark family ( [ boundb]-[boundd ] ) to discuss the decay of composite dirac fermions , and discussions can be generalized to other families including leptons . the composite dirac fermions @xmath168 ( [ boundd ] ) are more massive than their constituents of composite boson @xmath101 ( [ boundb ] ) , i.e. , @xmath257 . they decay to the sm elementary particles via the intermediate composite - boson states @xmath208 ( neutral decay channel ) or @xmath209 ( charged decay channel ) or quarkonium states . suppose that at the present energy of @xmath1 collisions at the lhc , composite dirac fermions produced are non - relativistic particles , and we consider them to be approximately at rest in the cm frame of @xmath1 collision . the neutral decay channels are , see fig . [ fermiond ] , @xmath258 \rightarrow ( \gamma \gamma , \gamma z^0 , w^+w^- , z^0z^0 ) + { \rm a~ jet } , \label{fdecay}\end{aligned}\ ] ] and @xmath259 . this indicates that a composite dirac fermion @xmath168 decays into a composite meson @xmath208 and a fundamental dirac fermion , @xmath122-quark @xmath260 $ ] or @xmath135-quark . the latter is an ultra - relativistic quark , @xmath261 , ending as a jet in the final states . and the former is a composite meson , @xmath262 , appearing as an intermediate and metastable state for a short time @xmath263 then decays into two photons of the energy @xmath264 or other diboson channels ( [ pidecay ] ) . in the cm frame of the @xmath1 collision , the composite fermion at rest decays into a quark ( jet ) and a composite boson moving apart in opposite directions ( @xmath265 ) . the kinematic distribution of final states is described by the invariant mass @xmath266 , @xmath267 where @xmath268 is the angle between the momenta @xmath269 and @xmath270 of two photons or other diboson states ( [ fdecay ] ) . the rate of composite fermion decay ( [ fdecay ] ) should be the product of @xmath271-decay rate ( [ 0pidecayr ] ) and the rate of @xmath168 decaying into a quark and @xmath271 , which will be presented in a lengthy article . however , the ratios @xmath272 , @xmath273 and @xmath274 are similar to the ratios ( [ br ] ) . if the bosonic resonance ( [ pidecay ] ) is verified and its invariant mass is determined , the resonance of composite fermion decay ( [ fdecay ] ) could possibly be identified by measuring such a peculiar final - state kinematics ( [ fdecayk ] ) of a single jet and two photons or other diboson states . the charged decay channel via the intermediate meson state @xmath275 ( [ pidecayr ] ) reads @xmath276 , \label{fdecayc}\end{aligned}\ ] ] and its charge conjugate . the weyl fields @xmath277 $ ] do not form a dirac fermion of @xmath122- or @xmath135-quark , but pick up @xmath278 and @xmath279 quark from the vacuum , and end as @xmath122 and @xmath135-quark ( jets ) in final states . on the other hand , the charged composite mesons @xmath275 ( [ pidecayr ] ) most probably decay into two quarks . therefore , the most probable final state of composite fermion ( [ fdecayc ] ) decay is expected to be four jets formed by four quarks , which was already discussed in ref . these discussions can be generalized to the decay channels whose intermediate state is a quarkonium state instead of a meson state . the same analysis and discussion can be generated to the case of composite fermion ( [ boundem ] ) that is composed by a hadronic composite boson and a charged lepton ( or a neutrino ) , see fig . [ fermiond ] , @xmath280 \rightarrow ( \gamma \gamma , \gamma z^0 , w^+w^- , z^0z^0 ) + { \rm an~ electron } , \label{fldecay}\end{aligned}\ ] ] or @xmath281 . this indicates that a composite dirac fermion @xmath168 ( [ boundem ] ) decays into a composite meson @xmath208 and a fundamental dirac fermion , a charged lepton @xmath282 $ ] or a neutrino @xmath283 . the latter is an ultra - relativistic lepton . the kinematic distribution is the same as the one in pure hadronic channel ( [ fdecayk ] ) . however , the neutrino in the final states carries away mixing energy and momentum . it should be noted that the ratios @xmath284 , @xmath285 and @xmath286 are similar to the ratios ( [ br ] ) . the decay processes ( [ fdecay ] ) and ( [ fldecay ] ) discussed in this section are relevant to the on - going tev experiments in the lhc @xmath1 collision . these give the examples for the verification or falsification of the composite boson and fermion scenario , see eq . ( [ boundb]-[boundl ] ) . to end this section , we mention other decay channels . analogously to the discussion of charged hadronic channel ( [ fdecayc ] ) , the charged dirac fermions ( [ boundem ] ) decay via the intermediate meson state @xmath275 ( [ pidecayr ] ) reads @xmath287 , \label{fdecaych}\end{aligned}\ ] ] and its charge conjugate . the final states are a lepton pair and a jet pair . as for the decay of leptonic composite fermions ( [ boundl ] ) , final states are either two gauge bosons and a lepton for the neutral channel via the intermediate @xmath208 state or four leptons for the charged channel via the intermediate @xmath209 state . however the decay rates are much smaller due to the `` missing '' of the factor @xmath255 and the smallness of the fine - structure constant @xmath256 relating to the production of leptons in the lhc @xmath1 collision . 0.1 cm analogously to the annihilation of electron and positron into two photons in the qed , a composite dirac fermion and its antiparticle annihilates to two photons , and other massive gauge bosons , see fig . [ fanni ] , @xmath288 where the composite dirac fermion @xmath168 has an electric charge @xmath289 and mass @xmath290 . apart from the diphoton final state , the final states of energetic dibosons @xmath4 and @xmath5 are two `` fat '' jets in opposite directions , each of them is made by two energetic quarks , or these two bosons can decay into the final states of two lepton pairs . while in the channel @xmath291 , the final states are an energetic photon and a `` fat '' jet or an energetic photon and a lepton pair . given by the mass - energy of two annihilating composite dirac fermions , the kinematic mass - energy of final states must be larger than @xmath292 . this is not an invariant energy - mass representing the resonance of an unstable composite particle . suppose that massive composite dirac fermions produced by the lhc @xmath1-collision at present energies are non - relativistic particles , @xmath293 , @xmath294 tev . two gauge bosons in the final state ( [ anni ] ) are ultra - relativistic , the cross - section @xmath295 of annihilating channel ( [ anni ] ) can be approximately estimated by using the dirac rate of point - like electron and positron annihilating to two photons with the replacement @xmath296 and @xmath297 , @xmath298 up to the form - factors of composite dirac fermions . the annihilation rate @xmath299 per unit time @xmath300 where the number density of the produced composite fermions @xmath301 is assumed . we can obtain the ratios @xmath302 , @xmath303 and @xmath304 , similarly to the ratios ( [ br ] ) . analogously to the positronium state of electron and positron pair in the qed , it is possible that the composite fermion and its antiparticle form an intermediate unstable coulomb bound state , then decaying to photons or other massive gauge bosons . the spin singlet of bound state decays into two photons ( even number of photons ) with the probability @xmath305 , the spin triplet of bound state decays into three photons ( odd number of photons ) with the probability @xmath306 . the following annihilation channels of two composite fermions with different charges or zero charge are also possible , @xmath307 and their charged conjugates . in addition , analogously to the annihilation of electron and positron , through an intermediate @xmath308-photon or @xmath239-boson , to produce a pair of particle and antiparticle in the sm , a composite fermion and its antiparticle annihilates , through an intermediate photon or @xmath239-boson , and produce a pair of elementray particle and its antiparticle in the sm , @xmath309 where the dirac fermion @xmath168 has an electric charge @xmath289 . the final state is an energetic lepton pair or quark pair , the former is the dilepton channel and the latter is the dijet channel . similarly to the channels ( [ anni1 ] ) two composite dirac fermions with different charges or zero charge can annihilate , through an intermediate charged boson @xmath237 or neutral boson @xmath239 , to the pair of two elementary fermions in the sm . we skip the discussions on ( i ) two composite dirac fermions annihilate to two composite bosons @xmath310 and @xmath311 ; ( ii ) the rare processes of @xmath310 and @xmath311 annihilation and decay into gauge bosons and other particles through the @xmath312 coupling . 0.1 cm for readers convenience , we first give a very brief review of effective four - fermion operators and the effective theory of composite particles , which were presented in previous papers . we then focus the discussions on the decay and annihilation channels of composite bosons and fermions into the final states of the sm gauge bosons , leptons and quarks , in connection with the on - going searches of the atlas and cms experiments at energy @xmath0 tev . the justifications are provided for the verification or falsification of such a theoretical framework of massive composite particles and sm gauge couplings in the uv - domain of effective four - fermion operators . we end this article by making some speculative considerations for experiments . in the lhc @xmath1 collisions , the most probably channels of producing composite particles are the composite bosons and fermions ( [ boundb]-[boundm ] ) in the first family via the four - fermion operators @xcite @xmath313 + g(\bar\psi^{ia}_l\nu_{r})(\bar \nu_{r}\psi^{ia}_{l } ) , \label{uhlx}\end{aligned}\ ] ] and the channels of producing composite particles by other quark and lepton families have smaller rate because of involving small gauge interactions . the last term of eq . ( [ uhlx ] ) represents the interaction of quarks and dark - matter particle @xmath169 , which stands for the right - handed sterile neutrino of the first fermion family . due to the @xmath237- and @xmath239-boson couplings @xmath314 to two constituent quarks ( @xmath315 ) of composite fermions , in particular @xmath237-boson coupling to @xmath36 doublet @xmath316 , massive composite dirac fermions have the decay channels of final states : ( i ) dijets tagged by two highly boosted bosons ww , wz or zz produced by high - energy constituent quarks ( @xmath315 ) of composite dirac fermions , together with additional quark jets ; ( ii ) four quark jets formed by four high - energy constituent quarks ( @xmath315 ) of composite dirac fermions with a peculiar kinematic distribution @xcite . the former should have smaller rate because of the gauge coupling @xmath314 , although we are not able to calculate the rates of these channels . the final states can also be high - energy leptons , however the branching ratio of @xmath237 and @xmath239 decaying to leptons is about several times smaller than that to hadrons ( jets ) @xcite . the composite dirac fermion can also decay in the channel wh @xcite , where the higgs boson is produced by @xmath315-quarks fusing into a top - quark pair via a gluon , and its production rate is then related to the qcd coupling @xmath317 . we can have similar discussions for the case of composite bosons . these composite dirac fermions could be experimentally verified by possibly observing the resonances in the invariant masses ( @xmath318 ) and kinematic distributions of final channels measured . if the cm energy @xmath319 of lhc @xmath1-collisions is close to the masses @xmath11 of composite dirac fermions ( @xmath320 ) , then the invariant masses @xmath321 . suppose that the recent atlas and cms preliminary results @xcite of diboson resonances ( dijets tagged by two bosons ) with invariant masses in the range from 1.3 to 3.0 tev could be further confirmed . these resonances are expected to be also seen in the channels of four quark jets , whose invariant mass and event rate should be larger , provided that these resonances are attributed to massive composite dirac fermions at tev scale . the cms result @xcite of resonances with final states being two jets could include the event of four quark jets , two of them are geometrically close together to form a `` wide jet '' , which should be tagged through a study of its substructure and flavor . moreover , if composite dirac fermions are formed by the last operator in eq . ( [ uhlx ] ) , in addition to jets in final state , dark - matter particles @xmath169 carry away missing energy - momentum @xcite . we can have similar discussions for the case of composite bosons . last but not least , all sterile neutrinos ( @xmath322 ) and gauge - singlet ( neutral ) states of massive composite dirac fermions , e.g. , @xmath323 $ ] , can be possible candidates of warm and cold dark matter @xcite . they can couple or decay to the sm elementary particles in the following ways . ( i ) gauge - singlet ( neutral ) states of massive composite dirac fermions become unstable and decay to sm elementary particles . ( ii ) sterile neutrinos interact with the sm elementary particles via the last term of eq . ( [ uhlx ] ) for @xmath324 being quark or lepton @xmath36-doublets . ( iii ) the terms in eqs . ( [ bhlxl ] ) and ( [ bhlbv ] ) give the interactions @xmath325 , \label{bhlbv'}\end{aligned}\ ] ] among sterile neutrinos @xmath326 ( dark matter ) and sm particles , where the lepton @xmath36 doublets @xmath43 , singlets @xmath44 and the conjugate fields of sterile neutrinos @xmath45 ( @xmath327 ) , and quark fields @xmath50 and @xmath51 . it is thus expected that at the same tev scale where composite bosons and fermions ( [ boundb]-[boundm ] ) appear as resonances in the lhc @xmath1 collisions , the composite boson @xmath328 or @xmath329 and the composite fermion @xmath330 $ ] or @xmath331 $ ] should be formed by high - energy sterile neutrino inelastic collisions , e.g. @xmath332 via the first or second term in eq . ( [ bhlbv ] ) , and then decay to produce an excess of cosmic ray electron and positrons @xcite in space laboratories . in addition , recent ams-02 results @xcite show that at tev scale the proton flux in terms of energy changes its power - law index . this implies that there would be `` excess '' tev protons whose origin can be also explained by the resonance of composite bosons and fermions due to the interactions ( [ uhlx ] ) and ( [ bhlbv ] ) of dark - matter and normal - matter particles at tev scale . we also expect that at the same tev scale via the last term in eq . ( [ uhlx ] ) , the third and fourth terms in eq . ( [ bhlbv ] ) , composite dirac fermions , e.g. , @xmath333 $ ] , @xmath334 $ ] or @xmath335 $ ] should be formed as resonances by high - energy sterile neutrinos inelastic collisions with nucleons ( xenon ) at the largest cross - section , then they decay to produce some other detectable sm particles in underground laboratories @xcite . similarly , in the icecube experiment @xcite , we expect the events that tev neutrinos change ( lose ) their directions ( energies ) by the first term of the interaction ( [ bhlxl ] ) to form the resonances of composite bosons and fermions at tev scale . in these inelastic collisions , if the accessible cm energy @xmath336 , the cross section for the allowed inelastic processes forming massive composite bosons and fermions will be geometrical in magnitude , of order @xmath337 in the cm frame where massive composite bosons and fermions are approximately at rest . we have to confess that the detailed theoretical analysis of the composite boson and fermion phenomenology has not completely performed yet , due to its nonperturbative nature . this is analogous to the hadron physics in nonperturbative qcd . 0.1 cm the author thanks prof . zhiqing zhang for discussions on the lhc physics , and prof . hagen kleinert for discussions on the ir- and uv - stable fixed points of quantum field theories . y. nambu and g. jona - lasinio , phys . rev . 122 ( 1961 ) 345 . f. englert , r. brout , phys . ( 1964 ) 321 ; + p. w. higgs , phys . 12 ( 1964 ) 132 ; phys . rev . lett . 13 ( 1964 ) 508 ; phys . rev . 145 ( 1966 ) 1156 ; + g. s. guralnik , c. r. hagen , t. w. b. kibble , phys . ( 1964 ) 585 ; and t. w. b. kibble , phys . rev . 155 ( 1967 ) 1554 . y. nambu , in proceedings of the 1989 workshop on dynamical symmetry breaking , edited by t. muta and k. yamawaki ( nagoya university , nagoya , japan , 1990);v.a . miranski , m. tanabashi and k. yamawaki , _ mod . phys . lett . _ * a4 * ( 1989 ) 1043 ; _ phys . lett . _ * b221 * ( 1989 ) 117 ; + h. kleinert , the su(3)-extension of their work in chapter 26 of the textbook http://klnrt.de/b6 , " particles an quantum field world scientific publishing company , 2016 . s. weinberg , in understanding the fundamental constituents of matter , edited by a. zichichi ( plenum press , new york , 1977 ) ; + h. kleinert , the chapter 30 of the textbook http://klnrt.de/b6 , " particles an quantum field world scientific publishing company , 2016 . in fact , an excess of cosmic ray electron and positrons is shown around tev scale , see for example , j. chang , _ et al . _ nature vol 456|20 nov . 2008 ; y - z . fan , b. zhang and j. chang , ijmpd vol . 19 , ( 2010 ) 2011 . |
the electron spin transport in condensed matter , * spintronics * , is a subject of rapidly growing interest both scientifically and from the point of view of applications to modern and future electronics . in many cases the electron spin transport can not be described adequately without accounting for the hyperfine interaction between electron and nuclear spins . under extreme conditions of high magnetic fileds , ultra - low temperatures , ultra high isotopical cleanness etc . , the nuclear spins in these sytems are very promising candidates for the qubits : the basic elements of emerging quantum memory , logics and hopefully quantum computers . here we review the progress in the * nuclear spintronics * i.e. in physics and applications of hyperfine interactions in such exotic systems , as superconducting , quantum hall , mesoscopic and nano - systems.*pacs * : 73.20.dx ; 31.30.gs ; 71.70.ej ; 74.80.fp ; 03.67.lx ; 76.60.-k | cond-mat0403087 | the electron spin transport is playing growing role in the fastly developing new directions of the modern high - tech electronics , the so - called * spintronics * @xcite . in most solids the conduction electron system is coupled by hyperfine interaction @xcite to the underlying system of nuclear spins , which may profoundly modify the electron spin transport at low temperatures , thus leading to * nuclear spintronics * effects , which will be described in this review . on the other hand , the application of high magnetic fields is a very powerful tool for studying the electronic properties of a large variety of metals , semiconductors and superconductors . due to the landau quantization of electron motion in sufficiently strong magnetic field , most of the transport properties , such as magnetization , conductivity etc . experience magnetic quantum oscillations ( qo ) @xcite . the landau quantization is most spectacularly manifest in the electronic magneto - transport in low dimensional conductors . striking examples are the celebrated quantum hall effects ( qhe ) @xcite . apart from the anomalous enhancement of the well known qo in two - dimensional electron systems ( 2des ) one expects also strong qo in physical properties which are not sensitive to the magnetic field in isotropic three dimensional metals . it was suggested in @xcite that in quasi - two - dimensional metals under strong magnetic fields the nuclear spin lattice relaxation rate @xmath0 should exhibit strong magnetic oscillations . this should be compared with the korringa relaxation law @xcite usually observed in three - dimensional normal metals , which results in a magnetic field independent nuclear spin - relaxation rate . this line of research seems to be useful in dense quasi - two - dimensional electronic systems , as is the case of synthetic metals ( gic s etc . ) and low - dimensional organic compounds . a completely new line of research , the hyperfine interaction between nuclear and electron spins in low dimensional and correlated electron systems , has been developed during the two last decades both theoretically [ 8 - 39 ] , and experimentally [ 40 - 59 ] . the growing attention is attracted by the hyperfine interactions in a ) qhe : theory [ 8 - 19 ] and experiment [ 40 - 53 ] ; b ) mesoscopics and nano - systems : theory [ 20 - 26 ] and experiment [ 54 - 55 ] ; c ) normal and superconducting metals : theory [ 27 - 31 ] and experiment [ 58 - 59 ] . very recently nuclear spin memory devices @xcite and the indirect hyperfine interaction via electrons between nuclear spin qubits in semiconductor based quantum computer proposals [ 33 - 39 ] attracts sharply growing attention . here i will outline the main theoretical concepts and some experimental achievements in the new and quickly developing field of * nuclear spintronics*. in metals and doped semiconductors , usually , the leading contribution to the spin - lattice relaxation process is due to the _ hyperfine fermi contact _ interaction between the nuclear spins and the conduction electron spins @xcite . this interaction is represented by the hamiltonian : @xmath1 where @xmath2 is the nuclear gyromagnetic ratio , @xmath3 is the nuclear spin and @xmath4 is the magnetic field on the nuclear site , produced by electron orbital and spin magnetic moments : @xmath5 here @xmath6 is the electron radius - vector , @xmath7 is the electron spin operator , @xmath8 is the bohr magneton and @xmath9 is the electronic @xmath9-factor . the nuclear spin - lattice relaxation rate @xmath10 , caused by the hyperfine fermi contact interaction between the nuclear spins and the conduction electron spins , is related to the local spin - spin correlation function through the equation : @xmath11 where @xmath12 , @xmath13 are the transverse components of the electron spin density operator at the nuclear position * r * , and @xmath14 is the nuclear magnetic resonance frequency . the rate of the nuclear spin - relaxation in metals is , usually , proportional to the temperature and to the square of the electronic density of states at the fermi energy ( the korringa law @xcite ) . this follows from the following expression : @xmath15\delta ( e_{f}-e_{i}+\gamma _ { n}h_{0}).\ ] ] at low temperatures : @xmath16\propto k_{b}t$ ] @xmath17 , where @xmath18 is the temperature in the energy units , we arrive at the linear in temperature dependence of @xmath19 @xmath20 which is the well known korringa law @xcite . here @xmath21 is the electron density of states at the fermi level . in high magnetic fields and in systems with reduced dimensionality this simple argumentation does not hold , since the electron spectrum acquires field induced @xcite or size quantized @xcite energy gaps . it was conjectured in @xcite and confirmed experimentally , see e.g. @xcite , that in qhe systems the nuclear spin relaxation rate should have an activation behavior @xmath22 where @xmath23 is either @xmath24 , the electron zeeman gap ( odd filling factors ) or @xmath25 , the landau levels gap ( even filling factors ) , instead of the usual korringa law . the discreteness of the electron spectrum manifests , at finite temperatures , in an activation type of the magnetic field dependence of the nuclear spin relaxation rate , @xmath26as it seen from eq . ( [ t1exp ] ) . this behaviour is similar to that of the magnetoresistance @xmath27 in the qhe , @xcite , see fig . 2 . .,width=480 ] this unusual magnetic field dependence of the nuclear spin relaxation reflects the fact that the energy gaps in the spectrum of two - dimensional electrons in strong magnetic fields ( either zeeman splitting or the landau levels gap ) are three orders of magnitude larger than the nuclear zeeman energy . indeed , the energy needed to reverse the spin of an electron in the external magnetic field @xmath28 is @xmath29 , which is much larger ( by a factor of @xmath30 , @xmath31 and @xmath32 being the nuclear and free electron masses ) than the energy @xmath33 provided by reversing the nuclear spin . therefore the delta function in eq . ( 2 ) can not be realized and the simultaneous spin flip of the nuclear and the electron spins ( flip - flop ) in landau levels , fig . 3 , is severely restricted by the energy conservation . on the other hand , in isotropic @xmath34 electron systems , in a strong magnetic field : @xmath35 , the kinetic energy of the electron motion parallel to the field should change in order to ensure the energy conservation of the process shown in fig . 4 . thus , in the isotropic @xmath34 model , the electron spin - flip will be accompanied by a simultaneous change of the landau level and of the kinetic energy parallel to the field , e@xmath36 according to @xmath37 while this is impossible for an ideal @xmath38 system in a strong magnetic field , it may take place in quasi - two - dimensional conductors , as is the case in superlattices , for certain regions of parameters . electron systems.,width=437 ] because of the existence of energy gaps in the electron spectrum of a 2des under strong magnetic fields ( the qhe systems ) , finite nuclear spin relaxation times @xmath39 could be expected only if 2des is subjected to different kinds of external potentials . broadened landau levels may overlap , thus providing a nonzero density of states of both spin projections at the fermi energy . this will result in finite relaxation time for excited nuclear spins interacting with the conduction electrons on the landau levels . description of these processes , performed in the framework of the finite temperature matzubara diagram techniques , is presented in @xcite . an important question in the quantum hall effect theory is the role of the edge states : electronic orbits magnetically confined to the sample boundary . while the number of edge states may be small compared to the bulk states , their contribution to @xmath39 can be very important , since they possess a homogeneous energy spectrum @xcite . in sufficiently clean heterojunctions , however , where the fractional qhe ( fqhe ) and wigner crystallization could be observed , the mechanisms mentioned above are extremely inefficient . at finite temperatures the energy conservation in the flip - flop process can be fulfilled by absorbing a phonon [ 13 - 15 ] . the mechanism , considered in @xcite , consists of two processes in which , first , electron and nuclear spins are simultaneously reversed by the hyperfine interaction , and , second , only electron spins are reversed by their coupling to the lattice strain . in these processes , a phonon is either absorbed or emitted to satisfy the energy conservation law . both the zeeman split - level energy ( @xmath40 ) and chemical potential ( @xmath41 ) depend on a magnetic field . the dominant contribution to nuclear relaxation comes from the transition between two spin - split levels that are near the chemical potential . the contributions from other landau levels that are far away from the chemical potential are proportional to @xmath42 $ ] and are negligible under the qhe condition . the value at a minimum depend strongly on the value of exchange interaction . at low temperatures ( @xmath43 ) , the nuclear spin relaxation rate varies exponentially with inverse temperature ( i.e. , @xmath44 $ ] ) because the number of thermally excited phonons in the reservoir becomes exponentially small as the temperature is decreased . at higher temperatures ( @xmath45 ) , on the other hand , the relaxation rate varies linearly with temperature ( i.e. , @xmath46 ) . nucleus - mediated spin - flip transitions in gaas quantum dots have been studied in @xcite . in these papers the electron spin relaxation times were calculated taking into account the mechanism , where the phonons provide the necessary energy and nuclear spins take care for the spin conservation during the electron transition between zeeman split states in quantum dots . the nuclear spin relaxation in 2deg , associated with the dipole - dipole interaction between the nuclear and the electronic spins was studied by ovchinnikov et al . the principle physical difference between these two mechanisms is that while in the contact interaction the total spin should be conserved and an electron has to flip its spin in order to relax the nuclear spin , the conserved quantity in the dipole - dipole interaction is the total ( spin plus orbital ) momentum . therefore an electron can relax a nuclear spin without changing its own spin state , just by shifting its center of orbit . this mechanism should be prevailing in clean samples , i.e. under the fqhe conditions . much of the recent attention paid to hyperfine interactions under conditions of the quantum hall effect is connected with correlation effects in 2des . this is based on the notion of a spin - exciton : the elementary excitation over the zeeman gap dressed by the coulomb interaction @xcite.this results in a strong enhancement ( up to a factor of 100 , as is the case in gaas ) of the effective @xmath47 -factor . spin exciton and skyrmion dispersion : the energy gap of large momentum excitons is two times larger than the skyrmion s one . in the insets the @xmath38 electron spin arrangements for spin excitons ( left ) and skyrmions ( right ) is shown.,width=443 ] due to the coulomb interaction the spin - excitons are bound states of electron - hole pairs which , unlike the individual electrons or holes , can propagate freely under the influence of a magnetic field due to their zero electric charge . these elementary excitations are , therefore , chargeless particles with a nearly parabolic dispersion in the low - k limit . at @xmath48 the gap is equal to the bare zeeman splitting . the energy spectrum of spin - excitons on the ground landau level : n=0 is @xcite : @xmath49{\frac{e^{2}}{\kappa a_{h } } } . \label{espex0}\ ] ] in the parabolic approximation ( small exciton momenta ) , the dispersion relation reads : @xmath50 where @xmath51 is the definition of the spin - exciton mass . the invariance of the energy gap with respect to the electron- electron interaction is associated with the fact that in creating a quasielectron - quasihole pair excitation at the same point in space ( i.e. with center of mass momentum @xmath48 ) the energy decrease due to the coulomb attraction is exactly cancelled by the increase in the exchange energy . thus the energy gap for the creation of a widely separated ( i.e. with @xmath52 ) quasielectron - quasihole pair ( large spin - exciton ) is equal to the exchange energy associated with the hole . external potentials , like long range potential fluctuations , periodically modulated heterostructures etc . , may reduce substantiality the spin - exciton gap . in @xcite the spin - exciton in a periodically modulated two - dimensional electron gas under strong magnetic fields was investigated . it was shown there that a periodic external potential , with a period in the submicron range , can remove the energy gap in the spin - exciton dispersion at certain values of the spin - exciton momenta . this may result in several interesting phenomena . for example , the rate of nuclear spin relaxation and depolarization processes in heterojunctions will be strongly enhanced . it is easy to see why the zeeman energy gap can be removed by the modulation potential in an ideal model of noninteracting electrons : consider a noninteracting electron - hole pair with opposite spins , which are subject to an external magnetic field and a periodic potential with amplitude @xmath53 and period @xmath54 by appropriately selecting the spatial separation between the electron and the hole along the modulation direction the energy difference between them can be reduced to zero . the minimal amplitude @xmath53 required to satisfy this condition is @xmath55 , the spatial separation between the electron and the hole involved is half of the modulation period @xmath56,and the corresponding wave number ( along the y - direction ) is @xmath57 . the picture is significantly more complicated , however , in the presence of the strong coulomb interaction , where bound spin - excitons exist . maniv et al . @xcite , have considered the effect of vacuum quantum fluctuations in the qh ferromagnetic state on the decoherence of nuclear spins . it was shown there that the virtual excitations of spin excitons , which have a large energy gap ( on the scale of the nuclear zeeman energy ) above the ferromagnetic ground state energy , lead to fast incomplete decoherence in the nuclear spin system . it is found that a system of many nuclear spins , coupled to the electronic spins in the @xmath38 electron gas through the fermi contact hyperfine interaction , partially loses its phase coherence during the short ( electronic ) time @xmath58 , even under the ideal conditions of the qhe , where both @xmath39 , and @xmath59 are infinitely long . the effect arises as a result of vacuum quantum fluctuations associated with virtual excitations of spin waves ( or spin excitons ) by the nuclear spins . the incompleteness of the resulting decoherence is due to the large energy gap of these excitations whereas the extreme weakness of the hyperfine interaction with the @xmath38 electron gas guarantees that the loss of coherence of a single nuclear spin is extremely small . the manipulation of the nuclear spins is carried out through spin flip - flop processes , associated with the transverse part of the interaction hamiltonian @xmath60 , i.e. @xmath61 $ ] , where @xmath62 are the transverse components of the nuclear spin operators , and @xmath63 are the corresponding components of the electron spin density operators . here @xmath64 , @xmath65 are the electron field operators with spin projections @xmath66 . the longitudinal part of @xmath60 , @xmath67 , which commutes with the hamiltonian @xmath68 , and so leaves the nuclear spin projections along @xmath69 unchanged , can still erode quantum coherence in the nuclear spin system @xcite . to simplify the analysis it is assumed that the nuclei under study have spin @xmath70 , and so the corresponding spin operators are expressed in terms of fermionic creation and annihilation operators , @xmath71 , as @xmath72 in this case the transverse components @xmath73 , @xmath74 , are up to a proportionality constant , just the off diagonal elements ( or coherences ) of the density matrix of a single nuclear spin ( qubit ) @xcite . the decay of these elements with time , which determines the rate of decoherence of a single qubit , can be thus found from the equations of motion for the operators @xmath71 in the heisenberg representation @xmath75 considering a single nuclear spin and evaluating its rate of decoherence due to the coupling with a bath of spin excitons , assuming that initially , at time @xmath76 , the electronic system is in its ground ( qh ferromagnetic ) state @xmath77 that the time dependence of the coherence @xmath78 is given by @xmath79 , where @xmath80.$ ] this result is similar to the expression found by palma et al . @xcite in an artificial model of pure decoherence , i.e. when energy transfer between the qubit and its environment is not allowed . the remarkable feature of this expression is due to the presence of the energy gap @xmath81 in the spin exciton spectrum , which is typically much larger than the nuclear zeeman energy @xmath82 . during a short time scale , of the order of @xmath58 , the coherence @xmath83 of a single nuclear spin diminishes and then saturates for a very long time ( i.e. of the order of the relaxation time @xmath59 ) at @xmath84 , where @xmath85 ^{2}}% \sim 1.\ ] ] for @xmath86 heterostructure the coupling constant @xmath87 is typically of the order of @xmath88 @xcite . iordanskii et al . @xcite have studied nuclear spin relaxation taking into account the creation of spin - excitons @xcite in the flip - flop process . the energy for the creation of a spin - exciton can be provided by the long range impurity potential in a process , where the electron turns its spin while its center of orbit is displaced to a region with lower potential energy . as shown in fig . 6 , the overlap of the initial and final location of the electron wave functions , centered at @xmath90 and @xmath91 respectively , is : @xmath92 $ ] . here @xmath93 is the nuclear position . nuclear spin relaxation by the conduction electron spin in the vicinity of a potential fluctuation is effective when the nuclear spin is positioned in the region of the overlapping initial and final states of the electron wave function . the energy conservation in the spin - exciton creation process can be written in the form : @xmath94 this expression defines the gradient of electric potential caused by the impurity , sufficient to create a spin - exciton during a flip - flop process . the probability of finding such a fluctuation is exponentially small : @xmath95.$ ] the momentum * p * of the spin - exciton is small and therefore the expansion in p can be performed everywhere in the final expression for electronic density of states ( dos ) : @xmath96 here @xmath97 depends on the following combination of physical parameters : @xmath98 @xmath99 reaches the maximum , usually , at @xmath100 . a complete expression for the nuclear depolarization rate from the golden rule formula at t=0 @xcite is given by @xmath101 where @xmath102 are filling factors for the electrons with spin up and spin down , respectively , n is the highest occupied landau level and @xmath103 are the laguerre polynomials . here the system of units is used , where @xmath104 and @xmath105 . the dynamics of the nuclear spins is governed by their interactions with each other and with their environment . in the regime of interest , these interactions are mediated by 2des . various time scales are associated with this dynamics . the relaxation time @xmath39 is related to energy exchange and thermalization of spins . quantum mechanical decoherence / dephasing will occur on the time scale @xmath59 . theoretical calculation of the nuclear - spin dephasing / decoherence time scale @xmath59 for in qhe systems with impurities was presented in @xcite . the hyperfine interaction between the electron and the nuclear spins , @xmath107 , can be split into two parts @xmath108 where @xmath109 corresponds to the coupling of the electrons to the diagonal part of nuclear spin operator @xmath110 , and @xmath111 to its off - diagonal part . it follows @xcite , that @xmath112 can be incorporated into the nuclear - spin energy splitting , redefining the hamiltonian of the nuclear spin as @xmath113 , where @xmath114 . the relevant terms in the full hamiltonian can be expressed solely in terms of the nuclear - spin operators and spin - excitation operators @xcite . it is assumed there that initially , at time @xmath76 , the nuclear spin is polarized , while the excitons are in the ground state , @xmath115 where @xmath116 is the ground state of spin - excitons . since the hamiltonian conserves the total number of elementary excitations in the system , the most general wave function can be written as @xmath117 corresponding to the nuclear spin in the ground state and @xmath118 describing the single - exciton state with the wave vector @xmath119 . equations of motion for the coefficients @xmath120 and @xmath121derived from the schr - dinger equation are : @xmath122 the relaxation rate and the added phase shift of the nuclear - spin excited - state probability amplitude @xmath123 are given by the real and imaginary parts of the pole , respectively : @xmath124 and @xmath125 so that @xmath126 $ ] . it is obvious that due to the large gap in the spin - exciton spectrum , @xmath127 , the energy conservation in the flip - flop process can not be satisfied , and so in the absence of interaction with impurities , @xmath128 . interactions with impurities will modify these solution , and , as a consequence , the energy conservation condition . in particular , if the impurity potential is strong enough , it can provide additional energy to spin - excitons , so that their energy can fluctuate on the scale of order @xmath129 thus making nuclear - spin relaxation possible . this mechanism was identified in @xcite , and it corresponds to large fluctuations of the impurity potential @xmath130 , which usually occur with a rather small probability , so @xmath39 is very large for such systems . the perturbative solution does not describe the energy relaxation ( @xmath39 ) , but it does yield the phase shift due to the impurity potential . this phase shift , when averaged over configurations of the impurity potential , produces a finite dephasing time , @xmath59,which can be calculated considering the reduced density matrix of the nuclear spin , given by @xmath131_{u}. \label{eq26moz}\ ] ] here the trace is partial , taken over the states of the spin - excitons , while the outer brackets denote averaging over the impurity potential . the trace over the spin - excitons can be carried out straightforwardly because within the leading - order perturbative approximation used here they remain in the ground state ; all excitations are virtual and contribute only to the phase shift . the diagonal elements of @xmath132 are not influenced by virtual excitations and remain constant . the off - diagonal elements of @xmath133 contain the factors @xmath134 $ ] . it is the averaging of these quantities over the white - noise impurity potential @xmath130 that yields dephasing of the nuclear spin . this averaging can be done by utilizing the relation @xmath135 _ { u}=\exp [ -{\frac{1}{2}}\left ( [ \phi ] _ { u}\right ) ^{2 } ] $ ] . apart from the direct nuclear spin relaxation , important information about the electron system can be obtained from nuclear spin diffusion processes . this is the case when the nuclear spins are polarized in a small part of a sample as it was experimentally observed in @xcite . to explain these experimental observations , bychkov et al @xcite have suggested a new mechanism for indirect nuclear spin coupling via the exchange of spin excitons . the spin diffusion rate from a given nuclear site @xmath136 within the polarized region is proportional to the rate of transition probability p(@xmath136 ) for the polarization of the nuclear spin @xmath137 , located at @xmath136 , to be transferred to a nuclear spin @xmath138 , positioned at @xmath139 , outside the polarized region , via the exchange of virtual spin excitons , fig . the virtual character of the spin - excitons , transferring the nuclear spin polarization , removes the problem of the energy conservation , typical for a single flip - flop process . furthermore , the virtual spin - excitons are neutral entities , which can propagate freely in the presence of a magnetic field . in this model the electron interactions play a crucial role : the kinetic energy of a spin exciton is due to the coulomb attraction between the electron and the hole . thus the proposed mechanism yields the possibility of transferring nuclear spin polarization over a distance much longer than the magnetic length @xmath140 . the long range nature of this mechanism is of considerable importance when the size of the region of excited nuclear spins , @xmath141 , is much larger than the magnetic length @xmath142 . as it is shown in @xcite , the potential of the nuclear spin - spin interaction , mediated by the exchange of spin - excitons , is a monotonic function of the distance between the two nuclei with the asymptotics : @xmath143 where @xmath144 this is typical for the interaction , mediated by the exchange of quasiparticles with an energy gap at @xmath145 , as is the case for the spin - exciton dispersion . the range @xmath146 of this potential is determined by the critical wave number @xmath147 as it follows from the uncertainty principle : @xmath148 . the negative sign of this interaction corresponds to attraction between the nuclear spins and may cause , at sufficiently low temperatures , a ferromagnetically ordered nuclear state in qhe systems . spin - excitons constitute the building blocks , from which more complex spin textures can be formed . for example , at finite densities the spin excitons condense @xcite into skyrmion - antiskyrmion pairs . these unusual topological excitations @xcite in the spin distribution of real 2des were observed , in nmr experiments , near filling factor @xmath149 @xcite . further evidences for skyrmions in 2des were found in transport and optical experiments . skyrmions , in qhe systems , are the topologically nontrivial spin excitations around filling factor @xmath149 @xcite which arise as a condensate of interacting spin excitons @xcite . the coulomb interaction acts to enlarge the skyrmion size while the zeeman splitting tends to collapse skyrmions . the interplay between these factors determines the final distribution of spins within a skyrmion , and its characteristic length scales . the resulting radius @xmath150 corresponds to the region where both these energies are of the same value , and grows weakly to infinity as the g - factor goes to zero @xcite , thus reflecting the importance of the long range coulomb repulsion associated with the skyrmion charge in the zero g - factor limit . a canonical @xmath151 transformation from the fully polarized ground state , where all spins are oriented along a single axis , to a state , consisting of a macroscopic number of differently oriented spins , each of which is slightly rotated with respect to its nearest neighbors in space was is developed in @xcite . it is found also that contrary to the lowest landau level , these energies are positive for both skyrmions and antiskyrmions . it should be noted that the spin - rotation transformation employed in @xcite is unitary and does not change the total number of electrons . thus by going to the new state @xmath152 from the fully polarized ground state @xmath153 , is exactly equal to half the total energy required to create a well separated electron - hole pair ( large spin exciton ) . if the zeeman splitting is not neglected , however , the increase in magnetic energy associated with the rotated spins determines a length scale for the spatial size of the defect . to take into account zeeman spin splitting the hartree - fock ( hf ) energy functional should be corrected by the additive term @xmath154 @xcite , where @xmath155 is the zeeman splitting energy . the corresponding equation for the vectorial field @xmath156 can be obtained by variation of the corrected energy functional with respect to @xmath157 under the constraint @xmath158 . the result of such a calculation is : @xmath159 , \label{eqforn}\ ] ] where @xmath160 and @xmath161 is the effective bohr radius ( note that @xmath162 is the free electron mass ) . in nmr experiments on skyrmions [ 43,46 - 48,51 ] the nuclear spins are strongly polarized . the sample inhomogeneity may result in a strong inhomogeneity of the hyperfine field , fig . 8 , and therefore spatially varying electron zeeman splitting . this may result in a strong localization of skyrmions @xcite , resulting in the combined topologically nontrivial electron - nuclear spin excitation , the skyr - nuon , fig . as we have seen in previous sections , the nuclear spin polarization , once created , remains finite for macroscopically long times @xcite . intensive experimental studies @xcite of this phenomenon have provided a more detailed knowledge on the hyperfine interaction between the nuclear and electron spins in heterojunctions and quantum wells . it was observed that the nuclear spin relaxation time is rather long ( up to 10@xmath163 ) and the hyperfine field acting on the charge carriers spins is extremely high , up to 10@xmath164 @xcite . the nontrivial physics in this subject is based on a fact that the discrete nature of the electron spectrum in these systems will result in exponentially long dependence of the nuclear spin relaxation times @xmath39 on the system parameters @xcite , i.e. t@xmath165 @xmath166 ( here @xmath167 is the electron energy level spacing and @xmath168 is the temperature ) . we assume here that similar law should take place also in the nanostructures with well defined size quantization of the electron spectrum . note that in this case @xmath39 is very sensitive to the potential fluctuations , caused by the inhomogeneous distribution of impurities in a heterojunction @xcite . a family of new physical effects in nanostructures with strong spin - orbital coupling appears when the electron spin degeneracy is lifted by a hyperfine field of polarized nuclei . indeed , the combined action of a strong nuclear polarization and the spin - orbit interaction , breaks the time reversal symmetry in a mesoscopic system . a good illustration for such * meso - nucleo - spinic * effects is the hyperfine aharonov - bohm effect ( habe ) in mesoscopic rings @xcite , caused by hyperfine interaction , as is explained in what follows . persistent currents ( pc ) in mesoscopic rings reflect the broken clock wise - anticlock wise symmetry of charge carriers momenta caused , usually , by the external vector potential . experimentally pcs are observed when an adiabatically slow time dependent external magnetic field is applied along the ring axis . the magnetic field variation results in the oscillatory , with the magnetic flux quantum @xmath169 ( or its harmonics ) period behavior of the diamagnetic moment ( the pc ) , which is the manifestation of the usual aharonov - bohm effect ( abe ) . it was suggested in @xcite that in a quantum ring with a nonequilibrium nuclear spin population the persistent current will exist , even in the absence of external magnetic field . it is shown there , that the abe like oscillations of pc with time will appear , during the time interval of the order of nuclear spin relaxation time @xmath39 . the physics of this phenomenon can be understood along the following lines . the hyperfine field , caused by the nonequilibrium nuclear spin population breaks the spin symmetry of charged carriers . combined with a strong spin - orbital ( so ) coupling , in systems without center of inversion @xcite , it results in the breaking of the rotational symmetry of diamagnetic currents in a ring . under the topologically nontrivial spatial nuclear spin distribution , the hyperfine field produces an adiabatically slow time variation of the berry phase of the electron wave function . the time variation of this topological phase results in observable oscillations of a diamagnetic moment ( the pc ) . it is one of a series of meso - nucleo - spinic effects , which may take place in mesoscopic systems with broken symmetry , due to the combined action of the hyperfine field and spin - orbital interaction . due to the contact hyperfine interaction eq . ( [ hint1 ] ) it follows that once the nuclear spins are polarized , i.e. if @xmath170 , the charge carriers spins feel the effective , hyperfine field @xmath171 which lifts the spin degeneracy even in the absence of external magnetic field . in gaas / algaas one may achieve the spin splitting due to hyperfine field of the order of the one tenth of the fermi energy @xcite . let us suppose therefore that the charge carriers spin orientation is partially polarized during the time interval of the order of @xmath39 . it is quite obvious that the topologically nontrivial spin texture combined with the spin - orbit interaction will result in a pc . taking the bychkov - rashba term @xcite @xmath172 \mathbf{\nu } , \label{hsorashba1}\ ] ] where @xmath173 for holes with @xmath174 ( @xmath175 is the free electron mass ) , and @xmath176 for electrons , @xmath177,@xmath178 are the charge carrier spin and momentum and @xmath179 is the normal to the surface . it can be rewritten in the form @xmath180 where @xmath181 @xmath182 stands for a nonequilibrium carriers spin population . under the conditions of a topologically nontrivial orientation of @xmath183 the wave function of a charge carrier encircling the ring gains the phase shift similar to the one in an external magnetic field like in the ordinary abe . this phase shift can be estimated as follows @xmath184 where @xmath185 is the ring perimeter . to observe the oscillatory persistent current connected with the adiabatically slow time - dependent @xmath186 @xmath185 is supposed to be less than the phase breaking length . taking the realistic values for @xmath187 and @xmath188 we estimate @xmath189 which shows the experimental feasibility of this effect . the standard definition of the spontaneous diamagnetic current is @xmath190 where f is the electron free energy and @xmath191 is the external ( probe ) magnetic flux . the oscillations of pc arise due to the exponential time dependence of the phase @xmath192 in eq . ( [ phi1 ] ) , with the time constant @xmath193from the analogy with the standard abe in a mesoscopic ring , we expect the following form for a persistent current in an one dimensional quantum ring at low enough ( t @xmath194 ) temperature @xmath195 here @xmath196 is the initial phase value . the marking difference between the periodical time dependence of standard aharonov - bohm oscillations , which are observed usually under the condition of linear time variation of the applied magnetic field and the hyperfine driven oscillations which die off due to the exponential time dependence of the nuclear polarization . in @xcite it was proposed to use the inhomogeneous hyperfine field to create so - called nuclear spin polarization induced ( nspi ) magnetic structures , such as magnetic quantum dots , wires , rings , superlattices etc . the spin splitting ( @xmath197 ) due to a hyperfine magnetic field can be comparable to the fermi energy of 2deg , so that the electrons in the region , where nuclear spins are polarized , will occupy the energetically more favorable states with the spins opposite to @xmath198 . the inhomogeneous nuclear polarization acts on the electrons as the effective confining potential @xmath199 . this effective confining potential can be used to create different nanostructures with polarized electrons in them . the method of local nuclear spin optical polarization allows to create different nspi quantum structures ( quantum dots , rings , wires , etc . ) using the same sample and different illumination masks . the time evolution of the hyperfine magnetic field in nspi quantum structures , due to the nuclear spin relaxation and the nuclear spin diffusion , leads to variation of the number of transverse modes and corresponding electron energies at a constant gate potential that can be directly measured by transport experiments . the dependence of the conductance at zero temperature on the number of transverse modes in the conductor is given by the landauer formula @xmath200 , where @xmath168 is the average electron transmission probability , @xmath201 is the number of the transverse modes and the factor @xmath202 stands for the spin degeneracy . it is assumed that the transition probability @xmath168 is independent of the energy in a small interval between the chemical potentials of the reservoirs . usually , the number of the transverse modes , defined by the effective width of the conductor , is controlled by the gate voltage and the conductance is changed in discrete steps @xmath203 . it is underlined in @xcite that due to the spin selective effective potential the height of the conductance steps is just @xmath204 which is a half of the conductance quantum @xmath205 . the experimental feasibility is based on the method of optical nuclear spin polarization @xcite . the resolution of the optical illumination of the sample can be high enough . usual optic technique allows to create the light beams of the width of the order of the wave length ( @xmath206 ) , by using near fields optics the beam width can be sufficiently reduced ( @xmath207 ) . hence a nspi qw of the width of @xmath208 can be created by the modern experimental technique . in semiconductor heterostructures having supreme quality , the electron mean free path can be as large as @xmath209 and nspi qw will operate in the quantum regime . in bulk semiconductors , the study of the conduction electron magnetotransport under the influence of the nonequilibrium nuclear spin polarization was started a long time ago by measurement of the magnetoresistance in bulk insb @xcite . very recently a refined detailed study of the influence of a weak hyperfine field on the hall effect in bulk insb under a strong magnetic field ( one occupied landau level ) was performed @xcite . apart of the usual hall effect , caused by the drift of electrons in the crossed electric and magnetic fields , there exist also another contribution to the nondiagonal components of the conductivity tensor , the anomalous hall effect @xcite ( ahe ) , which is caused , in nonmagnetic semiconductors , by the spin - orbit interaction combined with the carrier magnetization . it was proposed in @xcite that in the strong nuclear spin polarization limit and at low magnetic field this effect may give a leading contribution to the anomalous hall effect . due to this interaction , electrons with their spin polarization parallel to the magnetization axis will be deflected at right angles to the directions of the electric current and of the magnetization while electrons with antiparallel spin polarization will be deflected in the opposite direction . thus , if the two spin populations are not equal there appears a net current in the transverse direction . until now , studies of the anomalous hall effect have been limited to the case where the carrier magnetization is induced by the magnetic field . the magnetic field , however , produces a much larger normal hall effect which makes experimental studies quite difficult . recording the hall voltage in a sample with a dynamically polarized nuclear spins , during a time scale of the order of the nuclear spin relaxation time , which can be very long on the scale of typical electron equilibration processes , gives a direct measure of the spin dependent component of the hall voltage . if the experiment is designed in such a way that the hyperfine field is much higher then the external magnetic field the hyperfine field driven anomalous hall effect can be observed . insb is well suited for observation of the hahe effect because of the strong spin - orbit coupling and the strong , due to the high atomic numbers of the atoms involved , hyperfine interaction ( @xmath210 ev , @xmath211 ev , @xmath212 , @xmath213 ) . the maximal nuclear - spin polarization for all isotopes of insb yields @xmath214 @xcite . it is outlined in @xcite that the hahe _ _ _ _ gives the possibility to observe genuine spin dependent anomalous hall effect which relies on the natural relaxation of nuclear spins and facilitates the experiment since the anomalous hall voltage is not hidden by the much larger ordinary hall effect . the possibility that the hyperfine interaction between the conduction electron spins and nuclear spins may result in hyperfine residual resistivity ( hrr ) _ _ _ _ in clean conductors at very low temperatures was studied theoretically by dyugaev et al . @xcite . apart from the fundamental nature of this problem , the natural limitations on the mean free path are decisive in semiconductor based high speed electronic devices , like heterojunctions and quantum wells . the space periodicity of nuclei plays no specific role , as long as the nuclear spins are disordered and act as magnetic impurities @xcite with the concentration @xmath216 . this scattering is not operative at extremely low temperatures , in the @xmath217 region when the nuclear spins are ferro- or antiferromagnetically ordered . the hrr , arising due to the fermi ( contact ) hyperfine interaction between the nuclear and the conduction electron spins can be written in atomic units as : @xmath218 here @xmath219 @xmath220 are the electron and the nucleon masses , respectively ; @xmath221 ev and @xmath222 is the fine structure constant . in metals the effective electron - nuclear interaction constant is @xmath223 where the fermi energy @xmath224 varies in a wide interval @xmath225 . @xmath226 is @xmath227 for @xmath228 and @xmath229 for the rare earth metals . the total residual resistivity is therefore a sum of the impurity @xmath230 and the nuclear spin @xmath231 contributions : @xmath232 here @xmath233 in atomic units : @xmath234 sec . the nuclear contribution to resistivity starts to be operative when the impurity concentration is @xmath235 . in the limit of an ideally pure ( @xmath236 ) metal the universal residual resistivity @xmath237is , therefore @xmath238 and the mean free path is limited by @xmath239 cm . this yields @xmath240 cm in @xmath228 and @xmath241 cm for the rear earth metals . it is interesting to note that in materials with even - even nuclei ( zero spin ) , like @xmath242 , @xmath243 , @xmath244 , @xmath245 and isotopically clean graphite @xmath246 , where the electron - nuclear scattering is absent , the _ hrr _ would not be observed . the temperature and magnetic field dependence of the _ hrr _ contains reach information on the hyperfine interaction between the conduction electrons and the nuclear spins . usually , the temperature and the magnetic field dependence of residual resistivity due to nonmagnetic impurities is due mostly to the mesoscopic effects , and is vanishing in the limit @xmath247 . in a magnetic field such that @xmath248 the magnetic impurities freeze out and the kondo effect is quenched . in order to freeze out the nuclear spins however one should apply much higher magnetic fields , @xmath249 . therefore in the temperature interval @xmath250 @xmath251 the nuclear spin contribution may prevail even in metals with magnetic impurities . in metals like @xmath252 the nuclear magnetic moments @xmath253 and even without external magnetic field their @xmath254 degeneracy is lifted partially by the quadrupole effects ( in the case of cubic crystal symmetry the quadrupole splitting of the nuclear levels may happen due to the defects @xcite and dislocations @xcite ) . while the normal metals have a quite similar electronic structure , the experimentally observed temperature dependence of the dephasing time @xmath255 is quite different . this was shown in @xcite , where the value of @xmath256 was defined by the magnetoresistance measurements of long metallic wires @xmath257 in a wide temperature interval @xmath258k . in @xmath259 and @xmath260 wires @xcite@xmath261 saturates at low temperatures which contradicts the standard theory @xcite . strangely enough the ag wires do not show saturation @xcite at the lowest temperatures , in accordance with @xcite . it is conjectured in @xcite that the influence of the quadrupole nuclear spin splitting on the phase coherence time @xmath256 can be a clue to this puzzle . indeed , the nuclear spins of both @xmath259 and @xmath260 have a strong quadrupole moment ( @xmath262 ) and may act as inelastic two - level scatterers once their degeneracy is lifted by the static impurities @xcite , dislocations @xcite and other imperfections . it is known that ( see @xcite and references therein ) the nondegenerate two level scatterers may introduce inelastic phase breaking scattering of conduction electrons . this may be not the case for @xmath263 nuclei since their spin is @xmath264 . in this case the quadrupole splitting of nuclei spins by imperfections is negligible . in the absence of magnetic zeeman splitting therefore the nuclear spins in @xmath263 samples will act just as a set of elastic scatterers , and the temperature dependence of @xmath256 should obey the standard theory @xcite . the problem of coexistence of the superconducting and magnetic ordering , in spite of its long history @xcite , is still among the enigmas of modern condensed matter physics . most of the theoretical and experimental efforts were devoted to studies of the coexistence of electron ferromagnetism and superconductivity . the possibility of a reduction of @xmath265 by the nuclear ferromagnetism was outlined by dyugaev et al . in @xcite and theoretically studied in more details in @xcite . it was experimentally observed by several groups @xcite . it was recently conjectured @xcite that the hyperfine part of the nuclear - spin - electron interaction may result in the appearance of a nonuniform superconducting order parameter , the so called fulde - ferrel - larkin - ovchinni - kov state ( fflo ) @xcite . the fflo state was thought originally to take place in superconductors with magnetically ordered magnetic impurities @xcite . the main difficulty , however , in the observation of the fflo caused by magnetic impurity ordering is in the simultaneous action of the electromagnetic and exchange parts of the magnetic impurities on the superconducting order . in most of the known superconductors the part destroys the superconducting order before the exchange part modifies the bcs condensate to a nonuniform fflo state . the situation may change drastically in the case of nuclear spin ferromagnetic ordering . indeed , the nuclear magnetic moment @xmath266 is at least three orders of magnitude smaller than the electron bohr magneton @xmath267 so that the electromagnetic part of the nuclear spin fields is quite low , compared to that of the magnetic impurities . on the other hand the exchange part is strongly dependent on the nuclear charge @xmath268 . a growing number of models for electron and nuclear spin based memory cells @xcite and quantum information processing ( or quantum computing - qc ) has been recently proposed @xcite-@xcite . we will concentrate in what follows on the models based on the manipulation of nuclear spins in bulk semiconductor @xcite , heterostructures @xcite and quantum wires and rings @xcite . in @xcite we have proposed a quantum computer realization based on hyperfine interactions between the conduction electrons and nuclear spins embedded in a two - dimensional electron system in the quantum - hall effect . for modifications and improvements of this model see a recent review @xcite . maniv et al . @xcite have suggested a following physical process of preparing a coherent state in qh ferromagnets . let us assume that at time @xmath269 the filling factor was tuned to a fixed value @xmath270 and then kept constant until @xmath76 . if @xmath271 then at @xmath76 the nuclear spin system is in the ground state corresponding to the @xmath38 electron system at @xmath272 . suppose that at time @xmath76 the filling factor is quickly switched ( i.e. on a time scale much shorter than @xmath273 ) back to @xmath149 so that the nuclear spin system is suddenly trapped in its instantaneous configuration corresponding to @xmath274 . thus the nuclear spins for a long time @xmath275 ( @xmath276 @xmath273 ) will find themselves almost frozen in the ground state corresponding to the 2des at @xmath272 , since @xmath277 . the main idea behind this scenario stems from the experimental observation of a dramatic enhancement of the nuclear spin lattice relaxation rate @xmath278 , and of a sharp decrease of the knight shift in optically pumped nmr measurements , as the filling factor is shifted slightly away from @xmath279 . the prevailing interpretation of these closely related effects , associates them with the creation of skyrmions ( or antiskyrmions ) in the electron spin distribution as the @xmath38 electron system moves away from the quantum hall ferromagnetic state at @xmath149 . the off - diagonal long range magnetic order @xcite existing in this state , corresponds to a complex order parameter which is coupled locally through the hyperfine interaction to the nuclear spins @xcite . it is thus expected that the proposed manipulation of the nuclear spin system can be performed in a phase coherent fashion over a spatial region with size of the order of the skyrmion radius . the latter effect is still sufficiently weak to enable the survival of a coherent state of a large number of qubits in the computer memory at @xmath149 . to find an upper bound for the size of such a memory let us consider @xmath280 independent nuclear spins located at various sites @xmath281 in the quantum well . a number @xmath282 , stored in the computer memory , corresponds to the direct product of @xmath280 pure nuclear spin states @xmath283 @xmath284 @xmath285 , where @xmath286 , and @xmath287 is a nuclear state with spin projection @xmath288 located at the site @xmath289 . to be able to start a significant quantum computing process , however , a coherent superposition of such products , i.e. @xmath290 ( see e.g. @xcite ) , should be prepared at time @xmath291 . this superposition may be represented more transparently for our purposes by the direct product of mixed spin up and spin down states , @xmath292 , with the normalization @xmath293 . the mixing is expected to take place most efficiently via the flip - flop processes with the electron spins during the manipulation period when @xmath294 . let us further assume that the initial state @xmath295 is a completely coherent state , so that each of the numbers of length @xmath280 has an equal probability . this is a typical state required in carrying out efficient quantum calculations @xcite . in this state @xmath296 for any @xmath297 and @xmath298 it is evident that due to the complete coherence of the initial state @xmath299 the survival probability @xmath300 @xmath301 ^{n}\approx \exp \left\ { -\frac{1}{2}n\left [ 1-\func{re}j\left ( t\right ) \right ] \right\ } \approx e^{-\frac{1}{2}n\gamma \left ( t\right ) } \ ] ] depends only on the decoherence factor @xmath302 . the decay of @xmath303 therefore follows @xmath304 $ ] , saturating at @xmath305 $ ] for @xmath275 @xmath276 @xmath58 . despite the much larger drop in the level of coherence , the time scale over which the coherence diminishes is the same as in the case of a single qubit . the most popular and intensively developed is the kane model @xcite , where the qubits are the nuclear spin of the @xmath306p donor placed near to the surface of _ si_. the nuclear spins are manipulated by electric gates above the _ si _ surface via the atomic electrons . the two - qubit interaction is provided by the overlapping outer electrons of two neighboring @xmath306p donors . the kane model has stimulated detailed theoretical and experimental activity . the strength of the hyperfine interaction is proportional to the probability density of the electron wavefunction at the nucleus . in semiconductors , the electron wavefunction extends over large distances through the crystal lattice . two nuclear spins can consequently interact with the same electron , leading to electron - mediated or indirect nuclear spin coupling . voltages applied to metallic gates in a semiconductor device may be used tocontrol the hyperfine interaction and , correspondingly , the electron - mediated interaction between the nuclear spins . a quantum mechanical calculation proceeds by the precise control of three external parameters : \i ) gates above the donors control the strength of the hyperfine interactions and hence the resonance frequency of the nuclear spins beneath them ; \ii ) gates between the donors turn on and off electron - mediated coupling between the nuclear spins ; \iii ) a globally applied a.c . magnetic field b@xmath307 flips nuclear spins at resonance . custom adjustment of the coupling of each spin to its neighbors and to b@xmath308 enables different operations to be performed on each of the spins simultaneously . finally , measurements are performed by transferring nuclear spin polarization to the electrons and determining the electron spin state by its effect on the orbital wavefunction of the electrons , which can be probed using capacitance measurements between adjacent gates . an important requirement for a quantum computer is to isolate the qubits from any degrees of freedom that may lead to decoherence . if the qubits are spins on a donor in a semiconductor , nuclear spins in the host are a large reservoir with which the donor spins can interact . consequently , the host should contain only nuclei with spin _ i = 0_. this simple requirement unfortunately eliminates all _ iii - v _ semiconductors as host candidates , because none of their constituent elements possesses stable _ i=0 _ isotopes . iv _ semiconductors are composed primarily of _ i = 0 _ isotopes and can in principle be purified to contain only _ i = 0 _ isotopes . the only _ i = 1/2 _ shallow ( group _ v _ ) donor in si is @xmath306p . at sufficiently low @xmath306p concentrations at temperature _ t = 1.5 k _ , the electron spin relaxation time is thousands of seconds and the @xmath309p nuclear spin relaxation time exceeds 10 hours . it is likely that at millikelvin temperatures the phonon limited @xmath306p relaxation time is of the order of 10@xmath310 seconds @xcite , making this system ideal for quantum computation . recent calculations of relaxation and decoherence times for this system are presented in @xcite . in @xcite a new system is proposed , which consists of nuclear spins ( qubits ) embedded into a mesoscopic ring , fig . 11 , or a finite length quantum wire , fig . 12 . the hyperfine interaction of the electrons in the system with nuclear spins leads to an effective indirect nuclear spin interaction @xmath311 where @xmath312 is magnetic moment of a nucleus , @xmath313 and @xmath314 are functions of the system parameters presented in fig it is obtained in @xcite that the effective nuclear spins interaction exhibits sharp maxima , fig . 13 , as function of the magnetic field and nuclear spin positions which opens the way to manipulate qubits with almost atomic precision . the selective nuclear spin interaction can be obtained by changing external parameters of the system , as is shown in fig . 14 . in @xcite the expression for indirect nuclear spin interaction between nuclear spins in a mesoscopic system , consisting of a finite length quantum wire or ring with nuclear spins placed in the zero nuclear spin matrix , is obtained . interaction between any two qubits , which is necessary for two - qubit operations , is performed by the electrons in the wire . it is very sensitive to the system parameters : nuclear spin location , number of electrons , magnetic field and geometry of the system . its dependence on the system parameters is completely different from indirect nuclear spin interaction in @xmath38 and @xmath34 metals : by varying the external parameters ( magnetic field and number of electrons ) one can control with almost atomic precision the nuclear spin interaction strength by creating maxima of the amplitude of electron wave function on some qubits and zero on the other . the connections to the measuring system , preparation of the initial state and performing of the one - qubit operations using the nmr could be similar to the existing experimental suggestions @xcite . the decoherence time of the nuclear spins in mesoscopic systems is expected to be long enough to perform the quantum computation , since the discrete electron spectrum in mesoscopic systems imposes restriction on the flip - flop processes and the nuclear spin relaxation time at law temperatures is expected to have an activation behavior . ) , as function of the magnetic flux through the ring.,width=390 ] the measurement of the nuclear spin - relaxation in heterojunctions is a challenging experimental problem , since the direct detection of the nmr signals in solids requires usually @xmath315 nuclei . the number of nuclear spins interacting with the two - dimensional electrons is however , much smaller : @xmath316 . the first successful measurements of the magnetic field dependence of @xmath26 under qhe conditions were performed in a series of elegant experiments by the k. von klitzing group , @xcite . combining electron - spin resonance ( esr ) and resistivity measurement techniques they have observed the shifting of the esr resonance frequency by the hyperfine field of nonequilibrium nuclear spin population , which is the well known overhauser shift @xcite . in this experiment the @xmath38 electron zeeman splitting is tuned to the pumping frequency . the angular momentum gained by a 2deg electron , excited to the upper zeeman branch , is then transferred to the nuclear spins , thus creating a nonequilibrium nuclear spin population . these measurements show a close similarity between the magnetic field dependence of the nuclear spin - relaxation rate and the magnetoresistance in quantum hall effect , as it was suggested theoretically in @xcite , thus demonstrating clearly the importance of the coupling of nuclear spins to the conduction electron spins in the nuclear relaxation processes in these systems . various experimental techniques were used since and in what follows we will describe shortly the main developments and achievements in experimental studies of the hyperfine coupling between the nuclear spins and the electrons in qhe , mesoscopic and superconducting systems . another way of measuring the nuclear spin relaxation and diffusion in a heterojunction under strong magnetic field by transport techniques ( spin - diode ) was demonstrated by kane et al . they have reported measurements performed on spin diodes : junctions between two coplanar 2deg s in which @xmath317l on one side and @xmath318 1 on the other . the fermi level e@xmath319 crosses between spin levels at the junction . in such a device the 2deg is highly conducting except in a narrow region ( with a width of the order of several hundred angstroms ) where @xmath320 wald et al . @xcite presented the experimental evidence for the effects of nuclear spin diffusion and the electron - nuclear zeeman interaction on interedge state scattering . polarization of nuclear spins by dc current has proven to be a rich source for new , not yet understood completely , phenomena . this is the case , for example of anomalous spikes in resistivity around certain fractional filling factors , observed by kronmuller et al . @xcite and in resistively detected nmr in qhe regime reported by desrat et al . influence of nonequilibrium nuclear spin polarization on hall conductivity and magnetoresistance was observed and studied in detail by gauss et al . @xcite . in 1994 barrett et al . @xcite observed , for the first time , a sharp nmr signal in multi - quantum wells , using the lampel @xcite technique of polarizing the nuclear spins by optical pumping of interband transitions with near - infrared laser light _ ( opnmr)_. polarization of nuclei results in a significant enhancement of the nmr sensitivity , since the resonance in a two - level system results in equalizing their population . the difference in the population is , obviously , maximal when the spins are completely polarized . detailed studies of the knight shift data suggested @xcite that the usually accepted picture of electron spins , aligned parallel to the external field , should be modified to include the possibility of topologically nontrivial nuclear spin orientations , the skyrmions . optical polarization of nuclear spins was used also as a tool for reducing the zeeman splitting of @xmath38 electrons by kukushkin et al . this resulted in a noticeable enhancement of the skyrmionic excitations . similar results are reported by vitkalov et al . @xcite . very recently , a modern ultra - sensitive nmr spin - echo technique was employed to study the physics of qhe @xcite in gaas / algaas multi - quantum well heterostructures . the spin polarization of 2des in the quantum limit was investigated and the experimental data support the noninteracting composite fermion model in the vicinity of the filling factor @xmath321 . using the same technique the polarization of 2des near @xmath322 was investigated @xcite . it was shown there that a quantum phase transition from a partially polarized to a fully polarized state can be driven by increasing the ratio between the zeeman and coulomb energies . an amazing phenomenon , following from the hyperfine coupling between the electron and the nuclear spins , is the giant enhancement of the low temperature heat capacity of gaas quantum wells near the filling factor @xmath323 , discovered in 1996 by bayot et al . @xcite . as other thermodynamic properties , it experiences quantum oscillations , following from the oscillatory density of states @xmath324 at the fermi level . at about @xmath325 , and in clean samples , bayot et al @xcite have observed anomalous deviations from the free electron model , in the specific heat value ( up to four orders of magnitude ) for the filling factor in the range @xmath326 @xmath327 . their explanation is that in this interval of parameters , the electron system couples strongly to nuclear spin system with a concentration of several orders of magnitude larger than the electron one . this raises a question about the origin of the strong coupling between the electron and the nuclear spins in the interval @xmath326 @xmath327 . the guess is the skyrmions , since they are predicted to appear just in the same interval of the filling factor . additional support for this mechanism is in the results of @xcite , where the disappearance of the nuclear spin contribution to the heat capacity was reported , as the ratio between the zeeman and coulomb energies exceeds a certain critical value . the zeeman splitting of electrons was modified in these experiments by tilting the magnetic field . a new very promising technique for measuring spatially varying nuclear spin polarization within a gaas sample is reported in @xcite . in the force detected nmr ( fdnmr ) the sample is mounted on a microcantilever in an applied magnetic field . a nearby magnetic particle creates a gradient of magnetic field which exerts a force on the magnetized sample and triggers the cantilever oscillations . fdnmr is capable to perform the magnetic resonance imaging of the sample with a very high accuracy . this method can be useful in defining the spatial distribution of nuclear spin polarization in non homogeneous samples . this information may be crucial for understanding different peculiarities of data obtained by previously described methods . a new world of the low - temperature physics of the hyperfine interactions in superconducting metals opens in the @xmath328 region , where the nuclear spins start ordering , thus reducing the critical magnetic field of superconductors as has been recently discovered by the pobell group @xcite . they have studied the magnetic critical field @xmath329 of a metallic compound @xmath330 where the superconductivity sets up at @xmath331 they have observed , in @xmath332 , the nuclear spin ferromagnetic ordering at @xmath333 it was observed in these experiments that the magnetic critical field @xmath334 g is lowered by almost a factor of two at @xmath335 in writing this review the author benefited enormously from the colleagues and co - workers : p. averbuch , m. azbel , s.e . barret , c. berthier , yu . bychkov , a. dyugaev , e. ehrenfreund , v.i . falko , v. fleurov , l.p . gorkov , v.l . gurevich , m. horvati , s.v . iordanskii , a.g.m . jansen , k. von klitzing and his group , k. kikoin , ju.h . kim , b.i . lembrikov , d. loss , t. maniv , s. meshkov , d. mozyrsky , yu . ovchinnikov , yu . pershin , v. privman , f. pobell , m. potemski , late a. rozhavsky , a. shik , i. shlimak , b. spivak , p.c.e . stamp , p. wyder , a. zyuzin , whose knowledge , intuition and professional skills were crucial in going through the jungle of hyperfine interactions in such exotic systems , as superconducting , quantum hall , mesoscopic and nano - sytems . a. abragam , _ the principles of nuclear magnetism _ ( clarendon , oxford , 1961 ) ; c.p . slichter , _ principles of magnetic resonance _ ( springer verlag , berlin , 3rd ed , 1991 ) ; j. winter , _ magnetic resonance in metals _ ( clarendon press , oxford , 1971 ) . k. von klitzing , g. dorda , and m. pepper , phys . * 45 * , 494 ( 1980 ) ; d.c . tsui , h.l . strmer , and a.c . gossard , phys . . lett . * 48 * , 1559 ( 1982 ) ; r.b . laughlin , phys . lett . * 50 * , 1395 ( 1983 ) ; _ perspectives in quantum hall effect _ , eds : s. das sarma and a.p . pinczuk ( wiley , new york , 1997 ) . mani , w.b . johnson , v. narayanamurti , v. privman , and y.h . zhang , physica e * 12 * , 152 ( 2002 ) ; j.m . tailor , c.m . markus , and m.d . lukin , phys . * 90 * , 206803 ( 2003 ) ; a. imamoglu , e. knill , l. tian , and p. zoller , phys . lett . * 91 * , 017402 ( 2003 ) . p. khandelwal , n.n . kuzma , s.e . barrett , l.n . pfeiffer , and k.w . west , phys . lett . * 81 * , 673 ( 1998 ) ; a.e . dementyev , p. khandelwal , n.n . kuzma , s.e . barrett , l.n . pfeiffer , and k.w . west , phys . 83 * , 5074 ( 1999 ) ; n.n . kuzma , p. khandelwal , s.e . barrett , l.n . pfeiffer , and k.w . west , science * 281 * , 5377 , 686 ( 1999 ) . s. melinte , e. grivei , v. bayot , and m. shayegan , phys . lett . * 82 * , 2764 ( 1999 ) ; s. melinte , n. freytag , m. horvatic , c. berthier , l.p . levy , v. bayot , and m. shayegan , phys . lett . * 84 * , 354 ( 2000 ) . a. bychkov , t. maniv , i.d . vagner , and p. wyder , phys . rev . lett . * 73 * , 2911 ( 1994 ) ; t. maniv , yu . a. bychkov , a. kaplunovsky , and i.d . vagner , physica b * 204 * , 134 ( 1995 ) ; yu.a . bychkov , t. maniv . vagner , and p. wyder , europhys . lett . * 40 * , 557 ( 1997 ) . lee and c.l . kane , phys . * 64 * , 1313 ( 1990 ) ; s.l . sondhi , a. karlhede , s.a . kivelson , and e.h . rezayi , phys . b * 47 * , 16419 ( 1993 ) ; l. brey , h.a . fertig , r. cote , and a.h . macdonald , phys . lett . * 75 * , 2562 ( 1995 ) . for a review see s.m . girvin , in : _ topological aspects of low dimensional systems _ , ed . : a. comtet , t. jolicoeur , s. ouvry , and f. david ( springer - verlag , berlin and les editions de physique , les ulis , 2000 ) ; cond - mat/9907002 ( 1999 ) . b. altshuler , a. aronov , and d. khmelnitskii , j. phys . c * 15 * , 7367 ( 1982 ) ; b. altshuler and a.g . aronov , in : _ electron - electron interaction in disordered systems _ , eds : a.l . efros and m. pollak ( north - holland , amsterdam , 1985 ) . |
this paper introduces a markov chain approach that allows a rigorous analysis of agent based opinion dynamics as well as other related agent based models ( abm ) . by viewing the abm dynamics as a micro description of the process , we show how the corresponding macro description is obtained by a projection construction . then , well known conditions for lumpability make it possible to establish the cases where the macro model is still markov . in this case we obtain a complete picture of the dynamics including the transient stage , the most interesting phase in applications . for such a purpose a crucial role is played by the type of probability distribution used to implement the stochastic part of the model which defines the updating rule and governs the dynamics . in addition , we show how restrictions in communication leading to the co existence of different opinions follow from the emergence of new absorbing states . we describe our analysis in detail with some specific models of opinion dynamics . generalizations concerning different opinion representations as well as opinion models with other interaction mechanisms are also discussed . we find that our method may be an attractive alternative to mean field approaches and that this approach provides new perspectives on the modeling of opinion exchange dynamics , and more generally of other abm . : * agent based models * , * opinion dynamics * , * markov chains * , * micro macro * , * lumpability * , * transient dynamics * . 37l60 , 37n25 , 05c69 . ( ) mathematical physics , bielefeld university ( germany ) + ( ) dream & science factory , marseilles ( france ) + ( ) iseg , universidade tcnica de lisboa ( tulisbon ) and + research unit on complexity in economics ( uece ) + ( ) institute for complexity sciences ( icc ) , lisbon ( portugal ) + 2truecm | 1108.1716 | [ section - introduction ] recent improvements in multidisciplinary methods and , particularly , the availability of powerful computational tools are giving researchers an ever greater opportunity to investigate societies in their complex nature . the adoption of a complex systems approach allows the modeling of macro sociological or economic structures from a bottom up perspective understood as resulting from the repeated local interaction of socio economic agents without disregarding the consequences of the structures themselves on individual behavior , emergence of interaction patterns and social welfare . agent based models ( abm ) are at the leading edge of this endeavor . when designing an agent model , one is inevitably faced with the problem of finding an acceptable compromise between realism and simplicity . if many aspects are included into the agent description , the model might be plausible with regard to the individual behaviors , but it will be impossible to derive rigorous analytical results . in fact , it can even be very hard to perform systematic computations to understand the model dynamics if many parameters and rules are included . on the other hand , models that allow for an analytical treatment often oversimplify the problem at hand . in abm , we can find the whole spectrum between these two extremes . while simplicity is often favored by physicists in order to be able to apply their well developed tools from statistical physics , more realistic descriptions are often desired by researchers in the humanities because they are interested in incorporating into the model a reasonable part of their qualitative knowledge at the micro and macro scales . both views have , of course , their own merits . our paper is a contribution to interweaving two lines of research that have developed in almost separate ways : the markov chain approach and abms . the former represents the simplest form of a stochastic process while the latter puts a strong emphasis on heterogeneity and social interactions . the main expected output of our markov chain strategy applied to abm is a better understanding of the relationship between microscopic and macroscopic dynamical properties . moreover , we aim to contribute not only to the understanding of the asymptotic properties of abm but also to the transient mechanisms that rule the system on intermediate time scales . for practical purposes this is the most relevant information for two reasons : first , in our case the chains are absorbing , so the asymptotic dynamics is trivial and second , they describe the evolution of the system before external perturbations take place and possibly throw it into a new setting . agent based opinion models , a particular case of abm , are among the most simple models in the literature and are therefore a suitable starting point for the analysis . especially for _ binary opinion models _ several results have been obtained by previous authors using analytical tools , as shown in the review on social dynamics by castellano and co - workers @xcite . the most intensively studied model is the _ voter model _ , originally developed by kimura and weiss @xcite as a model for spatial conflict of two species ( see also refs . the analysis of binary opinion models is usually based on mean field arguments . the microscopic agent configuration is mapped onto an aggregate order parameter , and the system is reformulated on the macro scale as a differential equation which describes the temporal evolution of that parameter . a mean field analysis for the voter model on the complete graph was presented by slanina and lavicka in ref . @xcite , and naturally , we come across the same results using our method as the first part of approach . slanina and lavicka derive expressions for the asymptotic exit probabilities and the mean time needed to converge , but the partial differential equations that describe the full probability distribution for the time to reach the stationary state is too difficult to be solved analytically ( @xcite , pag.4 ) . analytical results based on the same methods have been obtained for the voter model on @xmath4dimensional lattices ( @xcite ) as well as for networks with uncorrelated degree distributions ( @xcite ) . one step to a more realistic agent description ( though still a caricature ) is achieved by allowing the agents to make n ary choices . among the most popular models that realize this is the axelrod model @xcite , which uses vectors as state variables and bounded confidence ( @xcite ) . in both models , the interaction probability is a function of the agent similarity such that similar agents tend to interact and so become more similar in the interaction . in the axelrod model as well as in other bounded confidence models , this leads to the emergence of clustering as agents converge in homogeneous subgroups while an appropriate distance in between these subgroups increases . an analytical treatment of this process is already quite difficult ( see ref . @xcite for an approximate mean field analysis ) . our markov chain approach shows how restrictions in the agent communication lead to the emergence of new absorbing states in the associated markov chain which correspond to system states where different opinions co exist . the usefulness of the markov chain formalism in the analysis of more sophisticated abms has been discussed by izquierdo and co - workers ( @xcite ) , who look at 10 well known social simulation models by representing them as a time homogeneous markov chain . among these models are the schelling segregation model ( @xcite , for which some analytical results are available , for example , in refs.@xcite ) , the axelrod model ( considered above ) and the sugarscape model from epstein and axtell @xcite . the main idea of izquierdo and co - workers @xcite is to consider all possible configurations of the system as the state space of the markov chain . despite the fact that all the information of the dynamics on the abm is encoded in a markov chain , it is difficult to learn directly from this fact , due to the huge dimension of the configuration space and its corresponding markov transition matrix . the work of izquierdo and co - workers mainly relies on numerical computations to estimate the stochastic transition matrices of the models . in our opinion , a well posed mathematical basis for these models may help the understanding of many of their observed properties . linking the micro description of an abm to a macro description in the form of a markov chain provides information about the transition from the interaction of individual actors to the complex macroscopic behaviors observed in social systems . in particular , well known conditions for lumpability make it possible to decide whether the macro model is still markov . conversely , this setting can also provide a suitable framework to understand the emergence of long range memory effects . in sociology , the concepts of micro and macro have long been an important subject of analysis . different but related meanings have been advocated by different authors ( see ref . @xcite and references therein ) , running from `` micro as dealing with individuals and macro as dealing with populations '' to `` micro as social processes that engender relations among individuals and macro as the structure of different positions in a population and their constraints on interaction '' . in any case the terms micro and macro relate in this context the action of individuals or small groups based on their mutual relations and the emergence of collective societal scopes . one of the first acknowledged synthetic formulations of this linkage between micro and macro in sociology studies is from max weber ( @xcite , pag.29 ) from where we quote the following basic observation : `` within the realm of social action , certain empirical uniformities can be observed , that is , courses of action that are repeated by the actor or ( simultaneously ) occur among numerous actors '' . we shall see how a stylized version of this belief is incorporated in our study when passing from micro to macro dynamics . talcott parsons @xcite later introduced the notion of internalization , posing an interesting question on the retroaction of the macro on the micro level of the models , to which we shall come back in section [ section - projected ] . @xcite for a discussion of this topic . on the other hand our work is certainly related to the social network literature , another old outstanding branch of sociology . sociograms , an important tool in social studies , had already been introduced by jacob levy moreno in the 30s , ( @xcite , @xcite ) inducing a graph representation of social relations , thus opening the way to social network based research . this approach has been developed since the seminal studies of john barnes @xcite and it expanded rapidly with the work of mark granovetter @xcite and many others . in our case , however , as in plenty of other models the emphasis is on the dynamics of the processes rather than on the structure alone ( @xcite and references therein ) . one of the key notions to understand the micro macro linkage is emergence . we follow ref . @xcite to describe various levels of emergence present in the sociological literature and comment on the links with our work . first the so called _ descriptive emergence _ problem : are `` the macroproperties a common property of many microunits '' or are they something fundamentally different ? this question is clearly addressed in our work . in fact , starting from a `` simple aggregative procedure '' @xcite on the individual attributes to define the macro level , we show how the nonlinearity features of dynamical process may naturally create `` something else '' at least in a stylized model . next , related to parson s phenomenon of internalization _ practical emergence _ stands for the possible discrepancy , `` if the macrostructural properties of a social system no longer correspond to the internalized rules and interactions of the individual actors '' , which leads to `` a practical problem for the individuals acting in this system '' . at the end of section [ subsection - delta2-transient ] we discuss a possible quantification , in a very rough sense , of this `` falling out '' ( @xcite , pag.339 ) of the micro and macro levels . finally , `` far more controversial '' ( @xcite , pag.339 ) _ explanatory emergence _ stands for a process leading to autonomous dynamics of structures ( @xcite , pags.280 - 303 ) . we shall comment on this possibility in section [ section - generalizations ] , without really answering the question , but indicating how it could be taken into account in the framework of our models . the paper is organized as follows : in section [ section - microdynamics ] , starting from the idea introduced by izquierdo and co - workers @xcite we rigorously treat a class of abm as markov chains . our main idea is to give a solid basis to the link from a micro to a macro description in the abm context and fully explore the potential of this construction , a task to which sections [ section - projected ] and [ section - opinion - projected ] are devoted . generalizations concerning different opinion representations as well as opinion models with other interaction mechanisms are then discussed in section [ section - generalizations ] . we find that our method may be an attractive alternative to mean - field approaches and this approach can provide new perspectives on the modeling of opinion exchange dynamics , and more generally of other abm . we end up with some final remarks and a prospective for further work in section [ section - conclusion ] . [ section - microdynamics ] here we consider the class of abm defined by a set n of agents , each one characterized by individual attributes that are taken in a finite list of possibilities . the meaning or the level of abstraction of the content of such attributes is not important as long as it can be codified in a finite , possibly very large , set of possibilities . the agents should also be able to regularly update their attributes according to the information conveyed from other agents as well as to their actual state . the collective updating process of the attributes of the agents at each time step consists of two parts . first a random choice of a subset of agents is made according to some probability distribution @xmath5 . the number of agents chosen at each time may also be random . then the attributes of the agents are changed ( updated ) according to some rule , called a dynamical rule , which depends on the the subset of agents selected at this time . we denote by @xmath6 the individual state or attribute space and we call the configuration space @xmath7 the set of all possible combination of attributes of the agents , i.e. @xmath8 . it is noteworthy pointing out the meaning of the distribution @xmath5 in cases where an abm is used to model human ( or animal , biological , etc . ) it should be thought as the idealized footprint of a collective structure , in the sense of constraints upon agent agency as well as enablers . see , for example , ( see ref . @xcite for a precise sense of structure as used in this context ) . notice that , at this level , the dynamics of the model is defined in the configuration space , which seeks to describe the dynamics of each agent in full detail . we shall refer to this as micro dynamics . let @xmath9 be a finite set ( ex . the configuration space of an abm ) and @xmath10 index a collection of maps @xmath11 and @xmath5 a probability distribution on @xmath10 . if @xmath12 is a sequence of independent random maps , each having distribution @xmath5 , and @xmath13 has distribution @xmath14 , then the sequence @xmath15 defined by @xmath16 is a markov chain on @xmath7 with transition matrix @xmath17 : @xmath18 conversely ( @xcite ) , any markov chain has a random map representation ( rmr ) . therefore ( [ rmr1 ] ) and ( [ rmr2 ] ) may be taken as an equivalent definition of a markov chain which is particularly useful in our case , because it shows that an abm that can be described as above is , from a mathematical point of view , a markov chain and can be treated following the approach developed in this work . this class includes the models described in ref . @xcite . in order to motivate and exemplify our point of view we introduce a model of opinion dynamics that is a particular case of the class of abm described above . we shall come back to this model during the paper to illustrate our results and to show how they lead up to final conclusions in this particular case . to define the opinion model , let us consider a population @xmath19 of @xmath20 agents and denote the opinion , which is the single attribute of agent @xmath21 at time @xmath22 as @xmath23 , where we assume that each agent can choose out of @xmath24 different alternative opinions so that the agent state space @xmath25 has @xmath24 states . let @xmath26 describe the opinions of all the agents at time @xmath22 . we refer to this as the _ opinion configuration _ or the _ opinion profile _ of the population . then the _ configuration space _ of the model is @xmath27 , the space of all possible configurations @xmath28 . in this very simple opinion model the willingness of two agents ( @xmath29 ) to communicate depends on the similarity of their opinions this can be encoded in form of a confidence matrix @xmath30 such that @xmath31 if the opinions @xmath32 and @xmath33 are sympathetic @xmath34 if they are not . the confidence matrix came from the fact that we can define this matrix by means of a confidence threshold , usually defined by means of a threshold on a distance on the set @xmath6 of the possible opinions . the confidence matrix is defined , for any pair ( x , y ) of elements of @xmath35 by @xmath36 if the relative confidence @xmath37 is below threshold and @xmath38 if not . notice that a confidence matrix is a generalization of the notion of confidence threshold . ] in the iteration process of this model , an agent pair @xmath39 is chosen at random `` to meet '' according to a probability distribution @xmath5 , ex . @xmath5 is the uniform distribution , @xmath40 , for all @xmath29 . if two agents @xmath41 meet and @xmath42 then agent @xmath21 imitates agent @xmath43 , that is @xmath44 , which brings us to define an updating function by @xmath45 . if , instead , @xmath46 the opinion configuration is not changed . because in this version of the model the first chosen agent always imitates the second , the model is sometimes called a directed opinion model . in fact we treat a more general class of abm with the following characteristics : ( h1 ) there is a finite number @xmath20 of agents . the set of all agents is denoted by @xmath19 . each agent is characterized by a single attribute which takes a value among @xmath47 possible alternatives . the set of @xmath47 elements of possible situations is denoted by @xmath35 . the configuration space is the set @xmath48 of all possible combinations of the situations for the @xmath20 agents . therefore , if @xmath49 we write @xmath50 with @xmath51 . ( h2 ) there is a @xmath52 matrix @xmath53 with entries @xmath54 or @xmath55 called the * confidence matrix * and an updating function denoted by @xmath56 that is a function from @xmath57 into @xmath57 . ( h3 ) the abm is defined as a discrete time random walk in the configuration space @xmath58 . if the walker is in a configuration @xmath59 at time @xmath22 , it will jump to the configuration @xmath60 under the following prescription : 1 . two agents , it is easy to generalize what follows in this case by just defining @xmath5 , @xmath53 and @xmath56 accordingly . ] @xmath21 and @xmath43 are chosen according to a given probability distribution @xmath5 on @xmath61 . 2 . compute @xmath62 . if @xmath63 the walker does nt move . if @xmath64 then the walker moves to a configuration @xmath65 which defers from @xmath59 only eventually in the value of the attributes of the agents @xmath21 and @xmath43 . that is : @xmath66 and @xmath67 [ no - alpha ] the use of a confidence matrix @xmath53 is just a matter of convenience . it would be possible to absorb @xmath53 in @xmath56 by forcing @xmath68 . we use this double encoding of the updating rule to follow the tradition in the field , assigning the updating rule @xmath56 to `` the model '' and the confidence matrix @xmath53 to `` a parameter of the model '' . as is clear from the previous discussion , under the hypotheses ( h1 ) , h(2 ) , ( h3 ) the abm is an homogeneous markov chain with state space @xmath58 being the configuration space and a transition probability @xmath69 as defined below . notice that , according to the transition rule described above , the possible jumps are constrained to a ball of `` hamming '' radius 2 around the actual configuration . we describe now the transition probability matrix of the markov chain according to eq . ( [ rmr2 ] ) . let us say that a pair of configurations @xmath59 and @xmath60 are * adjacent * if all the agents have the same attribute values in x and in y except possibly the agents @xmath21 and @xmath43 , i.e. @xmath70 if @xmath71 and @xmath72 . these are denoted @xmath73 . then we define a transition probability matrix @xmath17 of a markov chain by @xmath74 @xmath75 @xmath76 where @xmath77 notice that the summation in ( [ phat1 ] ) is needed because , in general , the updating function is not one to one in the configuration space so several links may join two given adjacent configurations . we shall see an example of such a situation in the opinion model . the transition probability matrix of a markov chain is of dimension @xmath78 and it describes the evolution of any initial distribution on the configuration ( state ) space . the comparison with a numerical simulation starting from some particular initial configuration may therefore be performed using the corresponding dirac distribution . but , even in this case , the markov analysis will spread this initial concentrated distribution in the course of the updating process . clearly it is possible to use other initial distributions for the markov chain , and each distribution will correspond to a set of numerical simulations with different initial configurations chosen according to the distribution . each row of the transition probability matrix has at most @xmath79 non zero elements ( including a loop ) . this is due to the adjacency criteria , which encodes the fact that only a pair of agents may change their state from one step to the other . in fact , the non zero elements inside the row @xmath59 are determined by the possible choices of couples @xmath80 of components ( agents ) of @xmath59 . these non zero elements will appear in the columns corresponding to the @xmath60 adjacent to @xmath59 for which @xmath64 . despite the fact that all the information of the dynamics of the abm is encoded in such a markov chain , it is not always easy to learn directly from this fact , because of the over large dimension of the configuration space and its corresponding markov transition matrix . as an example of a question that may be answered at this level , we mention the characterization of absorbing configurations . these are the @xmath50 such that , for any @xmath81 , @xmath63 or @xmath82 and they are easy to identify even in the case of bounded confidence models . it is clear that the properties of the transient dynamics are most relevant for markov chains with absorbing configurations . the analytical study of these properties is based upon the corresponding fundamental matrix @xcite . but such computation needs to invert a matrix of order @xmath83 that can be performed only numerically . therefore another strategy is lacking to go further in this direction . in the next section we shall present one possibility to overcome this difficulty . as already noticed , the directed opinion model is a special case of such a formalism . there , only one coordinate of the configuration may change at each step . the transition probability matrix @xmath17 is obtained from ( [ phat1 ] ) , ( [ phat2 ] ) , ( [ phat3 ] ) and([kappa ] ) by restricting the corresponding sums to the first index @xmath21 only . one immediate consequence of the topology of this transition matrix is that in the opinion model the only absorbing configurations are complete consensus ( all the agents having the same opinion ) or mutually antagonistic ( agents having opinions @xmath84 and @xmath85 for which @xmath86 for all pairs @xmath87 ) . recall that , with probability one , the system falls in one of the absorbing states in finite ( although not uniform ) time . therefore we see how in this model the lack of confidence allows for new absorbing states , stabilizing non consensual opinion profiles . [ section - projected ] a projection of a markov chain with state space @xmath58 is defined by a new state space @xmath88 and a ( projection ) map @xmath89 from @xmath58 to @xmath88 . the meaning of the projection @xmath89 is to lump sets of micro configurations in @xmath58 accordingly to some macro property in such a way that , for each @xmath90 , all the configurations of @xmath58 in @xmath91 share the same property . in fact such projections are important when catching the `` macroscopic '' properties of the corresponding abm because they are in complete correspondence with a classification based on an observable property of the system . to see how this correspondence works let us suppose that we are interested in some factual property of our agent based system . this means that we are able to assign to each configuration the specific value of its corresponding property . regardless of the kind of value used to specify the property(qualitative or quantitative ) , the set @xmath88 needed to describe the configurations with respect to the given property is a finite set , because the set of all configurations is also finite . let then @xmath92 be the function that assigns to any configuration @xmath93 the corresponding value of the considered property . it is natural to call such @xmath94 an observable of the system . now , any observable of the system naturally defines a projection @xmath89 by lumping the set of all the configurations with the same @xmath94 value . conversely any ( projection ) map @xmath89 from @xmath58 to @xmath88 defines an observable @xmath94 with values in the image set @xmath88 . therefore these two ways of describing the construction of a macrodynamics are equivalent and the choice of one or the other point of view is just a matter of taste . the price to pay in passing from the micro to the macrodynamics in this sense ( @xcite , @xcite ) is that the projected system is , in general , no longer a markov chain : long memory ( even infinite ) may appear in the projected system . given the transition probability and the initial distribution defining the first markov chain in @xmath58 , we are interested in the projected measure @xmath95 defined for all cylinders @xmath96 $ ] of @xmath88 by : @xmath97 = \hat\mu \pi^{-1 } [ x_{(1)},x_{(2 ) } , \dots , x_{(r)}]\ ] ] where @xmath98 denotes the corresponding probability for the initial markov chain . the conditions for a projection of a markov chain still to be a markov chain are known as lumpability ( or strong lumpability ) , and necessary and sufficient conditions for lumpability are known @xcite . in general it may happen that , for a given markov chain , some projections are markov and others not . therefore a judicious choice of the macro properties to be studied may help the analysis . in order to establish the lumpability in the cases of interest we shall use symmetries of the model . for further convenience , we state a result for which the proof is easily given thm . 6.3.2 of @xcite : [ proposition_lumpability ] let @xmath99 be a markov chain and @xmath100 a partition of @xmath58 . suppose that there exists a set @xmath101 of bijections of @xmath58 ( therefore a group of symmetries ) such that : \(1 ) @xmath101 preserves the partition : for any @xmath102 and any atom @xmath103 , we have @xmath104 . \(2 ) @xmath101 acts transitively on each @xmath105 , for some ( and then all ) @xmath106 . \(3 ) the markov transition probability @xmath17 is symmetric with respect to @xmath107 : @xmath108 then the partition @xmath100 is ( strongly ) lumpable . the opinion model is a nice example where such a projection construction is particularly meaningful . there , we consider the projection @xmath89 that maps each @xmath109 into @xmath110 where @xmath111 , @xmath112 , is the number of agents in @xmath59 with opinion @xmath84 . the projected configuration space is then made of the @xmath113 where @xmath114 , @xmath115 and @xmath116 . [ section - opinion - projected ] we shall now treat in detail the opinion model as an example of the previous ideas . [ subsection - delta2 ] the case of a binary opinion model , @xmath117 , is particularly simple and therefore well - suited for an analytical starting point . in this case bounded confidence is excluded . in binary state opinion models , the opinion of agent @xmath21 at time @xmath22 is a binary variable @xmath118 . the opinion profile is given by the bit string @xmath26 . the space of all possible configurations is @xmath119 . let us define a function @xmath120 on the configuration space @xmath7 such that @xmath121 it counts the number of agents in state @xmath55 . using @xmath122 , we define @xmath123 by @xmath124 each @xmath125 contains all the configurations ( @xmath28 ) in which exactly @xmath126 agents hold opinion @xmath55 ( and then @xmath127 hold opinion @xmath54 ) . in this way we obtain a partition of the configuration space @xmath7 . notice that @xmath128 and @xmath129 contain only one configuration , namely @xmath130 and @xmath131 . using the group @xmath132 of all the permutations of @xmath20 agents , it is clear that such a partition fulfills conditions ( 1 ) and ( 2 ) of proposition([proposition_lumpability ] ) . so lumpability of this partition leans on condition ( 3 ) of proposition([proposition_lumpability ] ) : the invariance of the markov transition matrix @xmath133 under the permutation group of agents . notice that no restriction on the confidence matrix is needed for it only depends on the opinions and not on the labeling of the agents . in fact , the probability distribution @xmath5 must be invariant under the permutation group @xmath132 and therefore uniform : @xmath40 , for all pair of agents ( @xmath29 ) . it is worthwhile noticing at this point that the uniform distribution , corresponding to the most unstructured dynamical rule , still entails emergent organized patterns in the system of opinions . because the set of opinions is dynamically organized , the homogeneity of the uniform distribution on the agent population has an implicit structure when viewed through the opinion content . moreover , for some other distributions @xmath5 , it may be possible to refine the partition so as to get lumpability . for instance , if the agents are divided in subsets in which @xmath5 is constant , then the partition defined by an equal number of opinions inside each subclass is lumpable . in this case @xmath53 will depend on the pair labeling of agents together with their respective opinions and not only on the latter . the block structure of @xmath53 then determines the projection scheme . for the model with complete confidence , @xmath134 for any @xmath87 , and uniform distribution @xmath5 , the markov chain is defined by the stochastic transition matrix : @xmath135 with @xmath136 and @xmath137 formulas ( [ eq : f1t2.p ] ) and([eq : f1t2.q ] ) follow from thm . 6.3.2 of @xcite and the fact that for any micro configuration in @xmath138 there are @xmath139 possible meetings taking it to @xmath140 and the same number of possible encounters taking it to @xmath141 . notice that in this case @xmath128 and @xmath129 are the only absorbing states of the process . the probability that any opinion change happens in the system is @xmath142 and then depends on the current opinion balance . but there is no general tendency of the system to be attracted by one of the extremes . due to the particular form of @xmath143 the prevalence of one opinion results in a reduced probability of further opinion change , contrary to the usual random walk with constant transition probabilities . for @xmath144 we have @xmath145 . by contrast , when @xmath126 is closed to @xmath54 or @xmath20 , there is a large probability for the system to stay unchanged . notice that for @xmath146 or @xmath147 this probability tends to @xmath55 when @xmath148 . this indicates that in this model once one opinion dominates over the other , public opinion as a whole becomes less dynamic , which also reveals a difficulty for new opinions to spread in the artificial society . [ subsection - delta2-transient ] in markov chains with absorbing states ( and therefore in abm ) the asymptotic status is quite trivial . as a result , it is the understanding of the transient that becomes the interesting issue . we shall now analyze the transient dynamics for the macro dynamics of the binary opinion model . in order to do so , all that is needed is to compute the fundamental matrix @xmath149 of the markov chain ( @xcite ) . let us express @xmath150 in its standard form in which the two first rows and columns stand for the absorbing states @xmath128 and @xmath129 and the remaining for the @xmath151 transient states : @xmath152 here @xmath153 is the @xmath154 matrix corresponding to the transient states ( without the first two rows and columns associated with @xmath128 and @xmath129 ) . the fundamental matrix @xmath155 is the inverse of @xmath156 where @xmath157 is the @xmath154 identity matrix . due to the structure of @xmath150 , @xmath156 is a tridiagonal matrix that can be inverted easily using for instance the tridiagonal matrix algorithm ( also known as thomas algorithm @xcite ) . we get : @xmath158 and @xmath159 equations ( [ eq : fij1 ] ) , ( [ eq : fij2 ] ) provide us with the fundamental matrix of the system for an arbitrary number of agents @xmath20 , giving information about mean quantities of the transient dynamics in this model . the corresponding matrix @xmath160 that encodes information about the variance @xcite of the same quantities reads : @xmath161 @xmath162 and @xmath163 an interesting quantity to characterize opinion dynamics is the time a process starting in @xmath138 takes to end in one of the two consensual absorbing states . defining @xmath164 and @xmath165 as the mean and the variance of the random variable for @xmath166 we got from ( [ eq : fij1 ] , [ eq : fij2 ] ) and @xcite : @xmath167\ ] ] and the corresponding expression for @xmath168 can explicitly be written from ( [ eq : gij1 ] , [ eq : gij2 ] , [ eq : gij3 ] ) using : @xmath169 where @xmath170 denotes the vector resulting from @xmath171 by squaring each entry . therefore @xmath172 + \\\nonumber + ( 2n-1 ) n \left ( \sum^{k-1}_{j=1 } \frac{(n - k ) } { ( n - j ) } + 1 + \sum^{n-1}_{j = k+1 } \frac{k}{j } \right ) + \\\nonumber + 2 n^2 k \left [ \sum^{n-1}_{i=1 } \frac{1}{k+i } \left ( \sum^{k+i-1}_{j=1 } \frac{(n - k - i ) } { ( n - j ) } + 1 + \sum^{n-1}_{j = k+i+1 } \frac{k+i } { j } \right ) -\right ] -\\\nonumber - n^2 \left ( \sum^{k-1}_{j=1 } \frac{(n - k)}{n - j } + 1 + \sum^{n-1}_{j= k+1 } \frac{k}{j}\right)^2\end{aligned}\ ] ] for a system of 1000 agents , fig . [ fig : tau1000 ] shows the mean times until absorption @xmath164 from each @xmath138 and fig . [ fig : tau21000 ] the corresponding variances @xmath173 . notice the contrast among the two scales showing how the variance is large compared with the mean . until absorption as a function of the initial configuration @xmath126 for @xmath174 . , scaledwidth=60.0% ] until absorption as a function of the initial configuration @xmath138 for @xmath174.,scaledwidth=60.0% ] there are interesting consequences of ( [ eq : tauk ] ) and ( [ eq : tupsilonk ] ) , in cases where the number of agents ( @xmath20 ) becomes large . first , as already pointed out , we see that the ratio between the variance and the mean is quite large and in fact it diverges with @xmath20 . hence , the means are fairly unreliable estimates in this system . this is often the case for absorbing markov chains @xcite making a direct interpretation of numerical simulations for this type of models tough . even more subtle , the time scale depends significantly on the starting configuration @xmath126 . in fact @xmath164 scales as @xmath175 for @xmath146 and @xmath176 but as @xmath79 for @xmath177 . we are therefore faced with a situation where to take the limit of asymptotic times first and then large number of agents or to do it in the reverse order is not equivalent . in other words , for a finite , even large , number of agents , there is a probability 1 of reaching one of the consensual configurations in finite time . by contrast , in the limit of an infinite number of agents this probability is 0 and the process will stay essentially in the configurations close to parity , @xmath177 . together with the presence of large fluctuations revealed in ( [ eq : tupsilonk ] ) this fact is the imprint of a ( dynamical ) phase transition . besides this analysis of the scaling law of the dynamics for large @xmath20 , it is also interesting to have an insight into the distributions of absorbing times for a system of fixed number of agents , the second item mentioned above . as known by the perron frobenius theorem @xcite this distribution is exponential for large @xmath22 with rate ( @xmath178 ) , @xmath179 being the maximal eigenvalue of the matrix @xmath153 . however , the correction to this distribution for intermediate times depends on the initial configuration . indeed in our case , the distribution of the times taken by the process to fall into one of the consensual configurations departs from the exponential in a way that is strongly dependent upon the initial state , as shown in figs . [ fig : absorbing_times_cdf ] and [ fig : absorbing_times_p ] . ( r.h.s . ) for @xmath180 and three starting configurations @xmath181 ( green ) , @xmath182 ( blue ) and @xmath183 ( red ) . vertical lines show the respective expected absorbing times @xmath171.,scaledwidth=60.0% ] for @xmath180 and three starting configurations @xmath181 ( green ) , @xmath182 ( blue ) and @xmath183 ( red ) . exponential functions ( dashed ) are shown to illustrate the exponential decay of the convergence times.,scaledwidth=60.0% ] the computation of the full time distribution is based on the fact that the powers @xmath184 of @xmath153 contain all the information about the probability that the process is still not absorbed after @xmath22 steps . to be precise , the sum over the @xmath126th row of @xmath184 equals the probability that the process starting at @xmath138 is not absorbed after @xmath22 iterations . this yields the cumulative distribution function shown in fig . [ fig : absorbing_times_cdf ] for a system of 100 agents and three starting configurations @xmath181 ( green ) , @xmath182 ( blue ) and @xmath183 ( red ) . the vertical dashed lines represent the respective mean values @xmath171 obtained using eq.[eq : tauk ] . for @xmath183 it becomes clear that around 60% of simulation runs are absorbed until the expected absorption time is reached . [ fig : absorbing_times_p ] shows the probability that the process is absorbed exactly at time @xmath185 . the three solid curves represent the respective probabilities for @xmath186 . the dashed curves are exponential functions that fit the distributions for large @xmath185 showing that the distributions decay with ( @xmath178 ) as claimed above . this leads to an interesting feature of the distribution of the absorption times coming from the fact that @xmath179 tends to one when @xmath187 . more precisely @xcite ( [ eq : pmat ] ) implies @xmath188 as a consequence , we see that the times for the system to get absorbed in the final states diverge with @xmath20 , and @xmath153 approaches a stochastic matrix . in fact in the limit of infinite @xmath20 consensus can not be reached . this is not the only reason why the dynamics inside the transient configurations is so important . in fact we might speculate that , in a more realistic description , exogenous events may interfere with the system and reset it from time to time , and then , in view of the previous analysis , even when the number of agents is finite but sufficiently large , the system will similarly never fall into a final absorbing unanimity configuration . notice that ( [ eq : fij1 ] , [ eq : fij2 ] ) and ( [ eq : gij1 ] , [ eq : gij2 ] , [ eq : gij3 ] ) can be used to gain new insight into the dynamics inside the transient . @xmath189 is the mean of the time the process is in the transient configuration @xmath138 when started in the configuration @xmath190 and @xmath191 is the corresponding variance [ fig : f_differentj ] and [ fig : f2_differentj ] show a quite different behavior depending on the initial situation . starting from @xmath190 close to @xmath192 or @xmath193 the strongly `` biased '' configurations the residence mean times in @xmath138 naturally decrease with the distance from @xmath21 but become almost independent of @xmath126 and @xmath20 for @xmath126 large whereas the corresponding variance diverges with @xmath20 . instead , starting from @xmath190 close to @xmath194 , the quasi - homogeneous configurations- the residence mean times and variance in @xmath138 always diverge . [ cols="^,^,^ " , ] the reason for such `` strange '' behavior is quite clear : as @xmath20 becomes large , almost all the realizations are trapped during very large times close to their initial configuration , see ( [ eq : pmat ] ) , and only very few realizations reach the opposite configurations but staying there for large times . that is , a complete overturn of the opinions is very rare but , when happened , the new situation naturally becomes as stable as the previous . therefore we are in a case where there is almost no realization behaving as the mean . on the other hand , starting from @xmath190 closed to @xmath194 , the `` homogeneous '' configurations , the mean times in @xmath138 also decrease with the distance from @xmath195 , but now the mean times all scale linearly with @xmath20 and the variances with @xmath79 . surprisingly these two behaviors , almost static on the border and very unstable `` back - and - forth '' on the center , compensate perfectly to end up in the same mean residence times and variance ( the diagonals of @xmath155 and @xmath196 ) for all the initial configurations . the same compensation appears when we compare the probabilities for a walk stating in @xmath190 to return in @xmath190 , which is independent of @xmath21 and almost sure for large @xmath20 : @xmath197 finally the probabilities for a process starting in @xmath190 to end up in @xmath128 or @xmath129 can be computed using also ( [ eq : fij1 ] , [ eq : fij2 ] ) . the result is as expected : @xmath198 and @xmath199 it is reasonable to hypothesize a correlation , if not a causal link , between fast changes in the agent opinion induced by the social process , here stylized in the dynamical rules , and the inconsistency experienced by agents between the micro and the macro level , described in section [ section - introduction ] . this conflict is referred to as practical emergence . it consists of a gradual separation of the individual mental patterns from the reality . the agent is then faced with a representation that is not always perfectly in keeping with the situation ( @xcite , pag.342 ) . in the opinion model , a possible rating of this practical emergence inconsistency is the mean time the macro process takes to change of state . indeed any change of state in this process corresponds to a change for a opinion of an agent . therefore the faster this rate , the smaller the switching mean time , and the more likely is the emergence of a practical disruption between picture and reality from the agent s point of view .. from ( [ eq : f1t2.p ] ) and ref . @xcite , thm . 3.5.6 , the mean time @xmath200 that the process remains in state @xmath138 once the state is entered ( including the entering step ) is : @xmath201 therefore , @xmath200 is of order @xmath202 for @xmath126 close to ( but smaller than ) @xmath20 and @xmath203 for @xmath126 close to @xmath202 . again , for @xmath20 large the process will be almost stationary in presence of a large majority supporting one of the opinions but extremely unstable when no opinion is clearly predominant . in the latter case practical emergence is plausible . we suggest correlating small values of @xmath200 with this phenomenon . ) . markov chain analysis shows that with probability @xmath204 the process is in the shaded region.,scaledwidth=60.0% ] to conclude this section fig . [ fig : timeevolution03 ] shows different realizations of the agent simulation along with the expected evolution in form of a confidence interval . the measure of the realizations inside a given confidence interval is an increasing function of time . however , since any individual realization may cross the border of this interval several times before falling in one of the final absorbing states a numerical evaluation of the convergence times may be quite delicate . [ subsection - any - delta ] for an opinion model with @xmath24 different opinions , the opinion of any agent @xmath21 at time @xmath22 is a variable @xmath205 . the opinion profile is given by the vector @xmath26 . the space of all possible configurations is then @xmath206 . following the same argument as for @xmath207 , we define @xmath208 to be the number of agents in the configuration @xmath28 with opinion @xmath84 , @xmath209 , and then @xmath210 as @xmath211 each @xmath212 contains all the configurations @xmath28 in which exactly @xmath111 agents hold opinion @xmath84 for any @xmath84 . as in the binary case , in this way we obtain a partition of the configuration space @xmath7 . it is clear that also in this case the group of all the permutations of @xmath20 agents labeling fulfills conditions ( 1 ) and ( 2 ) of proposition([proposition_lumpability ] ) and that condition ( 3 ) is verified if the probability distribution @xmath5 is permutation invariant and therefore uniform : @xmath40 , for all pair of agents ( @xmath29 ) . in this case , eq . ( [ eq : f1t2.p ] ) generalizes to : @xmath213 if @xmath214 and @xmath215 whereas @xmath216 for all other @xmath43 , and the probability that no opinion changes , eq . ( [ eq : f1t2.q ] ) , becomes @xmath217 the structure of ( [ eq : pp < > ] ) has an interesting consequence on the dynamics of the system . we see that , if one @xmath43 for which @xmath218 , the probability of transition to a state with @xmath219 is @xmath54 . in other words , to change the number of agents sharing opinion @xmath43 , at least one agent with such an opinion is needed . therefore , the state space is organized as a @xmath220 with absorbing faces ordered by inclusion , corresponding to increasing sets of opinions with no supporters . starting in some state with no null @xmath221 the process will finish at certain time in a state where , for the first time , @xmath222 for some @xmath43 ( notice that only one @xmath43 at each time can fall to zero since the sum of all @xmath221 is constant ) . from there , the given @xmath221 will stay equal to zero for ever , and ( [ eq : pp < > ] [ eq : pq < > ] ) tell us that the transition probabilities are now those of a system with @xmath223 opinions . because the condition @xmath224 is to be fulfilled by the remaining opinions , the system will then evolve exactly as if the n agents share @xmath225 opinions from the very beginning . after a certain time a new opinion will lose all its supporters and the system is now equivalent to a full system of @xmath226 opinions , and so on . the system will cascade up to the final absorbing state , with only one opinion shared by all the @xmath20 agents . we recall that each of such cascade transitions is achieved in finite ( random ) times . by computing the fundamental matrix of the subsystems it would be possible to access the mean and variance of the times the system evolves between two successive extinctions of group opinions . we conjecture the same scaling laws for a system of @xmath24 opinions as the ones already described for @xmath207 . [ subsection - reduction ] alternatively , we can make use of the symmetries in the structure of ( [ eq : pp < > ] ) and search for lumpable partitions to further reduce the problem . this can be done by considering the model from the perspective of a single `` party '' associated with ( say ) opinion @xmath54 . for that `` party '' , it may be important to know how many agents are supportive because they share the same opinion , and how many are not because they support one of the remaining opinions . thus , we reduce the model to a quasi binary variant with the supporter opinion @xmath54 on one side and and all other opinions ( @xmath227 ) on the other side , grouping together all the states with @xmath228 , @xmath229 . the corresponding partition reads : @xmath230 it is easy to verify that the chain ( on the @xmath231 ) is indeed lumpable with respect to @xmath232 and that @xmath233 it thus turns out that the chain formed by the @xmath234 is exactly the same as the chain derived for the binary model . therefore , the questions regarding the evolution of one opinion in relation to all the others taken together are addressed by the transient analysis performed in sec . [ subsection - delta2-transient ] . that is to say , from this point of view , each `` party '' may rely on the dynamics of a binary model as a coarse description of the evolution of its own status . there is however an important subtlety when doing such an analysis . the asymmetry of the partition one - against - all - others will be encoded in the initial condition . for instance , starting with an equally distributed profile of n agents corresponds to the initial condition @xmath235 in the detailed description but to @xmath236 in the coarse case . in such a way the asymmetry in the one - against - all - others description is recovered . another important issue concerns the effects of bounded confidence in the model , in other words if a certain number of pairs of opinions do not communicate . from a formal point of view bounded confidence is encoded in the confidence matrix @xmath53 just by putting @xmath237 if ( @xmath238 ) is one of such non communicating opinion pair , denoted to as @xmath239 . the consequence of bounded confidence is the emergence of non consensual absorbing states known as opinion clusters . in the following section , we treat in great detail the simplest case where bounded confidence is possible , namely @xmath240 . we postpone the general case to the last section since it is a simple generalization of this example . [ subsection - delta - three ] we are particularly interested in the @xmath241 case because it is the simplest version in which one can meaningfully consider _ bounded confidence _ effects . according to the general results of sec . [ subsection - any - delta ] in case of unbounded confidence , the projection from micro to macro dynamics is lumpable ( under the homogeneous hypothesis on @xmath5 of course).the markov chain topology obtained by this projection is shown in fig . [ fig : statetopologyandtransitions.t3.n8 ] along with the transition structure . the probabilities of the transitions are given by eqs . ( [ eq : pp < > ] ) and ( [ eq : pq < > ] ) which allows us to compute the complete transition matrix @xmath150 . . ] for the construction of @xmath150 , the nodes in the markov chain are labeled in increasing order from the absorbing to the central nodes , see fig . [ fig : statetopologyandtransitions.t3.n8 ] : labels 1 to 3 ( black ) for absorbing consensus states , labels 4 to 24 ( blue ) for two opinion states , labels 25 to 39 ( red ) for three - opinion states with one of the opinion supporters reduced to one element , and labels 40 to 45 ( red ) for the remainder states . it is possible to compute the fundamental matrix , at least numerically if @xmath20 is large , and this makes it possible to compute the significant statistical indicators of the model . for instance , if @xmath242 , the state space of the macro dynamics has 45 states and the mean times for the transient nodes to reach an absorbing state ( consensus ) range between 21 and 48 time steps , see fig . [ fig : t3.n8.absorbtimes ] . not surprisingly the mean transition times are a function of the distance to the absorbing states as measured on the graph of the state space ( fig . [ fig : statetopologyandtransitions.t3.n8 ] ) . from the fundamental matrix * f * it is also easy to compute the probabilities of ending up in each of the absorbing ( consensus ) states starting from any transient node using the matrix @xmath243 , where @xmath244 is defined in ( [ eq : canonicalform ] ) . for instance , for @xmath242 , the absorbing probabilities for any state are represented in fig . [ fig : absprobs.t3.n8 ] . notice that all three final states can be reached only from the inner nodes ( numbers 25 to 45 ) . ] let s now turn to the question of what happens if agents with a certain opinion do not accept to change their opinion after meeting an agent of another given opinion . in the opinion dynamics literature , this is referred to as _ bounded confidence_. from the markov chain perspective the emergence of opinion polarization becomes a simple consequence of the restrictions posed on the interaction process . as certain transitions are excluded , the state space topology of the markov chain changes in a way that new absorbing states become present . the respective states correspond to non consensus configurations , hence , they represent a population with opinion clustering . as an example , let us assume that agents in opinion state a are not willing to communicate with agents in state c and vice versa , that is to say @xmath245 . the corresponding markov transition matrix @xmath150 now reads : @xmath246 and @xmath247 the remaining entries are , as before , ( [ eq : pp < > ] ) and ( [ eq : pq < > ] ) . the resulting state space topology is shown in fig . [ fig : statetopology.bc ] , where all horizontal transition paths are removed , since those paths correspond to the @xmath248 opinion changes . . ] for the set of bordering nodes @xmath249 there is no longer any transition that leads away from them , so that all these nodes become absorbing states . the fact that these additional absorbing states @xmath250 represent opinion configurations with @xmath126 agents in state @xmath251 and @xmath252 agents in state @xmath253 explains why the introduction of interaction restrictions leads to possible final states with opinion polarization . it is noteworthy , however , that the opinion clustering would not be observed if only one of the two transitions , @xmath254 or @xmath255 , were excluded . in this case , there would still be a path leading away from the bordering nodes to one of the nodes ( @xmath256 or @xmath257 ) in the corner of the graph . such a set up corresponds to an asymmetric model where the bordering atoms @xmath249 become again transient , such that the process eventually leads to the final consensus configurations as previously described . however the final configuration @xmath258 would be much more likely than @xmath259 , as a consequence of the asymmetry of such a model variant . as for the case of unbounded confidence , the fundamental matrix can be computed here as well allows us to calculate the statistical quantities of the model such as absorbing probabilities and times . in fig.[fig : absprobs.t3.n8.bc ] the probabilities of a realization starting in one of the transient states ending up in each of the absorbing final states are shown for each initial node ( computed again by @xmath243 ) . if the process is in the first 10 nodes at @xmath260 , it will remain there forever as all these nodes are absorbing in the bounded confidence case . notice that nothing changes for the nodes 11 to 24 with respect to the unbounded case shown in fig.[fig : absprobs.t3.n8 ] . for a system in these configurations the communication constraint has no effect on the dynamics . the six absorbing non consensus states ( numbers 4 to 10 with only `` a '' and `` c '' opinion supporters ) are reachable only from the inner nodes , that is only if all opinions are present initially . it becomes clear that for some of these configurations , the probability of converging to consensus becomes very small ( e.g. nodes 25 to 30 ) . converging to opinion clustering or to the three consensus nodes . notice again that all final states and the non consensus states in particular can be reached only from the inner nodes ( numbers 25 to 45 ) . ] finally , we can compare the mean time before a realization starting in a transient state remains in the transient before absorption for the bounded and the unbounded case . this statistical indicator is represented in fig . [ fig : t3.n8.absorbtimes ] . notice that the times for the states 1 to 3 ( unbounded ) and 1 to 10 ( bounded ) are zero as in this case the process is absorbed from the very beginning . again , the non absorbing two opinion states ( 11 to 24 ) are not affected . . labels of the nodes are explained in the text . ] as in the general case of any @xmath24 we can search here for lumpable partitions to further reduce the problem taking the point of view of each `` party '' associated with opinions `` a '' , `` b '' or `` c '' . for the case of unbounded confidence we have shown in sec . [ subsection - reduction ] that the dynamics from any of these points of view reduces to the @xmath117 case . the status of the bounded confidence model is different . from the perspective of opinion `` b '' the partition in `` supporters '' and `` opponents '' is lumpable , therefore , the system evolves as a binary chain . this is not the case from the perspectives of opinions `` a '' or `` c '' . for instance , from the point of view of opinion `` a '' , the corresponding partition reads : @xmath261 and @xmath262 it turns out that the chain formed by the @xmath263 is not a markov chain since the r.h.s . of ( [ eq : yanolumpprob ] ) depends on @xmath264 and not only on @xmath265 @xcite . we see that the introduction of bounded confidence in this model leads to memory effects due to the fact that an agent switching from opinion `` a '' to opinion `` c '' necessarily goes through a visit to opinion `` b '' for at least one time step , therefore , the probability of this transfer will depend on the number of supporters of opinion `` b '' at that time . [ section - generalizations ] we first mention an easy generalization of the existence of absorbing states for the case of bounded confidence in a model with any number @xmath24 of different opinions . in order to get non consensual absorbing states it is necessary and sufficient that a subset of opinions is mutually incommunicable . in this case all the states belonging to the simplex generated by the mutually incommunicable opinions become absorbing . it is worthwhile noticing that absorbing states may appear in different clusters of simplexes provided that the corresponding opinions are related by chains of communicating links . an example of this type appears for @xmath241 if ( @xmath239 ) and ( @xmath266 ) but ( @xmath267 ) where the absorbing states are either the simplex with only @xmath251 and @xmath268 or with only @xmath251 and @xmath253 opinions . another interesting issue concerns agent models with vectorial ( or equivalently matrix or table ) individual state ( attribute ) space . suppose that at each time step each @xmath21 is characterized by a list of @xmath269 attributes , where the first attribute may take @xmath270 possible values , @xmath271 , and the @xmath272 attribute @xmath273 possible values . the corresponding abm can then be easily built as in sec . [ subsection - any - delta ] by taking @xmath274 . and , as long as one is interested in following the macrodynamics of the agents who all have @xmath24 attributes that are identical , , the reduction proposed in sec . [ subsection - reduction ] also applies . therefore , absorbing non consensual states will appear in exactly the same way as described above as a consequence of bounded confidence . for this vectorial opinion model there is , however , an unexpected subtlety when we are interested in the macrodynamics of the agents ranked by only one of their attributes , for instance if the agents are separated in @xmath270 different groups according to the number of agents sharing their first attribute . then the partition is no longer lumpable , and therefore the evolution of the correspondent random variables ( for instance , the number of elements of each group ) is not a markov chain . again , in this case , new memory effects may appear from this choice of aggregation to build the macrodynamics . the proof can be done as in ( [ eq : f1t3.ya ] ) and ( [ eq : yanolumpprob ] ) . we shall finish this section by noticing a trick that makes it possible to extend the previous analysis to systems where the agents are defined on a lattice or , more generally , on a graph . they may coexist in some nodes of the graph or , sometimes they may even move on it . then , in some cases , it is possible to reduce the model to the framework of our work by defining the agents as the nodes in the graph and including the presence or lack of an actor in the prescription of the attributes defining the elements of @xmath35 as well as the topology of the graph accordingly in@xmath53 . [ section - conclusion ] an important mark of abms is their ability to include arbitrary levels of heterogeneity and stochasticity ( or uncertainty ) into the description of a system of interacting agents . while computer simulations are often suited for making important dynamical trends of these models visible , a rigorous characterization of the different dynamical phases is difficult . in this work we analyze the dynamics of an abm from a markovian perspective and derive explicit statements about the possibility of linking a microscopic agent model to the dynamical processes of macroscopic observables that are useful for a precise understanding of the model dynamics . in this way the dynamics of collective variables may be studied , and a description of macro dynamics as emergent properties of micro dynamics , in particularly during transient times , is possible . in our context , the random map representation ( rmr ) of a markov process helps to understand the role devoted to the collection of ( deterministic ) dynamical rules used in the model from one side and of the probability distribution @xmath5 governing the sequential choice of the dynamical rule used to update the system at each time step from the other side . the importance of this probability distribution , often neglected , is to encode the design of the social structure of the exchange actions at the time of the analysis . not only , then , are features of this probability distribution concerned with the social context the model aims to describe , but also they are crucial in predicting the properties of the macro dynamics . if we decide to remain at a markovian level , then the partition , or equivalently the collective variables , to be used to build the model may be compatible with the symmetry of the probability distribution @xmath5 . in a sense the partition of the configuration space defining the macro level of the description has to be refined in order to account for an increased level of heterogeneity or a falloff in the symmetry of the probability distribution . it is , however , clear that , in absence of any symmetry , there is no other choice for this partition than to stay at the micro level and in this sense , no markovian description of a macro level , is possible in this case . at this point two possible approaches could be explored in the future : first to identify in specific models other natural symmetries inducing a markovian description of their macro counterpart and to study the evolution of the corresponding collective behavior ; and second , to tackle the problem of describing the dynamics of the macro variables in cases where they are not markovian . for this purpose the cases where @xmath5 is not homogeneous ( or symmetric ) but has special structural properties seem to be a crucial issue for further research . this point must have an important impact in the understanding of descriptive emergence since it is a ( the , in abm context ) source of long memory effects in the dynamics of the formation of collective social patterns . another important prospect concerns the measure of practical emergence or discrepancy , the gap between the macro - structural properties of a social system and internalized rules or intentions of the individual actors . the measure of this gap should lead to more elaborate gauges whose dynamics themselves call for new specific investigation . as pointed out in section [ section - introduction ] , under certain circumstances the macro process may undergo dynamical changes in its own structural rules . this fact is referred to as explanatory emergence . it can be understood either as a consequence of some external ( to the model ) inputs or on the basis of deep accelerations of the micro dynamics that in turn bring about the processes of change at the macro level . in both cases this question opens up to new theoretical as well as very interesting practical developments . in an abm as presented in section [ section - microdynamics ] , the macro - structural patterns of social relations are encoded in the probability distribution @xmath5 , or its generalization when the model allows a simultaneous updating depending on more than two agents . therefore a dynamics of this structure may be incorporated in the model as a dynamics of the probability distribution , allowing it to be time dependent : @xmath275 . this can be done at different levels . the simplest one occurs when the structural dynamics is assumed to be autonomous , in the form of a master - slave system where the dynamics of the probability distribution is defined independently of the evolution of the process . it is natural to suppose that this dynamics is slow with respect to a fast dynamics of the agent changes . a more sophisticated modeling consists of coupling the individual agent dynamics with the dynamics of the distribution . this corresponds to assuming a feedback of the agent dynamics on the evolution of the structural rules fixing the macro dynamics . in both cases a rigorous treatment of the problem is sufficiently compelling to deserve further research . the formalization of the relations between the micro and the macro levels in the description of the dynamics of agent based models as well as their mathematical characterization is a step towards a mathematical theory of emergence in complex adaptive systems . in this work we showed how a markov chain approach , in particular the use of the notion of lumpability , provides useful instruments for the analysis of the link from a microscopic abm to macroscopic observables . further research is needed to deepen our understanding of this link in a more general setting . we are grateful to ann henshall for a careful reading of the manuscript and to dima volchenkov for helpful discussions . financial support of the german federal ministry of education and research ( bmbf ) through the project _ linguistic networks _ ( http://project.linguistic-networks.net ) and the portuguese foundation for science and technology ( fct ) through the project pdct / ege/60193/2004 is also gratefully acknowledged . grauwin , s. , goffette - nagot , f. , and jensen , p. ( 2010 ) . dynamic models of residential segregation : an analytical solution . gate working paper no . available at ssrn : http://ssrn.com/abstract=1646260 |
second generation high energy neutrino telescopes are being built to reach sensitivities of neutrino emission from galactic and extragalactic sources . current neutrino detectors are already able to set limits which are in the range of some emission models . in particular , the * a*ntarctic * m*uon * a*nd * n*eutrino * d*etection * a*rray ( amanda ) has recently presented the so far most restrictive limit on diffuse neutrino emission . stacking limits which apply to agn point source classes rather than to single point sources are given as well . in this paper , the two different types of limits will be used to draw conclusions about different emission models . an interpretation of stacking limits as diffuse limits to the emission from considered point source class is presented . the limits can for instance be used to constrain the predicted correlation of egret - detected diffuse emission and neutrino emission . also , the correlation between x - ray and neutrino emission is constrained . further results for source classes like tev blazars and fr - ii galaxies are presented . starting from the source catalogs so - far examined for the stacking method , we discuss further potential catalogs and examine the possibilities of the second generation telescopes icecube and km3net by comparing catalogs with respect to northern and southern hemisphere total flux . ps . ^ 25_2.5 agn , neutrinos , stacking , diffuse spectra , limits , olbers paradox 95.55.vj , 95.80.+p , 95.85.pw , 95.85.ry | astro-ph0607427 | as is well known for hundreds of years , the sky would be infinitely bright in optical light , if all space were homogeneously full of stars , and space were infinitely extended without any change in properties @xcite . this reasoning can be applied to stars as well as to galaxies . this argument can also be applied to electromagnetic emissions and just as well to neutrino emission . the solution to olbers paradox is that space is not homogeneously full of identical distributions of stars or galaxies , and that the universe evolves quite strongly . this implies that the integral over the optical light of all galaxies has a finite sum , which does not exceed an observable level . clearly , given a proper sampling of galaxies , it is possible to determine this level . in some cases , there is a unique relationship between the optical light and the emission at some other wavelength . starting from this correlation , the summed emission from all galaxies over all history can be deduced and it is a simple function of the sum at optical wavelengths . the optical emission can certainly be replaced with the emission at some other electromagnetic wavelength , such as x - rays . equally , the correlated wavelength can be neutrino emission . if we now do not know what the sum of the emission is at the electromagnetic wavelength , but have limits , then a limit for the neutrino emission can be deduced a fortiori . vice versa , if we have a limit for the neutrino emission , a corresponding limit for the electromagnetic emission can be derived . in either case , it might happen , that we have a known background , and then can deduce whether this specific class of sources could possibly explain all background at the other `` wavelength '' . we start this argument with neutrinos , set a limit , and then derive a limit for the electromagnetic background . we could also start with the electromagnetic background , and then set a limit for the neutrino emission . so we have an _ olbers paradox for neutrinos_. this is the key point of this paper . more concretely , the interpretation of different amanda neutrino flux limits is done for different classes of active galactic nuclei ( agn ) . in the case of a stacking analysis in which the strongest sources from the same class are selected , a method for the interpretation of these limits as limits to the diffuse flux from the given source class is developed . additionally , future possibilities of second generation neutrino telescopes like icecube and km3net are examined . different source classes are discussed with respect to their distribution in the sky , i.e. which sources are in the northern and which are in the southern hemisphere . the field of view for icecube respectively km3net is the northern respectively the southern hemisphere and their view of the sky is complementary . this paper is organized as follows : in section [ nu_models ] , the prevailing models are discussed according to the normalization options given by experimental observations of cosmic rays ( crs ) and photons at different wavelengths . it is essential to discuss in this context which stacking limit can be applied to what prediction . current neutrino flux limits are investigated in section [ nu_limits : sec ] . a general ansatz for the interpretation of stacking limits as diffuse limits is presented here . in section [ stack_results ] , the method is applied to amanda s limits which have been derived by stacking agn according to their electromagnetic output . the question which stacking limits apply to which normalization scenario is discussed in section [ compare ] and it is examined how the limits constrain different models . in section [ class_analysis ] , further source classes are examined according to the possibility of applying the stacking method for high energy neutrinos , i.e. @xmath0 tev . in addition , a comparison of the contribution from the northern and southern hemisphere is done for various source classes in order to compare the capabilities of next generation s telescopes icecube and km3net . the results are summarized in section [ conclusions ] . in some hadronic acceleration models it is assumed that for each class of agn , the electromagnetic emission is correlated to a neutrino signal . apart from individual normalization factors , the corresponding cosmological integrations are basically mathematically identical . in this section , the correlation between the emission of neutrinos and photons at different wavelengths will be discussed according to neutrino flux models which are currently being discussed in the literature . sources of electromagnetic tev emission can be interpreted as optically thin to photon - neutron interactions , @xmath1 ( e.g. * ? ? ? * ; * ? ? ? * ) in hadronic acceleration models . in such a scenario , charged cosmic rays ( crs ) are produced in the vicinity of the source through the decay of the escaping neutrons . in this case , the resulting neutrino energy density would be proportional to the extragalactic cr component measured at earth . a theoretical upper bound of such a contribution to the diffuse neutrino flux has been derived by mannheim , protheroe and rachen ( in the following referred to as the _ mpr bound _ ) , see @xcite . within the framework of the proton - blazar model , the neutrino flux from high - frequency peaked bl lacs ( hbls ) has been calculated using the connection between cosmic rays and neutrino emission , see @xcite . in the case of optically thick sources ( @xmath2 ) , the photons interact with nucleons in the source before escaping at lower energies leading to the emission of sub - tev photons . therefore , the diffuse extragalactic background determined by the egret experiment @xcite ( @xmath3 mev ) can be interpreted as an avalanched tev signal from blazars and can thus be used to normalize the neutrino flux from egret or comptel - type sources . again , mpr give an upper bound to the contribution from such sources in @xcite , which is much less restrictive than the optically thin case . apart from the bound , a calculation of the maximum contribution from blazars is given in @xcite . in addition to the contribution from hbls within the framework of the proton - blazar model , a contribution from the optically thick low - frequency peaked bl lacs ( lbls ) can be calculated using the egret diffuse extragalactic background for a normalization of the neutrino spectrum , see @xcite . a model of @xmath4 interactions in agn and collisions of protons from the core with protons of the host galaxy is derived in @xcite . this model will be referred to as _ m95-a_. alternatively to optically thin sources with photon emission above @xmath5 mev , the environment can be optically thick allowing only photons below @xmath5 mev to escape . in that case , the neutrino signal can be normalized to the diffuse extragalactic contribution measured by comptel at energies in the range of @xmath6 mev @xcite . this would enhance the contribution of neutrinos from @xmath4 by almost an order of magnitude as shown in @xcite . we will refer to this model as _ m95-b _ in the following . a model by @xcite was originally using the diffuse cosmic x - ray background ( see sec . [ rosat_prediction ] ) and has recently been modified in a way that it is using the comptel diffuse background to normalize the neutrino spectrum @xcite . this reduces the formerly very high contribution by a factor of 10 . in addition , oscillations have been taken into account which leads to a further reduction by a factor of 2 . the measurement of the diffuse extragalactic contribution in x - rays by rosat has raised the question whether it is produced by radio - weak agn . assuming that the x - ray emission comes from the foot of the jet , the x - ray signal would be accompanied by a neutrino flux . a model by @xcite and an approach by @xcite have been presented . an alternative scenario would be the up - scattering of thermal electrons via the inverse compton effect . in that case , the x - ray component would not be accompanied by a neutrino signal , see e.g @xcite . until today , about @xmath7 of the diffuse x - ray signal has been resolved by rosat @xcite , with the help of chandra and xmm - newton data , this number can be updated to @xmath8 , see @xcite and references therein . more than @xmath9 of the diffuse background could be connected to the x - ray emission of active galactic nuclei most of which are radio weak . amanda s measurements of a diffuse neutrino flux did not yield a significant signal and constrains these models strongly . detailed examination of multi - wavelength observations have shown a correlation between the disk and jet power of active galactic nuclei , see @xcite . this correlation can be used to determine the diffuse neutrino flux from fr - ii galaxies and flat spectrum radio quasars within the framework of the jet - disk symbiosis model @xcite . the neutrino flux has been assumed to follow an @xmath10 spectrum which leads to a relatively high neutrino flux at energies above tev energies . assuming on the other hand a correlation between the radio spectral index and the index of the proton spectrum leads to an @xmath11 neutrino spectrum . since the flux is normalized at low energies , the contribution decreases significantly at higher energies with an increasing spectral index . the conversion of stacking limits into stacking diffuse limits is the main topic of this section . basically we use the contribution of the identified sources to the integrated background at a chosen electromagnetic wavelength to estimate by way of a physical model the corresponding ratio for neutrinos : what fraction of identified neutrino source candidates goes towards the integrated background ? it is important to note here that we use conservative estimates in order to obtain absolute , reliable upper limits . as limits to the flux of identified source classes we use the sample of agn classes from the amanda stacking analysis which is described in @xcite . to avoid any confusion of the notation , a paragraph on the limit conventions precedes the actual discussion of the limits . throughout this paper , several types of neutrino flux limits will appear . limits are usually given in the form @xmath12 with @xmath13 as the differential neutrino flux at earth and e as the neutrino energy . throughout the paper , the limits will be denoted as follows : * @xmath14 : diffuse limit ( dl ) derived by using data from the complete northern hemisphere . it is given in units of @xmath15 . * @xmath16 : stacking limit ( sl ) in units of @xmath17 , obtained for the point source flux from a certain class of agn . the principle is indicated schematically in fig . [ stacking_diffuse_fig ] . as an example , three source classes , fr - i and fr - ii galaxies as well as blazars , are displayed . the stacking limits are obtained by stacking the most luminous sources of the same class in the sky , indicated by sources with filled circles . weaker sources ( empty circles ) are not included . * @xmath18 : stacking diffuse limit ( sdl ) , derived from the stacking limit in the same units as the diffuse limit , @xmath15 . it is determined by taking into account the contribution from weaker sources as well as yet unidentified sources , present in a diffuse background . a similar convention is used to denote the corresponding sensitivities : * @xmath19 : diffuse sensitivity ( ds ) , giving the sensitivity of amanda towards a diffuse signal from the northern hemisphere . units are @xmath15 . * @xmath20 : the sensitivity to a single point source in amanda . units are @xmath17 . the unfolded spectrum of the diffuse neutrino flux as observed by the + amanda experiment is indicated in e.g. fig . [ blazars_thin ] . it can be seen that the measured flux follows the prediction of the atmospheric neutrino spectrum and no extragalactic contribution is observed @xcite . the unfolded spectrum is derived using data from the year 2000 . the current limit on the diffuse muon neutrino flux from extragalactic sources is given for four years of data ( 2000 - 2004 ) @xcite , @xmath21 in the energy range of @xmath22 gev to @xmath23 gev . in addition to a diffuse search , amanda examines a possible signal from single sources . no signal excess above the atmospheric background was observed yet . to maximize the significance of the signal to background expectation , a stacking method has been applied examining different source classes of agn @xcite . this method is commonly used in cases where a single source is not likely to have a significant signal above background ( e.g. * ? ? ? given the limit for a certain source class , @xmath16 , the differential flux renormalized to a diffuse signal is @xmath24 the transition from point source to diffuse flux is done by dividing the source limits by amanda s field of view , @xmath25 sr . the stacking factor @xmath26 is the ratio of the total photon signal in the sample and the photon signal which is included in the analysis . figure [ stacking_diffuse_fig ] gives a schematic representation . the stacking factor corrects for all sources with empty circles in the diffuse limit , since the signal from these sources is not included in the stacking analysis . the diffusive factor @xmath27 represents the ratio of the total diffuse flux to the contribution of resolved sources . sources which have not been resolved yet have to be considered as potentially contributing to the total diffuse neutrino signal as they contribute to the isotropic photon signal . in fig . [ stacking_diffuse_fig ] , this is indicated by the diffuse background which is different for each source class . this background is identified in a few cases by looking at the luminosity function of the objects , in others by measurements of the diffuse component . a detailed view on @xmath26 and @xmath27 will be given in section [ stack_results ] . recently , a stacking analysis has been published for the first time for a neutrino signal from active galactic nuclei for which eleven agn source classes have been defined . the source classes have been defined and selected as follows . the catalogs below have been selected in @xcite . 1 . gev blazars detected by egret ( _ gev _ ) . 2 . unidentified gev sources observed by egret ( _ unidgev _ ) . sources of possible galactic origin have been excluded . blazars that have been observed at infrared wavelengths ( _ ir _ ) . blazars that have been observed at kev wavelengths by the experiment heao - a ( _ kev(h ) _ ) . 5 . kev blazars observed by the rosat experiment ( _ kev ( r ) _ ) . 6 . the class of blazars with observed tev emission ( _ tev _ ) . 7 . compact steep sources ( css ) and ghz peaked sources ( gps ) ( _ css / gps _ ) as selected by @xcite . low luminosity faranoff riley galaxies ( fr - i ) with m-87 included ( _ fr - i(m ) _ ) . 9 . fr - i galaxies excluding m-87 ( _ fr - i _ ) . these two different cases are necessary to consider , since m-87 is the closest agn ( distance @xmath28 mpc ) and dominates the total signal of all fr - i galaxies . high luminosity faranoff riley galaxies ( _ fr - ii _ ) . 11 . quasi stellar objects ( _ qso _ ) . references to the corresponding catalogs as well as a detailed description of the analysis methods are given there . in brackets , the abbreviations for the source classes are given as they appear in fig . [ 5yr_stack_fig ] . the limits on the neutrino flux from the given source classes are displayed in fig . [ 5yr_stack_fig ] . they are compared to amanda s single source sensitivity of @xmath29 to point out the improvement that has been achieved in the stacking approach . in the following , the stacking factor @xmath26 and the diffusive factor @xmath27 will be discussed . it will be shown that seven of the eleven samples can effectively be used to determine a diffuse limit on the neutrino flux from the given class . the number of sources in a sample that is included in the stacking analysis is determined by optimizing the expected significance for signal and background . subsequently , the amount of the total photon signal of the analyzed sources is determined by the number of included objects . this is shown schematically in fig . [ stacking_diffuse_fig ] . the strongest fr - i galaxies ( filled circles ) are included in the stacking analysis . the stacking process for fr - i galaxies is indicated by circles around the stacked objects . fr - i galaxies having a too weak signal ( empty circles ) are not part of the stacking and have thus to be included in the _ stacking factor _ @xmath26 . quantitatively , the ratio of the total photon signal and the signal of the sources that are used in the stacking approach is represented by @xmath26 . the precise values of @xmath26 for each class are given in table [ 5yr_stack_tab ] . some of the stacking classes can not be used for a diffuse interpretation . firstly , the number of sources needs to be sufficiently high . an indication for an insufficient number of sources would be the use of all the sources in a sample in the stacking analysis . secondly , the luminosity function should be moderately falling : if only very few sources in our vicinity dominate the signal , no conclusions about the overall diffuse contribution can be justified . the complete sample was used for stacking in several cases due to the low number of detected sources in the given wavelength interval . in particular , these source classes are the infrared and heao - a sources as well as the tev blazars . since all twelve observed infrared sources have been used and only three sources are reported from heao - a , a diffuse limit to the source class can not be derived for these samples . there are five sources in the stacking sample of tev blazars . due to the small number , this class can not be used for a diffuse interpretation . due to the strong absorption of the tev signal in photons , the contribution of a neutrino flux from tev - resolved photon sources is expected to be much lower than the total flux . this is discussed in detail in section [ compare ] . apart from the three classes already excluded , the class of fr - i galaxies including m-87 is not suitable for an interpretation of a stacking limit as a diffuse limit : m-87 makes up most of the photon signal from all fr - i galaxies . this is also the reason why the stacking limit is very close to the single point source sensitivity . this result must therefore rather be interpreted as a point source result of m-87 . the fr - i sample without m-87 on the other hand is applicable for the diffuse interpretation . .table of the source class limits obtained with the stacking method . five years of data , 2000 - 2004 , have been used for the analysis with amanda in @xcite . listed are the source class , the selection wavelength , the stacking and diffusive factors , the stacking limit @xmath16 in units of @xmath30 and the possibility of interpreting the stacking result as a diffuse limit . [ cols= " < , < , < , < , < , < " , ] apart from the contribution of resolved sources to a diffuse background , a component of unresolved sources has to be considered . the ratio of the total diffuse signal to the signal of resolved sources is called the _ diffusive factor _ @xmath27 in the following . in fig . [ stacking_diffuse_fig ] , this is indicated as the diffuse photon background of unidentified sources . a truly diffuse contribution from e.g. dark matter decays is discussed as a component of the diffuse photon background @xcite , but it will be considered as negligible here . any contribution of a diffuse component not connected to any source population would improve the limits accomplished . @xmath9 of the diffuse x - ray background has been identified as agn in the rosat data sample @xcite . thus , the diffusive factor @xmath27 , which is the inverse of the fraction of resolved sources , for the kev rosat data sample is given as @xmath31 . note that today , xmm - newton and chandra provide values of @xmath32 resolved sources . but since both the neutrino flux predictions are based on and the stacking limit has been derived from rosat data , a fraction of @xmath9 resolved sources is used . for egret sources , it could be shown that only 1/12 of the total diffuse background can be made up by resolved sources , as derived by @xcite . the claim from @xcite that about @xmath33 of the diffuse background of blazars is made up by resolved agn , while the remaining contribution is from unresolved sources which are not blazars , would reduce @xmath27 significantly to @xmath34 . however , @xcite point out that this estimate is based on assumptions that do not hold for egret blazars . thus , @xmath35 will used as a conservative number in the following . with the launch of glast , the question of the contribution from resolved blazars will most likely be settled . for the remaining source classes , there is no explicit estimate of the contribution of resolved sources to a diffuse background . in this case , an estimate of @xmath27 can be possible for source populations where there is an estimate of the weakest sources in the sky . in this case , the luminosity evolution of the population has to be considered . in the euclidean case , sources evolve as @xmath36 with @xmath37 as the observed flux and @xmath38 as the number of sources for fluxes greater than @xmath37 . in such a case , the integral of the distribution diverges for a vanishing minimal source strength , @xmath39 . for cosmological sources , the behavior is not euclidean and in some cases , it is possible to determine a lower threshold for the flux , see ( * ? ? ? * for instance , such determination can be done in the case of fr - ii galaxies which are high luminous agn with a defined luminosity limit of @xmath40 erg / s @xcite . the determination of a lower luminosity limit enables the calculation of the total number of sources in a population @xmath41 . the total number of sources is given by integrating the number per flux interval , dn / ds over the flux : @xmath42 for the case of divergence for @xmath43 , the lower integration limit is essential to know as mentioned before . this is not trivial , since measurements for most source classes do not reveal the low luminosity cutoff . the luminosity functions of different agn classes at @xmath44 are shown in fig . [ lumi_fcts ] . in this case , the diffusive factor can be estimated by assuming that sources which are not in the catalog are not stronger than the weakest source in the catalog . thus , the flux of any additional source @xmath45 is smaller than the flux of the weakest source , @xmath46 , @xmath47 for all non - resolved sources @xmath48 . the total flux not considered in the calculation is thus @xmath49 here , @xmath50 is the number of additional sources which is calculated from the total number of sources expected in the sky , @xmath41 , subtracting the number of sources in the catalog , @xmath51 . then , @xmath27 is calculated to @xmath52 where @xmath53 is the total flux in the catalog . figure [ ximax_ntot ] shows the behavior of the maximum value of the diffusive factor @xmath54 according to equ . ( [ ximax ] ) at the example of egret sources . the figure emphasizes the challenge in such a representation . it is essential to have an estimate of the total number of sources in the class to get a good estimate for @xmath54 . this implies the necessity of the knowledge of the absolute lower luminosity limit . for fr - ii galaxies , this is relatively simple , since the number of sources decreases for low luminosities @xcite . fr - i galaxies on the other hand have a high number of low luminosity sources and it is not known at which luminosity the function turns . in such a case , the diffusive factor can be quite large , since a high contribution from low luminosity sources is expected . thus , @xmath27 has explicitly been determined according to the neutrino flux model which is tested by the limits . for fr - ii galaxies , the investigated neutrino flux model presented by @xcite uses the luminosity function by @xcite , labeled fr - ii in fig . [ lumi_fcts ] . in this case , a total number of @xmath55 sources is expected and the maximum diffusive factor is determined to @xmath56 . for the remaining source classes , fr - i galaxies , gps / css and qsos , the value of @xmath54 will not be given , since no explicit neutrino model for these source classes is examined here . note that @xmath54 gives an absolute upper limit : every source not included in the stacking analysis is assumed to have the flux of the weakest included source , @xmath46 . this results in an overestimate of @xmath27 , since it is likely that only a small fraction of the sources has such a high flux . for egret sources , equ . ( [ ximax ] ) can be used to determine a lower limit of sources contributing to the diffuse egret background . figure [ ximax_ntot ] shows the dependence of the maximum value of @xmath27 on the total number of sources in the source class of blazars emitting at @xmath57 mev . using a diffusive factor of @xmath58 , the total number of sources contributing to the total diffuse egret background is @xmath59 . figure [ stack_diff_5yr ] shows the diffuse limits that could be derived from the stacking approach for each source class . it is compared to the diffuse limit of amanda ( solid line ) for the operation time considered ( 2000 - 2003 ) , + @xmath60 , given in @xcite . several stacking diffuse limits are more restrictive than the general diffuse limits . the exact numbers are given in table [ 5yr_diff_tab ] , where the stacking limit , the diffuse bound derived from the stacking method and the ratio of the latter and the general diffuse limit are compared . note that in the case of css / gps , qsos and fr - i galaxies , the diffusive factor still has to be applied . figure [ stack_diff_5yr ] also indicates the waxman / bahcall bound ( wb bound , dashed line ) . this bound gives a theoretical estimate of the maximum contribution to be expected in neutrinos from active galactic nuclei @xcite . this bound is valid for optically thin sources at an energy of @xmath61 ev , where it is most restrictive . the description for all energies is given by @xcite . five of the seven source class limits lie below the indicated wb bound . this shows that the sensitivity reached with this method in amanda is already extremely high . next generation s neutrino telescopes such as icecube and km3net which will have a cubic kilometer of ice resp . water instrumented will therefore be able to provide more information either by giving very strong restrictions on prevailing models or by confirming a positive neutrino signal . note that the class of tev blazars is on the other hand an optimal candidate for a general diffuse search . the contribution of photon - observable tev blazars is very small , since many sources are hidden in photons due to the strong absorption , while they would still be visible in neutrinos . this will be discussed in section [ compare ] . the diffuse limits discussed in sections [ introduction ] and [ stack_results ] constrain some of the currently discussed neutrino flux models . however , it is known that these models bear different uncertainties in both normalization and spectral shape due to a lack of knowledge of the conditions at the source . in this section , the neutrino flux models having been discussed in section [ introduction ] will be examined with respect to the limits obtained . the detection of tev photon emission from distant sources is limited due to the absorption of high energy photons by the extragalactic background light ( e.g. * ? ? ? * ; * ? ? ? the absorption factor @xmath62 describes the ratio of the total emitted tev photon flux from hbls and the tev flux from hbls up to a redshift @xmath63 . it is a measure of the absorption of tev photons . the ratio of a diffuse neutrino signal from photon - observable tev blazars will be calculated using the general diffuse limit and taking into account the absorption factor for tev photons : only a relatively small fraction of all tev blazars can be identified in tev photons , since sources at high redshifts are hidden due to the strong absorption . in this paragraph , the absorption factor and the general diffuse neutrino limit are used to derive the maximum contribution of tev photon - observable sources to the total diffuse neutrino flux . in the case of tev blazars , the absorption factor @xmath62 is much greater than unity due to the strong absorption of tev photons by the extragalactic background light . hbls seem to have no or even a slightly negative evolution ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? for a no - evolution scenario as it is discussed for bl lacs , the co - moving density @xmath64 is considered to be constant with redshift , @xmath65 . using a negative evolution with less than @xmath66 does not change the results significantly . we also neglect the positive source evolution which has to be present up to a certain redshift @xmath67 : both effects positive evolution up to @xmath67 and negative evolution at higher redshifts are believed to cancel . each effect for itself alters the result less than @xmath68 in opposite directions . the normalization of the co - moving density is not important in this calculation , since only ratios of concrete values of @xmath69 are considered and the constant of proportionality cancels . the neutrino flux from sources up to a certain redshift @xmath63 is given by @xmath70 where @xmath71 is the energy of the neutrino at the source . the @xmath72-weighted flux is thus given as @xmath73 here , @xmath74 is the evolution factor depending on the upper redshift integration limit @xmath63 . if all present sources are considered , @xmath75 is a reasonable value . recent searches for luminous galaxies at redshifts between @xmath76 imply that only few ultra - luminous objects only exist beyond @xmath77 @xcite . at higher redshifts , no large galaxies are observed , leading to the conclusion that only smaller galaxies can be present . also , the results to not change significantly when going up to higher redshifts : the major contribution comes from redshifts of @xmath78 . the absorption factor @xmath62 is given by the ratio of the total flux and the contribution that is observed by tev - photon experiments : @xmath79 here , @xmath63 is the maximum redshift at which tev photon sources can be identified by present imaging air cherenkov telescopes ( iacts ) . tev sources have up to today been observed up to a redshift of z=0.186 - 1es1101 - 232 @xcite . as a conservative upper limit , the maximum redshift of observation with iacts is taken to be @xmath80 , since an upper bound is derived from these values . the absorption factor depending on @xmath63 is shown in fig . [ absorption ] . at @xmath80 , the numerical value is @xmath81 it is important to note that for a lower energy threshold ( e.g. 30 gev ) , iacts could be able to detect sources up to @xmath82 . this would lead to a much lower absorption factor of @xmath83 using the above reflections , one can argue that a diffuse limit can be used to estimate the maximum diffuse signal from hbls resolved in tev photons : from the existing general diffuse limit , it is known that no source type and , thus particularly also not the class of tev blazars , can contribute more than @xmath84 : @xmath85 an upper limit to the normalization constant @xmath86 of the neutrino flux can be derived by inserting equ . ( [ all ] ) with @xmath87 in equ . ( [ all_dl ] ) : @xmath88 the contribution of sources at @xmath89 is given as @xmath90 with the final result of @xmath91 the upper limit on the neutrino flux up to @xmath63 is shown in fig . [ flux_fraction ] . the curve shows the no - evolution scenario . the contribution of photon - observable tev blazars to the diffuse neutrino flux is therefore limited to @xmath92 where @xmath93 is used as a lower limit . the contribution of sources observed by iacts is thus about three orders of magnitude lower than the possible total contribution . this is displayed in fig . [ blazars_thin ] together with the general diffuse limit . the limit to the maximum contribution from photon - observable tev blazars is shown ( _ obs . tev _ ) as well . the contribution is a factor of @xmath94 lower than the diffuse limit . this indicates that a diffuse analysis of tev sources is most effective with an overall diffuse approach , since most of the tev sources are hidden due to the strong absorption at such high photon energies . the stacking of the strongest sources in the sky in comparison only comprise a small fraction of the total diffuse flux as these calculations show . this result stands in contrast to the other sources samples where the selection of the strongest sources yields stronger restrictions than a general search for a diffuse signal , see following paragraphs . the importance of reaching down to low energies with iacts is demonstrated in this calculation , since lowering the energy threshold to observe sources up to @xmath95 would already include sources with a one order of magnitude higher maximum neutrino flux of @xmath96 if the sources are optically thick to photon - neutron interactions , @xmath98 , the photon signal which is emitted from the sources lies in the mev to gev range . thus , egret and comptel diffuse cosmic photon fluxes are used to determine the expected neutrino contribution . figure [ blazars_egret ] shows neutrino flux predictions in connection with egret data and the derived stacking diffuse limits for a neutrino signal from egret sources . since the indicated models are all normalized to the egret diffuse spectrum , the limit of identified egret blazars applies . the gray line between @xmath99 tev and @xmath100 tev indicates the most conservative calculation where it is assumed that the total diffuse background as measured by egret it produced by agn @xmath101 . the limit falls short of the mpr bound and also touches the maximum prediction for the source class predicted by mpr and it is possible to reach sensitivities at the level of predictions concerning a neutrino signal from blazars . the same is valid for the prediction by mannheim ( m(95 ) , a ) : the proton - proton contribution from optically thick blazars can be constrained even though the atmospheric flux exceeds the prediction at these energies . with an overall diffuse analysis , it would not be possible to extract any information on this low - energy part of the spectrum from agn . the contribution from lbl within the proton - blazar model dominates the total observable flux at ultra high energies . with the current data , it is not possible to get to any reliable conclusion from the stacking method , since it is valid at energies @xmath102 pev while the flux is present at much higher energies . a stacking analysis of cascade events with second generation telescopes such as icecube would be an option to explore regions of energies at @xmath103 pev . a stacking approach for cascades is however accompanied by the challenge of getting a reasonable directional reconstruction of the neutrino events , since the cascade signal is generally not as boosted as a signal from neutrino - induced muons . even in the most conservative case of a diffusive factor of @xmath35 , the model of proton - proton energies can be constrained at low energies . with the stacking method , the diffuse limit can be extended to sensitivities far below the flux of atmospheric neutrinos . a stacking analysis using comptel blazars has not been done yet . the option is discussed in section [ class_analysis ] . current diffuse neutrino flux limits are still about a factor of 2 or more above the predictions . the hypothesis of an avalanched tev signal down to mev energies will be tested by icecube . the x - ray component of radio weak agn can be correlated to the emission of neutrinos at the foot of the jet . this has been investigated in calculations of @xcite , @xcite as well as @xcite . all models are constrained by the current general diffuse amanda limit . the resolved rosat sources which have been used to determine the stacking limit are radio loud and thus , the limit is not as easily applicable to these predictions . if the same production mechanism is assumed in radio loud as in radio weak sources , the stacking diffuse limit of rosat sources would apply in this case . however , the limit to radio weak objects @xmath104 is about a factor of 10 higher than the calculated limit for radio loud sources @xmath105 , since radio loud objects are about a factor of 10 less frequent than radio weak sources , @xmath106 the more restrictive limit in this case is therefore the general diffuse limit . figure [ rosat ] shows the prediction of neutrinos being produced coincidentally with x - rays at the foot of agn jets . the diffuse rosat - measured background has been used in two models @xcite to normalize the neutrino spectrum . in the third case presented by @xcite , the luminosity evolution function of radio quiet agn has been used so that the result also applies to the correlation with the x - ray diffuse background . in the case of the prediction by @xcite , amanda s diffuse neutrino flux limit lies an order of magnitude below the flux . stecker et al . updated their calculation by normalizing to the diffuse flux as measured by comptel @xcite . this is discussed in the context of the mev photon emission in correlation with neutrinos in section [ thick ] . the absence of neutrinos from rosat - detected sources implies strongly that the x - ray emission from agn can not be directly correlated to neutrino emission . while the missing emission of neutrinos disfavors a hadronic scenario for the x - ray emission from agn , it is consistent with inverse compton ( ic ) models for the x - ray emission and underlines leptonic models for the production of x - rays in agn . calculations by @xcite show a flux peaking at higher energies compared to the prediction from @xcite . the main reason is that @xcite use a more conservative estimate of the maximum energy and a simpler approach for the spectral behavior . the spectrum of the model by @xcite is not exactly @xmath10-shaped , while the amanda limits are typically derived assuming an @xmath10 spectrum . the modeling of different spectral shapes , including @xcite , has been done recently , see @xcite . the curved blue line indicates the limit modeled according to the model of @xcite . the limit lies half an order of magnitude below the prediction . while the previously described models normalize the diffuse spectrum by using rosat data , calculations by @xcite use the radio luminosity function for radio quiet agn and a single source normalization . further differences with respect to model 1 and 2 are the maximum energy and the break in the neutrino spectral shape . the underlying idea is the same by assuming proton - proton and proton - photon interactions at the foot of agn jets , leading to x - rays in coincidence with neutrinos . this model overproduces neutrinos as well as the previous ones . three independent calculations on the correlation of x - ray emission from radio quiet agn and neutrino emission have been examined . the absence of neutrino emission from the foot of relativistic jets implies that the particles there are not accelerated to high energy , and then interact . however , the concept that the innermost ring of the accretion disk , just underneath the jet , turns into an advection dominated accretion flow ( adaf ) with a very high temperature ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , is consistent with the results . this implies that the spin parameter of the black hole is high , above 0.95 , and that hadronic interaction at weakly relativistic temperatures produce charged pions , which decay and provide an energetic particle seed population for further acceleration downstream for the radio emission of the jet @xcite . this is then consistent with an inverse compton explanation of the x - ray emission @xcite , and predicts that there should be a large production of energetic neutrinos at an energy commensurate with the pion mass . the radio emission of active galactic nuclei is likely to be directly correlated to neutrino emission in the jet . this has been discussed by @xcite at the example of fr - ii radio galaxies and flat spectrum radio quasars ( fsrqs ) . in both cases , the normalization of the neutrino spectrum depends on many different intrinsic factors , i.e. the correlation between radio and disk luminosity and the parameterization of the maximum proton acceleration energy . the jet - disk correlation of agn has been worked out by @xcite . although the model includes several parameters , all are fixed by the comparison with the data except the accretion rate . within that model , it is shown that the parameter reach extreme values and are not strongly scattered . the jet - disk symbiosis model has been proven for different source types , reaching from microquasars to quasars . a parameter in the calculation of the neutrino flux is the optical depth of the source which is unknown . in @xcite , the effective optical depth of the source , @xmath107 is defined as product of the proton - photon optical depth @xmath108 and a reduction factor due to contributions from bethe - heitler production @xmath109 , @xmath110 it can be derived from the solution of the transport equation , the neutrino flux is proportional to @xmath111 . for large depths , the neutrino flux is therefor proportional to @xmath107 . the effective optical depth has been chosen to be @xmath112 in the calculations described above . the optical depth of proton - photon interactions is determined by the product of the photo - hadronic cross section @xmath113 and the photon density in the source , @xmath114 . these in turn depend on the disk luminosity @xmath115 and the extent @xmath116 of the source such as @xmath117 with @xmath118 erg / s and @xmath119 cm . this shows that the photo - hadronic optical depth itself uncertain , since low values around unity are possible as well as extremely high numbers . amanda limits will be used to derive limits on the optical depth of fr - ii galaxies and fsrqs . note that the determination of @xmath107 happens only within the specific model of neutrino production as it is described above . the difficulty of drawing more general conclusions lies in the uncertainty of the spectral index . this highly depends on the original spectral behavior of the protons . this can only be simulated and is not directly observable . using a generic @xmath10 spectrum for the neutrino flux from fr - ii galaxies , the flux normalization has been derived to be @xmath120 comparing the normalization in equ . ( [ bbr_norm ] ) with the diffuse limit shows that a flux with the chosen parameter settings is not detected . an upper limit to the optical depth can be derived from the limit yielding @xmath121 within three years of icecube , it will be possible to explore sources with @xmath122 . figure [ fr2 ] shows the the prediction of a neutrino flux from fr - ii radio galaxies . model ( 1 ) is the original calculation of @xcite modified by a factor @xmath107=0.5 . this is the maximum contribution for an @xmath10 signal from those sources not violating the diffuse limit of amanda . however , there is another production scenario which leads to a significant reduction of a contribution from fr - ii galaxies . the synchrotron spectral index of fr - ii galaxies @xmath123 is correlated to the neutrino spectral index @xmath124 by @xmath125 ( * ? ? ? the mean synchrotron spectral index is @xmath126 for large scale emission and therefore , the neutrino spectrum can be as steep as @xmath11 . this prediction is shown in fig . [ fr2 ] as model ( 2 ) . in that case , the contribution is very low and does neither violate the current diffuse limit , nor is it within reach of icecube s sensitivity for three ( or even five ) years of operation @xcite . in such a case , an observation of a neutrino signal from fr - ii galaxies would be very difficult . if the spectrum were this steep for real , then all ultra high energy cosmic ray ( uhecr ) arguments would fail and so it may be more appropriate to assume a flatter spectrum . the radio spectral index of the hot spots are usually about @xmath127 and so correspond to a particle spectral index of about @xmath128 . using fig . [ fr2 ] it can be estimated that a spectrum with an index of @xmath129 could already be explored by the three year sensitivity of icecube . using a similar correlation between radio and neutrino flux as for fr - ii galaxies , the contribution from fsrqs to the diffuse neutrino flux can be calculated to @xmath130 the flux lies well beyond the current diffuse limit and even below the sensitivity of icecube for an optical depth around @xmath112 . equivalent as for the case of fr - ii radio galaxies , the upper limit to the optical depth in fsrqs can be determined to be @xmath131 by using the diffuse amanda limit . the effective optical depth is quite uncertain in the case of flat spectrum radio quasars . just as an example , a sub - class of fsrqs are gps , which are compact agn where it is assumed that the jet runs into dense matter , yielding a high potential for proton - photon interaction which results in a high optical depth . however , other sources like tev blazars contribute to the source class , being at the other end with a very low optical depth . with icecube , it will be possible to examine sources with @xmath132 . consequently , neutrino emission in coincidence with the radio signal from fsrqs will only be observable in the near future , if the optical depth of fsrqs is significantly higher than the one for fr - ii galaxies . in the previous sections it has been shown that the stacking method can be used to increase the diffuse sensitivity of high energy neutrino telescopes to certain source classes . further catalogs published recently which are not yet part of the stacking search in amanda will be discussed in this section . the references for the catalogs are given throughout the text . they will be examined with respect to the potential neutrino signal and the effectiveness of the method . also , a diffuse component from unresolved sources is expected to show deviations from a purely isotropic flux , since it is expected to follow the source distribution in the sky . the total signal from a certain source class to be observed by a neutrino telescope is highly dependent on the detector s field of view . we will generically examine source classes concerning the total flux from the northern respectively from the southern hemisphere . this gives a qualitative examination of the capabilities of icecube and km3net which is planned to observe the southern hemisphere . the local supercluster for example is observable from the northern hemisphere and so in some cases , a significant fraction of the total flux in a sample comes from that hemisphere . it is shown , however , that there are classes with the dominant contribution in the southern sky . a further constraint is the limitation of the source catalogs themselves : radio data are mainly given for the northern hemisphere and there are only a few southern identified sources in the case of very sensitive radio - selected samples . here , high energy photon catalogs which are mostly provided by satellite experiments are much more complete in the sense of directionality . this subsection gives an overview of four additional source catalogs which yield high capabilities for neutrino searches with the stacking approach . the point source sensitivity is likely to be increased in all the cases , diffuse limits can be derived in two of the cases . a summary of the basic properties of these catalogs is given in table [ new_cats_basics ] . the maximum diffusive factor for the examined catalogs can be determined as described in section [ stack_results ] , equ . ( [ ximax ] ) , assuming that the total number of sources in that class can be determined . this is quite challenging , since a lower luminosity limit is difficult to determine . figure [ xi_new_cats ] shows the increase of the diffusive factor with @xmath41 for the whole sky for all four examined catalogs . table [ new_cats_prop ] reviews the parameters used for the four catalogs . neutrino flux predictions using the diffuse background as measured in hard x - rays by comptel are based on the assumption that the cosmic diffuse photon flux below 100 mev is directly correlated to the neutrino flux . about 10 blazars could be resolved by comptel underlining the assumption that intrinsically weaker agn are responsible for the diffuse signal . no neutrino stacking analysis of the resolved comptel blazars has been done yet . using the same assumptions as @xcite ( _ m95-b _ ) and @xcite ( _ stsa05 _ ) would challenge such an analysis . the derivation of a diffuse limit from a stacking limit would however be difficult at this stage , since only a negligible fraction of the sources potentially responsible for a diffuse signal is resolved into sources . thus , the major fraction of the diffuse signal would not be included in a stacking analysis and the limit would not apply to diffuse predictions . the planned launch of satellites such as glast @xcite and mega @xcite encourages to pursue the examination of this source class , since it gives the prospect of increased statistics of resolved sources in the considered energy range . as mentioned before , the limit to a neutrino flux from rosat - detected sources would be improved significantly by selecting radio - weak sources instead of radio strong ones . a significant neutrino flux would not be expected as discussed in section [ compare ] , but the improvement of the limit would underline the fact that x - ray emission from the foot of the jet is not correlated with a neutrino signal . in hard x - rays , 15 sources have been identified in the integral survey by jem - x @xcite and could be interesting to examine . with the current analysis , heao - a results were used , while the more recent measurements by integral are bound to be more accurate . since a direct correlation between hard x - ray sources as detected by integral and soft x - ray sources as seen by rosat does not seem to exist , an independent analysis would be reasonable . the emission of hard x - rays is likely to be correlated with tev photon emission at least via the up - scattering of synchrotron photons by inverse compton scattering . therefore , under the assumption of a correlated emission of hard x - rays and tev photons , an analysis of the hard x - ray signal gives an indication of the flux from high - frequency peaked bl lacs . a catalog of 42 agn with emission in soft @xmath97-rays , ( @xmath133 ) kev has been released by integral @xcite . here , some of the sources detected by osse in the energy range of ( @xmath134 ) kev could be confirmed , others were not seen . we suggest to use integral data to define a source class for the stacking of potential neutrino sources . the emission of neutrinos from starburst galaxies has been suggested by @xcite ( l&w in the following ) . it is assumed that 1 . relativistic protons are accelerated along with relativistic electrons , 2 . the observed radio emission results from pion - induced electrons . the same pions produce neutrinos , 3 . protons lose all their energy in @xmath4 interactions before reaching the diffusion time . it has been pointed out by @xcite that l&w overestimate the fraction of the diffuse far infrared ( fir ) flux coming from starbursts . while it is @xmath135 on average , l&w assume that @xmath136 of the detected signal comes from starbursts . on the whole , the diffuse flux from starbursts is presumable much lower than predicted : l&w assume that starbursts are loss dominated , which means that most primaries interact and do not escape the source . this enhances the neutrino flux , since basically all protons lose their energy in @xmath4 or @xmath137 interactions and produce neutrinos . observations of the spectral radio index of the sources ( @xmath138 ) indicate however that starbursts are in the diffusion limit , indicating that a negligible fraction of protons interact and only few neutrinos are produced . this contribution can not be expected to be observed by icecube . there is another possibility to expect enhanced neutrino emission from starbursts . in the past few years , it could be shown that long grbs are typically connected to the explosion of wolf - rayet stars into a supernova ic . these occur preferably in star forming regions . thus , a diffuse flux of grbs similar to the prediction of @xcite should originate from the direction of these galaxies . there are two different ways to normalize the diffuse grb spectrum . one method is to assume that the observed kev - photon flux is proportional to the neutrino flux . in that case , the normalization is dependent on the number of observed grbs per year . this number is strongly dependent on the instrument and the number is not very exact . under the assumption that grbs accelerate protons up to the highest energies , @xmath139 ev , the neutrino spectrum can also be normalized to the flux of ultra high energy cosmic rays ( uhecrs ) . in this case , the normalization is independent of grb observations . it should be kept in mind that the spectral index of the spectrum still varies from burst to burst - in the model of waxman&bahcall , an average spectral index has been used . it is possible to look for a neutrino signal from grbs by stacking star burst galaxies . this method has one advantage over a triggered - grb search : it is a systematic search , since independent of grb data . a disadvantage is that only nearby events are included , since the sample of starbursts only reaches out to redshifts of @xmath140 . it should however be possible to use iras data to identify starburst galaxies at higher redshift . the search for a grb signal from starbursts should be considered as a systematic search for choked and undetected grbs . the sources can be selected according to their fir - flux , since this is a measure of the sn rate in a starbursts . a higher fir flux indicates a high star formation rate , thus more sne and therefore , also more grbs . in order to get diffuse limits from source stacking , it is important to choose source classes which have information on both resolved sources and diffuse background . that way , the parameters @xmath26 and @xmath27 are easily and correctly determined . a good example is the egret catalog , where both diffuse and resolved emission could be proven to be correlated . alternatively , the determination of an upper limit to @xmath27 is possible when working with a relatively complete catalog of the strongest sources in the sky by using a generic number of total sources in the source class as it is described in section [ stack_results ] , equ . ( [ ximax ] ) . the diffuse interpretation of source classes like tev blazars on the other hand is difficult , since iacts can barely look for diffuse emission and the sensitivity of all - sky monitors such as milagro is not high enough yet to detect extragalactic diffuse emission . future projects like hawc experiment as a successor of milagro @xcite and cta following in the footsteps of h.e.s.s . and magic , but aiming at a large field of view ( * ? ? ? * e.g. ) , will help enhancing the completeness of the sample . in the first approach of agn stacking in order to examine a potential excess in neutrinos , the number of used sources was determined by optimizing the significance in the detector . the optimal number of sources was in that case typically around @xmath141 . in order to achieve the best stacking diffuse limit it is interesting to optimize the number of sources by taking into account @xmath27 as well . it needs to be tested if , by reducing @xmath27 as much as possible and taking the penalty of a possibly increased point source stacking limit instead , the diffuse limit can be increased . also , it needs to be considered carefully if an optimization to a possible detection or to a limit is the most reasonable choice . for the standard analysis of muon neutrino signatures in high energy neutrino detectors , the field of view is @xmath25 sr . the two experiments icecube and km3net will be observing the northern respectively the southern hemisphere in the near future and whole sky coverage is achieved by the combination of the two experiments . previously , in the stacking analysis of amanda data , sources with @xmath142 were excluded due to the decreasing sensitivity and high muon background towards the horizon . here , we include all sources , for both northern and southern hemisphere down to @xmath143 , expecting a much better sensitivity and better muon rejection near the horizon for both icecube and km3net due to improved directional reconstruction . in this section , the different source classes discussed above will be examined with respect to their luminosity evolution . the total sample will be compared to the contribution from the northern and from the southern sky . for each sample , two figures will be discussed . ( a ) represents the differential source counts , @xmath144 versus the logarithm of the flux @xmath37 . ( b ) shows the integral source evolution with the total flux of @xmath145 sources , organizing the sources according to their strength , starting with the brightest one . open ( black ) sources display in both cases the whole sky . red stars consider only the southern hemisphere and blue triangles show the northern hemisphere contribution . figure [ egret ] shows the luminosity evolution of egret sources , for the whole sky as well as for the northern and southern populations . the three most luminous sources in the egret sky are in the southern hemisphere . that is why the main contribution from gev @xmath97 rays is located in the southern hemisphere where the neutrino signal contribution is thus expected to be much higher . given the restrictive limit which could already be derived from amanda data , this source sample is interesting for icecube . considering the source distribution in the sky , the class of gev blazars is particularly useful for km3net given the high total flux . the comptel catalog of sources with @xmath146 mev is displayed in fig . [ comptel ] . in this case , the main contribution lies in the northern hemisphere . the small number of only 11 identified comptel sources make a diffuse interpretation of a possible limit difficult . satellites like mega would make the investigation even more interesting . the 11 identified sources are however still useful to investigate with respect to a point source signal . the maximum diffusive factor lies around @xmath147 for 10,000 sources and @xmath148 for 100,000 sources in the class . depending on the steepness of the evolution function , this factor can be significantly smaller , so that doubling the number of sources could already help to draw a conclusion about a diffuse limit . + a catalog of soft @xmath97-ray - detected agn is given by integral @xcite . 19 of 42 sources are in the northern sky , among the 23 southern sources are the three most luminous ones . the catalog is a good candidate for diffuse interpretation of the neutrino results , given the relatively high number of sources . the luminosity evolution is displayed in fig . [ integral_flux ] . it can be seen that the evolution is still rising and that there is still a significant fraction of signal missing . this is mirrored in the numerical value of the maximum diffusive factor which is calculated to @xmath149 for @xmath150 sources and @xmath151 for @xmath152 sources . these relatively large numbers increase a potential limit on the neutrino flux from integral sources by about one order of magnitude it is already discussed that even with a source class of @xmath153 sources , the strongest @xmath154 agn make up the dominant contribution . thus @xmath155 is the more realistic value in this case . ] . figure [ rosat_cat ] shows the luminosity evolution for rosat - detected sources . 84 sources are in the sample totally , of which 35 are northern and 49 are southern sources . the three most luminous sources are in the northern hemisphere . however , these are objects which have such a large flux , since they are extremely close to earth and not because of their high intrinsic luminosity . this is why these sources were excluded in the stacking analysis of @xcite . without these sources , the contributions from both hemispheres are comparable . a sample of hard x - ray agn is available from the integral satellite , see fig . [ integral_hx ] . integral detected 15 sources in the energy band of @xmath156 kev to @xmath157 kev . among the 10 sources in the southern hemisphere are the two strongest sources in the sample . there are only 5 northern sources and the main contribution comes from the southern hemisphere . the integral sample can improve the stacking analysis of hard x - ray sources which has been done with heao - a information . only three northern sources were reported from heao - a . for a diffuse interpretation , there are still too few sources in the sample , though . the catalog fr - i and fr - ii sources is restricted to values of @xmath158 . therefore , only few southern sources ( 7 fr - i and 14 fr - ii galaxies ) are in the complete sample which is seen in fig . [ fr_i_cat ] for fr - i galaxies and in fig . [ fr_ii_cat ] for fr - ii galaxies . the flux is totally dominated by the northern hemisphere . the problem for the catalog of css and gps sources is similar to the situation of the fr - i and fr - ii catalog . css have been selected at @xmath159 , while gps include data with @xmath160 . therefore , there are no southern sources in the case of css . figure [ gps_cat ] shows the gps sample . 8 southern and 20 northern sources have been identified with the main contribution from the northern hemisphere . in the sample of qsos as presented by @xcite , there are only 3 sources in the southern hemisphere with a measured flux at the selection wavelength , @xmath161 m . the sample selection of the sample was done for @xmath162 , which only leaves a small window on the southern sky . while this source class is well - suited for the analysis of the northern hemisphere , there is too little data available in the southern hemisphere . the catalog of starburst galaxies includes 199 sources . while the most luminous sources are in the southern , the total flux is higher in the northern hemisphere , since 153 of 199 sources are located north . a stacking analysis has good potential for both hemispheres . in this paper , limits to the diffuse flux from seven different agn source classes could be derived from point source stacking limits from amanda . there is in many cases an additional constraint on flux models at the lowest energies : while the general diffuse limit is restricted to the energy band of @xmath22 gev to @xmath23 gev , the stacking limits reach down to @xmath163 gev . in many cases , the diffusive factor can conservatively only be estimated . it is expected , that the additional diffuse component is much smaller than the values derived here as upper limits . a reduction of the diffusive factor would result in the improvement of the stacking diffuse limits . neutrino flux predictions normalized to the diffuse egret signal above @xmath5 mev use the assumption that the diffuse component is produced by unresolved agn , see @xcite . thus , the egret stacking limit applies to such calculations . the limit falls short of the outline of two of the models , @xcite . the model of proton - proton interactions normalized to the egret diffuse results violates the limit at low energies around @xmath163 gev . at the highest energies , the stacking diffuse limit is more restrictive than the general diffuse limit and reaches below sensitivities of the maximum contribution from proton - photon interactions . a detection of a flux from optically thick blazars should be visible with second generation neutrino telescopes such as icecube . in the case of x - ray agn , the limit strongly disfavors a hadronic model . it is an order of magnitude below the flux predictions . according to the strongly restrictive limit , neutrinos are not produced at the foot of agn jets in coincidence with x - rays . thus , the diffuse emission as measured by rosat is likely to be due to inverse compton radiation . analyzing the correlation between the radio and neutrino emission from fr - ii galaxies and blazars , the limit was used to constrain the optical depth of the sources . for fr - ii galaxies , @xmath164 could be derived , while the upper limit for blazars is given as @xmath165 . thus , the detection potential for such a source class in icecube is very high , since within three years , icecube is sensitive to sources of optical depth with @xmath166 ( fr - ii ) resp . @xmath167 ( blazars ) . it was shown that a general diffuse analysis gives a high discovery potential for hidden tev sources : while tev photons are absorbed at high redshifts , neutrinos propagate freely . an upper limit to the neutrino flux from photon - resolved tev blazars is determined to be @xmath168 the considerations above show that an investigation of a neutrino signal can be used to help determining different intrinsic parameters of the source type considered . it could be shown that a stacking analysis yields valuable information on the diffuse contribution from theses sources . the detection probability is enlarged significantly in such an approach and there is a high detection potential with icecube and km3net . a cascade stacking analysis by these neutrino detection arrays should be considered in order to increase the sensitivity to ultra high energy neutrino fluxes as they are described by @xcite . apart from the catalogs which have already been used in the stacking analysis of amanda , further source classes are investigated here . the possibility of a stacking analysis of comptel sources is discussed in this paper as a first approach to examine optically thick sources with photon emission of @xmath146 mev . although no stacking diffuse limit can be derived yet due to the small percentage of resolved sources , future experiments such as glast at an energy range of @xmath169 gev @xcite and mega at @xmath170 mev @xcite give hope to resolve many more sources in the mev range . another interesting source class is the catalog of starburst galaxies . with the correlation between starforming regions and long duration gamma ray bursts , an enhanced neutrino signal from starbursts can be expected . the prospects of icecube and km3net are different considering source stacking . most catalogs show a bias to one of the hemispheres . both neutrino detection arrays can be used to extract complementary information about the different source classes . the upper limits derived in this paper constraint several prevailing neutrino flux models . this underlines the necessity of calculation on the basis of new developments within astroparticle physics and numerical approaches . a unified model concerning acceleration processes in agn would give the opportunity to examine a potential neutrino signal with respect to different agn classes within the same framework . the numerical results could be applied to individual sources and to diffuse flux measurements . the interaction between the observation of resolved sources and diffuse photon and proton components of crs is of high significance with respect as it could be shown in this paper . we would like to thank francis halzen , tanja kneiske , anita reimer , reinhard schlickeiser , karl mannheim , marek kowalski , elisa resconi , ty deyoung , soeb razzaque and werner collmar for inspiring discussions and helpful comments on this work . many thanks also to the entire icecube collaboration for useful comments and suggestions . + support for plb is coming from the auger membership and theory grant 05cu 5pd 1/2 via desy / bmbf . jkb and wr acknowledge the support by the bmbf , grant 05 ci5pe1/0 and by the dfg , grant rh 35/2 - 3 . , h. w. m. , 1826 . ber die durchsichtigkeit des weltraums . fr das jahr 1826 . later puplished in the collection _ olbers . sein leben und seine werke _ , _ editor : _ c. schilling , _ publisher : _ julius springer , _ year : _ 1894 . , g. , smith , a. , mcenery , j. e. , 2005 . in : novello , m. , perez bergliaffa , s. , ruffini , r. ( eds . ) , the tenth marcel grossmann meeting . on recent developments in theoretical and experimental general relativity , gravitation and relativistic field theories . p. 1068 . , a. , et al . , august 2006 . the next generation all - sky gamma - ray telescope : mini - hawc . in : tev particle astrophysics ii . http://icecube.wisc.edu/tev/presentations.php . , p. , et al . astrophys . j. 494 , 523 . |
we present a fundamental plane ( fp ) analysis of 141 early - type galaxies in the shapley supercluster at based on spectroscopy from the aaomega spectrograph at the aat and photometry from the wfi on the eso / mpi 2.2 m telescope . the key feature of the survey is its coverage of low - mass galaxies down tokms . we obtain a best - fitting fp relation in the band . the shallow exponent of is a result of the extension of our sample to low velocity dispersions . considering the subsample of km s galaxies , the fp relation is , consistent with previous studies in the high - luminosity regime . we investigate the origin of the intrinsic fp scatter , using estimates of age , metallicity and/fe . we find that the fp residuals anti - correlate ( ) with the mean stellar age in agreement with previous work . however , a stronger ( ) correlation with/fe is also found . these correlations indicate that galaxies with effective radii smaller than those predicted by the fp have stellar populations systematically older and with over - abundances larger than average , for their . including/fe as a fourth parameter in the fp , the total scatter decreases from 0.088dex to 0.075dex and the estimated intrinsic scatter decreases from 0.068 dex to 0.049dex . thus , variations in/fe account for of the total variance around the fp , and of the estimated intrinsic variance . this result indicates that the distribution of galaxies around the fp are tightly related to the enrichment , and hence to the timescale of star - formation . our results appear to be consistent with the merger hypothesis for the formation of ellipticals which predicts that a significant fraction of the scatter is due to variations in the importance of dissipation in forming merger remnants of a given mass . [ firstpage ] galaxies : abundances , galaxies : ellipticals , galaxies : formation , galaxies : fundamental parameters , galaxies : structure | 0902.4383 | early - type galaxies are observed to obey a set of scaling relations that connect their photometric and kinematic properties ( e.g. kormendy relation , kormendy 1977 , faber - jackson relation , faber @xmath13 jackson 1976 ) . among these , the most notable , due to its surprising small scatter ( @xmath14dex ) , is the relation between effective radius @xmath15 , mean surface brightness within the effective radius @xmath16 and central velocity dispersion @xmath5 @xcite . in the three - dimensional space ( @xmath17@xmath15 , @xmath18 , @xmath19 ) , elliptical galaxies populate a tight plane known as the fundamental plane ( fp ) and usually expressed in the form : @xmath20 if elliptical galaxies formed a homologous family , i.e. systems with density , luminosity and kinematical structures equal over the entire early - type sequence and with constant mass - to - light ratios , then the virial theorem predicts a correlation with @xmath21 , @xmath22 . however , observations show that the plane is somewhat `` tilted '' with respect to virial expectations , with best - fit scalings @xmath23 , @xmath24 ( e.g. * ? ? ? the origin of the fp tilt has been much debated and can be interpreted as the breakdown of either of the two assumptions in the virial expectation . a systematic variation in the mass - to - light ratio along the fp could be due to variations in the stellar content ( age , metallicity or imf ) and/or the amount of dark matter among ellipticals . performing detailed dynamical analyses of 25 galaxies with sauron integral - field stellar kinematics to @xmath15 , @xcite find the `` tilt '' almost exclusively due to real m / l variations of the form @xmath25 , while structural and dynamical non - homologies have negligible effects . they also find the variation of the dynamical m / l ratio to correlate with the h@xmath26 line - strength , and ascribe most of the tilt to stellar population ( age ) effects . on the other hand ( e.g. la barbera et al . @xcite , trujillo et al . @xcite ) other authors find that the tilt is not primarily driven by stellar populations , but instead results from other effects , such as non - homology . although the fp relation is quite tight , there is none the less a significant scatter around the plane that can not be attributed to measurement errors . the origin of this intrinsic component has been investigated by many authors . @xcite found that they were unable to reduce the scatter by introducing additional parameters , such as ellipticity or isophotal shape of the galaxies , into the fp relation . variations in stellar populations along the sequence of early - type galaxies are found to be partially responsible for the intrinsic scatter ( e.g. gregg 1992 ; guzman @xmath13 lucey 1993 ; guzman , lucey , bower 1993 ) . prugniel and simien ( 1996 ) studying the correlation between the residuals from the fp and the residuals from the colour and mg@xmath27 line - strength vs. luminosity relations , found that blue and low - mg@xmath27 elliptical galaxies deviate systematically from the value predicted by the fp . following this evidence forbes et al . ( 1998 ) , studying a sample of non - cluster galaxies , found that the residuals of the fp correlate with the ages of the galaxies , i.e. that the scatter of the fp is partly due to variation in galaxy age at a given mass , and in particular to variations in the time of the last starburst . on the contrary , they found that the effect of changes in metallicity is negligible . similar results were obtained by reda et al . ( 2005 ) analysing a sample of isolated galaxies : some objects deviate from the fp relation having lower m / l ratio and this was interpreted as due to their younger stellar populations , probably induced by recent gaseous merger . the same conclusions were reached by wuyts et al . ( 2004 ) for two high - redshift clusters . they found that the residuals from the fp correlate with the residuals from the h@xmath26 - @xmath5 relation . this confirms the role played by stellar populations in determining the appearance of the fp , with relations appearing more dispersed for samples of galaxies that are more dispersed in age . the existence of the fp , its small observed scatter and the tilt have presented a long standing challenge to theoretical models explaining the origin of early - type galaxies . in fact , whatever the scenario of formation and evolution of early - type galaxies is , it has to be able to explain the existence of such a tight correlation and its deviation from virial expectations and therefore to link galaxy structure and dynamics with their star - formation histories . in the recent years , observations and simulations have broadly supported the galaxy merging scenario which fits naturally into the @xmath28cdm hierarchical cosmology ( e.g. steinmetz & navarro , 2002 ) . in the hierarchical scenario ellipticals form through the merging of disk galaxies @xcite . in the merging context , significant new insights have been made through large - scale gas dynamical simulations of galaxy mergers @xcite , indicating that for lower mass galaxies , dissipation becomes increasingly important , driving nuclear starbursts that contribute larger mass fractions , and producing systematic trends with mass in both the structures and stellar populations of the remnant ellipticals @xcite . the fp tilt then arises as a direct consequence of the systematic trends with mass of the importance of dissipation during mergers . in the same scenario of galaxy formation , the origin of the intrinsic scatter in the fp arises as a combination of the scatter in the total baryon - to - dark matter content of the progenitor galaxies , and variations in the dissipational fractions at fixed stellar mass . this latter factor should be observable as correlations between the residuals from the fp and the stellar population parameters , and represents a critical test of the merger scenario @xcite , through the predicted coevolution of the stellar populations and structures of elliptical galaxies . the increasing importance of dissipation in the formation of low - mass galaxies and the different mechanisms that drive the evolution and the star - formation histories for low- and high - mass galaxies ( haines et al . 2007 ) should be reflect in variations with mass of the structural and kinematical properties and hence in variations both in the shape and orientation of the fp for these two families of galaxies . it should also be noted that non - merger origins may be important for lower mass galaxies , whose evolution turns out to be primarly driven by the mass of their host halo , probably through the combined effects of tidal forces and ram - pressure stripping ( haines et al . 2007 ) . + to date no large homogeneous sample of galaxies covering both the giant and dwarf regime exists . although many fp datasets for systems as different as bcgs , normal es , des , dsphs have been analysed and compared to look for changes in @xmath8 @xcite , these studies suffer from the non - homogenity of the samples both in terms of differences in measuring the quantities entering the fp ( for example different fits to derive structural parameters , different apertures to measure the velocity dispersion ) as well as the selection of galaxies themselves . the form of the fp obtained can be influenced by all of these criteria . in this paper we present a fp analysis of 141 early - type galaxies from the shapley supercluster ( @xmath29 ) with both new @xmath4-band surface photometry measurements and published velocity dispersion measurements from @xcite . the sample is randomly selected down to @xmath30 and represents the largest homogeneous sample of low - mass early - type galaxies with reliable velocity dispersions down to @xmath31kms@xmath2 . in section 2 , we present the photometric and spectroscopic data ( including velocity dispersion measurements and spectral indices ) and the catalogue of `` newly derived '' structural parameters ; the morphological classification is described in section 3 . our fp fits for the overall and high-@xmath10 ( @xmath32kms@xmath2 ) samples are presented in section 4 , which discusses how the selection criteria can affect the values of the fp coefficients and mimic a possible curvature of the plane . in section 5 we quantify the contribution of stellar population to the intrinsic scatter . we discuss our results in section 6 and give a summary in section 7 . the origin of the tilt of the fp will be investigated in a forthcoming paper . throughout this paper we use h@xmath33kms@xmath2mpc@xmath2 , @xmath34 and @xmath35 . with this cosmology 1 arcsec@xmath36kpc at @xmath0 and the distance modulus is 36.69 . the sample of galaxies used in this work consists of 141 early - type @xmath37 galaxies ( red points in figure [ map ] ) distributed throughout entire shapley optical survey ( sos , mercurio et al . @xcite ) area , but mainly located in high - density regions ( i.e. cluster cores ) . the sample selection is summarized in table 1 . in the region covered by sos there are 378 confirmed supercluster members . here we analyse the sample of 141 galaxies which have all the following characteristics : ( i ) an early - type morphology ( see section 3 ) , ( ii ) reliable surface photometry , ( iii ) a measured velocity dispersion and ( iv ) an insignificant h@xmath8 emission ( ew(h@xmath38 ) . spectra were obtained using the aaomega fibre - fed spectrograph at the anglo - australian telescope . a full description is given by ( * ? ? ? * hereafter slh ) ; we summarize some key points here . the spectroscopic sample limit is @xmath39 in the cluster cores , with galaxies selected from the noao fundamental plane survey ( nfps ; * ? ? ? * ) images ( green boxes in figure [ map ] ) . outside of these regions , brighter galaxies were selected from the 2mass extended source catalogue ( @xmath40 ) . the fibre diameter of 2arcsec corresponds to 1.9kpc . long integrations resulted in high signal - to - noise ratios , s / n@xmath1145 per for @xmath41kms@xmath2 and s / n@xmath1190 per for @xmath32kms@xmath2 . velocity dispersions were measured with respect to the best - matching simple stellar population ( ssp ) templates . the errors were estimated from monte carlo simulations , and are @xmath42dex at @xmath41kms@xmath2 and @xmath43dex for @xmath32kms@xmath2 . the spectral resolution of 3.2 ( 82kms@xmath2 instrumental dispersion ) allows the recovery of velocity dispersions as low as @xmath1140kms@xmath2 . however , some galaxies in the sample remain kinematically unresolved , i.e. have velocity dispersions consistent with zero . when comparing the new velocity dispersion measurements with those previously obtained from the nfps ( having a factor @xmath44 lower s / n ) , slh find the new velocity dispersions of low-@xmath10 objects to be systematically lower by @xmath45dex , which they attribute to a combination of higher signal - to - noise and the use of a range of ssp templates rather than individual k - giant stars . the absorption index data are tabulated by slh . here we also make use of single - burst equivalent ages , metallicities ( z / h ) and abundance ratios ( @xmath8/fe ) estimated using the models of thomas , maraston @xmath13 bender ( 2003 ) . details of this process , and the stellar population parameters for each galaxy are provided by smith , lucey @xmath13 hudson ( 2009 ) . the resulting scaling relations between the three stellar population parameters and velocity dispersion are : @xmath46 typical errors for galaxies with @xmath10 in the range 50100kms@xmath2 are 14@xmath12 in age , 0.05dex in [ z / h ] and 0.04dex in @xmath8/fe , while they reduce to half these values for galaxies with @xmath47kms@xmath2 . in this work we refer to velocity dispersions measured into an aperture of r@xmath48/8 . we have corrected our velocity dispersions ( @xmath10@xmath49 ) acquired with fibres of 1 arcsec radius ( r@xmath50 ) to the apertures of r@xmath48/8 following @xcite : @xmath51 hereafter , we adopt the notation @xmath10@xmath52 = @xmath10@xmath53 for the velocity dispersion corrected to this fiducial aperture . starting from the sample of galaxies observed by slh we select those 396 galaxies belonging spectroscopically to the shapley supercluster ( @xmath54 , blue and red points in figure [ map ] ) . for photometry we refer to the sos survey @xcite . the sos is based on data acquired with the wfi camera ( 4 @xmath55 2 mosaic of 2k @xmath55 4k ccds with a pixel scale of 0.238@xmath56 pixels ) mounted on the eso / mpi 2.2-m telescope at la silla observatory . @xmath4-band imaging was acquired in good seeing conditions ( @xmath57arcsec ) for eight contiguous fields covering a region of 2deg@xmath58 centred on the shapley supercluster core ( see figure [ map ] ) . five exposures were obtained for each field giving a total exposure time of 1200s for the mosaic images ( 240s @xmath55 5 ) . the reduction was carried out with the alambic pipeline ( version 1.0 , vandame 2004 ) and the catalogue was produced with the sextractor package @xcite , plus a set of procedures designed @xmath59 @xmath60 to remove spurious detections ( bad pixels , cosmic rays , etc ) and correct the photometry for blended sources . the survey is complete to @xmath61 ( @xmath62 ) . for more details see @xcite . sos photometry is available for 378 supercluster galaxies observed by slh , all of which are detected at signal - to - noise levels greater than 100 in each exposure , such that reliable structural parameters can be derived . structural parameters were derived using the software 2dphot described by @xcite . this is an automated tool measuring both integrated and surface photometry of galaxies and is furnished with several tasks to carry out reliable star - galaxy separation , psf modelling and estimation of catalogue completeness . the main steps of the 2dphot algorithm are : i ) creation of a clean catalogue of the input image with sextractor ; ii ) estimation of the fwhm and the definition of `` sure stars '' ; iii ) construction of an accurate psf model taking into account both possible spatial variations as well as deviation of stellar isophotes from circularity ; iv ) derivation of structural parameters ( effective radius @xmath15 , mean surface brightness @xmath63 , sersic index @xmath64 , total magnitude m@xmath65 , etc . ) by fitting galaxy images with 2d psf - convolved sersic models , as well as the measure of the fit accuracy ( @xmath66 ) . the measurement of structural parameters is strictly dependent both on the psf model and on the signal - to - noise ratio . since the mosaic sos images are obtained by stacking jittered images to cover the gap regions between the eight ccds , the signal - to - noise ratio is not constant among the images being lower in the overlapping gap area . moreover , in the gap area the psf turns out to be poorly defined due to its spatial variations . we removed the galaxies in these regions from our sample and performed the surface photometry only for galaxies in highest s / n regions ( 224 galaxies ) , where the psf is well defined . we correct our mean surface brightnesses for cosmological dimming ( @xmath67= 0.208 mag / arcsec@xmath58 ) , galactic extinction ( @xmath68= 0.147 mag , schlegel et al . 1998 ) and k - correction ( @xmath68 = 0.05 , poggianti 1997 ) and convert from magarcsec@xmath69 to log@xmath16 expressed in physical units @xmath70pc@xmath69 through @xmath71log(206265pc/10pc ) ) where m@xmath72=4.42 is the solar absolute magnitude ( binney @xmath13 merrifield 1998 ) . + + to derive the errors , we measure the structural parameters on the five single exposures ( @xmath73 , @xmath74 with @xmath75 ) , assuming that the observed spread of their values is directly related to the s / n on the final mosaic used to derive the structural parameters entering the fp . due to the size of the sample ( five exposures ) only two scale estimators can be considered reliable @xcite : the classical standard deviation and the that obtained by the gapper algorithm @xmath76 . the gapper is a robust scale indicator ( wainer & thissen , 1976 ) based on the gaps between ordered statistics which has a high level of efficiency for samples as small as five objects . if we have @xmath64 measures of a quantity x ranked in increasing order ( x@xmath77,x@xmath27, ... x@xmath78,x@xmath79 ) , according to the gapper algorithm we can measure a robust scale indicator as : @xmath80 where @xmath81 to avoid an overestimation of the errors due to the presence of outliers , we first compute the @xmath76 of the @xmath82 and @xmath83 distributions and reject all values that deviate more than 3@xmath76 from the median value , before subsequently computing the classical standard deviation as well as the covariance matrices of the clipped sample for both variables ( @xmath84 , @xmath85 , cov[@xmath86 ) . the errors on the mosaic value of @xmath87 and @xmath88 ( @xmath89 and @xmath90 respectively ) are given by @xmath91 and @xmath92 , where @xmath64 is the number of measures available . the typical errors on the logarithms of effective radius ( @xmath87 ) and mean surface brightness ( @xmath88 ) are 0.03 and 0.04 , respectively . these errors explicitly include the effects of noise in the galaxy , but not the presence of neighbouring objects . we have estimated the effect of the latter , by repeatedly placing copies of the galaxies one - by - one at random positions across the same ccd image ( where the psf should remain constant ) and reapplying 2dphot , finding the variations in the structural parameters to be consistent with the previously obtained errors , albeit with a small number ( @xmath93% ) of @xmath94 outliers when the galaxy is placed very close to a bright star or galaxy . we have morphologically classified the 224 galaxies with available surface photometry , by inspection of the residual maps provided by 2dphot . we denote as late - type " all galaxies showing signs of spiral arms or asymmetric disturbance , and as early - type " all those with no such structures . the resolution of sos images does not allow any finer classification , for example into elliptical and lenticular galaxies , since the presence of a residual disk can be seen only in particular cases , i.e. when it is very bright , widespread or edge - on . in figure [ morpho ] we show some illustrative examples of galaxies classified as early- and late - types , covering the full range of magnitudes studied here . we have checked this classification by comparing our results with those of @xcite for a subsample of 54 galaxies belonging to a3558 and a3562 , finding perfect agreement . in their classification galaxies were subdivided into e , s0 , se and sl classes ( the latter being early and late spirals ) . our early - type sample contain only galaxies classified as e and s0 by @xcite . the strong correlation between the sersic index @xmath64 and luminosity @xcite prevents any morphological classification based on sersic index alone . @xcite show that the values of sersic index @xmath64 of de , ordinary e / s0 galaxies and bcgs follow a @xmath95 trend from @xmath96 to @xmath97 , whereby brighter galaxies have larger values of @xmath64 . in figure [ n_dist ] we demonstrate the equivalent trend for our galaxy sample , + plotting the sersic index distribution as function of morphological type and magnitude . the median value of @xmath64 for early - type ( filled circles ) and late - type ( open circles ) galaxies are shown for each magnitude bin ( chosen to contain the same number of galaxies ) , while the error bars " indicate the inter - quartile values . as pointed out by @xcite the distribution of @xmath64 for early - type galaxies is shifted towards higher values with increasing luminosity . late - type galaxies have lower values of @xmath64 at all magnitudes , but there is not a clean _ a priori _ separation between the two morphological classes . the morphological classification results in a sample of 141 early - type galaxies and 44 late - type or irregular galaxies , while 31 galaxies are removed from the sample since they are not kinematically resolved , and 8 are removed as having significant h@xmath8 emission ( ew(h@xmath98 ) . we summarize the characteristics of the sample in table [ sample ] , while in table [ sample2 ] we report structural and kinematical parameters with the corresponding errors for the 141 galaxies used in the following section . c c sample selection + galaxies with spectra & 565 + cluster members ( @xmath99 ) & 396 + cluster members in sos region & 378 + cluster members in sos region with surface photometry & 224 + galaxy classification + early - type & 141 + late - type & 44 + unresolved kinematically & 31 + with ew(h@xmath98@xmath100 & 8 + [ sample2 ] lccccccccccc i d & ra(j2000 ) & dec(j2000 ) & r@xmath65 & @xmath18 & @xmath101 & @xmath87 & @xmath102 & @xmath103 & @xmath104 & @xmath64 & @xmath105 + & & & & ( km / s ) & & ( kpc ) & & @xmath106pc@xmath69 & & & + tmassj13274662 - 3059237 & 13:27:46.6 & -30:59:24 & 15.166 & 2.201 & 0.010 & 0.904 & 0.042 & 1.766 & 0.070 & 10.708 & 1.25 + nfpj132828.6 - 313205 & 13:28:28.6 & -31:32:04 & 15.404 & 2.112 & 0.012 & 0.308 & 0.002 & 2.861 & 0.004 & 1.589 & 6.43 + nfpj132738.0 - 313041 & 13:27:38.0 & -31:30:41 & 16.549 & 2.044 & 0.016 & 0.038 & 0.004 & 2.943 & 0.008 & 2.283 & 1.31 + nfpj133408.2 - 314735 & 13:34:08.1 & -31:47:34 & 16.083 & 1.757 & 0.097 & 0.682 & 0.044 & 1.842 & 0.059 & 3.518 & 2.03 + tmassj13295423 - 3201001 & 13:29:54.2 & -32:00:59 & 15.229 & 2.153 & 0.010 & 0.475 & 0.005 & 2.597 & 0.007 & 4.617 & 1.17 + nfpj132656.0 - 312528 & 13:26:56.0 & -31:25:27 & 14.813 & 2.031 & 0.007 & 0.797 & 0.007 & 2.118 & 0.011 & 5.452 & 1.99 + tmassj13294834 - 3115580 & 13:29:48.3 & -31:15:58 & 14.075 & 2.296 & 0.004 & 0.984 & 0.023 & 2.043 & 0.035 & 6.945 & 1.15 + nfpj132810.5 - 312310 & 13:28:10.5 & -31:23:09 & 13.868 & 2.267 & 0.004 & 1.108 & 0.043 & 1.875 & 0.067 & 7.796 & 1.48 + nfpj132426.5 - 315153 & 13:24:26.5 & -31:51:53 & 14.857 & 2.302 & 0.006 & 0.506 & 0.036 & 2.685 & 0.051 & 8.414 & 1.38 + nfpj132923.8 - 314832 & 13:29:23.2 & -31:48:32 & 15.060 & 1.718 & 0.062 & 0.972 & 0.046 & 1.673 & 0.073 & 9.545 & 1.77 + for the fp we use the representation of equation [ formula ] where @xmath15 is the effective radius measured in kpc , @xmath5 is the central velocity dispersion corrected to an aperture of radius r@xmath108/8 in kms@xmath2 ( see section 2.1 ) and @xmath16 is the mean surface brightness within @xmath15 expressed in @xmath106pc@xmath69 . to derive the value of @xmath8 , @xmath26 and @xmath109 we adopted the orthogonal fit which minimizes the quantity : @xmath110 i.e. the sum of the absolute residuals perpendicular to the plane . this is less sensitive to outliers than the classic least - squares method ( * ? ? ? * hereafter jfk96 ) . for the full sample of 141 galaxies we found : @xmath111 where the errors on coefficient are computed via a bootstrap procedure . the above relation presents a scatter of 0.088dex in the log r@xmath48 direction . we estimate the intrinsic contribution to the scatter by subtracting in quadrature from the rms scatter , the observational errors on each variable taking into account the correlation between the errors on the effective radius and surface brightness . we obtain an intrinsic scatter @xmath112 = 0.068 . when deriving the @xmath113 the errors on distance estimated through the standard deviation of the redshift distribution are also considered . to take into account the different uncertainties on @xmath5 , @xmath15 , @xmath16 along the plane , a weighted fit is adopted . following @xcite we add in quadrature to the measurement errors the intrinsic scatter orthogonal to the fp , @xmath114 . the fitting function to be minimized is : @xmath115 where the sum is extended over all sample galaxies and where @xmath116.\ ] ] the intrinsic scatter , @xmath114 , is adjusted to give the expected value of 0.8 per degree of freedom ( the mean absolute value of a standardised gaussian distribution is 0.8 ) . the resulting equation for the fp is : @xmath117 + + the orthogonal and weighted fits are consistent , demonstrating that the results are robust to the effects of the larger @xmath5 uncertainties ( 0.01 dex for @xmath118 km s@xmath2 and 0.05 for @xmath119 km s@xmath2 ) for the lowest mass galaxies . the scatter in the log r@xmath48 direction is now equal to 0.092 , while its intrinsic component is 0.070 . figure [ fp_tot ] shows the two edge - on fp views ( left and central panels ) and its trend along the direction of luminosity ( right panel ) . the value of the @xmath8 coefficient obtained for the shapley sample , is significantly lower than the typical values of 1.21.3 reported for samples dominated by giant galaxies ( e.g. jfk96 ) . if we restrict our sample to galaxies with @xmath120 100kms@xmath2 ( high-@xmath5 sample , 91 galaxies ) we obtain the fp relation : @xmath121 with the overall scatter in @xmath87 reduced to 0.090 . in figure [ fp_high ] we plot the fp as fitted for the high-@xmath5 sample ( black points ) . galaxies with @xmath122 kms@xmath2 are also reported for comparison ( red dots ) . the @xmath8 value of the high-@xmath5 sample is closer to those of other authors . the lower @xmath8 value found for the total sample is thus probably due to the extension of our sample to very low mass galaxies , down to @xmath12350kms@xmath2 . the improved method used to obtain the velocity dispersions , as well as the high signal - to - noise levels of the spectra , can produce systematic effects on the resultant values of @xmath5 , particularly for low-@xmath5 objects . slh find that velocity dispersions obtained using single , old , solar - metallicity models instead of templates with different metallicities ( as used in this paper ) are overestimated by @xmath124% for @xmath125kms@xmath2 galaxies and by @xmath126% for @xmath127kms@xmath2 . as a result , in our sample the low-@xmath5 values are systematically lower than those of previous samples ( e.g. nfps sample ) , causing an increase of the fp tilt of about 1015@xmath12 . the impact of the low-@xmath5 limits is explored further in section 4.2 . we notice that results presented in section 5 and discussed in section 6 has been checked to be independent of @xmath5 cuts . besides the range of velocity dispersions , the determination of the three fp coefficients is strictly dependent on other factors such as the selection criteria of the sample , the fit algorithm and the procedure used to derive structural and kinematical parameters @xcite . we address this problem comparing our fp with the previous work by jfk96 . their sample consists of 81 early - type galaxies down to gunn @xmath128 in the central region of coma cluster . for this section , we have corrected our velocity dispersions to their standard fixed aperture of 0.595@xmath129kpc of radius . to analyse the consistency of the two samples , in figure [ fc_coma ] + we plot the spectroscopic completeness functions ( cfs ) of the coma ( open circles ) and shapley ( filled circles ) fp samples as a function of magnitude . the magnitudes of the shapley galaxies are converted from @xmath4 to the gunn-@xmath130 photometric system according to the typical colour of @xmath131 for elliptical galaxies @xcite . the coma cf is computed using the jfk96 fp sample and the complete photometric catalogue of early - type coma galaxies published by @xcite . the limiting magnitude of the coma sample is m@xmath132 ( dashed line ) . the coma sample is complete at bright magnitudes , but the completeness declines rapidly towards zero at the faint magnitude limit . in contrast , about 30% of shapley galaxies are spectroscopically observed and have available velocity dispersion measurements independent of magnitude down to m@xmath132 . in the magnitude bins fainter than m@xmath133 , the shapley cf declines gently due to an increasing fraction of galaxies with surface brightnesses and velocity dispersions too low for successful @xmath5 measurements , before dropping rapidly in the faintest bin . green triangles represent the cf of shapley galaxies observed spectroscopically with no regards to successful velocity dispersion measurements : the decline at faintest magnitude has now disappeared except for the faintest bin ( m@xmath134 ) . the distribution reflects well the random selection criteria of our spectroscopic survey for @xmath37 galaxies . to compare the fps of the two samples we have selected from shapley only early - type galaxies with m@xmath135 and @xmath136kms@xmath2 ( hereafter `` matched '' sample ) corresponding roughly to the limits of the coma fp sample . this new sample consists of 88 galaxies . the fp of the `` matched '' sample is : @xmath137 with a scatter equal to 0.08dex in log r@xmath48 direction . in figure [ fpcoma ] and @xmath136kms@xmath2 ( shapley `` matched '' sample , black points ) . the solid line represents the shapley fp fit . for comparison are reported also coma galaxies from jfk96 ( open points).,title="fig:",width=264 ] + we plot the edge - on view of the fp for the `` matched '' sample ( filled circles ) . coma galaxies ( open circles ) are reported for comparison . since in jfk96 the value of h@xmath138 is set equal to 50kms@xmath2mpc@xmath2 , the effective radii of coma galaxies have been shifted by a factor 0.146dex . both the distributions and the dispersions of the two samples are consistent , in fact the fp for coma sample , as found by jfk96 is : @xmath139 the possible reasons for the remaining slight differences between the two fps could be the different sampling at the brightest magnitudes of the two datasets , and the different procedures used to obtain the structural parameters , since jfk96 fit their early - type galaxies with de vaucouleurs profiles . observing the high-@xmath5 fit ( see figure [ fp_high ] ) it is notable that all galaxies with velocity dispersions less than 100kms@xmath2 ( low - mass galaxies ) are systematically displaced above ( left panel ) or below ( central panel ) the best - fit plane as defined by the high-@xmath5 sample . in particular , the log@xmath5 edge - on view suggests that low - mass galaxies do not follow the same relation as massive galaxies . thus , at face value , the fp appears curved , as suggested also by other studies ( jfk96 , zaritsky et al . @xcite , donofrio 2008 , nigoche - nietro 2008 ) investigating the curvature of the fp as a function of mass and/or luminosity . studying a sample of 69 faint early - type galaxies in the core of coma cluster , @xcite find a similar behaviour for the @xmath140 relation , with dwarf galaxies having a trend ( @xmath141 ) shallower than that observed for elliptical giant systems ( @xmath142 ) . in figure [ coeff_alpha ] , we investigate this possible curvature by analysing the dependence of @xmath8 as a function of different low-@xmath143 limits . + we cut the galaxy sample at various values of @xmath5 and estimate the fp coefficients through the weighted orthogonal fit procedure . one can see that @xmath8 increases systematically as the @xmath5 cut moves to higher values , removing more galaxies from the sample . the values of @xmath8 for the whole sample ( @[email protected] ) and for the highest-@xmath5 cuts ( @[email protected] ) differ by more than twice the standard errors , indicating a possible curvature of fp . to investigate if these observed variations of @xmath8 with the sample selection are due to a real curvature of the fp , or induced by selection effects , a set of simulations were performed . we constructed 1000 mock catalogues using the observed values of log@xmath16 and log@xmath5 and the value of log@xmath15 assigned using the relation : @xmath145 where , in this case , @xmath8 = 0.92 , @xmath26 = 0.78 , and @xmath109 = 0.496 are the coefficients obtained by fitting the fp of the overall shapley sample with log @xmath15 as the dependent variable , @xmath146 = 0.099 is the observed scatter in the log @xmath15 direction , and n(0,@xmath146 ) is a gaussian random variable with zero mean and standard deviation of @xmath146 . for each mock catalogue we evaluate the relative change in @xmath8 ( @xmath147 ) between the orthogonal fits obtained for the whole sample and after applying a cut at @xmath148kms@xmath2 . the resulting histogram of @xmath147 values is shown in figure [ sim ] , + where the dashed line indicates the @xmath147 computed directly for the shapley sample . this shows that the observed change in @xmath8 is fully consistent with that expected for a linear relation . moreover , in figure [ conf_sim ] + we compare the fp edge - on projection ( eq . 11 ) for the observed high-@xmath5 sample ( solid symbols ) and a simulated sample obeying the overall shapley fp relation ( open symbols ) . the two samples show similar behaviour , with the same apparent curvature such that galaxies with low velocity dispersions are placed systematically below the fp relation . given that the simulated sample is defined to follow a linear relation , this suggests that the apparent curvature in figure [ fp_high ] can be explained solely by selection biases . hence , in spite of the wide - range in velocity dispersions covered by our sample , this is not sufficient to distinguish between either a linear or curved fp relation , but there is no convincing evidence for the latter in our data . we find that for the shapley fp , the contribution of the intrinsic scatter ( 0.070 ) to the total scatter ( 0.092 ) is larger than that of the measurement errors ( 0.060 ) . taking advantage of the wide variety of indices derived by slh we investigate the possible origin of the intrinsic scatter of the fp analysing the correlations between both spectral indices ( mgb , fe5015 , fe4383 , h@xmath26 , hgf , hdf ) and stellar population parameters ( age , metallicity , @xmath8-enhancement ) and the residuals from the fp . all of the many spectral indices are known to correlate with @xmath5 , which , if not corrected for , could produce spurious correlations with the fp residuals . to avoid this problem , rather than use the spectral indices themselves , we consider their residuals with respect to the _ index_-log @xmath5 relation . for each index , we fitted first the _ index_-log @xmath5 relation , shown in figures [ corrindfe ] and [ corrindba ] ( bottom panels ) for metallicity- and age- sensitive indices respectively , and then determine the residuals with respect to that relation ( middle panels ) . in the figures , the upper panels show the _ index_-log @xmath5 residuals versus the residuals from the fp in the log r@xmath48 direction . for each relation , we quantify the product - moment correlations between the residuals and spectral index through the correlation factor @xmath130 and the bisector least - squares fit , assuming that the distributions are both gaussian . uncertainties in @xmath130 and the fits are estimated through 10000 monte carlo realizations , taking into account the errors on each value . these values together with the probability @xmath149 that two quantities with correlation factor @xmath130 are not correlated are reported in figures [ corrindfe ] and [ corrindba ] . + + among the primarily metal - sensitive indices ( figure 11 ) , the strongest dependence is found for fe4383 , which is correlated with fp residuals at the 4@xmath10 level . a similar but weaker correlation is obtained for fe5015 . for mgb , however , a weak correlation is obtained in the opposite sense . thus galaxies which are more compact than expected from the fp have lower fe4383 indices than expected for their @xmath5 , but _ higher _ mgb indices than expected for their @xmath5 . this already provides a hint that the physical driver for the residual trends is not the total metallicity , but instead the ratio between mg and fe abundances . the balmer lines show weak positive correlations , such that galaxies with strong index values ( for their @xmath5 ) are more diffuse than predicted by the fp . this could be caused either by age or by metallicity effects , since the balmer lines are sensitive to both parameters to some extent . using a single spectral index it is not possible to disentangle between the effects of age and metallicity . in this section , we use estimates of age , metallicity and @xmath8-enhancement ( @xmath8/fe ) derived from the index measurements by means of a multi - index procedure , to provide more physically meaningful information ( smith et al . 2009 ) . as for the single spectral indices , the three stellar population parameters age , metallicity and @xmath8/fe correlate strongly with @xmath5 as shown by the scaling relations of equation [ eqage ] , therefore we first compute the residuals of stellar population parameters with respect to these relations . the lower panels of figure [ correlations ] show the stellar population residuals against @xmath150 characterized , by construction , by the absence of correlations between the residuals and log @xmath5 itself . the top panels show the resultant correlations between the age , metallicity and @xmath8/fe residuals ( left , middle and right panel respectively ) and the fp residuals along the log @xmath15 projections . + both age and @xmath8/fe are seen to be strongly anti - correlated ( at @xmath93@xmath10 and @xmath94@xmath10 respectively ) with the residuals computed in the @xmath87 direction , while metallicity shows a _ positive _ correlation with the fp residuals along the same axis . we find these correlations to be independent of the range of @xmath10 covered , for example obtaining consistent results when using only the high-@xmath5 sample . in table [ percentuale ] we summarize the probability @xmath149 for all the cases . we note that in the figures 11 , 12 and 13 , the velocity dispersion enters into both axes , since each quantity is a residual from a relation involving @xmath5 . the errors are hence correlated which will produce a bias towards positive correlations . for each of the primary ssp parameters we measure the level of this bias through monte carlo simulations , based on fake data in which there are no intrinsic correlations . firstly , for each galaxy we assign fake ssp values ( age , z / h , @xmath8/fe ) following the previously derived index-@xmath5 relations , and having the same rms intrinsic scatter in the ssp value about the relation . we then perturb each of the values log @xmath5 , log @xmath15 , and log @xmath151 by their uncertainties for each galaxy , and recalculate the residuals about the fp and the index-@xmath5 relation as before . through these simulations , we measure this bias ( @xmath152 r ) , i.e. the difference in the correlation coefficient , due to the correlated measurement errors in @xmath5 and finds : @xmath153 we see that this bias contributes to the observed correlation between the metallicty and fp residuals , but acts to oppose the anti - correlations seen for age and @xmath8-enhancement . in each case the effect of the bias is small , being at a level of @xmath1110% of the observed correlation . the relation between residuals obtained taking into account this bias is shown in figure 13 by the red dot - dashed line . it should be remembered that the stellar population parameters presented here are derived through fixed apertures covering only the galaxy centres . therefore , if galaxies have significant population gradients , the observed trend could be due to sampling the central stellar populations at different radii . in the case of metallicity , this may be a factor , as early - type galaxies are observed to have negative metallicity gradients ( kuntschner et al . 2006 ; rawle et al . 2008 ) of the order @xmath154dex in [ z / h ] . however , spatially - resolved spectroscopy of early - type galaxies show that they generally have flat radial trends in age and @xmath8/fe ( snchez - blzquez et al . 2007 ; rawle et al . 2008 ) , indicating that the anti - correlation of age and abundance ratios with the fp residuals in the log @xmath15 and log @xmath16 directions is robust . the strong correlations found between the stellar population parameters , principally the @xmath8/fe , and the residuals of the fp suggest that its scatter is in part due to variations in the stellar populations at fixed galaxy @xmath143 . to this aim we fit the modified fp relation : @xmath155 where the stellar population parameter spp@xmath156 is , in turn , @xmath8 enhancement , age and metallicity of the galaxies , and study the variation induced in the scatter by keeping @xmath8 and @xmath26 fixed as in eq . [ simulazione ] and allowing @xmath157 to vary to minimize the fp scatter . it is necessary to keep @xmath8 and @xmath26 fixed as each of the stellar population parameters correlate strongly with @xmath5 , and hence would introduce spurious variations in @xmath8 . the results obtained are listed in table [ fit_4dim ] , where the second column indicates the estimated strength of the spp term @xmath157 , the following two columns indicate the overall scatter in the log @xmath15 direction after the addition of the spp term , and the relative intrinsic scatter , and the final column the rms contribution of the ssp term to the scatter in log @xmath15 . [ cols="^,^,^,^,^ " , ] the fit with the most significant @xmath157 coefficient ( @xmath158 ) is the fit with the @xmath8/fe as fourth parameter . in the case of age and metallicity , the new coefficient is non - zero at just the @xmath113@xmath10 level , indicating the weaker impact of these parameters on the fundamental relation . while it seems that none of the additional spp terms reduce the overall scatter significantly , the relative impact of the three spp terms becomes clearer when considering the intrinsic scatter ( i.e. after accounting for the measurement uncertainties ) . both the relation including the effect of age and metallicity have an intrinsic scatter of @xmath159 , only marginally lower than the value without the spp term , while in the case of @xmath8/fe the intrinsic scatter is reduced to 0.049 . this is comparable to the rms contribution from the @xmath8/fe term ( 0.047 ) , i.e. @xmath8/fe contributes around half of the intrinsic scatter , and indicates that the distribution of galaxies around the fp are tightly related to the enrichment , and hence to the timescale of star - formation . the @xmath8/fe dependence of the fp scatter is illustrated further in figure [ split ] which demonstrates also the reduction in total scatter despite the introduction of measurement errors in @xmath8/fe . we split the shapley sample into three groups ( shown as red , grey and blue points in figure [ split ] ) according to the position of each galaxy with respect to the @xmath8/fe - log@xmath5 relation ( see left panel ) : the 50@xmath12 of galaxies closest to the relation are plotted in grey , while the 25@xmath12 with @xmath8/fe higher ( lower ) than expected for their @xmath10 are plotted in red ( blue ) . a modest reduction in total scatter is seen , despite the introduction of uncertainty due to @xmath8/fe measurement errors . + furthermore the total fp scatter is reduced by including the extra terms , in spite of the fact that the ssp parameters are subject to their own substantial uncertainties . this implies there must be a still greater reduction in the _ intrinsic _ fp scatter . jfk96 pointed out that the dispersion around the fp relation is not completely due to the measurement errors but has an intrinsic scatter whose nature is not yet understood . the existence of this intrinsic scatter was interpreted as due to another `` fundamental '' parameter characterising the family of early - type galaxies . in many of the previous works on the fp ( see section 1 ) , a strong effort has been made to find correlations between the fp residuals and different line - indices considered to be representative of a particular stellar population parameter ( for example h@xmath26 for age , mg@xmath160 for metallicity etc ) . in section 5 we showed that the fp residuals are strongly correlated with estimated stellar population parameters especially with the @xmath8-abundance ratio ( @xmath8/fe ) . given that age and @xmath8/fe are known to correlate strongly for galaxies of a given mass , while age and metallicity anti - correlate ( e.g. proctor & sansom 2002 ; smith et al . 2007b ) , it is not clear which of the three parameters drives the correlations . however , the fact that the strongest correlation is observed for the @xmath8/fe suggests that this parameter is playing the major role , while age and metallicity trends just reflect their mutual correlations . the trends observed for the metallicity indices also suggest that the @xmath8/fe is driving the correlations . in fact , the positive correlations between the z / h dependent indices fe4383 and fe5015 combined with the negative correlation for mgb , is consistent with an anti - correlation for the abundance ratio @xmath8/fe . moreover , we found that about half of the fp intrinsic variance is due to variations in the stellar populations as described by the @xmath8-enhancement trend , while the other stellar population parameters seem to contribute rather less to the intrinsic scatter . it has long been realised ( tinsley 1980 ) that in the study of galaxy formation and evolution , a major role can be played by the analysis of the abundance ratios , due to them being relatively model independent and primarily affected by stellar nucleosynthesis and the initial mass function ( matteucci 1996 ) . in particular , the ratio between the so - called @xmath8 elements ( synthesized in type ii supernovae ) and iron ( mainly from delayed type ia events ) is widely accepted to be affected by the characteristic timescale of star - formation , the @xmath8 elements and iron having different production timescales ( greggio & renzini 1983 ) . according to the simplest and widely accepted scenario , a galaxy with a high value of @xmath8/fe has experienced many type ii sne events , but almost no sne of type ia during the major epoch of star - formation : this constrains the timescale for this epoch of stellar formation to be shorter than 3@xmath5510@xmath161 years . our results suggest that the galaxies which formed their stars over a shorter duration ( high @xmath8/fe ) are also those which are more compact . such a pattern may be expected in a dissipational merger inducing a nuclear sturburst . vazdekis , trujillo and yamada ( 2004 ) , studing 21 early - type galaxies already claimed the existence of a correlation between sersic index and mg / fe ratio ( i.e. between galaxy structure and stellar population ) , even if they did not correct for correlations with @xmath10 . they interpreted their results as more massive galaxies having their star formation quenched on shorter time - scales . although , in agreement with our finding , their analysis is based on to a smaller sample spanning a narrower range of velocity disperisons and luminosities with compared to our shapley sample making any comparisons difficult . @xcite make specific predictions for the effects of varying the dissipational fraction on the remnant spheroids of a given mass , resulting in correlations between the structural and stellar population parameters . by considering identical progenitor disks ( at @xmath162 ) with initial gas fractions @xmath163 following an exponential star - formation history with time - scale @xmath164 , we have that the gas fraction at the time of the merger @xmath165 ( and hence dissipational fraction in the merger - induced starburst ) will scale as @xmath166 . if the remaining gas is then consumed in the central star - burst , producing a passively - evolving spheroid remnant , then the dissipational fraction will increase for earlier mergers , and hence mean stellar age . @xcite also show that ellipticals formed through mergers with higher dissipation fractions should be more @xmath8-enriched . as merger remnants involving more dissipation , like those between more gas - rich disks , are expected to yield larger mass fractions formed in nuclear starbursts , which reduce significantly the effective radii of the remnant , we should expect both mean stellar age and @xmath8/fe to anti - correlate with the residuals in the @xmath15 direction . we observe clear correlations between the fp residuals and age and @xmath8/fe , with those galaxies of a given mass with effective radii smaller than predicted by the fp to have stellar populations systematically older and with higher abundances than average , fully consistent with the predictions of @xcite . we have derived the fp of a sample of 141 early - type @xmath37 galaxies in the shapley supercluster at @xmath0 . velocity dispersions and stellar population parameters were derived from the spectroscopic data of smith et al . ( 2007 ) , while r - band photometry is from the shapley optical survey @xcite . the final sample extends down to m@xmath167 + 3 in magnitude and 50 km s@xmath2 in @xmath5 . using the software 2dphot ( la barbera et al . 2008b ) we derived for each galaxy the structural parameters @xmath15 , @xmath63 , @xmath64 , and @xmath168 by fitting a 2d psf - convolved sersic model . the morphological classification was performed by eye and checked with those of @xcite for a subsample of 54 galaxies finding complete agreement . adopting a weighted fit ( see eq . [ fp_weighted ] ) we derived the fp : @xmath15@xmath169@xmath170 for the 141 early - type r@xmath10718 galaxies . the low value of @xmath8 can be related to the extension of our sample towards very low values of velocity dispersion and and also to the method used to determine @xmath10 . considering the subsample of @xmath6 km s@xmath2 galaxies the fp turns out to follow the relation @xmath171@xmath169@xmath172 which is consistent with the coma fp obtained by jfk96 . observing the significant change in the @xmath8 value between the total and the high-@xmath5 samples , we investigate the possible curvature of the fp . applying the same cuts to a simulated sample following a linear fp relation , we ascertain that the observed curvature can be explained by selection effects . departure from a linear trend of the fp relation has been claimed by donofrio et al . ( 2008 ) , and desroches et al . ( 2001 ) found that the faint - end luminosity cut influences the coefficient @xmath8 of the fp , in agreement with nigoche - netro ( 2008 ) . similar results were found by hyde @xmath13 bernardi ( 2008 ) studying the fp residuals along the plane . however , these works pointed out that conclusive evidence of the fp curvature needs either more robust statistics or higher accuracy in the velocity dispersion estimates for low mass systems . the most important result of this work is our demonstration that the fp residuals are correlated with stellar population characteristics ( line- strength indices and derived ssp parameters ) . in particular , fp residuals are anti - correlated both with the @xmath8-element abundance ratio , @xmath8/fe , and with galaxy age resulting in trends whereby galaxies more compact than expected from the fp relation have stellar pupulations systematically older and with higher abundances than average . previous studies have reported correlations of the fp residuals with stellar age ( e.g. forbes et al . 1998 ; reda et al . although our fp residual do show a correlation with age , we recover a much stronger signal for @xmath8/fe than for age , suggesting that this is the more fundamental dependence . indeed , a multiple regression analysis suggests there is no age correlation at fixed @xmath8/fe , but a very strong @xmath8/fe correlation at fixed age . the correlation between @xmath8/fe and fp residuals does not indicate a direct _ causal _ effect , since varying @xmath8/fe at fixed age and z / h has little effect on the stellar mass - to - light . instead , the correlation suggests that the structural properties and the star - formation history are both dependent on some unobserved aspect of the galaxy assembly process . at face value , our results are broadly consistent with recent galaxy merger simulations , which predict a sequence of formation mechanisms governed by the varying importance of dissipation ( hopkins et al . 2008 ) . in this scenario , mergers with a higher initial gas - fraction trigger more centrally - concentrated starbursts , and higher @xmath8 abundances due to the short duration of star - formation in the burst . ag gratefully acknowledges the hospitality of the university of birmingham during her stays there , where some of the work was performed . cph acknowledges financial support from stfc . rjs is supported by stfc rolling grant pp / c501568/1 extragalactic astronomy and cosmology at durham 20052010 " . 99 beers t. c. , flynn k. , gebhardt k. , 1990 , aj , 100 , 32 bender r. , burstein d. , faber s. m. , 1992 , apj , 399 , 462 bernardi m. , et al . , 2003 , aj , 125 , 1866 bertin e. , arnouts s. , 1996 , a&a , 117 , 393 bertin g. , ciotti l. , del principe m. , 2002 , a&a , 386 , 149 binney & merrifield , 1998 , in _ galactic astronomy _ , princeton univeristy press caon , n. , capaccioli m. , donofrio m. , 1993 , mnras , 265 , 1013 cappellari m. , et al . , 2006 , mnras , 366 , 1126 djorgovski s. , davis m. , 1987 , apj , 313 , 59 donofrio et al . , 2008 , arxiv:0804.1892 dressler et al . , 1987 , apj , 313 , 42 faber s. m. @xmath13 jackson r. e. , 1976 , apj , 204 , 668 forbes et al . , 1998 , apj , 508 , 43 fukugita m. , shimasaku k. , ichikawa t. a. w. , 1995 , pasp , 107 , 945f gerhard o. , kronawitter a. , saglia r. p. , bender r. 2001 , aj , 121 , 1936 graham a. w. , guzmn r. , 2003 , aj , 125 , 2936 gregg m. , 1992 , apj , 384 , 43 greggio l @xmath13 renzini a. , 1983 , a.a , 118 , 217 . guzman r. @xmath13 lucey j. r. , 1993 , mnras , 263 , l47 . guzman r. , lucey j. r. , bower r. g. 1993 , mnras , 265 , 731 . haines c. p. , merluzzi p. , mercurio a. , gargiulo a. , krusanova n. , busarello g. , la barbera f. , capaccioli m. , 2006 , mnras , 371 , 55 haines c. p. , gargiulo a. , la barbera f. , mercurio a. , merluzzi p. , busarello g. , 2007 , mnras , 381 , 7 hopkins p. f. , cox t. j. , hernquist l. , 2008 , apj , 689 , 17 jrgensen i. , franx m. , 1994 , apj , 433 , 553 jrgensen i. , franx m. , kjgaard p. , 1995 , mnras , 276 , 1341 jrgensen i. , franx m. , kjgaard p. , 1996 , mnras , 280 , 167 ( jfk96 ) kelson d. d. , illingworth g. d. , van dokkum p. g. , franx m. , 2000 , apj , 531 , 184 kormendy j. , 1976 , apj , 218 , 333 kroupa p. , 2001 , mnras , 322 , 231 kuntschner et al . , 2006 , mnras , 369 , 497 la barbera f. , busarello g. , merluzzi p. , de la rosa i. , coppola g. , haines c. p. , 2008a , apj , 689 , 913 la barbera f. , de carvalho r. r. , kohl - moreira j. l. , gal r. r. , soares - santos m. , capaccioli m. , santos r. , santanna n. , 2008b , pasp , 120 , 681 maraston c. , 2005 , mnras , 362 , 799 matkovi a. guzmn r. , 2005 , mnras , 362 , 289 matteucci f. , 1996 , _ fundamentals of cosmic physics _ , vol . 283 - 396 mercurio a. , merluzzi p. , haines c. p. , gargiulo a. , krusanova n. , busarello g. , la barbera f. , capaccioli m. , covone g. , 2006 , mnras , 368 , 109 nelan j. e. , smith r. j. , hudson m. j. , wegner g. a. , lucey j. r. , moore s. a. w. , quinney s. j. , suntzeff n. b. , 2005 , apj , 632 , 137 nigoche - netro a. et al . , 2008 , preprint , asrto - ph 08051142 pahre m. a. , djorgovski s. g. , de carvalho r. r. , 1999 , asp conference series , vol . 163 , 17 poggianti b. m. , 1997 , a@xmath13as , 122 , 399 proctor r. n. & sansom a. e. , 2002 , mnras , 333 , 517 prugniel p. & simien f. , 2002 , a@xmath13a , 309 , 749 rawle t. d. , smith r. j. , lucey j. r. , swinbank a. , 2008 , mnras , 389 , 1891 reda f. et al . , 2005 , mnras , 360 , 693 robertson b. , cox t. j. , henquist l. , franx m. , hopkins p. f. , martini p. , springel v. , 2006 , apj , 641 , 21 snchez - blzquez p. , et al . , 2006 , mnras , 371 , 703 snchez - blzquez p. , forbes d. a. ; strader j. , brodie j. , proctor r. , 2007 , mnras , 377 , 759 schlegel d. j. et al . , 1998 , apj , 500 , 525 smith r. j. , et al . , 2004 , aj , 128 , 1558 smith r. j. , lucey j. r. , hudson m. j. , 2007 , mnras , 381 , 1035 ( slh ) smith r. j. , lucey j. r. , hudson m. j. 2009 , mnras , submitted steinmetz m. @xmath13 navarro j. f. , 2002 , newa , 7 , 155 thomas d. , maraston c. , bender r. , 2003 , mnras , 339 , 897 thomas d. , maraston c. , korn a. , 2004 , mnras , 351 , l19 thomas j. , saglia r. p. , bender r. , thomas d. , gebhardt k. , magorrian j. , corsini e. m. , wegner g. , 2005 , mnras , 360 , 1355 thomas t. , katgert p. , 2006 , a&a , 446 , 31 tinsley b. m. , 1980 , _ fund cosmic phys . _ , 5 , 287 toomre a. , toomre j. , 1972 , apj , 178 , 623 toomre a. , 1977 , in evolution of galaxies and stellar populations , ed . b. m. tinsley & r. b. larson , 401 trujillo i. , burkert a. , bell e. f. , 2004 , apjl , 600 , 39 vandame b. , 2004 , phd thesis vazdekis a. , trujillo i. , yamada y. 2004 , apj , 601l young c. k. , currie m. j. , 1994 , mnras , 268 , l11 wainer h. & thissen d. , 1976 , psychometrika , 41 , 9 wuyts s. et al . , 2004 , apj , 605 , 677 zaritsky d. , gonzalez a. h. , zabludoff a. i. , 2006 , apj , 638 , 725 |
we present a catalog and imaging atlas of classical ( collisional ) ring galaxies distilled from the arp - madore _ atlas of southern peculiar galaxies and associations _ and supplemented with other known ring galaxies from the published literature . the catalog lists the original host object , compiles available redshifts and presents newly determined positions for the central ( target ) galaxy and its nearest companion(s ) . 127 collisional ring systems are illustrated and their components identified . all of the rings have plausible colliders identified ; many are radial - velocity confirmed companions . finally , we make note of the existence of a rare sub - class of ring galaxies exemplified by am 2136 - 492 , double / concentric ring galaxies . these objects are predicted by numerical simulations , but they appear to be quite rare and/or short - lived in nature . | 0902.4629 | arp & madore ( 1987 , hereafter am87 ) undertook a fairly comprehensive survey of the southern sky in search of peculiar and interacting galaxies visible on the iiia - j photographic survey plates as part of the uk schmidt sky survey . they inspected some 100,000 galaxies and culled out about 7,000 objects that were noteworthy enough to be cataloged , and in a number of cases , individually illustrated . based on this statistically robust sample am87 went beyond simple cataloging and added a broad list of descriptors covering about two dozen properties . one of these descriptor / classifications was category 6 : ring galaxy . as already noted by am87 `` ring galaxy '' , in the isolated context of this particular catalog , was a purely morphologically defined descriptor ; as a consequence this descriptor encompasses several physically distinct types of galaxies . first , there are relatively normal galaxies with `` outer rings '' . in the rc3 system of de vaulcouleurs et al . these are typed as * rs * galaxies , and called `` outer ringed '' galaxies in the compilation by buta ( 1995 ) . outer rings are not uncommon and are generally well understood ( see buta for an overview ) . second , there are `` polar ring galaxies '' ; systems thought to be advanced mergers with the disrupted satellite now orbiting the parent galaxy over the pole and forming the ring - like structure . these are exceedingly rare objects ( see whitmore et al . 1990 for a recent compilation and references therein . ) finally , there are the collisional systems ( called `` ring '' in rc3 ) ; and they are the focus of this paper . in an earlier paper ( which was based on an incomplete , early - release version of the arp - madore catalogue ) few , madore & arp ( 1986 ) ( hereafter fma ) pointed out the two dominant types of rings in the catalog and separated them into two distinct classes : * o - types * for the outer ( `` resonant '' ) rings , and * p - types * for the classical ( `` collisional '' ) rings . the main conclusion of that paper was that the * p - type * ( ring ) galaxies were consistent with the head - on collision scenario , being championed in that same time period by theys & spiegel ( 1976 ) and independently by lynds & toomre ( 1976 ) . the full success of that theory has broader implications which are explored in madore et al . all three of the current authors independently inspected and reclassified each of the 500 + category 6 ( ring galaxies ) in am87 . only * p - type * galaxies ( on the fma classification system ) were retained for this study . that is , any object with a crisp , relatively high - surface - brightness ring , with the nucleus often asymmetrically placed ( or in some cases absent ) , was given the p classification and retained . objects with more diffuse or feathery outer rings ( of generally somewhat lower surface brightness ) , with fairly large nuclear regions or bulges symmetrically placed with respect to the ring , were deemed to be the * o types*. these objects were returned to the general pool . out of 552 class 6 ( ring ) objects in am87 we culled out 104 * p - type * ring galaxies ; and for only a handful of objects was there any disagreement among the authors as to their classification . erring on the conservative / inclusive side , all of the contentious objects were retained in this study . table 1 lists all 104 pure * p - type * ( ring ) galaxies found in am87 . we note here that pure ring galaxies constitute less than 2% of the am87 peculiar galaxy population , and as such they are found in the general population of galaxies only 0.01% of the time ( as normalized by the entire population of @xmath0100,000 galaxies inspected by am87 in the course of constructing the catalog as a whole . ) ring galaxies are a rare type of collision with short - lived evidence . it is of course interesting to contemplate what ring galaxies look like after the collisionally - induced burst of star formation fades . we leave that as an exercise for the future . rather than simply illustrating a few of the most spectacular examples of ring galaxies , as was done in am87 , we have chosen to re - catalog and present the entire sample . however , before doing that we decided to additionally survey the literature for previously cataloged ring galaxies and add them to the _ atlas _ for completeness sake . the am87 sample is given in table 1 , while the ring galaxies from the literature are collected in table 2 . clearly the northern and equatorial sky is incompletely sampled . we estimate that about 200 more ring galaxies await discovery once those regions are inspected to the same apparent - size and surface - brightness limit as am87 . table 1 and 2 share the same basic formating . column 1 contains the name of the host ring galaxy complex ; the second column contains the name or names of the ring nucleus followed by as many companions as are suspected of being colliders . the name of the ring nucleus is composed of the name of the host object followed by rn ( e.g. , am 0147 - 350:rn ) ; by way of contrast , companion / colliders are identified by the host object name followed by c1 , c2 , etc . ( e.g. , am 0147 - 350:c1 , am 0147 - 350:c2 , am 0147 - 350:c3 ) . columns 4 and 5 respectively contain ra(2000 ) and dec(2000 ) positions , each measured to a precision of 5 arcsec . in column 5 the first entry ( d = diameter ) , corresponds to the diameter of the ring , measured in arcsec , while the following entries in that same column but now for the companions are marked by s , and give the separation of the companion from the nucleus of the ring ( again , measured in arcsec ) . finally , column 6 gives the heliocentric velocity for the object / component when available from the literature as captured by ned in its december 2008 release . all of the objects in this new listing consist of at least two objects in the immediate field of view that can reasonably be identified with the collider and the ( collidee ) ring . we mark those objects in each of the illustrations and for uniformity sake we have measured the positions of all objects given in tables 1 and 2 using the display capabilities of aladin interacting with the ned image archive . the presence of nearby companions was by no means a foregone conclusion given that the rings were chosen purely on the basis of their morphology , not on the basis of their nearby or distant environment . we consider it highly likely that these optical companions are the colliders , plausibly responsible for the `` gravitational splash '' and the ensuing ring - like , star - formation response predicted / post - dicted by the simulations . indeed , of the 64 companions with published redshifts ( found from a search of the january 2008 release of the ned database ) 60 have redshifts that place them at the same distance as the ring itself . in all four of the other instances where one of the apparent companions proved to be a background object there was always an additional candidate collider in the same field of view . although common radial velocities are not conclusive proof that any given object is the collider it does clearly demonstrate that the vast majority of those objects are not chance line - of - sight background ( or foreground ) objects . 111 apparent companions still need redshifts ; 43 of the rings themselves are without published velocities . we are working to ameliorate that situation . while all of our ring galaxies have one or more apparent companions , as a class ring galaxies tend not to be found in high - density ( cluster ) environments . indeed , only 7 of the 132 rings cataloged here have more than three companions in their immediate vicinity . they are am 0401 - 641 , am 0417 - 391 , am 0455 - 465 , am 1003 - 215 , am 1251 - 283 , am 2100 - 725 and am 2200 - 715 . 83 of the rings have only one nearby optical companion , 29 have two companions , and only 13 have three . it seems reasonable to suppose that in dense systems the more frequent , but off - center encounters with other group members will disrupt and/or destroy any ring that might form from a direct hit more quickly than in the quiescent field . perhaps the increased velocity dispersion of all the members associated with the cluster environment also plays a role . and it should not be forgotten that very rich clusters are _ currently _ depleted of disk galaxies in any case : with no disks to be hit , no rings can be formed . however , at high redshift the situation is expected to be quite different . a search for rings in various `` deep fields '' ( udf , goods and gems ) has been made ( elmegreen & elmegreen 2006 ) . according to those authors and judging from the published images of their pure `` rings '' ( their figure 1 ) these are almost exclusively * o - type * ( outer - ring ) galaxies with well - centered bulges and no obvious companions . some collisional objects may be found in their sample of fifteen `` partial rings '' ; however , the area shown around each partial ring is too small to judge whether these objects have interacting companions or not . figure 1 shows the wide range of absolute ( k - band ) magnitudes that the ring galaxies in this catalog have . spanning more than a factor of 100 in luminosity galaxies with ring - like morphology can be as faint as m@xmath1 = -19 mag and as bright as m@xmath1 = -25 mag ; typical ring galaxies have m@xmath1 = -23 mag which is somewhat fainter than the knee in the general - field luminosity function . the colliders are typically half a magnitude fainter than the central ( target ) galaxy ( figure 2 ) , but their dispersion is large ( @xmath20.75 mag ) . the current diameters of the rings cataloged here span a wide range being broadly peaked at 30 - 40 kpc , but extending up to values as large as 70 kpc ( figure 3 ) . none of these plots argue for separate populations of ring galaxies based solely on these physical properies . some ring galaxies ( e.g. , am 0058 - 220 , am 1953 - 260 , arp 147 & vii zw 466 ) do not have any obvious component that can unambiguously be identified as a nucleus . the morphology of such objects has been aptly likened to `` smoke rings . '' the apparent lack of a nucleus may be because the target galaxy simply never had a particularly noteworthy nucleus to begin with ( i.e. , even before the collision ) or it may be that the timing of the interaction , combined with our particular viewing angle places the nucleus in the ring ( see ngc 4774 , arp 146 ) and/or even confused with the intruder . in other instances the nucleus may have been disrupted by the collision . detailed modeling would illuminate these possibilities . we draw attention to a sub - class of objects in this compilation of ring galaxies : double rings . am 2136 - 492 is noteworthy . the cartwheel ( am 0035 - 335 ) is another example ( with its celebrated compact nuclear ring ) ; while am 0339 - 625 , am 1323 - 222 and am 1354 - 250 may be other less obvious examples of this class . figure 6 shows a deeply stretched image of am 0339 - 625 illustrating its fragmented , outer ring - like feature . although multiple rings structures seem to be a common feature in the numerical simulations ( e.g. antunes & wallin 2007 ) they appears not to be not so common in nature . a simple calculation concerning the apparent frequency of ring galaxies leads to a rather surprising / interesting result . the relative frequency of the collisonal ring phenomenon is at least 1 in 1000 , given that am87 inspected approximately 100,000 galaxies and out of those cataloged only 100 rings . ( and this is a lower limit given that edge - on rings are probably under - represented in the catalog . ) if the ring structure can unambiguously be detected as such for one dynamical timescale ( @xmath3 yrs , say ) and if the collision rate is constant over cosmic time , then about one in every ten galaxies in this same volume is expected to have been perturbed in this same way over a hubble time . given the restricted parameter space available for a head - on collision to occur this strongly suggests that , if the orbits of these types of companions are isotropic ( and only a small fraction of them produce rings ) then the majority of galaxies in the local volume have presumably been hit ( with a different impact parameter ) by a companion comparable in mass to that of a typical ring - galaxy collider . in figure 4 we show the histogram of relative velocities from those ring galaxies with redshift measurements for both the ring itself and at least one of its putative colliders . the median velocity difference for this sample is rather small ( @xmath0100 km / s ) . excluding obvious background objects with @xmath4v @xmath5 1,000 km / s , the largest two remaining velocity differences are 293 and 419 km / s . we conclude that the majority of these systems are bound and that if simulations are a guide then there will be a rapid merging of these pairs in the course of the next passage . computer simulations of galaxy interactions first showed that ring galaxies could convinciungly be created by the head - on collision of a satellite galaxy with a primary disk . observations of prototypical ring galaxies , such as the cartwheel ( fosbury & hawarden 1977 ) and the lindsay - shapley ring , each confirmed the `` prediction '' in as much as colliders were found in evidence , and within the range of distances , velocities and position angles expected for head - on collisional events . the present study follows the early work of few , madore & arp ( 1986 ) and extends this confirmation of theory from a handful of galaxies to a relative complete sample of over 100 suspected and previously cataloged ring galaxies . however , there is another prediction that is fundamentally at odds with what is observed for this sample of ring galaxies . that prediction comes from @xmath6-cdm cosmology in which many more satellite galaxies are expected to be orbiting central galaxies than are in fact observed . making these satellites non - luminous ( be it by supernova events , by interactions , by initial conditions , or by fiat ) does not prevent them from interacting gravitationally , and as such they too should be responsible for creating rings through collisions in direct proportion to their numbers with respect to their luminous colliding counterparts . the astonishing fact is that _ there is not a single compelling example of a ring galaxy without a plausible optical collider . _ there is no isolated cartwheel in the sky ; there are no smoke rings without the smoking gun . and there should be many . we state this simple fact and leave a more detailed discussion of this point and its implications for @xmath6-cdm cosmology to the companion paper by madore et al . ( 2009 ) . we wish to acknowledge and thank the creators of saoimage ds9 and aladin both of which were used extensively in the preparation of this _ atlas_. and , of course , this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . finally , we thank ian steer for bringing the ring galaxy ic 1908 to our attention . am 0034 - 351&am 0034 - 351:rn & 00:37:09.3 & -34:56:15 & d = 31 & 8,891 + & am 0034 - 351:c1 & 00:37:04.7 & -34:56:24 & s = 58 & . . + & & & & & + am 0035 - 335&am 0035 - 335:rn & 00:37:41.1 & -33:42:59 & d = 58 & 9,050 + & am 0035 - 335:c1 & 00:37:44.7 & -33:42:20 & s = 61 & 9,104 + & am 0035 - 335:c2 & 00:37:43.1 & -33:42:08 & s = 60 & 8,639 + & & & & & + am 0053 - 353&am 0053 - 353:rn & 00:56:19.3 & -35:14:53 & d = 47 & 14,494 + & am 0053 - 353:c1 & 00:56:15.7 & -35:15:15 & s = 43 & 14,420 + & & & & & + am 0054 - 634&am 0054 - 634:rn & 00:56:51.1 & -63:28:51 & d = 27 & 11,557 + & am 0054 - 634:c1 & 00:56:45.5 & -63:29:16 & s = 20 & 11,526 + & am 0054 - 634:c2 & 00:56:48.8 & -63:28:38 & s = 45 & . . . + & & & & & + am 0058 - 220&am 0058 - 220:rn & 01:00:46.0 & -21:43:30 & d = 17 & . . . + & am 0058 - 220:c1 & 01:00:47.2 & -21:43:26 & s = 17 & 12,524 + & & & & + am 0058 - 311&am 0058 - 311:rn & 01:00:44.5 & -33:55:49 & d = 34 & 23,650 + & am 0058 - 311:c1 & 01:00:48.1 & -33:55:49 & s = 46 & 23,684 + & & & & & + am 0058 - 665&am 0058 - 665:rn & 01:00:15.1 & -66:37:07 & d = 36 & 21,055 + & am 0058 - 665:c1 & 01:00:32.3 & -66:37:49 & s = 112 & . . . + & am 0058 - 665:c2 & 01:00:29.4 & -66:36:11 & s = 103 & . . . + & & & & & + am 0102 - 610&am 0102 - 610:rn & 01:04:56.3 & -60:48:58 & d = 22 & . . + & am 0102 - 610:c1 & 01:04:53.8 & -60:49:17 & s = 26 & . . . + & & & & & + am 0115 - 463&am 0115 - 463:rn & 01:18:09.9 & -46:14:32 & d = 33 & 17,688 + & am 0115 - 463:c1 & 01:18:09.7 & -46:15:07 & s = 34 & . . + & & & & & + am 0126 - 525&am 0126 - 525:rn & 01:28:24.6 & -52:38:08 & d = 29 & 3,700 + & am 0126 - 525:c1 & 01:28:25.2 & -52:38:11 & s = 06 & 3,450 + & & & & & + am 0147 - 350&am 0147 - 350:rn & 01:49:43.5 & -34:49:53 & d = 52 & 8,244 + & am 0147 - 350:c1 & 01:49:47.2 & -34:48:57 & s = 73 & . . . + & am 0147 - 350:c2 & 01:49:41.4 & -34:49:25 & s = 39 & . . . + & am 0147 - 350:c3 & 01:49:39.2 & -34:50:08 & s = 55 & bckgrnd + & & & & & + am 0154 - 353&am 0154 - 353:rn & 01:57:08.8 & -35:18:47 & d = 23 & . . . + & am 0154 - 353:c1 & 01:57:08.9 & -35:18:51 & s = 06 & . . . + am 0155 - 815&am 0155 - 815:rn & 01:54:41.4 & -81:37:53 & d = 38 & . . + & am 0155 - 815:c1 & 01:53:00.8 & -81:37:57 & s = 43 & . . + & am 0155 - 815:c2 & 01:52:39.7 & -81:37:50 & s = 05 & . . . + & & & & & + am 0157 - 311&am 0157 - 311:rn & 01:59:39.5 & -31:01:11 & d = 29 & 17,742 + & am 0157 - 311:c1 & 01:59:37.8 & -31:01:22 & s = 24 & . . + & & & & & + am 0200 - 581&am 0200 - 581:rn & 02:01:50.8 & -57:56:52 & d = 27 & . . + & am 0200 - 581:c1 & 02:01:50.7 & -57:57:01 & s = 10 & . . . + & am 0200 - 581:c2 & 02:01:51.4 & -57:56:56 & s = 06 & . . . + & & & & & + am 0240 - 242&am 0240 - 242:rn & 02:42:32.0 & -24:13:52 & d = 26 & 25,310 + & am 0240 - 242:c1 & 02:42:32.0 & -24:13:57 & s = 05 & . . + & am 0240 - 242:c2 & 02:42:33.4 & -24:13:05 & s = 46 & . . . + & & & & & + am 0305 - 824&am 0305 - 824:rn & 03:01:44.3 & -82:35:30 & d = 34 & . . + & am 0305 - 824:c1 & 03:02:01.8 & -82:35:26 & s = 34 & . . . + & & & & & + am 0318 - 610&am 0318 - 610:rn & 03:19:37.5 & -60:57:01 & d = 17 & . . + & am 0318 - 610:c1 & 03:19:37.9 & -60:57:10 & s = 09 & . . . + am 0322 - 374&am 0322 - 374:rn & 03:23:54.6 & -37:30:36 & d = 43 & 1,405 + & am 0322 - 374:c1 & 03:23:53.7 & -37:30:47 & s = 15 & 1,397 + & & & & & + am 0339 - 625&am 0339 - 625:rn & 03:40:37.8 & -62:46:49 & d = 30/62 & 14,860 + & am 0339 - 625:c1 & 03:40:39.6 & -62:46:36 & s = 18 & . . + & & & & & + am 0401 - 641&am 0401 - 641:rn & 04:01:36.9 & -64:04:03 & d = 40 & . . + & am 0401 - 641:c1 & 04:01:43.4 & -64:05:29 & s = 16 & . . . + & am 0401 - 641:c2 & 04:01:54.9 & -64:06:10 & s = 85 & 12,457 + & am 0401 - 641:c3 & 04:01:37.2 & -64:04:38 & s = 78 & . . . + & am 0401 - 641:c4 & 04:01:46.7 & -64:05:54 & s = 28 & . . + & & & & & + am 0403 - 555&am 0403 - 555:rn & 04:04:16.4 & -55:46:23 & d = 40 & . . + & am 0403 - 555:c1 & 04:04:17.0 & -55:46:06 & s = 18 & 17,825 + & am 0403 - 555:c2 & 04:04:12.1 & -55:45:20 & s = 80 & . . . + & & & & & + am 0404 - 271&am 0404 - 271:rn & 04:06:28.8 & -27:09:21 & d = 25 & 26,443 + & am 0404 - 271:c1 & 04:06:29.4 & -27:09:24 & s = 08 & . . . + & & & & & + & & & & & + am 0413 - 590&am 0413 - 590:rn & 04:14:25.9 & -58:57:45 & d = 43 & 17,143 + & am 0413 - 590:c1 & 04:14:27.0 & -58:57:40 & s = 09 & . . + & & & & & + am 0415 - 475&am 0415 - 475:rn & 04:16:29.9 & -47:50:54 & d gt 84 & 10,079 + & am 0415 - 475:c1 & 04:16:18.6 & -47:49:12 & s = 152 & . . + & am 0415 - 475:c2 & 04:16:31.6 & -47:50:43 & s = 23 & . . . + & & & & & + am 0417 - 391&am 0417 - 391:rn & 04:19:39.6 & -39:10:31 & d = 31 & 15,255 + & am 0417 - 391:c1 & 04:19:38.9 & -39:10:26 & s = 09 & . . . + & am 0417 - 391:c2 & 04:19:39.9 & -39:10:23 & s = 10 & 15,255 + & am 0417 - 391:c3 & 04:19:40.5 & -39:08:40 & s = 114 & 14,820 + & am 0417 - 391:c4 & 04:19:40.4 & -39:11:39 & s = 70 & . . . + & am 0417 - 391:c5 & 04:19:51.3 & -39:10:05 & s = 140 & 15,185 + & & & & & + am 0425 - 421&am 0425 - 421:rn & 04:26:36.6 & -42:05:38 & d = 55 & 4,491 + & am 0425 - 421:c1 & 04:26:44.6 & -42:05:42 & s = 91 & . . . + & & & & & + am 0437 - 314&am 0437 - 314:rn & 04:39:40.0 & -31:42:39 & d = 28 & . . + & am 0437 - 314:c1 & 04:39:40.3 & -31:42:38 & s = 05 & . . . + & & & & & + am 0437 - 644&am 0437 - 644:rn & 04:37:54.9 & -64:38:59 & d = 14 & . . + & am 0437 - 644:c1 & 04:37:56.9 & -64:39:06 & s = 33 & . . . + & & & & & + am 0438 - 503&am 0438 - 503:rn & 04:39:20.5 & -50:31:50 & d = 38 & . . + & am 0438 - 503:c1 & 04:39:22.1 & -50:31:45 & s = 15 & . . . + & & & & & + am 0438 - 661&am 0438 - 661:rn & 04:38:37.1 & -66:13:03 & d = 34 & 14,650 + & am 0438 - 661:c1 & 04:38:37.9 & -66:13:49 & d = 12 & . . + & & & & & + am 0455 - 465&am 0455 - 465:rn & 04:57:01.7 & -46:45:00 & d = 15 & . . + & am 0455 - 465:c1 & 04:57:04.1 & -46:44:54 & s = 26 & . . . + & am 0455 - 465:c2 & 04:57:06.5 & -46:45:03 & s = 49 & . . . + & am 0455 - 465:c3 & 04:57:00.1 & -46:46:00 & s = 60 & . . . + & am 0455 - 465:c4 & 04:56:55.8 & -46:44:10 & s = 83 & . . . + & am 0455 - 465:c5 & 04:57:06.5 & -46:46:08 & s = 80 & . . . + & & & & & + am 0507 - 512&am 0507 - 512:rn & 05:08:28.0 & -51:21:29 & d = 34 & . . + & am 0507 - 512:c1 & 05:08:27.6 & -51:20:38 & s = 43 & . . . + & am 0507 - 512:c2 & 05:08:27.3 & -51:21:22 & s = 11 & . . . + & am 0507 - 512:c3 & 05:08:29.5 & -51:21:42 & s = 20 & . . . + am 0520 - 390&am 0520 - 390:rn & 05:22:42.5 & -39:03:48 & d = 43 & 15,114 + & am 0520 - 390:c1 & 05:22:43.4 & -39:03:25 & s = 26 & . . + & am 0520 - 390:c2 & 05:22:48.3 & -39:03:38 & s = 68 & . . . + & am 0520 - 390:c3 & 05:22:44.3 & -39:03:41 & s = 23 & . . . + & & & & & + am 0521 - 434&am 0521 - 434:rn & 05:22:30.5 & -43:46:17 & d = 30 & 24,291 + & am 0521 - 434:c1 & 05:22:30.4 & -43:45:51 & s = 26 & . . + & & & & & + am 0545 - 355&am 0545 - 355:rn & 05:47:22.0 & -35:49:29 & d = 42 & 13,885 + & am 0545 - 355:c1 & 05:47:19.4 & -35:49:45 & s = 36 & . . . + & & & & & + am 0545 - 434&am 0545 - 434:rn & 05:47:00.0 & -43:44:56 & d = 38 & . . + & am 0545 - 434:c1 & 05:47:01.2 & -43:44:33 & s = 26 & . . + & & & & & + am 0642 - 645&am 0642 - 645:rn & 06:42:34.2 & -64:58:28 & d = 20 & . . + & am 0642 - 645:c1 & 06:42:32.7 & -64:59:24 & s = 05 & . . . + & am 0642 - 645:c2 & 06:42:33.4 & -64:59:28 & s = 05 & . . . + & & & & & + & & & & & + & & & & & + am 0642 - 801&am 0642 - 801:rn & 06:38:38.4 & -80:14:50 & d = 67 & 4,777 + & am 0642 - 801:c1 & 06:38:36.6 & -80:14:49 & s = 08 & . . . + & am 0642 - 801:c2 & 06:38:27.9 & -80:14:59 & s = 29 & . . . + & & & & & + am 0643 - 462&am 0643 - 462:rn & 06:45:03.3 & -46:26:56 & d = 60 & 12,056 + & am 0643 - 462:c1 & 06:45:00.2 & -46:26:43 & s = 36 & . . . + & & & & & + am 0644 - 741&am 0644 - 741:rn & 06:43:05.6 & -74:14:12 & d = 97 & 6,505 + & am 0644 - 741:c1 & 06:43:06.1 & -74:12:55 & s = 78 & 7,000 + & am 0644 - 741:c2 & 06:43:25.5 & -74:15:27 & s = 110 & 6,750 + & am 0644 - 741:c3 & 06:44:17.5 & -74:16:37 & s = 326 & 6,430 + & & & & & + am 0755 - 785&am 0755 - 785:rn & 07:53:31.5 & -79:06:28 & d = 27 & . . . + & am 0755 - 785:c1 & 07:53:35.6 & -79:06:28 & s = 13 & . . . + & & & & & + am 0814 - 760&am 0814 - 760:rn & 08:13:31.6 & -76:17:29 & d = 28 & . . + & am 0814 - 760:c1 & 08:13:30.9 & -76:17:41 & s = 13 & . . . + & & & & & + & & & & & + & & & & & + am 1003 - 215&am 1003 - 215:rn & 10:05:43.2 & -22:05:43 & d = 28 & 26,329 + & am 1003 - 215:c1 & 10:05:43.3 & -22:05:53 & s = 12 & . . . + & am 1003 - 215:c2 & 10:05:43.4 & -22:05:03 & s = 41 & . . + & am 1003 - 215:c3 & 10:05:44.1 & -22:06:19 & s = 38 & . . . + & am 1003 - 215:c4 & 10:05:46.2 & -22:05:52 & s = 43 & . . . + & & & & & + am 1006 - 380&am 1006 - 380:rn & 10:09:05.2 & -38:24:35 & d = 80 & 4,943 + & am 1006 - 380:c1 & 10:09:08.0 & -38:23:48 & s = 58 & 4,481 + & & & & & + am 1025 - 370&am 1025 - 370:rn & 10:27:13.8 & -37:25:17 & d = 27 & . . + & am 1025 - 370:c1 & 10:27:13.8 & -37:25:33 & s = 16 & . . . + & & & & & + am 1133 - 245&am 1133 - 245:rn & 11:35:30.8 & -25:08:44 & d = 67 & 11,704 + & am 1133 - 245:c1 & 11:35:17.6 & -25:08:45 & s = 180 & 11,682 + & am 1133 - 245:c2 & 11:35:34.5 & -25:05:41 & s = 189 & 11,601 + & & & & & + am 1135 - 284&am 1135 - 284:rn & 11:37:42.2 & -29:05:01 & d = 24 & 5,558 + & am 1135 - 284:c1 & 11:37:36.2 & -29:04:11 & s = 97 & . . + & & & & & + & & & & & + am 1152 - 421&am 1152 - 241:rn & 11:55:21.7 & -42:29:19 & d = 46 & 4,617 + & am 1152 - 241:c1 & 11:55:21.3 & -42:29:17 & s = 07 & . . + & & & & + am 1159 - 530&am 1159 - 530:rn & 12:01:35.9 & -53:21:28 & d = 55 & . . + & am 1159 - 530:c1 & 12:01:32.3 & -53:20:47 & s = 52 & . . . + & & & & + am 1249 - 462&am 1249 - 462:rn & 12:52:38.7 & -46:38:59 & d = 50 & . . + & am 1249 - 462:c1 & 12:52:34.9 & -46:40:04 & s = 78 & . . . + & & & & & + am 1251 - 283&am 1251 - 283:rn & 12:54:42.4 & -28:51:57 & d = 29 & 17,003 + & am 1251 - 283:c1 & 12:54:43.1 & -28:52:17 & s = 24 & 16,242 + & am 1251 - 283:c2 & 12:54:43.1 & -28:50:17 & s = 100 & 16,359 + & am 1251 - 283:c3 & 12:54:36.5 & -28:51:28 & s = 82 & bckgrnd + & am 1251 - 283:c4 & 12:54:40.9 & -28:53:03 & s = 69 & 16,576 + & am 1251 - 283:c5 & 12:54:49.7 & -28:52:37 & s = 105 & foregrnd + & & & & & + am 1300 - 412&am 1300 - 412:rn & 13:03:00.3 & -41:42:16 & d = 18 & 3,422 + & am 1300 - 412:c1 & 13:03:00.4 & -41:42:20 & s = 05 & . . + & & & & & + & & & & & + am 1308 - 253&am 1308 - 253:rn & 13:11:08.7 & -25:54:00 & d = 44 & 13,556 + & am 1308 - 253:c1 & 13:11:08.5 & -25:54:00 & s = 05 & . . + & am 1308 - 253:c2 & 13:11:07.4 & -25:54:28 & s = 85 & . . . + & & & & & + am 1323 - 222&am 1323 - 222:rn & 13:26:20.3 & -22:37:54 & d = 88 & 4,627 + & am 1323 - 222:c1 & 13:26:29.4 & -22:33:49 & s = 272 & bckgrnd + & & & & & + am 1325 - 251&am 1325 - 251:rn & 13:27:58.1 & -25:31:49 & d = 18 & 10,936 + & am 1325 - 251:c1 & 13:27:57.7 & -25:31:44 & s = 07 & . . + & am 1325 - 251:c2 & 13:27:57.1 & -25:30:50 & s = 52 & . . . + & & & & & + am 1354 - 250&am 1354 - 250:rn & 13:57:13.7 & -25:14:44 & d = 53 & 6,150 + & am 1354 - 250:c1 & 13:57:17.8 & -25:13:30 & s = 97 & 6,175 + & am 1354 - 250:c2 & 13:57:24.6 & -25:13:59 & s = 157 & . . . + & & & & & + am 1358 - 221&am 1358 - 221:rn & 14:01:08.0 & -22:33:35 & d = 53 & 10,940 + & am 1358 - 221:c1 & 14:01:13.2 & -22:32:45 & s = 87 & . . . + & am 1358 - 221:c2 & 14:01:12.8 & -22:31:52 & s = 121 & . . . + & am 1358 - 221:c3 & 14:01:25.6 & -22:34:20 & s = 247 & 10,786 + & & & & & + am 1413 - 243&am 1413 - 243:rn & 14:16:09.3 & -24:50:30 & d = 35 & 13,740 + & am 1413 - 243:c1 & 14:16:04.9 & -24:49:21 & s = 94 & . . + & am 1413 - 243:c2 & 14:16:09.3 & -24:51:09 & s = 37 & . . . + & am 1413 - 243:c3 & 14:16:13.3 & -24:50:47 & s = 56 & . . . + & & & & + am 1425 - 234&am 1425 - 234:rn & 14:55:09.8 & -23:58:09 & d = 61 & . . + & am 1425 - 234:c1 & 14:55:09.4 & -23:58:16 & s = 09 & . . . + & am 1425 - 234:c2 & 14:55:08.0 & -23:56:24 & s = 112 & . . . + & am 1425 - 234:c3 & 14:55:00.6 & -23:55:44 & s = 195 & . . . + & & & & & + am 1434 - 783&am 1434 - 783:rn & 14:40:27.5 & -78:48:35 & d = 40 & 4,625 + & am 1434 - 783:c1 & 14:40:32.2 & -78:48:28 & s = 40 & . . + & & & & & + am 1452 - 234&am 1452 - 234:rn & 14:55:09.8 & -23:58:09 & d = 61 & 12,304 + & am 1452 - 234:c1 & 14:55:09.4 & -23:58:16 & s = 09 & . . + & am 1452 - 234:c2 & 14:55:08.0 & -23:56:24 & s = 112 & . . . + & am 1452 - 234:c3 & 14:55:00.6 & -23:55:44 & s = 195 & . . . + & & & & & + am 1514 - 362&am 1514 - 362:rn & 15:17:48.2 & -36:34:56 & d = 24 & . . + & am 1514 - 362:c1 & 15:17:47.8 & -36:35:00 & s = 06 & 7,108 + am 1627 - 824&am 1627 - 824:rn & 16:35:56.7 & -82:50:17 & d = 32 & . . + & am 1627 - 824:c1 & 16:36:01.0 & -82:50:12 & s = 11 & . . . + & am 1627 - 824:c2 & 16:35:17.2 & -82:51:17 & s = 94 & . . + & & & & & + am 1724 - 622&am 1724 - 622:rn & 17:29:09.6 & -62:26:45 & d = 125 & 4,641 + & am 1724 - 622:c1 & 17:29:25.3 & -62:28:50 & s = 167 & 4,800 + & & & & & + am 1827 - 625&am 1827 - 625:rn & 18:32:02.7 & -62:55:44 & d = 33 & 4,086 + & am 1827 - 625:c1 & 18:32:02.6 & -62:55:47 & s = 04 & . . . + & & & & & + am 1854 - 490&am 1854 - 490:rn & 18:58:06.8 & -49:00:41 & d = 14 & 4,250 + & am 1854 - 490:c1 & 18:58:05.5 & -49:00:52 & s = 18 & . . + & & & & & + am 1947 - 445&am 1947 - 445:rn & 19:51:19.8 & -44:52:40 & d = 62 & 5,805 + & am 1947 - 445:c1 & 19:51:26.3 & -44:52:36 & s = 141 & 5,672 + & & & & & + am 1953 - 260&am 1953 - 260:rn & 19:56:28.4 & -25:54:56 & d = 40 & . . . + & am 1953 - 260:c1 & 19:56:29.0 & -25:55:16 & s = 24 & 14,802 + & & & & & + & & & & & + am 1957 - 394&am 1957 - 394:rn & 20:00:36.0 & -39:39:02 & d = 51 & . . + & am 1957 - 394:c1 & 20:00:35.4 & -39:39:02 & s = 05 & . . . + & & & & & + am 2012 - 282&am 2012 - 282:rn & 20:15:13.9 & -28:18:25 & d = 25 & 7,051 + & am 2012 - 282:c1 & 20:15:13.6 & -28:18:24 & s = 06 & . . + & am 2012 - 282:c2 & 20:15:14.2 & -28:18:19 & s = 08 & . . . + & & & & & + am 2021 - 724&am 2021 - 724:rn & 20:26:29.2 & -72:36:46 & d = 28 & . . + & am 2021 - 724:c1 & 20:26:33.7 & -72:36:08 & s = 42 & 22,706 + & & & & & + am 2024 - 544&am 2024 - 544:rn & 20:27:53.9 & -54:38:00 & d = 61 & 8,100 + & am 2024 - 544:c1 & 20:27:57.0 & -54:37:51 & s = 28 & . . . + & am 2024 - 544:c2 & 20:27:48.5 & -54:38:28 & s = 55 & . . . + & & & & & + am 2026 - 424&am 2026 - 424:rn & 20:29:32.2 & -42:30:24 & d = 61 & . . + & am 2026 - 424:c1 & 20:29:33.5 & -42:30:24 & s = 17 & 15,486 + & & & & & + am 2033 - 260&am 2033 - 260:rn & 20:36:20.5 & -25:56:52 & d = 23 & 12,500 + & am 2033 - 260:c1 & 20:36:23.9 & -25:57:29 & s = 60 & 12,257 + & am 2033 - 260:c2 & 20:36:20.7 & -25:56:41 & s = 10 & . . . + am 2034 - 483&am 2034 - 483:rn & 20:38:08.7 & -48:23:16 & d = 22 & . . + & am 2034 - 483:c1 & 20:38:06.2 & -48:23:16 & s = 26 & . . . + & am 2034 - 483:c2 & 20:38:02.8 & -48:22:55 & s = 63 & . . . + & & & & & + am 2044 - 691&am 2044 - 691:rn & 20:48:56.8 & -69:05:32 & d = 42 & 11,413 + & am 2044 - 691:c1 & 20:48:56.6 & -69:05:29 & s = 03 & . . + & & & & & + am 2056 - 392&am 2056 - 392:rn & 21:00:07.3 & -39:17:57 & d = 48 & . . + & am 2056 - 392:c1 & 21:00:15.8 & -39:20:05 & s = 150 & . . . + & & & & & + am 2100 - 725&am 2056 - 392:rn & 21:05:38.2 & -72:47:09 & d = 27 & . . + & am 2056 - 392:c1 & 21:05:54.5 & -72:47:20 & s = 72 & 20,984 + & am 2056 - 392:c2 & 21:05:52.9 & -72:47:22 & s = 65 & . . . + & am 2056 - 392:c3 & 21:05:49.5 & -72:46:21 & s = 69 & . . . + & am 2056 - 392:c4 & 21:05:58.8 & -72:47:59 & s = 103 & . . . + & am 2056 - 392:c5 & 21:05:16.2 & -72:47:58 & s = 109 & . . + & am 2056 - 392:c6 & 21:05:11.5 & -72:47:04 & s = 119 & . . . + & am 2056 - 392:c7 & 21:05:48.2 & -72:47:29 & s = 48 & . . . + & & & & & + & & & & & + am 2107 - 474&am 2107 - 472:rn & 21:10:31.0 & -47:30:32 & d = 31 & 5,060 + & am 2107 - 472:c1 & 21:10:31.4 & -47:30:36 & s = 05 & . . . + & & & & & + am 2128 - 302&am 2128 - 302:rn & 21:31:08.2 & -30:16:27 & d = 84 & 10,037 + & am 2128 - 302:c1 & 21:31:01.5 & -30:19:40 & s = 208 & 9,199 + & & & & & + am 2131 - 254&am 2131 - 254:rn & 21:34:29.2 & -25:28:38 & d = 33 & 16,218 + & am 2131 - 254:c1 & 21:34:31.1 & -25:28:46 & s = 29 & . . . + & & & & & + am 2131 - 495&am 2131 - 495:rn & 21:34:37.6 & -49:42:46 & d = 27 & 21,645 + & am 2131 - 495:c1 & 21:34:37.2 & -49:42:32 & s = 15 & . . + & & & & & + am 2132 - 535&am 2132 - 535:rn & 21:36:11.1 & -53:41:26 & d = 32 & . . + & am 2132 - 535:c1 & 21:36:13.0 & -53:42:54 & s = 91 & . . . + & am 2132 - 535:c2 & 21:36:27.1 & -53:43:09 & s = 176 & . . . + & am 2132 - 535:c3 & 21:36:26.7 & -53:43:31 & s = 188 & . . . + & & & & & + am 2134 - 471&am 2134 - 471:rn & 21:37:28.0 & -47:02:09 & d = 52 & 9,254 + & am 2134 - 471:c1 & 21:37:31.1 & -47:00:38 & s = 98 & . . . + & am 2134 - 471:c2 & 21:37:35.9 & -47:02:40 & s = 90 & . . . + am 2136 - 492&am 2136 - 492:rn & 21:39:25.6 & -49:09:36 & d = 40/72 & 15,990 + & am 2136 - 492:c1 & 21:39:31.6 & -49:08:11 & s = 104 & . . + & & & & & + am 2141 - 515&am 2141 - 515:rn & 21:45:15.1 & -51:41:54 & d = 33 & . . + & am 2141 - 515:c1 & 21:45:14.9 & -51:41:49 & s = 05 & . . . + & & & & & + am 2145 - 543&am 2145 - 543:rn & 21:49:08.5 & -54:18:14 & d = 22 & 19,544 + & am 2145 - 543:c1 & 21:49:08.2 & -54:17:53 & s = 22 & . . + & & & & & + am 2152 - 592&am 2152 - 592:rn & 21:56:19.4 & -59:07:15 & d = 15 & 22,351 + & am 2152 - 592:c1 & 21:56:17.5 & -59:07:17 & s = 15 & . . . + & am 2152 - 592:c2 & 21:56:20.8 & -59:06:51 & s = 27 & . . . + & & & & & + am 2155 - 263&am 2155 - 263:rn & 21:58:38.4 & -26:22:41 & d = 27 & 21,255 + & am 2155 - 263:c1 & 21:58:37.2 & -26:22:25 & s = 23 & . . . + & & & & & + & & & & & + & & & & & + & & & & & + & & & & & + am 2200 - 715&am 2200 - 715:rn & 22:04:57.5 & -71:42:14 & d = 32 & . . . + & am 2200 - 715:c1 & 22:04:22.4 & -71:43:20 & s = 177 & 9,350 + & am 2200 - 715:c2 & 22:04:12.7 & -71:35:08 & s = 478 & . . + & am 2200 - 715:c3 & 22:04:47.8 & -71:35:59 & s = 379 & . . + & am 2200 - 715:c4 & 22:05:51.5 & -71:39:51 & s = 288 & . . + & am 2200 - 715:c5 & 22:06:23.9 & -71:43:53 & s = 418 & . . + & am 2200 - 715:c6 & 22:05:39.0 & -71:46:45 & s = 330 & . . + & & & & & + am 2201 - 230&am 2201 - 230:rn & 22:04:19.3 & -22:47:33 & d = 24 & 21,732 + & am 2201 - 230:c1 & 22:04:19.5 & -22:47:48 & s = 17 & 21,675 + & & & & & + am 2220 - 493&am 2220 - 493:rn & 22:23:13.4 & -49:17:46 & d = 32 & . . . + & am 2220 - 493:c1 & 22:23:12.7 & -49:17:16 & s = 31 & 17,947 + & & & & & + am 2230 - 481&am 2230 - 481:rn & 22:33:47.0 & -48:01:30 & d = 52 & 10,476 + & am 2230 - 481:c1 & 22:33:43.3 & -48:01:31 & s = 37 & 10,361 + & & & & & + am 2238 - 541&am 2238 - 541:rn & 22:41:55.1 & -53:58:43 & d = 32 & . . . + & am 2238 - 541:c1 & 22:41:57.5 & -53:59:45 & s = 67 & . . . + & & & & & + am 2240 - 304&am 2240 - 304:rn & 22:43:36.9 & -30:28:50 & d = 22 & 17,449 + & am 2240 - 304:c1 & 22:43:36.3 & -30:28:55 & s = 08 & . . + & & & & & + am 2302 - 322&am 2302 - 322:rn & 23:05:39.9 & -32:09:07 & d = 25 & 18,137 + & am 2302 - 322:c1 & 23:05:41.9 & -32:08:11 & s = 66 & 18,077 + & am 2302 - 322:c2 & 23:05:43.9 & -32:09:46 & s = 59 & bckgrnd + & & & & & + am 2308 - 324&am 2308 - 324:rn & 23:11:17.3 & -32:27:07 & d = 33 & 11,327 + & am 2308 - 324:c1 & 23:11:16.9 & -32:27:21 & s = 15 & . . . + & & & & & + am 2313 - 261&am 2313 - 261:rn & 23:15:46.5 & -25:54:20 & d = 20 & 6,472 + & am 2313 - 261:c1 & 23:15:46.5 & -25:54:28 & s = 09 & . . . + & & & & & + am 2317 - 672&am 2317 - 672:rn & 23:20:09.7 & -67:08:58 & d = 32 & . . + & am 2317 - 672:c1 & 23:20:11.2 & -67:09:00 & s = 10 & . . . + & am 2317 - 672:c2 & 23:20:11.6 & -67:08:28 & s = 33 & . . . + & & & & & + am 2322 - 671&am 2322 - 671:rn & 23:25:42.8 & -67:03:29 & d = 19 & . . + & am 2322 - 671:c1 & 23:25:45.3 & -67:03:33 & s = 15 & . . . + & & & & & + am 2323 - 512&am 2323 - 512:rn & 23:26:31.8 & -51:08:07 & d = 21 & . . + & am 2323 - 512:c1 & 23:26:31.2 & -51:07:45 & s = 23 & . . . + & & & & & + am 2338 - 312&am 2338 - 312:rn & 23:41:26.0 & -31:03:22 & d = 54 & 17,538 + & am 2338 - 312:c1 & 23:41:19.2 & -31:03:01 & s = 91 & 17,268 + & & & & & + am 2343 - 703&am 2343 - 703:rn & 23:46:39.3 & -70:22:51 & d = 44 & . . . + & am 2343 - 703:c1 & 23:46:38.5 & -70:22:25 & s = 27 & . . . + & & & & & + am 2353 - 291&am 2353 - 291:rn & 23:56:23.8 & -29:01:25 & d = 36 & 8,931 + & am 2353 - 291:c1 & 23:56:25.1 & -29:01:24 & s = 17 & 8,953 + arp 146&arp 146:rn & 00:06:44.3 & -06:38:08 & d = 19 & 22,616 + & arp 146:c1 & 00:06:44.7 & -06:38:13 & d = 09 & . . . + & & & & & + arp 318&arp 318:rn & 02:09:24.4 & -10:08:10 & d = 40 & 4,073 + & arp 318:c1 & 02:09:20.7 & -10:08:00 & s = 55 & 3,864 + & arp 318:c2 & 02:09:38.3 & -10:08:48 & s = 208 & 3,851 + & arp 318:c3 & 02:09:42.7 & -10:11:02 & s = 319 & 3,934 + & & & & & + arp 010&arp 010:rn & 02:18:26.2 & + 05:39:14 & d = 44 & 9,108 + & arp 010:c1 & 02:18:28.6 & + 05:40:08 & s = 65 & . . + & & & & & + arp 273&arp 273:rn & 02:21:28.6 & + 39:22:32 & d = 100 & 7,563 + & arp 273:c1 & 02:21:32.7 & + 39:21:24 & s = 82 & 7,335 + & & & & & + arp 145&arp 145:rn & 02:23:07.7 & + 41:21:11 & d = 57 & 5,425 + & arp 145:c1 & 02:23:11.1 & + 41:22:04 & s = 36 & 5,852 + & & & & & + ngc 0985&ngc 0985:rn & 02:34:37.6 & -08:47:15 & d = 36 & 13,168 + & ngc 0985:c1 & 02:34:38.6 & -08:48:03 & s = 47 & . . . + & & & & & + arp 118&arp 118:rn & 02:55:12.0 & -00:11:03 & d = 30 & 8,648 + & arp 118:c1 & 02:55:09.5 & -00:10:40 & s = 45 & 8,459 + & arp 118:c2 & 02:55:06.5 & -00:09:46 & s = 114 & 8,417 + & & & & & + arp 147&arp 147:rn & 03:11:18.6 & + 01:18:51 & d = 17 & 9,656 + & arp 147:c1 & 03:11:19.4 & + 01:18:48 & s = 14 & 9,415 + & & & & & + ic 1908&ic 1908:rn & 03:15:05.4 & -54:49:09 & d = 30 & 8,234 + & ic 1908:c1 & 03:15:05.7 & -54:49:20 & s = 13 & . . + & & & & & + eso 200-ig 009&eso 200-ig 009:rn & 03:21:40.5 & -51:39:31 & d = 29 & 17,328 + & eso 200-ig 009:c1 & 03:21:41.0 & -51:39:49 & s = 21 & 20,980 + & eso 200-ig 009:c2 & 03:21:43.6 & -51:39:42 & s = 32 & . . + & & & & & + arp 219&arp 219:rn & 03:39:53.2 & -02:06:47 & d = 46 & 10,488 + & arp 219:c1 & 03:39:54.8 & -02:07:25 & s = 44 & . . + & & & & & + arp 141&arp 141:rn & 07:14:20.0 & + 73:28:25 & d = 75 & 2,728 + & arp 141:c1 & 07:14:20.2 & + 73:28:51 & s = 26 & 2,735 + & & & & & + arp 143&arp 143:rn & 07:46:55.0 & + 39:00:56 & d = 87 & 4,257 + & arp 143:c1 & 07:46:52.9 & + 39:01:55 & d = 65 & 4,048 + & arp 143:c2 & 07:46:55.5 & + 39:00:27 & d = 30 & 4,002 + & & & & & + ngc 2793&ngc 2793:rn & 09:16:47.1 & + 34:25:56 & d = 46 & . . . + & ngc 2793:c1 & 09:16:47.2 & + 34:25:48 & s = 11 & 1,687 + & ngc 2793:c2 & 09:16:40.8 & + 34:26:52 & s = 97 & bckgrnd + & ngc 2793:c3 & 09:16:46.6 & + 34:26:14 & s = 18 & 1,667 + & & & & & + arp 142&arp 142:rn & 09:37:44.1 & + 02:45:39 & d = 53 & 7,225 + & arp 142:c1 & 09:37:45.0 & + 02:44:51 & s = 52 & 6,806 + & arp 142:c2 & 09:37:41.2 & + 02:46:45 & s = 81 & . . + & & & & & + ic 0614&ic 0614:rn & 10:26:51.8 & -03:27:53 & d = 34 & 10,258 + & ic 0614:c1 & 10:26:51.9 & -03:27:39 & s = 16 & . . + & ic 0614:c2 & 10:26:49.0 & -03:29:19 & s = 97 & 18,097 + & & & & & + arp 107&arp 107:rn & 10:52:14.8 & + 30:03:28 & d = 55 & 10,372 + & arp 107:c1 & 10:52:18.4 & + 30:04:20 & d = 70 & 10,665 + & & & & & + arp 148&arp 148:rn & 11:03:52.4 & + 40:50:54 & d = 18 & 10,551 + & arp 148:c1 & 11:03:53.8 & + 40:51:00 & s = 17 & . . + & & & & & + arp 335&arp 335:rn & 11:04:23.5 & + 04:49:44 & d = 130 & 7,738 + & arp 335:c1 & 11:04:23.5 & + 04:49:29 & s = 15 & 7,763 + & arp 335:c2 & 11:04:16.5 & + 04:52:10 & s = 180 & 7,647 + & & & & & + vii zw 466&vii zw 466:rn & 12:32:04.9 & + 66:24:13 & d = 21 & 14,490 + & vii zw 466:c1 & 12:32:10.6 & + 66:24:19 & s = 34 & . . + & vii zw 466:c2 & 12:32:13.1 & + 66:23:59 & s = 50 & 14,100 + & vii zw 466:c3 & 12:32:11.6 & + 66:23:21 & s = 65 & 14,360 + & & & & & + ngc 4774&ngc 4774:rn & 12:53:06.4 & + 36:49:10 & d = 23 & 8,373 + & ngc 4774:c1 & 12:53:05.8 & + 36:49:34 & s = 25 & 8,373 + & & & & & + vv 256&vv 256:rn & 14:00:54.4 & + 40:59:19 & d = 64 & 3,738 + & vv 256:c1 & 14:00:56.2 & + 41:00:22 & s = 68 & 3,834 + & & & & & + kig 0686&kig 0686:rn & 15:32:56.9 & + 46:27:07 & d = 11 & . . + & kig 0686:c1 & 15:32:57.5 & + 46:27:14 & s = 09 & 656 + & & & & & + kig 0700&kig 0700:rn & 15:45:14.8 & + 00:46:24 & d = 19 & . . + & kig 0700:c1 & 15:45:14.3 & + 00:46:21 & s = 08 & 3,839 + & & & & & + arp 125&arp 125:rn & 16:38:13.9 & + 41:56:19 & d = 45 & 8,572 + & arp 125:c1 & 16:38:13.6 & + 41:55:51 & s = 31 & 8,404 + & & & & & + arp 150&arp 150:rn & 23:19:30.0 & + 09:30:30 & d = 34 & 11,879 + & arp 150:c1 & 23:19:31.0 & + 09:30:11 & s = 26 & 11,562 + & arp 150:c2 & 23:19:33.7 & + 09:29:42 & s = 75 & 11,637 + & arp 150:c3 & 23:19:27.7 & + 09:29:40 & s = 59 & 12,350 + & & & & & + arp 284&arp 284:rn & 23:36:14.2 & + 02:09:17 & d = 44 & 2,798 + & arp 284:c1 & 23:36:21.8 & + 02:09:25 & s = 111 & 2,771 + |
strong stellar magnetic fields are believed to truncate the inner accretion disks around young stars , redirecting the accreting material to the high latitude regions of the stellar surface . in the past few years , observations of strong stellar fields on t tauri stars with field strengths in general agreement with the predictions of magnetospheric accretion theory have bolstered this picture . currently , nothing is known about the magnetic field properties of younger , more embedded class i young stellar objects ( ysos ) . it is believed that protostars accrete much of their final mass during the class i phase , but the physics governing this process remains poorly understood . here , we use high resolution near infrared spectra obtained with nirspec on keck and with phoenix on gemini south to measure the magnetic field properties of the class i protostar wl 17 . we find clear signatures of a strong stellar magnetic field . analysis of this data suggests a surface average field strength of kg on wl 17 . we present our field measurements and discuss how they fit with the general model of magnetospheric accretion in young stars . | 0905.4784 | it is now generally accepted that accretion of circumstellar disk material onto the surface of classical t tauri stars ( cttss ) is controlled by strong stellar magnetic fields ( e.g. see review by bouvier et al . cttss represent class ii sources in the classification system defined by lada ( 1987 ) . the definition is based on a gradually falling spectral energy distribution ( sed ) beyond @xmath1 m . this sed shape is believed to arise from a geometrically thin , optically thick accretion disk containing a high concentration of submicron sized dust grains ( e.g. bertout et al . 1988 ) . at some level , the final mass of these forming stars is determined by how much of this disk material accretes onto the central star . additionally , it is within the disks around these low mass pre - main sequence stars that solar systems similar to our own form . it is critical to understand how the central young star interacts with and disperses its disk in order to understand star , and particularly planet , formation . the class i sources defined by lada ( 1987 ) represent one of the earliest stages of star formation and are identified by a rising sed . these sources are deeply embedded within molecular clouds and are very faint or undetectable at optical wavelengths because of a thick envelope of circumstellar dust . it has been commonly thought that class i objects represent an earlier evolutionary stage relative to class ii sources , with the paradigm emerging that class i sources are young protostars near the end of their bulk mass accretion phase . this paradigm is bolstered by the very weak photospheric absorption features in near infrared ( ir ) spectra of these objects ( e.g. casali & matthews 1992 ; greene & lada 1996 , 2000 ) . the lack of absorption lines was interpreted by these authors as the result of strong veiling produced by emission originating in a vigorously accreting circumstellar disk which is being fed by an infalling envelope . this emission is reprocessed by the dusty envelope which results in both the observed featureless continuum and the rising sed . as the accretion rate in the disk weakens and the thick circumstellar envelope either accretes onto the star plus disk system or is disrupted by strong outflows , it is generally thought that class i objects evolve into class ii sources . this general paradigm has been recently challenged by white & hillenbrand ( 2004 ) who find no strong differences in the properties of the central stellar source between a sample of optically selected class i and class ii sources in taurus . on the other hand , doppmann et al . ( 2005 ) argue that the white & hillenbrand ( 2004 ) results are biased by their optical selection of these class i young stellar objects ( ysos ) . doppmann et al . ( 2005 ) perform an extensive ir study of class i ysos in several star forming regions and conclude that , while there is a fair amount of spread in the stellar and accretion properties of these objects , the general paradigm of class i sources representing an earlier , higher accretion rate phase of stellar evolution relative to class ii sources is borne out ( see also prato et al . 2009 ) . some of the confusion and disagreement over the true nature of the class i ysos may be due to variability . it has been suggested that the bulk of a star s final mass is accreted through episodic events where the accretion rate through the disk increases by a factor of @xmath2 for some period of time ( e.g. , hartmann 1998 ) . these episodes of rapid disk accretion may be what we recognize as fu orionis events ( hartmann & kenyon 1996 ) , with these events occuring more frequently during the class i stage . as a result , class i objects should display a large range of accretion behavior , with some objects accreting at close to typical ctts rates , while others are accreting much more rapidly than this . qualitatively , such a picture matches the range of behavior found in these sources in recent studies ( doppmann et al . 2005 , white et al . 2007 , prato et al . 2009 ) . since class i ysos are often rapidly accreting material , the question arises as to how this process occurs . there is substantial evidence to show that fu ori outbursts are the result of very rapid disk accretion ( for a review see hartmann & kenyon 1996 ) . the evidence also suggests that when these objects are not in outburst , accretion onto the central protostar occurs through a disk with infalling material from the envelope piling up in the disk ( e.g. bell 1993 ) . such a scenario can explain the apparent low luminosity of some class i sources relative to what is expected if the infalling material from the envelope were to land initially on the central object ( kenyon et al . 1993 , 1994 ) . these observations of fu ori and lower luminosity class i ysos suggest that accretion onto the central source occurs primarily through a disk whether a particular yso is in a high or low accretion state . for the class ii sources ( cttss ) this accretion process appears to be well described by the magnetospheric accretion paradigm ( see bouvier et al . 2007 for a review ) , but it is currently unclear to what extent this model is appropriate for class i protostars . the magnetospheric accretion model is successful at explaining a number of obsverations of cttss . a key question in the study of these stars is to understand how they can accrete large amounts of disk material with high specific angular momentum , yet maintain rotation rates that are observed to be relatively slow ( e.g. hartmann & stauffer 1989 , edwards et al . this problem is solved in current magnetospheric accretion models by having the stellar magnetic field truncate the inner disk , typically near the corotation radius , and channel the disk material onto the stellar surface , most often at high stellar latitude . the angular momentum of the accreting material is either transferred back to the disk ( e.g. , knigl 1991 ; cameron & campbell 1993 ; shu et al . 1994 ) or is carried away by some sort of accretion powered stellar wind ( e.g. , matt & pudritz 2005 ) . greene and lada ( 2002 ) analyzed the stellar parameters and mass accretion rate of the class i source oph irs 43 and showed that these were consistent with magnetospheric accretion models provided the magnetic field on this source is on a order a kg in strength . covey et al . ( 2005 ) analyzed the rotational properties of class i sources and found that while they are rotating more rapidly than cttss on average , they are not rotating at breakup velocities . these observations could be interpreted in the standard magnetospheric accretion paradigm if the accretion rates of class i sources are larger on average than those of cttss . magnetospheric accretion naturally requires a strong stellar magnetic field . several ttss have now been observed to have strong surface magnetic fields ( basri et al . 1992 ; guenther et al . 1999 ; johns krull 2007 ; johns krull et al . 1999b , 2004 ; yang et al . 2005 , 2008 ) , and strong magnetic fields have been observed in the formation region of the emission line at 5876 ( johns krull et al . 1999a ; valenti & johns krull 2004 ; symington et al . 2005 ; donati et al . 2007 , 2008 ) , which is believed to be produced in a shock near the stellar surface as the disk material impacts the star ( beristain , edwards , & kwan 2001 ) . while a considerable amount is now known about the magnetic field properties of class ii ysos , almost nothing is known directly about the magnetic field properties of class i sources . while not a class i source , fu ori has recently shown evidence for a magnetic field in its disk , revealed through high resolution spectropolarimetry ( donati et al . 2005 ) ; however , there are currently no observations of magnetic fields on the surface of a class i protostar . this is in part due to their faintness and the need for substantial observing time on @xmath3 m class telescopes equipped with high resolution near ir spectrometers in order to obtain the necessary data . in order to begin to address the magnetic field properties of class i sources , we have begun an observational program to survey the magnetic field properties of several class i ysos in the @xmath4-ophiuchi star forming region . here , we report on our first field detection on the class i source wl 17 ( 2mass j16270677 - 2438149 , iso - oph 103 ) . this source has a rising ir sed ( wilking et al . 1989 ) with a spectral index of @xmath5 over the 2 24 @xmath6 m region ( evans et al . wl 17 has been detected in x - rays ( imanishi et al . 2001 ; ozawa et al . 2005 ) suggesting the star is magnetically active . the temperature ( @xmath7@xmath8 k ) and luminosity ( @xmath9 ) of wl 17 ( doppmann et al . 2005 ) give it a mass of @xmath10 and an age @xmath11 years using the tracks of siess et al . ( 2000 ) . in this paper , we look for zeeman broadening of k - band lines in high resolution spectra of wl 17 to diagnose its magnetic field properties . magnetic broadening is easiest to detect when other sources of line broadening are minimized , and the small rotation velocity of wl 17 ( @xmath12sin@xmath13 km s@xmath14 ; doppmann et al . 2005 ) is a great advantage for this work . muzerolle et al . ( 1998 ) derive an accretion luminosity of @xmath15 based on the br-@xmath16 line luminosity measurements of greene and lada ( 1996 ) . this accretion luminosity implies a mass accretion rate of @xmath17 yr@xmath14 using equation ( 8) of gullbring et al . ( 1998 ) with the disk truncation radius assumed to be at @xmath18 . while there are reasons to be concerned about this mass accretion rate estimate ( which are discussed in 4 ) , such a truncation radius implies a stellar field of about a kilogauss . detecting and measuring that field is the goal of the current work . in 2 we describe our observations and data reduction . the magnetic field analysis is described in 3 . in 4 we give a discussion of our results , and 5 summarizes our findings . spectra analyed here for magnetic fields on wl 17 come from two sources . keck nirspec data taken on ut 10 july 2001 is analyzed along with gemini phoenix data . the keck data have already been published by doppmann et al . ( 2005 ) and the reader is referred to that paper for observing and data reduction details . for reference , the resolution of the nirspec data is @xmath19 ( 16.7 km s@xmath14 ) . spectra of wl 17 were acquired on ut 03 april 2006 with the phoenix near - ir spectrograph ( hinkle et al . 2002 ) on the 8-m gemini south telescope on cerro pachon , chile . spectra were acquired with a @xmath20 ( 4-pixel ) wide slit , providing spectroscopic resolution @xmath21 ( 7.5 km s@xmath14 ) . the grating was oriented to observe the spectral range @xmath22 = 2.21942.2300 @xmath6 m in a single long - slit spectral order , and a slit position angle of @xmath23 was used . the seeing was approximately @xmath24 in @xmath25 band through light clouds , and wl 17 data were acquired in two pairs of exposures of 1200 s duration each . the telescope was nodded @xmath26 along the slit between integrations so that object spectra were acquired in all exposures for a total of 80 minutes of integration time on wl 17 . the b1v star hr 5993 was observed similarly but with shorter exposures for telluric correction of the wl 17 spectra . both wl 17 and hr 5993 were observed at similar airmasses , @xmath27 = 1.01 - 1.05 . observations of a continuum lamp were acquired for flat fielding . all data were reduced with iraf . first , pairs of stellar spectra taken at the two nod positions were differenced in order to remove bias , dark current , and sky emission . these differenced images were then divided by a normalized flat field . next , bad pixels were fixed via interpolation with the _ cosmicrays _ task , and spectra were extracted with the _ apall _ task . spectra were wavelength calibrated using low - order fits to 7 telluric absorption lines observed in each spectrum , and spectra at each slit position were co - added . instrumental and atmospheric features were removed by dividing wavelength - calibrated spectra of wl 17 by spectra of hr 5993 for each of the two slit positions . final spectra were produced by combining the corrected spectra of both slit positions and then normalizing the resultant spectrum to have a mean continuum flux of 1.0 . the most successful approach for measuring magnetic fields on late type stars in general has been to measure zeeman broadening of spectral lines in unpolarized light ( stokes @xmath28 e.g. , robinson 1980 ; saar 1988 ; valenti et al . 1995 ; johns krull & valenti 1996 ; johns krull 2007 ) . in the presence of a magnetic field , a given spectral line will split up into a combination of both @xmath29 and @xmath30 components . the @xmath29 components are linearly polarized parallel to the magnetic field direction ; and the @xmath30 components are circularly polarized when viewed along the magnetic field , and linearly polarized perpendicular to the field when viewed from that direction . the exact appearance of a line depends then on the details of the field strength and direction , even when viewed in unpolarized light . for any given zeeman component ( @xmath29 or @xmath30 ) , the splitting resulting from the magnetic field is @xmath31 where @xmath32 is the land-@xmath32 factor of the transition , @xmath33 is the strength of the magnetic field ( given in kg ) , and @xmath22 is the wavelength of the transition ( specified in ) . class i ysos are relatively rapid rotators ( e.g. , covey et al . 2005 ) compared to most main sequence stars and most ttss in which zeeman broadening has been detected , though wl 17 in particular has a relatively low @xmath12sin@xmath34 . equation ( 1 ) shows the broadening due to the zeeman effect depends on the second power of the wavelength , whereas doppler broadening due to rotation or turbulent motions depends on wavelength to the first power . thus , observations at longer wavelength are generally more sensitive to stellar magnetic fields . there are several lines in the k band which are excellent probes of magnetic fields in late - type stars ( e.g. saar & linsky 1985 ; johns krull 2007 ) , and here we observe 4 of them with nirspec : ( air wavelengths ) 2.22112 @xmath6 m with @xmath35 , 2.22328 @xmath6 m with @xmath36 , 2.22740 @xmath6 m with @xmath37 , and 2.23106 @xmath6 m with @xmath38 . we observe the first 3 of these with phoenix . in addition to the strongly zeeman sensitive lines , our wavelength settings also record a few additional atomic lines that are weaker and less zeeman sensitive ( lower land-@xmath32 values ) as well as the @xmath39 co bandhead at 2.294 @xmath6 m for the case of the nirspec data . the co lines are much less magnetically sensitive than the lines and provide a good check on other line broadening mechanisms . in order to measure the magnetic field properties of wl 17 , we directly model the profiles of several k band photospheric absorption lines . our spectrum synthesis code and detailed analysis technique for measuring magnetic fields using these k band lines is described elsewhere ( johns krull et al . 1999b ; johns krull et al . 2004 , yang et al . 2005 ) . here , we simply review some of the specific details relevant to the analysis presented here . in order to synthesize the stellar spectrum , we must first specify the atmospheric parameters : effective temperature ( @xmath7 ) , gravity ( @xmath40 ) , metallicity ( [ m / h ] ) , microturbulence ( ) , macroturbulence ( ) , and rotational velocity ( ) . rotational broadening in ysos is large compared to the effects of macroturbulence . this makes it difficult to solve for separately , so a fixed value of 2 km s@xmath14 is adopted here as it was in the above mentioned papers . valenti et al . ( 1998 ) found that microturbulence and macroturbulence were degenerate in m dwarfs , even with very high quality spectra . therefore , for the low turbulent velocities considered here , microturbulence is neglected , allowing to be a proxy for all turbulent broadening in the photosphere . while and can in principle have different effects of the spectral lines ( potentially affecting the line equivalent width ) at the relatively low resolution and signal - to - noise used here , the effect of the two mechanisms on the _ shape _ of the spectral lines is equivalent , and the corresponding broadening is significantly less than that due to the resolution or magnetic fields . any errors in the intrinsic line equivalent widths that result from an inaccurate value of can in principle be compensated for by small errors @xmath7 , @xmath40 , [ m / h ] , or the derived k band veiling . with the turbulent broadening specified , estimates are still needed for @xmath7 , @xmath40 , , and [ m / h ] for wl 17 . doppmann et al . ( 2005 ) find @xmath7@xmath41 k and a gravity of @xmath40@xmath42 for wl 17 . for the stellar atmosphere then , we take the model from a grid of nextgen " atmospheres ( allard & hauschildt 1995 ) which is equal in effective temperature ( 3400 k ) and closest in gravity ( @xmath40@xmath43 ) to the values determined for wl 17 . yang et al . ( 2005 ) , using the 4 lines covered in the nirspec data here , performed several tests of the magnetic analysis methods used here to see how sensitive the results are to small errors in the effective temperature and gravity assumed in the magnetic analysis . they find that a 200 k error in @xmath7 or 0.5 dex error in @xmath40 typically results in less than a 10% error in the derived magnetic field strength . therefore , we are confident that our particular choice for the stellar atmosphere will not lead to significant error in the magnetic field properties we estimate for wl 17 . finally , solar metallicity is often assumed for young stars , and this assumption is supported by the few detailed analyses that have been performed ( e.g. padgett 1996 ) . we assume solar metallicity here for wl 17 . with the above quantities specified , we can then synthesize spectra for our lines of interest using the polarized radiative code synthmag ( piskunov 1999 ) . the rotational broadening of wl 17 has been meaured by doppmann et al . ( 2005 ) , where they find @xmath44 km s@xmath14 . the analysis of doppmann et al . ( 2005 ) used the co bandhead data used here to measure . since the co lines are very insensitive to magnetic fields , we expect this to be an accurate estimate of the rotational broadening of wl 17 ; however , we let be a free parameter of our fits described below . as mentioned above , class i sources are often observed to have substantial k band veiling ( e.g. , greene & lada 1996 ) . veiling is an excess continuum emission which , when added to the stellar spectrum , has the effect of weakening the spectral lines of the star in continuum normalized spectra . near infrared veiling is assumed to arise from the disk around young stars . veiling is measured in units of the local stellar continuum , and doppmann et al . ( 2005 ) found a k band veiling of @xmath45 for wl 17 using the same nirspec data we use in part here . doppmann et al . ( 2003 ) showed that when magnetic fields are unaccounted for in spectroscopic analysis , the results can be somewhat biased . therefore , we choose to let the k band veiling be an additional free parameter of the spectral fitting performed here . in addition , we attempt to simultaneously fit both keck nirspec and gemini phoenix data which were observed at different times . cttss regularly show significant variations in their k band flux on timescales as short as a day ( and occasionally shorter ) , likely as a result of accretion variability ( skrutskie et al . 1996 , carpenter et al . 2001 , eiroa et al . additionally , barsony et al . ( 2005 ) have shown the the near - ir brightness of wl 17 is variable , suggesting the k band veiling of wl 17 may be variable . therefore , we separately solve for the k band veiling in the keck and gemini data . in previous studies of the magnetic field properties of ttss it was found that the zeeman sensitive lines could not be well fit with a single value of the magnetic field strength . instead , a distribution of magnetic field strengths provide a better fit ( johns krull et al . 1999b , 2004 ; yang et al . 2005 ; johns - krull 2007 ) . it was also found that fits to the spectra are degenerate in the derived field values unless the fit is limited to specific values of the magnetic field strength , separated by @xmath46 kg , which is the approximate magnetic resolution " of the data . while the nirspec data used here are slightly lower in resolution , the phoenix data are a bit higher in spectral resolution than that used in the studies cited above . therefore , we use the same limitations when fitting the spectra of wl 17 . we assume the star is composed of regions of 0 , 2 , 4 , and 6 kg magentic field , and we solve for the filling factor , @xmath47 , of each of these components subject to the constraint @xmath48 . the different regions are assumed to be well mixed over the surface of the star different components are not divided up into well defined spots or other surface features . the field geometry is assumed to be radial in all regions . another key assumption is that the temperature structure in all the field regions is assumed to be identical for the purpose of spectrum synthesis : the fields are not confined to cool spots or hot plage like regions . a final assumption we make here is that the photospheric magnetic field properties of the star are the same between the two observing epochs . this may or may not be a good assumption . substantial variability is seen in cttss , both photometrically and spectroscopically , which has been interpreted as rotational modulation of a non axisymmetric stellar magnetosphere ( e.g. , see bouvier et al . 2007 for a review ) . this certainly suggests variation of the field geometry above the star , but not necessarily as much variation of the photospheric field itself . very little is known about variations over timescales of months to years in the photospheric field properties of cttss ( and nothing is known regarding class i sources ) . two field measurements exist for bp tau ( johns krull et al . 1999b ; johns krull 2007 ) and for t tau ( guenther et al . 1999 ; johns krull 2007 ) , and for both stars , the mean field strengths recovered from the two epochs agree to within the quoted uncertainties . we therefore assume identical field properties between the two epochs for wl 17 and show below that this provides a good match to the data within the uncertainties . there are then 6 free parameters in our model fits : the k band veiling , @xmath49 , for each observing epoch ; the value of @xmath47 for the 2 , 4 and 6 kg regions ( @xmath50 , so the filling factor of the 0 kg field region is set once @xmath47 is determined for the other 3 regions ) ; and the value of . synthetic spectra are convolved with a gaussian of the appropriate width to represent the spectral resolution before comparison with the data . we have compared both calibration lamp lines and non - saturated telluric lines to gaussian fits and find that the line profiles are well matched by this assumed line shape . we are therefore confident that a gaussian is a good approximation for the instrumental profile . we solve for our 6 free parameters using the nonlinear least squares technique of marquardt ( see bevington & robinson 1992 ) to fit the model spectra to the observed spectra shown in figure [ profiles ] . in our first attempt , labelled f1 in table [ fitpars ] , the entire observed region shown in figure [ profiles ] is used in the fit . the parameters derived from all our fits are listed in table [ fitpars ] . in figure [ profiles ] we show the spectra of wl 17 in the regions of the zeeman sensitive lines and the co bandhead along with our best fitting model spectrum ( f1 ) . also included in the figure is a model with no magnetic field for comparison . the zeeman sensitive lines at 2.2211 , 2.2233 , and 2.2274 @xmath6 m ( land-@xmath51 , respectively ) are significantly broader in the phoenix data ( @xmath21 ) than is predicted by the model with no magnetic field . on the other hand , the width of the zeeman insensitive ( land-@xmath52 ) line at 2.2266 @xmath6 m is accurately predicted by both models due to its much weaker magnetic broadening . this suggests that the excess broadening seen in the lines is not due to an error in our assumed instrumental profile . in the lower resolution ( @xmath53 ) nirspec data , the lines again appear broader than predicted by the model with no magnetic field , though the higher veiling associated with that data makes the lines weaker which combined with the noise in the data makes the reality of the excess broadening less certain than in the gemini phoenix data . however , the nirspec data are fully consistent with the magnetic broadening clearly seen in the phoenix data . the mean field , @xmath54 , that we find for wl 17 is 2.9 kg . in the spectral regions used for this analysis there are some relatively strong telluric absorption lines . the spectra shown in figure [ profiles ] have been corrected for telluric absorption ; however , the regions affected by this absorption are likely more uncertain than regions not affected by such absorption . in order to test the sensitivity of our results to possible errors in the telluric correction , we increased the uncertainty of spectral regions where the telluric absorption lines went below 97% of the continuum during the fitting process . these regions are shown in the bold lines above the spectra in figure [ profiles ] . in these regions , we increased the uncertainty by a factor of 3 and reperformed the fit . the fit parameters derived in this way are also given in table [ fitpars ] as those derived from constraints f2 . the values of the fitted parameters are almost identical to those derived above ( constraints f1 ) . a second concern is that the numerical fitting method might be misled by small changes in @xmath55 due to the inclusion of large amounts of continuum regions on which the model has no real effect . therefore , we performed a third fit ( labelled f3 in table [ fitpars ] ) in which we eliminated much of the continuum and focussed down on the lines for fit constraints . the regions of the spectra used for this fit are shown in bold in figure [ profiles ] , and the fit parameters are again reported in table [ fitpars ] . for this fit , we maintained the uncertainty of the telluric affected regions at 3 times their nominal values . again , the fitted parameters are nearly identical to those determined in fits f1 and f2 . our detection of a mean field of @xmath56 kg on wl 17 is the first magnetic field measurement on a class i protostar . previous studies using k band data of comparable resolution and signal - to - noise level have shown that the field uncertainties are dominated by systematic effects associated with the choice of magnetic model ( johns krull et al . 1999b , 2004 ; yang et al . 2008 ) and possible errors in the stellar parameters used to model the star ( yang et al . 2005 ) . based on these studies , while the formal uncertainty in the mean field determination is quite low , we estimate the true uncertainty in our mean field measurement of be @xmath57 % . johns krull ( 2007 ) measured the mean magnetic field on a sample of 14 cttss in the taurus star forming region , finding field strengths which ranged from 1.1 to 2.9 kg , with a mean of 2.2 kg . yang et al . ( 2005 , 2008 ) measured the mean field on a total of 5 stars in the tw hydrae association ( twa ) finding values that range from 2.3 to 3.5 kg with a mean field of 3.0 kg . yang et al . ( 2008 ) find that this difference in mean field strength between the two samples is marginally significant , with the older stars ( twa ) having a larger field strength on average . however , yang et al . ( 2008 ) point out that the twa stars have smaller radii on average due to their older age ( @xmath58 myr compared to @xmath46 myr for taurus ) , and that the mean magnetic flux in the twa stars is actually smaller than that in the taurus stars . wl 17 has a field strength that is large relative to the taurus stars studied by johns krull ( 2007 ) and it also has a relatively large radius and corresponding high magnetic flux . on the other hand , wl 17 is in many ways similar to df tau , which has both a large radius ( 3.4 @xmath59 ) and a mean field of 2.9 kg , equal to that of wl 17 . observations of a statistically significant sample of class i sources will be required to see how their magnetic field properties compare as a group to older populations of class ii and iii ( diskless t tauri stars ) objects . our derived @xmath60 km s@xmath14 , with a formal uncertainty of @xmath61 km s@xmath14 , is consistent with the value of 12 km s@xmath14 reported by doppmann et al . the veiling we derive for wl 17 is quite different in the two epochs . for the nirspec data , we find @xmath62 . using the same nirspec data set , doppmann et al . ( 2005 ) quote a veiling value of @xmath63 , which includes a correction for a systematic effect seen in the best fit synthesis models to observations of mk standards . the measured veiling from doppmann et al . ( 2005 ) without the correction for the systematic effect was @xmath64 , based on fits to two wavelength regions containing strong lines of al , mg , and na . the co bandhead was the third wavelength region used in the doppmann et al . ( 2005 ) study , but was only used to derive the rotation value . as a result , there are actually no wavelength regions in common between this study and that of doppmann et al . ( 2005 ) for the purpose of determining @xmath49 . we do note that when using only the co bandhead region of wl 17 , doppmann et al . find a value of @xmath65 ( doppmann , private communication ) , though this region was not actually used in their final veiling determination . another difference between this study and that of doppmann et al . ( 2005 ) for the determination of the veiling is the inclusion of magnetic fields . all the atomic lines used in this analysis and that of doppmann et al . have some zeeman sensitivity . at the resolution of the nirspec data ( @xmath53 ) and for the strong , broad ( e.g. ) lines used by doppmann ( 2005 ) , magnetic fields primarily increase the equivalent width of the lines compared to models which do not include fields ( doppmann et al . 2003 ) . as a result , the somewhat stronger lines produced by magnetic models must be diluted by more veiling flux than that required for these same ( weaker ) lines produced by non - magnetic models in order to match a given set of observed line strengths . the veiling , @xmath49 , inferred from an observed spectrum will thus be larger when derived by comparison to magnetic models and smaller when derived by comparison to non - magnetic models . this is the effect we see when comparing the results here with those of doppmann et al . our veiling estimate of @xmath66 from nirspec is a little less than doppmann et al . find from the co bandhead alone , but a little larger than the @xmath67 they find from their atomic lines , though accounting for magnetic fields would likely bring their @xmath67 value up by some amount . we note that we use a single value of @xmath49 to fit both the co bandhead and the line region in the nirspec data ( lower two panels of figure [ profiles ] ) . some of the difference in @xmath49 found by doppmann et al . ( 2005 ) in different wavelength regions is likely due to model atmosphere and line data uncertainties ( see also a discussion of this in doppmann et al . 2005 ) . as a result , it is likely that our formal uncertainty of 0.1 on the veiling is too low , so we arbitrarily increase it by a factor of 3 and estimate a veiling for our nirspec data of @xmath68 . we adopt the same uncertainty for our phoenix data , which is at a comparable signal - to - noise , giving @xmath69 for this observation time . assuming the veiling differences quoted above result in a corresponding change to the k band source brightness , and that only the strength of the veiling continuum changed between these observations ( i.e. that the underlying star remained constant ) , wl 17 should have been a factor of @xmath70 brighter in the k band at the time of the keck nirspec observation relative to the gemini phoenix observation . interestingly , if we adopt the k band veiling correction of doppmann et al . ( 2005 ) , the individual veilings for each epoch are lowered , but the predicted flux ratio between the two observing epochs remains almost the same ( 3.4 instead of 3.5 ) . the k band flux factor variation of 3.5 calculated above corresponds to a variation of 1.36 magnitudes . there are relatively few studies of the near - ir variability of class i sources ; however , the few that exist suggest that while the implied k band variability of wl 17 is large , it may not be too extreme for such a source . kenyon and hartmann ( 1995 ) plot a histogram of the standard deviation for protostars in their sample that have two or more k band photometric measurements , finding values that reach as high as @xmath71 . in their study , park and kenyon ( 2002 ) find values of @xmath72 up to 0.52 . our two veiling meaurements give @xmath73 , while barsony et al . ( 1997 ) report @xmath74 for wl 17 on the basis of 6 different k band photometric measurements . barsony et al . ( 2005 ) tabulate @xmath49 variations for several sources in @xmath4 oph , and while none show quite as large a variation as we find for wl 17 , a few other show large @xmath49 variations with some ranging in value from 1 4 . barsony et al . ( 2005 ) also study the mid - ir variability of their sources , finding that wl 17 varies in 12.5 @xmath6 m flux by a factor of 2.4 which is similar in magnitude to the factor of 3.5 change in the k band brightness found here . values of @xmath75 are usually taken to indicate active accretion from a circumstellar disk , and large variable k band veiling such as that shown by wl 17 and several other sources in @xmath4 oph is usually taken as evidence that this accretion can be highly time variable ( e.g. barsony et al . the exact cause of this high degree of variability is not yet clear however . the veiling analysis and implied k band photometric variations described above assumes the underlying star remains constant ; however , it is likely that the star also possesses cool starspots which could contribute some k band variation due to rotational modulation . veiling is usually attributed to a source producing a featureless continuum . starspots themselves contribute very weakly to veiling variations since the spectrum of the spots contains many of the lines in present in the non - spotted photosphere . the potential effect of this can be estimated by measuring the veiling of a non - accreting t tauri star using another non - accreting star as a template . in the optical , this level of veiling is @xmath76 ( hartigan et al . 1991 ) and should be smaller in the k band since the spot quiet atmosphere contrast is much weaker . weakly or non - accreting t tauri stars do show k band brightness variations with a peak - to - peak amplitude @xmath77 magnitudes as a result of spots ( e.g. skrutskie et al . 1996 , carpenter et al . 2001 ) which is substantially less than the k band variations implied by our veliling measurements or observed by barsony et al . thus , the majority of the k band veiling variation here must be due to changes in the accretion properties of wl 17 . one of the motivations for this study is to see to what extent the magnetospheric accretion paradigm in place for class ii ysos may be applicable to class i sources . krull et al . ( 1999b ) give equations for predicting the stellar magnetic field strength required for 3 different prescriptions ( knigl 1991 , cameron & campbell 1993 , shu et al . 1994 ) of magnetospheric accretion theory . the work of knigl ( 1991 ) and shu et al . ( 1994 , see also camenzind et al . 1990 ) give the same scaling with stellar and accretion parameters : @xmath78 . the scaling from cameron and campbell ( 1993 ) is very similar : @xmath79 . using these equations to predict the field on wl 17 is uncertain due to difficulties with estimating the luminosity of wl 17 ( which affects the derived mass , radius , and accretion rate ) . luminosity estimates for wl 17 range from 1.8 @xmath80 ( doppmann et al . 2005 ) to 0.12 @xmath80 ( bontemps et al . additional uncertainties also affect the mass accretion rate , and the rotation period of wl 17 is unknown . assuming @xmath7@xmath81 k from doppmann et al . ( 2005 ) is a fairly robust estimate , we use this in combination with the two quoted stellar luminosities to derive the quantities needed to predict the stellar field strength from magnetospheric accretion theory . these stellar properties are reported in table [ magaccretion ] along with the field predictions for the studies mentioned above . we note that the stellar luminosity from bontemps et al . ( 2001 ) in combination with the 3400 k effective temperature would give wl 17 an age of @xmath82 myr using the pre - main sequence tracks of siess et al . such an age would be unusual for a class i source . the doppmann et al . ( 2005 ) stellar luminosity may be more accurate than the bontemps et al . ( 2001 ) value . this is because the former was derived by using photometric measurements to estimate and correct for the extinction and veiling seen toward the photosphere of the spectroscopically determined effective temperature , while the latter was determined by de - reddening near - ir photometry of wl 17 to intrinsic ctts colors and assuming an intrinsic j band flux to stellar luminosity relationship for a typical ctts . the doppmann et al . ( 2005 ) approach corrects for the veiling and effective temperature measured specifically for wl 17 , while the bontemps et al . ( 2001 ) technique does not . nevertheless , we report magnetospheric accretion estimates based on both values of the stellar luminosity in order to illustrate the sensitivity of the expected fields to various stellar parameters , notably the luminosity . in addition to issues related to the correct luminosity for wl 17 , there is additional uncertainty regarding the accretion rate estimate . the value of @xmath83 yr@xmath14 quoted in 1 is based on the accretion luminosity estimate of muzerolle et al . ( 1998 ) which is in turn based on the br-@xmath16 line luminosity estimate from greene and lada ( 1996 ) . a major concern in this process is the extinction correction used to recover the br-@xmath16 line luminosity . for example , greene and lada ( 1996 ) corrected their data for wl 17 back to the ctts locus in the jhk color color diagram , not necessarily back to the stellar photosphere . as an example of the sensitivity to the details of extinction corrections and the photometric data used , we note that doppmann et al . ( 2005 ) also compute br-@xmath16 line luminosities ( their figure 11 ) . these authors de - redden the h - k color to an intrinsic value of 0.6 and then add in a correction for scattered light based on the models of whitney et al . this results in a br-@xmath16 line luminosity of @xmath84 ( doppmann , private communication ) , which in turn gives an accretion luminosity of 2.7 @xmath80 using the muzerolle et al . ( 1998 ) relationship . calculating the mass accretion rate as given in the introduction results is a value of @xmath85 yr@xmath14 . we include the field estimates resulting from this accretion rate in table [ magaccretion ] , and we note that such an accretion rate is about the level needed to accrete a 0.5 @xmath86 star in a @xmath87 yr . this large accretion rate estimate for a class i source is also supported by the estimate of @xmath88 yr@xmath14 found by greene and lada ( 2002 ) for oph irs 43 . obviously , the accretion rate and implied magnetic field are fairly sensitive to the details of the br-@xmath16 line luminosity calculation . we therefore recomputed the br-@xmath16 line luminosity from the measured equivalent width value of 4.3 ( doppmann et al . 2005 ) , the 2mass photometry ( skrutskie et al . 2006 ) , and correcting for extinction by de - reddening the jhk colors to the ctts locus and correcting for an extra @xmath89 mag ( see doppmann et al . we also used a distance of 135 pc to the @xmath4 oph cloud ( mamajek 2008 ) . this produced a br-@xmath16 line luminosity of @xmath90 , about 2.5 times higher than that given by greene & lada due mostly to the more recent photometry and extinction correction technique . this new line luminosity indicates an accretion luminosity of 0.9 @xmath80 from the muzerolle et al . ( 1998 ) relationship , which implies a mass accretion rate of @xmath91 yr@xmath14 , which is well bounded by the accretion rates given in table [ magaccretion ] . all of the above discussion implicitly assumes the muzerolle et al . ( 1998 ) br-@xmath16 accretion luminosity relationship holds for class i sources ; however , this relationship was derived based on a sample of class ii objects . if there are any systematic differences between these and class i sources , the resulting accretion rate will be in error . for example , if the br-@xmath16 line is more optically thick in class i sources due to systematically higher accretion rates , the accretion rate we derive above for wl 17 will be lower than the true value . this suggests that the derived accretion rates for wl 17 may be too low . the rotation period reported in table [ magaccretion ] is an upper limit based on the stellar radius and measured . if the rotation period is actually shorter , the derived magnetic field values will be smaller . in this sense , the values reported in the table are an upper limit depending on the inclination of the source . the field strengths reported in table [ magaccretion ] are those corresponding to the equatorial field strength for an assumed dipolar field geometry . the polar field strength in such a geometry is twice this value . the mean field of 2.9 kg found for wl 17 is well above the predicted values for the larger luminosity found by doppmann et al . ( 2005 ) ; while for the lower luminosity of bontemps et al . ( 2001 ) , the measured field value may not be strong enough , particularly if the field is not dominated by the dipole component . in most ttss , the dipole component is found to be weak ( johns krull et al . 1999a ; daou et al . 2006 ; yang et al . 2007 ; donati et al . it is important to note that the data presented here only probe the photospheric field strength , while providing few constraints on the field geometry . high spectral resolution near - ir circular spectropolarimetry will likely be required to explore the field geometry on class i ysos such as wl 17 . in summary , the magnetic field we measure on wl 17 is roughly consistent with predicted magnetic field strength required for magnetospheric accretion . however , detailed quantitative comparisons are greatly hampered due to a number of uncertainties related to other relevant stellar parameters that are currently difficult to estimate for class i sources ( see also the discussion in prato et al . class i sources by their nature are deeply embedded and therefore suffer substantial extinction . uncertainties involved with the proper way to de - redden observations of these sources , combined with variable accretion luminosity ( continuum and line ) , a general lack of measured rotation periods for class i objects , and uncertainties in the methods used to measure accretion rates suggest that much work is still left to do before the magnetospheric accretion paradigm can be firmly established or refuted for class i sources . on the sun and active stars , it is expected that magnetic flux tubes are confined at photospheric levels by the gas pressure in the external non - magnetic atmosphere . for example , spruit & zweibel ( 1979 ) computed flux tube equilibrium models , showing that the magnetic field strength is expected to scale with gas pressure in the surrounding non - magnetic photosphere . similar results were found by rajaguru et al . field strengths set by such pressure equipartition considerations appear to be observed in active g and k dwarfs ( e.g. saar 1990 , 1994 , 1996 ) and possibly in m dwarfs ( e.g. johns krull & valenti 1996 ) . class i ysos have relatively low surface gravities and hence low photospheric gas pressures , so that equipartition flux tubes would have relatively low magnetic field strengths compared to cool dwarfs . the maximum field strength allowed for a confined magnetic flux tube is @xmath92 where @xmath93 is the gas pressure at the observed level in the stellar atmosphere . here , we take as a lower limit in the atmosphere ( upper limit in pressure ) the level where the local temperature is equal to the effective temperature ( 3400 k ) in the nextgen models of the appropriate gravity . this is the approximate level at which the continuum forms , with the lines forming over a range of atmospheric layers above this level at lower pressure . the values of @xmath94 are given in table [ magaccretion ] . the mean field we measure for wl 17 is well above the value of @xmath94 for either assumed luminosity and resulting gravity . this suggests that pressure equipartition does not hold in the case of class i ysos , and it also suggests the gas pressure in the atmospheres of these young stars is dominated by their magnetic fields . indeed , our fit for the magnetic field of wl 17 has only 3% of the surface as field free ; however , the uncertainty on the filling factor of this component is such that this is not a significant measurement . it may well be that the entire surface of wl 17 is covered with strong magnetic fields . what is certain is that the field we measure is too strong to be accounted for by current models of dynamo action on fully convective stars , where the field strength is equal to the equipartition value ( bercik et al . 2005 , chabrier & kker 2006 , dobler et al . 2006 ) and the filling factor of this field is typically very low ( cattaneo 1999 , bercik et al . rapid rotation can enhance field production and may help organize magnetic fields into large scale dipolar or quadrapolar geometries ( e.g. dobler et al . 2006 , brown et al . 2007 ) ; however , it is not yet clear that such models can produce mean fields on young stars in the range of @xmath95 kg . stellar magnetic fields give rise to activity " in late - type stars . activity is typically traced by line emission or broad band emission at high energy wavelengths such as x - rays . pevtsov et al . ( 2003 ) find an excellent correlation between the x - ray luminosity , @xmath96 , and magnetic flux in solar active regions and dwarf type stars with each of these two quantities ranging over almost 11 orders of magnitude . using the x - ray luminosity magnetic flux relationship of pevtsov et al . ( 2003 ) with our mean magnetic field measurement of 2.9 kg for wl 17 , we can predict @xmath96 for this yso if the radius is known . in table [ magaccretion ] we report the predicted x - ray luminosity values for the two stellar radii implied by the different stellar luminosity estimates of bontemps et al . ( 2001 ) and doppmann et al . wl 17 has been detected in x - rays by chandra ( imanishi et al . 2001 ) and by xmm ( ozawa et al . 2005 ) , and possibly by asca , though crowding was an issue for this latter observation ( kamata et al . the chandra observation gives an x - ray luminosity of @xmath97 erg s@xmath14 while the xmm observation found @xmath98 erg s@xmath14 . both values are lower than either of the predicted values in table [ magaccretion ] , though the xmm value of ozawa et al . ( 2005 ) is not much lower than the lower estimate which is based on the bontemps et al . ( 2001 ) luminosity for wl 17 . we do note though that the pevtsov et al . ( 2003 ) x - ray luminosity magnetic flux relationship is based on observed x - ray emission outside obvious flares : the so - called quiescent emission level . the xmm x - ray lightcurve of wl 17 presented by ozawa et al . ( 2005 ) is dominated by a strong , long lasting flare . thus , the quiescent x - ray emission of wl 17 is likely significantly weaker than the predicted x - ray emission , independent of the exact value of the the stellar radius for wl 17 . pevtsov et al . ( 2003 ) found evidence that ttss were underluminous in x - rays relative to the relationship defined by dwarf stars , and this result appears to hold for a majority of these stars ( johns krull 2007 ; yang et al . 2008 ) . in that sense , wl 17 follows the same pattern seen in most class ii sources . johns krull ( 2007 ) suggested that this might be due to the fact that the magnetic pressure at the surface of these stars is so strong . perhaps the ubiquitous , strong fields of these young stars inhibit ( relative to dwarf stars ) photospheric gas motions from moving flux tubes around and building up magnetic stresses which eventually results in coronal heating . alternatively , wl 17 is deeply embedded and may suffer substantial x - ray attenuation , though this in nominally accounted for in the x - ray measurements . we have measured the magnetic field on the class i source wl 17 , and we find the star is largely covered by strong magnetic fields . the surface averaged mean field on the star is @xmath0 kg . comparing this field with predictions from magnetospheric accretion theory or with x - ray measurements depends fairly sensitively on the value of the stellar radius appropriate for wl 17 . for the relatively large luminosity and associated stellar radius from doppmann et al . ( 2005 ) , the measured field values are more than strong enough to be consistent with magnetospheric accretion ; however , for the lower radius implied by the bontemps et al . ( 2001 ) luminosity , the field may in fact be too weak . for either radius , the measured fields are stronger than expected by pressure balance arguments for magnetic flux tubes confined by a non - magnetic atmosphere , suggesting that the star is likely fully covered by fields . in addition , independent of the radius assumed , the measured x - ray emission is lower than is to be expected based on correlations established between these two quantities on the sun and late - type dwarf stars . we are pleased to acknowledge numerous , stimulating discussions with j. valenti on all aspects of the work reported here . we also acknowledge many useful comments and suggestions from an anonymous referee . cmj - k wishes to acknowledge partial support for this research from the nasa origins of solar systems program through grant numbers nag5 - 13103 and nng06gd85 g made to rice university . tpg acknowledges support from nasa s origins of solar systems program via wbs 811073.02.07.01.89 . krc acknowledges support for this work by nasa through the spitzer space telescope fellowship program , through a contract issued by the jet propulsion laboratory , california institute of technology under a contract with nasa . this work made use of the simbad reference database , the nasa astrophysics data system , and vald line database . allard , f. & hauschildt , p.h . 1995 , in bottom of the main sequence and beyond , ed . tinney ( berlin : springer ) , 32 basri , g. , marcy , g. w. , & valenti , j. a. 1992 , apj , 390 , 622 barsony , m. , kenyon , s. j. , lada , e. a. , & teuben , p. j. 1997 , , 112 , 109 barsony , m. , ressler , m. e. , & marsh , k. a. 2005 , , 630 , 381 bell , k. r. 1993 , ph.d . thesis , university of california , santa cruz beristain , g. , edwards , s. , & kwan , j. 2001 , , 551 , 1037 bercik , d. j. , fisher , g. h. , johns - krull , c. m. , & abbett , w. p. 2005 , , 631 , 529 bertout , c. , basri , g. , & bouvier , j. 1988 , , 330 , 350 bevington , p. r. & robinson , d. k. 1992 , data reduction and error analysis for the physical sciences , ( new york : mcgraw hill ) bontemps , s. , et al . 2001 , , 372 , 173 bouvier , j. , alencar , s.h.p . , harries , t.j . , johns krull , c.m . , & romanova , m.m . 2007 , in protostars and planets v , b. reipurth , d. jewitt , & k. keil ( eds . ) , ( tucson : univ . of arizona press ) brown , b. p. , browning , m. k. , brun , a. s. , miesch , m. s. , nelson , n. j. , & toomre , j. 2007 , unsolved problems in stellar physics : a conference in honor of douglas gough , 948 , 271 carpenter , j. m. , hillenbrand , l. a. , & skrutskie , m. f. 2001 , , 121 , 3160 casali , m. m. & matthews , h. e. 1992 , mnras , 258 , 399 cattaneo , f. 1999 , , 515 , l39 chabrier , g. , kker , m. 2006 , , 446 , 1027 cameron , a. c. & campbell , c. g. 1993 , , 274 , 309 covey , k. r. , greene , t. p. , doppmann , g. w. , & lada , c. j. 2005 , , 129 , 2765 daou , a. g. , johns krull , c. m. , & valenti , j. a. 2006 , , 131 , 520 dobler , w. , stix , m. , & brandenburg , a. 2006 , , 638 , 336 donati , j .- f . , et al . 2007 , , 380 , 1297 donati , j .- f . , 2008 , , 386 , 1234 donati , j .- f . , paletou , f. , bouvier , j. , & ferreira , j. 2005 , , 438 , 466 doppmann , g. w. , greene , t. p. , covey , k. r. , & lada , c. j. 2005 , , 130 , 1145 doppmann , g. w. , jaffe , d. t. , & white , r. j. 2003 , , 126 , 3043 edwards , s. , hartigan , p. , ghandour , l. , & andrulis , c. 1994 , , 108 , 1056 evans , n. j. , et al . 2009 , , 181 , 321 greene , t. p. & lada , c. j. 1996 , aj , 112 , 2184 greene , t. p. & lada , c. j. 2000 , aj , 120 , 430 greene , t. p. & lada , c. j. 2002 , aj , 124 , 2185 guenther , e. w. , lehmann , h. , emerson , j. p. , & staude , j. 1999 , a&a , 341 , 768 gullbring , e. , hartmann , l. , briceo , c. , & calvet , n. 1998 , apj , 492 , 323 hartigan , p. , kenyon , s. j. , hartmann , l. , strom , s. e. , edwards , s. , welty , a. d. , & stauffer , j. 1991 , , 382 , 617 hartmann , l. 1998 , accretion processes in star formation ( cambridge university press ) hartmann , l. & kenyon , s. j. 1996 , araa , 34 , 207 hartmann , l. & stauffer , j. r. 1989 , aj , 97 , 873 hinkle , k. h. , blum , r. , joyce , r. r. , ridgway , s. t. , rodgers , b. , sharp , n. , smith , v. , valenti , j. , & van der bliek , n. 2002 , , 4834 , 353 imanishi , k. , koyama , k. , & tsuboi , y. 2001 , , 557 , 747 johns krull , c. m. 2007 , , 664 , 975 johns krull , c. m. & valenti , j. a. 1996 , , 459 , l95 johns krull , c. m. , valenti , j. a. , hatzes , a. p. , & kanaan , a. 1999a , 510 , l41 johns krull , c. m. , valenti , j. a. , & koresko , c. 1999b , apj , 516 , 900 johns krull , c. m. , valenti , j. a. , & saar , s. h. 2004 , , 617 , 1204 kamata , y. , koyama , k. , tsuboi , y. , & yamauchi , s. 1997 , , 49 , 461 kenyon , s. j. , calvet , n. , & hartmann , l. 1993 , , 414 , 676 kenyon , s. j. , gomez , m. , marzke , r. o. , & hartmann , l. 1994 , , 108 , 251 kenyon , s. j. , & hartmann , l. 1995 , , 101 , 117 knigl , a. 1991 , apj , 370 , l39 lada , c. j. 1987 , star forming regions , 115 , 1 matt , s. , & pudritz , r. e. 2005 , , 632 , l135 mamajek , e. e. 2008 , astronomische nachrichten , 329 , 10 muzerolle , j. , calvet , n. , & hartmann , l. 1998 , apj , 492 , 743 ozawa , h. , grosso , n. , & montmerle , t. 2005 , , 438 , 661 padgett , d. l. 1996 , , 471 , 847 park , s. , & kenyon , s. j. 2002 , , 123 , 3370 pevtsov , a. a. , fisher , g. h. , acton , l. w. , longcope , d. w. , johns - krull , c. m. , kankelborg , c. c. , & metcalf , t. r. 2003 , , 598 , 1387 piskunov , n. 1999 , astrophysics and space science library , 243 , 515 prato , l. , lockhart , k. e. , johns krull , c. m. , & rayner , j. t. 2009 , , 137 , 3931 rajaguru , s. p. , kurucz , r. l. , & hasan , s. s. 2002 , , 565 , l101 saar , s. h. 1990 , iau symp . 138 , solar photosphere : structure , convection , and magnetic fields , ed . j. o. stenflo ( dordrecht : kluwer ) , 427 saar , s. h. 1994 , iau symp . 154 , infrared solar physics , eds . d. m. rabin et al . ( dordrecht : kluwer ) , 493 saar , s. h. 1996 , iau symp . 176 , stellar surface structure , eds . k. g. strassmeier & j. l. linsky ( dordrecht : kluwer ) , 237 saar , s. h. , & linsky , j. l. 1985 , , 299 , l47 shu , f. h. , najita , j. , ostriker , e. , wilkin , f. , ruden , s. , & lizano , s. 1994 , apj , 429 , 781 siess , l. , dufour , e. , & forestini , m. 2000 , , 358 , 593 skrutskie , m. f. , et al . 2006 , , 131 , 1163 skrutskie , m. f. , meyer , m. r. , whalen , d. , & hamilton , c. 1996 , , 112 , 2168 spruit , h. c. , & zweibel , e. g. 1979 , , 62 , 15 symington , n. h. , harries , t. j. , kurosawa , r. , & naylor , t. 2005 , , 358 , 977 valenti , j. a. & johns krull , c. m. 2004 , ap&ss , 292 , 619 valenti , j. a. , marcy , g. w. , & basri , g. 1995 , , 439 , 939 valenti , j. a. , piskunov , n. , & johns - krull , c. m. 1998 , , 498 , 851 white , r. j. & hillenbrand , l. a. 2004 , apj , 616 , 998 white , r. j. , greene , t. p. , doppmann , g. w. , covey , k. r. , & hillenbrand , l. a. 2007 , protostars and planets v , 117 whitney , b. a. , kenyon , s. j. , & gomez , m. 1997 , , 485 , 703 wilking , b. a. , lada , c. j. , & young , e. t. 1989 , , 340 , 823 yang , h. , johns krull , c. m. , & valenti , j. a. 2005 , , 635 , 466 yang , h. , johns krull , c. m. , & valenti , j. a. 2007 , , 133 , 73 yang , h. , johns krull , c. m. , & valenti , j. a. 2008 , , in press lccccccccccc 3400&1.8&3.9&0.31&16.9&@xmath99&0.72&0.33&0.85&2.7&1.06 & 29.0 + 3400&0.12&1.0&0.29&4.33&@xmath100&4.43&1.74&5.22&3.9&1.44 & 1.3 + 3400&1.8&3.9&0.31&16.9&@xmath101&2.20&1.18&2.60&2.7&1.06 & 29.0 + |
a compact , highly efficient single - particle counting detector for ions of kev / u kinetic energy , movable by a long - stroke mechanical translation stage , has been developed at the max - planck - institut fr kernphysik ( max planck institute for nuclear physics , mpik ) . both , detector and translation mechanics , can operate at ambient temperatures down to k and consist fully of ultra - high vacuum ( uhv ) compatible , high - temperature bakeable and non - magnetic materials . the set - up is designed to meet the technical demands of mpik s cryogenic storage ring ( csr ) . we present a series of functional tests that demonstrate full suitability for this application and characterise the set - up with regard to its particle detection efficiency . | 1411.4530 | in the field of ion storage rings dedicated to molecular and atomic physics experiments , the electrostatic cryogenic storage ring ( csr ) @xcite of the max planck institute for nuclear physics ( mpik ) in heidelberg , germany , bridges the gap between traditional medium - energy magnetic heavy - ion synchrotrons , like tsr @xcite , astrid @xcite or cryring @xcite , and smaller low - energy electrostatic storage rings . most of the latter are off - springs of the elisa facility @xcite and target very low ion energies around 20 kev @xcite . the most advanced designs , like the recently commissioned desiree facility in stockholm @xcite , the future riken storage ring in saitama @xcite , and also the above mentioned csr , feature beam guiding vacuum systems that can be cooled to deeply cryogenic temperatures , which results in very low residual gas density and very long storage times@xcite . at a maximum kinetic energy of 300 kev per unit charge , ions stored in the csr are still low energetic compared to traditional magnetic storage rings . nevertheless , ions with charge - to - mass ratios up to 1/160 e / u will have storage velocities high enough to allow phase - space cooling using an electron cooler @xcite . consequently , csr will be the first electrostatic storage ring to enable ion - neutral , ion - photon , and ion - electron collision experiments at high precision using cooled , low - emittance stored beams . collision experiments with stored ions generally produce daughter beams of different ion - optical rigidity compared to their parent , which leave the closed orbit of the storage ring at the next bending element @xcite . the daughter particles can e.g. be charge - changed products from electron collisions of highly - charged atomic ions , neutral products from recombination of singly charged cations , or fragments from dissociation of stored molecular or cluster ions . in the case of low - emittance ion beams as stored in an electron cooler ring , movable , single - particle counting detectors of narrow aperture are a well - established way to selectively and sensitively measure the production rates of specific daughters @xcite . the bending elements of the storage ring then act as mass spectrometers that separate different product beams according to their charge - to - mass ratio @xmath1 and , thereby , allow to determine the corresponding reaction cross - sections . in the case of csr , such a detector has to meet additional technical criteria imposed by the design of the storage ring . a first limitation arises from the characteristic storage energy of csr . among its prime experimental targets are singly - charged molecular and weakly - charged atomic cations . given the maximum kinetic energy of 300 kev per unit charge , limited by the deflection fields of the csr bending electrodes , the specific kinetic energies of stored ions and of daughter beams produced in collision experiments can thus be as low as a few kev / u . this can easily result in penetration depths in solids as short as 100 nm @xcite and , hence , excludes any detection principle that requires the particles to traverse significant layers of passive material before reaching the detection volume , as is notably the case for surface - barrier semiconductor counters . second , the detector has to operate in the rather harsh csr environment . for the above mentioned advantage of extremely good vacuum conditions , but also in order to enable experiments on rotationally - ground - state molecular ions @xcite , csr is an all - cryogenic storage ring , where the beam - guiding vacuum chamber as well as all ion beam optics are cooled to @xmath0 k by a superfluid helium circuit @xcite . both , the detection principle and the mechanics enabling free positioning of the detector at the output of a csr bending element , must work reliably at this low temperature . in addition , in order to achieve uhv vacuum conditions ( @xmath2 mbar ) also in room or intermediate temperature operation , the beam pipe and all beam - facing experimental equipment are required to be bakeable to 250@xmath3c . in cryogenic mode , a room - temperature equivalent pressure of @xmath4 mbar is then achieved @xcite . last , in order to enable the csr to store ions of very low rigidity , e.g. low - energetic protons or light molecular ions like h@xmath5 or h@xmath6 , disturbances of the ion trajectories induced by stray magnetic fields must be kept at a minimum . all materials used for equipment installed in vicinity of the stored - beam orbit are therefore required to be of very low relative magnetic permeability @xmath7 . in this paper , we present a movable single - particle detector developed at mpik for counting of reaction products from ion collision experiments at csr . in sect . [ design ] we discuss the technical design of the set - up , consisting of the actual particle sensor and the cryogenic translation mechanism . in sect . [ tests ] we characterise the detector with regard to its counting efficiency and reliability of operation in a cryogenic vacuum environment . in sect . [ commissioning ] we report the the first operational test at csr in room temperature mode . section [ summary ] provides a summary and outlook onto possible future upgrades of the set - up . the detector ( ( 4b ) in fig . [ csr ] ) is installed behind the 6@xmath3 bending electrode of the storage ring following the straight section dedicated to the future electron cooler@xcite of csr . limited by the geometry of the beam - guiding vacuum chamber , the detector is , in theory , able to collect daughter particles in the range @xmath8 , where @xmath9 is the relative difference in charge - to - mass ratio @xmath1 of the daughter ( index d ) compared to the stored parent beam ( index p ) . for the experimentally important case of a positively charged atomic parent beam , the range @xmath10 is most relevant , with the lower and upper boundaries corresponding , respectively , to full neutralisation or ionisation to double charge of the stored parent . the general detector concept was based on the following considerations : 1 . detectors relying on surface secondary electron ejection and subsequent multiplication have been shown to achieve very good quantum efficiency for ions in the specific energy range of a few kev / u @xcite . as noted above , detection principles involving non - sensitive layers that have to be penetrated by the particles before these reach the counting volume are prohibited by the low ion energies at the csr . micro - channel plates ( mcps ) were considered to be the only compact electron multipliers known to work reliably at temperatures of 20 k and below@xcite . especially the so - called ` extended dynamic range ' ( edr ) variants were expected to perform quite well in cryogenic environment thanks to their lower electric resistance . single - channel electron multipliers ( cems ) were expected to cause problems as their resistance rises exponentially towards low temperatures , causing electron depletion of their single channel wall at high count rates . this assumption turned out to be partly unjustified , as an edr cem was successfully tested at 25 k within the test series described in sect . [ testcryo ] . still , the proven edr mcp back end was chosen as electron multiplier in the here reported work . 3 . in order to be able to detect a wide spectrum of possible daughter beams in the csr , the detector must be movable across the diameter of the beam - guiding vacuum chamber , which is approximately 300 mm wide . the standard technique used to move mechanical components across long travel ranges inside uhv chambers are linear actuators , driven from the atmosphere side of the set - up and sealed by long edge - welded bellows . manufacturers provide no data on their performance at cryogenic temperatures , but excessive compression or expansion of the bellows at low temperatures is explicitly not recommended as embrittlement of the stainless steel welding joints could lead to the formation of vacuum leaks . at the desiree project , this problem has been addressed by mounting movable detectors onto sleds that are actuated by stepper motors directly inside the vacuum cryostat . while the magnetic fields of these motors might have been tolerable at csr as well , provided a certain distance to the stored beam orbit could have been maintained , the electric power used for their operation also leads to significant heating and disturbs the cryogenic environment @xcite . hence , it was decided to move the csr detector using a cryogenic worm drive operated from the atmosphere side of the set - up . to this end , a rotary uhv feed - through specified for cryogenic temperatures could be obtained commercially , as described later . the csr vacuum system consists of two distinct vacuum vessels : an outer isolation vacuum chamber and enclosed by the latter an inner cryogenic beam pipe ( cf . [ csr ] ) . rectangular , 80@xmath11200 mm@xmath12 , vacuum ports along the beam line are foreseen for mounting particle detectors off - beam - axis at csr ( cf . figs . [ csr ] and [ sensor ] ) . the first operational specimen that is described here is affectionately named ` compact ' , the ` cold movable particle counter ' , by its developers . the beam - facing components of the detector , including all mechanical support and all electric connections , were designed to fit entirely through one of the rectangular flanges ( ( 4b ) in fig . [ csr ] and ( 9 ) in fig . [ sensor ] ) . further components in the isolation vacuum chamber are necessary to mechanically and electrically operate the detector from the atmosphere side of the set - up . these had to be designed with low thermal conductance in mind in order not to inflict excessive heat load to the 10-k cold stage of the csr cryostat . ( colour online ) schematic view of the storage ring csr and close - up of the detector position . ions are injected by fast switching of a 6@xmath3 bending electrode ( 1 ) . following the electron cooler ( 2 ) , the third quadrant of the ring ( 3 ) features four vacuum ports for the installation of particle detectors ( 4a4b ) . these ports are placed in - between the 6@xmath3 and 39@xmath3 bending electrodes ( 5 and 6 ) , which act as charge - to - mass selectors that separate daughter beams ( dotted red lines ) from the stored ions ( solid red ) . the here described detector is situated at position ( 4b).,width=302 ] ( colour online ) the particle sensor of the detector . the top left and right panels show schematic side and front views , respectively . particles ( denoted by a@xmath13 ) enter the sensor through the 20@xmath1150 mm@xmath12 entrance window ( 1 ) and impact onto a curved converter cathode ( 2 ) ( biased at @xmath14 v ) where they release secondary electrons ( e ) . the latter are collected and multiplied by the mcp stack ( 3 ) ( biased at @xmath15 v and @xmath16 kv on the input and output sides , respectively ) . the resulting electron pulse is collected on an anode ( 4 ) ( @xmath17 kv ) from which the counting signal is decoupled . the mcp - anode stack is equipped with a resistive heating module ( 5 ) . a grounded housing ( 6 ) surrounds the electrodes . the upper right panel ( 7 ) shows the simulated secondary electron collection efficiency as a function of the particle impact position in the plane of the entrance window . the lower left ( 8) and right ( 9 ) panels show , respectively , photographic images of the sensor internals and of the fully assembled uhv set - up including the translation stage.,width=302 ] the particle sensor is shown in fig . [ sensor ] . it consists of an aluminium secondary - electron converter cathode ( ( 2 ) in fig . [ sensor ] ) , a micro - channel plate and anode stack ( ( 3 ) and ( 4 ) in fig . [ sensor ] ) , and a grounded housing encapsulating the latter two ( ( 6 ) in fig . [ sensor ] ) . incoming heavy particles ( labelled a@xmath13 in fig . [ sensor ] ) hit the converter cathode , biased at a potential of @xmath14 v , and produce secondary electrons . these are accelerated towards the input side of the mcp stack , biased at @xmath15 v. the initially slow ( @xmath18 ev ) secondary electrons thus impinge onto the mcp at a kinetic energy of 1.2 kev . the mcp stack consists of two edr channel plates ( _ photonis _ , type 18/12/10/12d40:1 edr , ms ) of 18 mm useful diameter , assembled in chevron configuration . a potential difference of 1.7 to 2.0 kv can be applied across both plates for electron multiplication . on the output side of the mcps , a stainless steel anode , biased at @xmath17 kv with respect to ground , collects the electron bunches . each primary incident particle generally releases several secondary electrons @xcite . the overall detection probability is hence not limited to the open area ratio of the mcp surface ( @xmath1960% ) , as would be the case for direct heavy particle impact onto the channel plate . in fact , it can reach unity for particles that release many secondaries ( cf . [ testpulse ] ) . a positive side effect of the secondary - electron detection scheme is that the mcp surface is not exposed to direct impact of high - energy massive particles , which could lead to sputtering damage at high flux densities . traditionally , charged particle detectors in storage rings have been equipped with apertures a little larger than the expected daughter beam diameter ( up to a few mm ) and are movable across the storage beam pipe in the two directions transverse to the ion beam axis . given the difficulties arising from the csr environment , motion of the sensor along the vertical axis was not considered . instead , a vertically - elongated sensitive aperture ( ( 1 ) in fig . [ sensor ] ) of @xmath20 mm@xmath12 acceptance was chosen . a highly efficient ion detector , based on secondary - electron emission , has been described in detail by rinn et al.@xcite , and has been used successfully at mpik for many years . their design is suitable for a @xmath18-cm - diameter circular entrance window . for our purpose , that concept had to be extended towards the above described 20@xmath1150 mm@xmath12 aperture , which is not only larger in spite of the similar overall size of the device , but also lacks any cylindrical symmetry in the secondary - electron collector optics . moreover , the sensor has to be enclosed in a grounded housing to prevent disturbances of the ion optics of csr by stray electric fields . the resulting non - sensitive rim laterally surrounding the detection aperture must be as narrow as possible in order to allow the detector to collect product particles close to the stored - ion orbit without intersecting the latter . given these boundary conditions , simulations of the secondary - electron trajectories were performed using the simion 8 software package @xcite in order to determine an optimum electrode geometry . a curved conversion cathode , as shown in fig . [ sensor ] , turned out to provide the highest efficiency in transporting the secondary electrons to the mcp stack . as shown in the upper right panel ( 7 ) of fig . [ sensor ] , the secondary - electron collection efficiency is @xmath21% over practically the entire area of the sensor aperture . the lateral non - sensitive detector rim could be narrowed down to 2.5 mm ( ( 6 ) in fig . [ sensor ] ) . in order to ensure good performance of the electron multiplier stage also at very low ( @xmath0 k ) ambient temperature , a miniature ohmic heating module is included on top of the mcp stack ( ( 5 ) in fig . [ sensor ] ) . its purpose is to keep the mcps warmer by 10 to 20 k relative to the experimental environment , thus preventing excessively high channel plate resistance in cold operation of csr . the heating element is made from an 0.1-mm - thick and 27-cm - long constantan wire wound around ceramic spacers . it has an electric resistance of @xmath22 at 300 k , which changes only slightly upon cooling to cryogenic temperature . the various electrodes of the mcp - anode stack are electrically insulated from the heating element by sapphire washers ensuring good thermal coupling . operational tests of this mcp heating are presented in sect . [ testcryo ] . ( colour online ) schematic cut - open view of the detector within csr . shown are the 10-k beam pipe ( 1 ) , the room temperature isolation vacuum chamber ( 2 ) and the two thermal radiation shields ( 3 ) of the csr cryostat . the detector mechanics consist of the cryogenic translation stage in the uhv chamber ( 4 ) , a cryogenic bellow - sealed rotary feed - through ( 5 ) , a thermally decoupling drive ( 6 ) supported by the 40-k stage in the isolation vacuum of the cryostat , and a standard rotary feed - through on the room - temperature chamber wall ( 7 ) . the movable sensor ( 8) is connected to the cryogenic high - voltage feed - through via a set of five flexible cube2 bands ( not shown ) . the dotted red line represents the closed - orbit trajectory of particles stored in csr , whereas the dash - dotted red line represents a charged - changed product beam intercepted by the detector.,width=302 ] the translation mechanics of the detector consist of four main components ( cf . [ csrdetail ] ) : the cryogenic translation stage mounted within the uhv beam pipe of csr ( ( 4 ) in fig . [ csrdetail ] ) , a bellow - sealed cryogenic rotary feed - through ( ( 5 ) in fig . [ csrdetail ] ) in the experimental chamber wall , a standard high - vacuum rotary feed - through ( ( 7 ) in fig . [ csrdetail ] ) on the room - temperature isolation vacuum chamber of the storage ring , and an intermediate rotary drive ( ( 6 ) in fig . [ csrdetail ] ) contained within the isolation vacuum vessel of the csr cryostat . the cryogenic translation stage allows to move the particle sensor across the 300-mm - wide beam - guiding uhv chamber while maintaining secure high - voltage connections between the detector electrodes and electric feed - throughs in the experimental vacuum chamber flange . motion is induced using a worm drive consisting of a 316l stainless steel tr10@xmath112-threaded rod , rotating inside a matching nut made of polyether ether ketone ( peek ) and embedded into a carriage supporting the particle sensor . the threaded rod is fixed to the cryogenic rotary feed - through at its one end and loosely contained in an all - ceramic ball bearing at the other . identical ceramic ball bearings ensure low - friction guidance of the detector carriage along an aluminium rail also attached to the cryogenic vacuum chamber flange . to account for the significantly stronger thermal shrinking of peek compared to stainless steel , a very loose fitting ( @xmath23 mm ) of the male and female tr10@xmath112 threads at room temperature had to be chosen . the corresponding loss in positioning precision is irrelevant for the purpose of selecting specific daughter ion beams that are usually separated by several cm . for compatibility with cryogenic operation and uhv requirements , no liquid lubricants are used in any part of the drive . the cryogenic rotary feed - through ( ( 5 ) in fig . [ csrdetail ] ) in the experimental vacuum flange is provided off - the - shelf by _ agilent technologies _ ( model l6691301 , on cf16 flange ) . electric connections within the beam - guiding vacuum chamber are done via five flexible , 0.1-mm - thick , 4-mm - broad , and approximately 250-mm - long bands of cube2 that bend following the motion of the sensor carriage , while being securely fixed to the detector electrodes and heating as well as to the cryogenic high - voltage feed - throughs in the experimental vacuum chamber flange . spacers made from peek ensure electric high - voltage insulation among the cube2 bands and towards a grounded faraday cage surrounding the flexible high voltage lines . within the isolation vacuum chamber , an intermediate drive ( ( 6 ) in fig . [ csrdetail ] ) , installed on the 40-k thermal radiation shield ( ( 3 ) in fig . [ csrdetail ] ) of the csr cryostat , provides transmission of rotary motion from the atmosphere side actuator to the cryogenic translation mechanism while ensuring thermal decoupling of the @xmath0-k beam - guiding vacuum chamber from the 300-k outer cryostat wall . the 40-k drive contains a spring - loaded sliding bearing with relatively high friction and interconnects to the 10-k and room - temperature rotary feed - throughs via nosings of very loose fitting ( approximately @xmath24 ) . by inducing a proper sequence of rotations from the atmosphere side , it is hence always possible to position the 40-k drive in a way that , at rest , it touches neither the room - temperature components nor the 10-k cold stage of the csr , and hence maintains near - perfect thermal isolation among the different temperature stages of the cryostat . the friction bearing itself is electrically insulating , such that a sensor wire connected to the drive allows the user to check whether it is not touching ground potential at either vacuum chamber . all electronics and power supplies are kept on the atmosphere side of the set - up . hence electric lines within the isolation vacuum chamber have to connect the room - temperature and cryogenic electric feed - throughs . as electric and thermal conductance strongly correlate for almost all materials , this requires a trade - off between thermal insulation and signal quality . the ( static ) high - voltage potentials of the converter cathode and of the mcp input and output electrodes are delivered via kapton - insulated 0.25-mm thick manganin wires , which have very low thermal conductance , but also high ohmic resistance . for the anode connection , from which the fast counting signal is decoupled , a kapton - insulated co - axial copper cable ( _ allectra _ 311-kap-50-rad , 50 @xmath25 ) is used . in order to limit the resulting thermal load on the 10-k beam pipe of the csr , the cable is thermally anchored to the intermediate 40-k stage of the cryostat . in addition , the screen of the cable is locally reduced to a single strand at the exit of the 40-k volume . in this way the heat conductivity of the screen is greatly reduced while galvanic contact is maintained . still , due to the high thermal conductance of copper , the resulting detector - related heat input to the cryogenic beam line is dominated by the co - axial anode connection . it can be estimated to @xmath26mw using tabulated heat conductivity values @xcite . the high - voltage potentials for the converter , anode , and mcp input electrodes are provided by standard low - ripple power supplies ( _ iseg _ nhq 204 m ) . the gain voltage is generated by a floating high - voltage supply ( _ spellman _ mcp series ) stacked onto the mcp input potential . the electron current pulses arriving at the anode are separated from the high potential by a decoupling capacitor directly outside the isolation vacuum feed - through and driven into a 50-@xmath25 current - sensitive fast front - end amplifier ( _ ortec _ vt120a ) with gain factor 200 . normally no further amplification stage is needed before discrimination of the pre - amplified pulses . in order to ensure suitability for operation at csr , a series of tests of the detector set - up were performed . the assembled sensor and translation stage were baked to 250@xmath3c without resulting damage . functional tests of the sensor with regard to signal quality and counting efficiency were performed . furthermore , electrical and mechanical performance of the set - up at low temperature was tested . the detection principle based on conversion of the primary incident particle into several secondary electrons , together with good collection efficiency of the secondaries by the positively biased mcp stack , suggests good overall particle counting performance of the set - up . this section provides a theoretical and experimental description of the detection efficiency of the sensor . we assume that the entire sensitive aperture of the sensor is irradiated by a homogeneous flux of primary particles . let @xmath27 be the average secondary electron yield per primary heavy particle in a certain spot of the converter cathode , and @xmath28 the likelihood for any given secondary electron to reach the mcp input surface . both @xmath27 and @xmath28 may vary spatially across the area of the converter cathode . it is well established that the number @xmath29 of secondary electrons arriving at the mcp can then be described by a plya distribution@xcite @xmath30 given by @xmath31 where @xmath32 is the mean of the secondary electron yield @xmath27 of the cathode , weighted with the probability @xmath28 for a given electron to reach the mcp input surface , i.e. @xmath33 . the dimensionless parameter @xmath34 with @xmath35 is the relative variance@xcite of the product @xmath36 . we assume further that each electron impinging on the mcp has a probability @xmath37 of triggering an avalanche in a channel . the total number @xmath38 of mcp cascades resulting from one incident primary particle is then distributed according to @xmath39 since the binomial distribution @xmath40 gives the probability for @xmath38 electrons out of @xmath29 triggering an mcp cascade at individual triggering probability @xmath37 , and as the number @xmath29 of impinging secondaries is itself plya - distributed according to eq . ( [ polya ] ) . a large variance of the product @xmath41 is reflected by a high value of @xmath34 . in the limit @xmath42 , @xmath43 converges to the geometric distribution . for zero variance , i.e. for @xmath44 , @xmath43 becomes a poisson distribution @xcite of mean @xmath32 . hence , if only a sufficiently small , specific spot of the cathode is irradiated by primary particles , eq . ( [ pulseheightpolya ] ) simplifies to @xmath45 where we have set @xmath46 as , in contrast to eq . ( [ pulseheightpolya ] ) , @xmath27 and @xmath28 can now be considered constant over the cathode area of interest . it is worth noting that the @xmath38 secondary electrons that multiply successfully should be assumed to impinge into _ different _ mcp channels . in contrast to a single - channel electron multiplier , whose gain is damped by the space - charge limit of the electron avalanche for high numbers of impinging electrons , the output signal amplitude of the mcp is therefore proportional to @xmath38 . for a detector irradiated by a product beam from an electron - cooled , low - emittance ion beam @xcite , the irradiated area of the converter cathode can be smaller than @xmath47 mm@xmath12 . hence , under the assumption that the spatial variation of @xmath41 is negligible across this characteristic beam size , eq . ( [ pulseheight ] ) is applicable . under the further assumption that a primary particle incident onto the sensor is successfully detected as long as it triggers _ at least one _ electron avalanche , the overall detection efficiency @xmath48 is then given by @xmath49 as @xmath50 over practically the entire detector aperture according to simulations ( cf . [ design ] and fig . [ sensor ] ) , @xmath48 depends mainly on the local secondary electron yield @xmath27 and the mcp efficiency @xmath37 for 1.2-kev electrons . both numbers are difficult to predict theoretically . it is common practice to use the open area ratio of the mcp of 0.6 as estimate for @xmath37 . the few direct measurements for electron energies around 1 kev that have been published largely confirm this value , but come with quite large uncertainties @xcite . in the following , we thus assume @xmath51 . the value of @xmath27 is even harder to estimate . while electron ejection from pure metal targets can be modelled quite reliably , it is known that , in practice , the yield of secondaries depends not only on the velocity of the heavy projectile and the material of the cathode ( al ) , but also on the chemical composition and thickness of any oxide or other adsorption layers that may have formed on the al surface @xcite . these latter parameters are difficult to control . it is however safe to assume that , in average , several electrons are ejected if the velocity of the incident particle is high enough . from eq . ( [ efficiency ] ) one calculates that starting from an electron yield of @xmath52 the overall detection efficiency @xmath48 should be greater than 95% , as shown in fig . [ effiplot ] . ( colour online ) theoretical detection efficiency @xmath48 ( solid curve ) expected according to eq . ( [ efficiency ] ) as a function of the converter secondary electron yield @xmath27 for fixed values @xmath53 and @xmath54 . the dashed lines represent the systematic error based on an estimated relative uncertainty of 33% on @xmath37 ( see text ) . the filled circles represent the measured values @xmath55 with their experimental uncertainties.,width=302 ] the above theoretical model for the detection efficiency was tested empirically using the ` pinhole ' faraday cup technique that has been described in detail by other authors @xcite . a beam of @xmath56ar@xmath57 ions was produced from a penning - type ion source , momentum - selected in a bending dipole magnet , and accelerated to a total kinetic energy of 80 kev ( 2 kev / u ) using the csr injector platform @xcite . after focusing and collimation to a cross - section of [email protected] mm@xmath12 , the remaining beam current of approximately 0.5 na was dumped onto a faraday cup . a circular pinhole aperture of 10.5(5 ) @xmath58 m diameter at the bottom of the cup allowed a very small ( @xmath59% ) portion of the ion beam to pass and impinge onto the sensor . in this way , a rate of ar@xmath57 ions suitable for single particle counting ( a few @xmath60 s@xmath61 ) arrived at the detector , while the effect of the pinhole aperture on the overall ion current measured at the faraday cup was negligible . the pinhole also made sure that only a very small spot on the conversion cathode was irradiated by ions , and its position was approximately centred onto the sensitive area of the detector . the ion beam was then steered across the pinhole aperture in two dimensions ( @xmath62 and @xmath63 ) using small dipole magnets following the collimator . the faraday cup aperture was much larger than the beam cross section . hence the cup always measured the full beam current @xmath64 , independent on the steerer settings . the rate of particles @xmath65 detected by the sensor after the pinhole , and normalised to the corresponding beam current @xmath66 , was measured for a matrix of beam deflections @xmath67 . the experimental detection efficiency can then be calculated@xcite as @xmath68 where @xmath69 is the area of the pinhole aperture , @xmath66 the measured ion current in the cup , @xmath70 the elementary charge , @xmath71 the charge state of the ions , and @xmath72 the mesh area of the scanning matrix @xmath67 : @xmath73 in this way , an experimental detection efficiency for 80-kev ar@xmath57 of @xmath74 was obtained , and could be reproduced several times within the given uncertainty . the 3% statistical error bar stems mainly from the ion current measurement @xmath66 at the faraday cup . larger systematic uncertainties arise from determination of @xmath69 : direct microscopic measurement of the pinhole diameter ( @xmath75 m ) leads to a relative uncertainty of @xmath76 . due to the thickness of the pinhole aperture of 12.7 @xmath58 m , the angle of incidence of the ion beam onto the cup , estimated to @xmath77 , leads to an additional , asymmetric error bar of @xmath78 on @xmath69 , and , hence , of @xmath79 on the measured detection efficiency @xmath48 . a second , similar , experiment was performed using a very slow beam of ar@xmath80 of 6 kev energy ( 150 ev / u ) . as the csr injector was then not available , another , simpler , ion source and extraction beam line was used . in the same way as described above , a detection efficiency @xmath81 was obtained . the lesser statistical precision of this experiment results mainly from the fact that the beam delivered by the small ion source suffered from lower intensity and worse collimation . the former led to increased uncertainties of @xmath66 , the latter to a larger error bar on @xmath65 , as background count rate from stray ions in the vacuum chamber had to be subtracted . the main benefit of the experiment with 6-kev ions was that a direct measurement of the secondary electron yield @xmath27 could be performed : the aperture at the bottom of the faraday cup was enhanced such as to allow the full ion beam of 0.1 to 1 na to impinge on the converter cathode . the resulting total ( impinging ions plus emitted electrons ) current flowing at the cathode could be directly measured by a floating nano - amperemeter . by comparison to the independently measured primary ion current a value of the secondary electron yield of @xmath82 could be determined . as noted above , the anode pulse heights are proportional to the number @xmath38 of secondary electrons that are successfully multiplied in the mcp , and are hence distributed according to @xmath83 from eq . ( [ pulseheight ] ) , whose mean is @xmath84 . the ratio of mean anode pulse heights observed in the experiments on ar@xmath57 ( 80 kev ) and ar@xmath85 ( 6 kev ) suggested that the secondary electron yield for the former was higher by a factor 3.6(7 ) , i.e. @xmath86 . the observed secondary electron yields are lower than what has been reported for argon impacting onto adsorption - contaminated al surfaces @xcite , but also significantly higher than the values published for atomically clean al targets @xcite . the results of both measurements of @xmath55 are shown in fig . [ effiplot ] with error bars representing the quadrature sums of their statistical and systematic uncertainties . while the latter are quite large , the agreement with the simple model from eqs . ( [ pulseheight ] ) and ( [ efficiency ] ) is reasonable . we note that for the higher energetic 80 kev ( 2 kev / u ) ar ions , a very good overall detection efficiency of practically 1 is observed . in csr operation , with typical ion energies of a few kev / u , a similar performance can be expected . compatibility with cryogenic operation was tested , separately for the particle sensor and translation stage , using a test cryostat equipped with a _ leybold _ rpk-800 dual - stage pressurised helium refrigerator . a thermal shield attached to the first cold head stage cools down to @xmath87 k and prevents radiative heat input onto the experimental platform . the latter consists of a 400@xmath1180 mm@xmath12 copper support fixed to the second stage of the cold head . it is sufficiently spacious to hold the detector translation stage or the particle sensor , and cools to @xmath88 k without heat load . the set - up is evacuated to a base pressure of @xmath89 mbar by a turbo - molecular pump prior to cooling down . the particle sensor was cooled to a temperature of approximately 20 k while being irradiated by an @xmath90am @xmath91 emitter . the latter had an activity of 3.4 mbq , but was mounted in the room - temperature part of the vacuum chamber and collimated in such a way that @xmath91 particles entered the detection aperture at an estimated rate of @xmath92 . upon impact onto the converter cathode , the 5.4-mev @xmath91 particles generated secondary electrons which were multiplied in the mcps . after decoupling and amplification , as described in sect . [ design ] , the resulting anode current pulses were converted into digital counts using a linear discriminator . the discrimination threshold was deliberately set high to make sure that no electric noise , that could have occurred during the experiment , would be misinterpreted as particle hits . ( colour online ) rate of discriminated alpha particle hits ( black dots ) during cool down of the sensor in the cryogenic test chamber . the horizontal axis shows the time since the beginning of the experiment , which lasted for approximately two weeks . the red dashed curve represents the measured temperature of the detector housing . the inset shows the resistance @xmath93 of the mcp stack in the low - temperature region . the blue dashed curve is an exponential fit to the data ( black dots).,width=302 ] ( colour online ) effect of localised heating of the mcps . the long - dashed red curve represents the temperature of the sensor housing . the short - dashed blue line is the temperature of the mcp set , obtained by calibration of the bias current . the black dots represent the discriminated alpha particle pulse rate . during the indicated time intervals , the mcp heating operated at powers of 85 mw and 200 mw , respectively.,width=302 ] the evolution of the discriminated pulse rate during cool - down of the sensor from 300 k to 20 k is shown in fig . [ cooldown ] . the temperatures in various places of the test cryostat were measured using pt1000 thermo - sensors , one of which was attached to the outside of the grounded detector housing . during the cooling process , the measured count rate decreased by 8% , from 2.5 to 2.3@xmath94 s@xmath61 . during the same time interval the resistance of the mcp stack , derived from the measured bias current at an operating voltage of 1750 v , rose from 56 m@xmath25 at 300 k to 85 g@xmath25 at 22 k. hence , the initial decrease in the discriminated count rate can be interpreted as a reduction of the average pulse height due to the rise in mcp resistance . it was followed by a second , very slow , decrease of the discriminator rate at constant temperature that could be observed for several days of measurement ( cf . [ cooldown ] ) . this can not be explained by further increase of the mcp resistance , as the measured bias current was then constant . instead , we believe that this behaviour is , at least partly , due to a variation in gain related to cryo - adsorption of residual gas onto the mcp . this explanation is supported by the response of the set - up to localised heating of the mcp as discussed below . using a 5-w heater attached to the cold stage of the test cryostat , the temperature of the entire set - up could be varied between 22 and 45 k. in this way , the mcp bias current could be precisely calibrated against temperature in that range . in good approximation , we found the resistance of the channel plates to rise exponentially with decreasing temperature , as shown in the inset of fig . [ cooldown ] , and as reported previously by other authors @xcite . the behaviour of the mcps below 22 k could not be probed due to the limited power of the cryo cooler and accuracy of the pt1000 sensors . at csr , significantly lower temperatures of @xmath0 k are expected , which will lead to even higher resistance of the channel plates than observed in this experiment . in order to guarantee good performance in csr operation , a means to locally raise the mcp temperature relative to the cryogenic environment is required . this possibility is provided by the heating element included in the mcp - anode stack as described in sect . [ design ] . the response of the sensor to the mcp heating is shown in fig . [ mcpheating ] . during the indicated time intervals , electric current was fed through the constantan heating wire , first at a power of 85 mw , then of 200 mw . using the mcp resistance as a proxy for temperature ( cf . [ cooldown ] ) a prompt warming of the channel plates was observed after switching on the heating . the temperature would typically stabilise after @xmath95 h. for a heating power of 200 mw , the mcp temperature rose by more than 20 k , while the temperature of the detector housing was practically unaffected . this suggests that the heating is an efficient way of warming the mcp locally to a known - good operating temperature without disturbing the surrounding cryostat . no deterioration of the anode signal quality by the heating current was observed . the effect of the warming of the mcp on the pulse heights is shown in fig . [ mcpheating ] : a significant rise in the discriminator rate was visible as soon as the heating element was switched on . moreover , part of this rate enhancement was irreversible : even after the heating was switched off and the mcp cooled down again , the discriminator rate remained at a @xmath96% higher level than before the heating cycle . this supports our interpretation that the decrease in mcp pulse heights is partly due to residual gas adsorption onto the channel plates as noted above : as soon as the mcp is locally warmed , the surface contamination evaporates and is adsorbed onto the cold environment instead . at the csr , which provides base vacuum conditions that are orders of magnitude better compared to the test cryostat , these effects are expected to be much less pronounced . using the same cryostat set - up , the mechanical components of the detector were tested for reliable operation at low temperature . the translation stage successfully performed 100 motion cycles over the full 300-mm stroke without apparent deterioration of the peek / stainless - steel worm drive or the flexible electric connections . furthermore , independent tests of the reliability of the isolation vacuum drive and of the cryogenic rotary feed - through were performed at 38 k and 22 k , respectively . although the manufacturer ( _ agilent technologies _ ) specifies a minimum operating temperature of 77 k , the cold feed - through performed a total of 30000 revolutions at 22 k without resulting vacuum leak . as stated in sect . [ design ] , an edr micro - channel plate was chosen as secondary electron multiplier stage for the detector , as its cryogenic behaviour could be expected to be favourable from previously published data . however the possibility to use a single - channel electron multiplier was also briefly investigated . cems provide higher gain and stronger pulse height saturation compared even to a chevron mcp stack . hence they could be an interesting upgrade for future revisions of the detector . two channel electron multipliers were tested : a standard , high - resistance ( @xmath97 m@xmath25 at 300 k ) cem by _ philips _ ( type x 719 bl ) and a lower - resistive ( @xmath98 m@xmath25 ) enhanced dynamic range cem by _ photonis _ ( type 7010 m c edr ) . the cems were incorporated into a spare particle sensor as described by rinn et al.@xcite , which operates on the same principle of secondary - electron ejection from a conversion electrode as the detector described in this work . the sensor was mounted onto the cold stage of the test cryostat and cooled to a temperature of @xmath99 k while being irradiated by the @xmath90am source as described above . it turned out that the edr cem did produce anode pulses at 25 k , although they were of much lower intensity than in room temperature operation . the loss in gain related to cooling was significantly stronger compared to the edr mcp stack , but could be compensated by raising the operating voltage of the cem from initially @xmath100 kv to @xmath101 kv . the standard resistance cem , however did not produce measurable anode pulses at 25 k , although it operated normally before the cool - down and after rewarming to room temperature . ( colour online ) electric resistance of the three tested electron multipliers , measured as a function of temperature . the edr mcp ( long - dashed blue curve ) and edr cem ( solid green ) display approximately the same behaviour and both could be operated at 25 k. the standard cem ( short - dashed red ) has got a resistance that is higher by a factor of @xmath102 across the entire temperature range and did not produce measurable electron pulses at 25 k ( see text).,width=302 ] figure [ cemres ] shows the electric resistance of all three electron multipliers , measured using a high - voltage ohmmeter as a function of detector temperature in the range from 25 to 270 k. both edr mcp and edr cem display approximately the same behaviour of resistance versus temperature . the standard cem , however , has a resistance that is larger by more than an order of magnitude compared to the edr detectors , amounting to @xmath103 t@xmath25 at 25 k. we suspect that this extremely high resistance limits the bias current of the cem to the point where the channel wall can effectively not recharge such that sustained electron multiplication is not possible any more . we note that operation of an edr cem in a cryogenic detector for the csr storage ring seems possible , especially if the electron multiplier was equipped with a local heating element as described above . as all secondary electrons emitted from the conversion cathode enter the same amplification channel , a cem has the potential of producing saturated , narrow pulse height distributions that can be easier to discriminate than the output signals of mcps . however , more detailed tests of the cem behaviour are necessary and the proven mcp stack was used as electron multiplier stage in this work . the full detector set - up as shown in fig . [ csrdetail ] including all mechanics and electrical wiring was tested in operation during room - temperature commissioning @xcite of csr in april 2014 . for detector tests , the storage ring operated using a beam of 50-kev n@xmath104 , produced from the injection accelerator . most of the csr pumping facilities had still to be installed , and cryogenic operation was not yet foreseen . due to the limited pumping speed , the residual gas pressure in the ring was not better than @xmath105 mbar , even after an intermediate bakeout at 120@xmath3c had been performed . consequently , the storage life time @xmath106 of the n@xmath104 ion beam was @xmath191 ms only . by collisions with residual gas particles ( x ) , the n@xmath104 molecular ion most frequently dissociated into a neutral / charged fragment pair @xmath107 the products n and n@xmath80 had a kinetic energy of 25 kev each and were separated from the stored n@xmath104 beam by the 6@xmath3 deflector element preceding the detector in the beam line as shown in fig . [ csrdetail ] . the corresponding parameters @xmath108 from eq . ( [ etaequation ] ) are @xmath109 and @xmath110 for the charged and neutral fragment , respectively . hence , both daughters could be detected selectively by moving the sensor to the corresponding position in the csr beam line . using the electronics scheme described above , the anode pulses triggered by the impinging atomic particles were recorded . despite the long and complicated electric line which connects the sensor anode to the decoupling and amplification circuit ( cf . [ design ] ) , necessarily involving several changes in impedance , the signal quality proved to be quite good . as shown in fig . [ pulse ] , the anode pulses were followed by only little ringing , allowing for easy and efficient signal discrimination . as the stored n@xmath5 beam was uncooled , the n and n@xmath80 products could be expected to irradiate the full sensitive area of the sensor . in contrast to sect . [ testpulse ] , the number @xmath38 of secondary electrons multiplied in the mcps is then gouverned by plya statistics according to eq . ( [ pulseheightpolya ] ) . the pulse height distribution is defined by the electric response of the mcp stack to those @xmath38 electron avalanches . an ultraviolet ( uv ) algan light emitting diode ( led , _ seti _ uvtop240 ) is installed in the csr for detector testing purposes . it allows to irradiate particle detectors with an uncollimated beam of @xmath111-nm photons . their energy of 5.1(1 ) ev allows the photons to eject a single electron _ at maximum _ upon absorption at the conversion cathode . the uv led hence provides a means to relate the measured pulse height distribution for heavy particle impact to the average number @xmath32 of secondary electrons reaching the mcp . ( colour online ) measured pulse height spectra for uv irradiation ( circles ) and for 25-kev n@xmath80 ions ( triangles ) . the dashed blue curve ( @xmath112 ) is a fit to the uv - induced single secondary electron spectrum . the solid red line ( @xmath113 ) is modelled from the latter assuming plya - distributed secondary electron numbers @xmath38 . some individual contributions @xmath114 are labelled 1 to 15 ( see text ) . the inset displays an averaged anode pulse shape obtained for 25-kev n@xmath80 . the anode pulses were driven into 50 @xmath25 and amplified by a factor 200 ( cf . [ design]).,width=302 ] the pulse height distribution @xmath112 measured for uv photon irradiation is shown in fig . [ pulse ] . the mcp stack operated at a voltage of 950 v per plate . if @xmath112 is the pulse height spectrum corresponding to precisely one initial electron multiplied at the mcp set , then the response for precisely two electron cascades is given by the convolution @xmath115 , and for precisely @xmath38 cascades by @xmath116 . the pulse height spectrum @xmath113 for n@xmath80 ions can thus be modelled from the measured single - electron detector response @xmath112 as @xmath117 the weights @xmath118 are taken from eq . ( [ pulseheightpolya ] ) as a function of the parameters @xmath32 and @xmath34 . the correction factor @xmath119 ensures that the resulting distribution @xmath113 is normalised to 1 , although there is a non - vanishing probability ( @xmath120 ) for an ion to produce no signal . figure [ pulse ] shows the measured pulse height spectrum for n@xmath80 products impinging onto the detector , along with a model spectrum @xmath113 derived from the measured uv - photon response @xmath112 . empirically , the uv response was found to be well reproduced by a sum of an exponential decay and a gaussian centred at zero . subsequently , the convolutions @xmath121 were computed numerically . the best fit for @xmath113 from eq . ( [ fdistr ] ) to the n@xmath80 data is obtained for @xmath122 and @xmath123 ( cf . ( [ pulseheightpolya ] ) ) . the fit model includes no other free parameters . as a large area of the converter cathode was irradiated in this experiment , a large value of @xmath34 was to be expected . since the plya parameter @xmath34 is the relative variance@xcite of the distribution of @xmath124 , this corresponds roughly to values @xmath125 across the entire converter cathode area for 25-kev n@xmath80 particles . note however that the fit function @xmath113 depends quite strongly on the low - energy part ( @xmath126 mv ) of @xmath112 which , as visible in fig . [ pulse ] , has not been measured but extrapolated . hence , this result is likely bound to a large systematic uncertainty . still , it seems that a basic understanding of the pulse height spectra can be obtained from eq . ( [ pulseheightpolya ] ) . the measured pulse height distribution for 25-kev n@xmath80 also shows that the loss in detection efficiency by the discrimination threshold which can be as low as a few 10 mv , thanks to the good signal quality is small ( a few 10@xmath127 ) . ( colour online ) measured rate of n@xmath80 impacts ( black dots ) onto the detector as a function of storage time of the n@xmath104 beam in csr . the red line indicates the exponential decay of the ion beam with lifetime @xmath128 ms . the second and third frames , labelled a and b , are close - ups of the signal as indicated by the dashed blue lines.,width=302 ] figure [ rates ] shows the measured detector count rate as a function of storage time of the n@xmath104 beam in csr . the data is averaged over several thousand individual ion injections into the storage ring . the sharp modulation of the signal , clearly visible in frame a of fig . [ rates ] , stems from the fact that the injected ion bunch had a length of only 29 m , as opposed to the full closed orbit circumference of csr which is of 35.4 m. hence , the train of particles passed by the detector only for @xmath129% of the time of one revolution period , which was of 59.4 @xmath58s . for both , neutral and charged products , count rates up to several 10@xmath130 s@xmath61 were recorded directly after ion injection , followed by an exponential decrease with decay constant @xmath131 ms , which matched the anticipated lifetime of the ions due to residual gas collisions . after @xmath96 ms , corresponding to @xmath132 revolutions of the ion bunch in the storage ring , the measured product rate vanished , showing that the edr mcp stack is effectively free of any dark count rate , even in room - temperature operation . we have designed a highly efficient , movable single - particle counter that is suitable for operation in a cryogenic ultra - high vacuum environment . while the device has been developed specifically for operation at the csr storage ring , we note that the latter shares many of its characteristics with similar electrostatic and cryogenic storage devices in operation@xcite or in construction@xcite . hence , we believe that our development work is of interest also to others . the detector has been successfully commissioned at csr in room - temperature operation . the set - up is bakeable to high ( 250@xmath3c ) temperature , and off - line performance tests have shown that it is fully suited for cryogenic operation in the @xmath0-k environment of upcoming csr experiments . the particle sensor geometry has been optimised using simulations of the secondary - electron optics , and direct measurements have confirmed an overall detection efficiency close to 100% for particle energies of a few kev / u . while the set - up has been designed primarily with future electron cooler experiments in mind , we expect to develop variants of the detector also for other experimental sections of the csr storage ring . in that process , also the possibility of using a cem as alternative electron multiplication stage will be further investigated . we gratefully acknowledge the technical support this work has obtained from the mpik mechanical workshop and accelerator staff as well as the financial support it has received from the max planck society ( mpg ) . o. n. was supported , in part , by the united states national science foundation ( nsf ) and national aeronautics and space administration ( nasa ) . the work of a. b. as well as that of k. s. was funded by the german research foundation ( dfg ) within the priority programme 1573 , under contract numbers wo 1481/2 - 1 and schi 378/9 - 1 , respectively . 100 r. von hahn , k. blaum , f. fellenberger , m. froese , m. grieser , c. krantz , m. lange , s. menk , f. laux , d. a. orlov , r. repnow , a. shornikov , t. sieber , and a. wolf , nucl . instrum . methods phys . b * 269 * , 2871 ( 2011 ) c. krantz , f. berg , k. blaum , f. fellenberger , m. froese , m. grieser , r. von hahn , m. lange , f. laux , s. menk , r. repnow , a. shornikov , and a. wolf , j. phys . ser . * 300 * , 012010 ( 2011 ) m. grieser , yu . a. litvinov , r. raabe , k. blaum , y. blumenfeld , p. a. butler , f. wenander , p. j. woods , m. aliotta , a. andreyev , a. artemyev , d. atanasov , t. aumann , d. balabanski , a. barzakh , l. batist , a .- p . bernardes , d. bernhardt , j. billowes , s. bishop , m. borge , i. borzov , f. bosch , a. j. boston , c. brandau , w. catford , r. catherall , j. cederkll , d. cullen , t. davinson , i. dillmann , c. dimopoulou , g. dracoulis , ch . e. dllmann , p. egelhof , a. estrade , d. fischer , k. flanagan , l. fraile , m. a. fraser , s. j. freeman , h. geissel , j. gerl , p. greenlees , r. e. grisenti , d. habs , r. von hahn , s. hagmann , m. hausmann , j. j. he , m. heil , m. huyse , d. jenkins , a. jokinen , b. jonson , d. t. joss , y. kadi , n. kalantar - nayestanaki , b. p. kay , o. kiselev , h .- j . kluge , m. kowalska , c. kozhuharov , s. kreim , t. krll , j. kurcewicz , m. labiche , r. c. lemmon , m. lestinsky , g. lotay , x. w. ma , m. marta , j. meng d. mcher , i. mukha , a. mller , a. st j. murphy , g. neyens , t. nilsson , c. nociforo , w. nrtershuser , r. d. page , m. pasini , n. petridis , n. pietralla , m. pftzner , z. podolyk , p. regan , m. w. reed , r. reifarth , p. reiter , r. repnow , k. riisager , b. rubio , m. s. sanjari , d. w. savin , c. scheidenberger , s. schippers , d. schneider , r. schuch , d. schwalm , l. schweikhard , d. shubina , e. siesling , h. simon , j. simpson , j. smith , k. sonnabend , m. steck , t. stora , t. sthlker , b. sun , a. surzhykov , f. suzaki , o. tarasov , s. trotsenko , x. l. tu , p. van duppen , c. volpe , d. voulot , p. m. walker , e. wildner , n. winckler , d. f. a. winters , a. wolf , h. s. xu , a. yakushev , t. yamaguchi , y. j. yuan , y. h. zhang , and k. zuber , eur . j. special topics * 207 * , 1 ( 2012 ) r. stensgaard , phys . * t22 * , 315 ( 1988 ) k. abrahamsson , g. andler , l. bagge , e. beebe , p. carl , h. danared , s. egnell , k. ehrnstn , m. engstrm , c. j. herrlander , j. hilke , j. jeansson , a. kllberg , s. leontein , l. liljeby , a. nilsson , a. paal , k .- rensfelt , u. rosengrd , a. simonsson , a. soltan , j. starker , m. af ugglas , and a. filevich , nucl . instrum . methods phys . res . , sect . b * 79 * , 269 ( 1993 ) s. p. mller , nucl . instrum . methods phys . res . , sect . a * 394 * , 281 ( 1997 ) t. tanabe , k. chida , k. noda , and i. watanabe , nucl . instrum . methods phys . res . , sect . a * 482 * , 595 ( 2002 ) s. jinno , t. takao , y. omata , a. satou , h. tanuma , t. azuma , h. shiromaru , k. okuno , n. kobayashi , and i. watanabe , nucl . instrum . methods phys . a * 532 * , 477 ( 2004 ) h. t. schmidt , r. d. thomas , m. gatchell , s. rosn , p. reinhed , p. lfgren , l. brnnholm , m. blom , m. bjrkhage , e. bckstrm , j. d. alexander , s. leontein , d. hanstorp , h. zettergren , l. liljeby , a. kllberg , a. simonsson , f. hellberg , s. mannervik , m. larsson , w. d. geppert , k. g. rensfelt , h. danared , a. pal , m. masuda , p. halldn , g. andler , m. h. stockett , t. chen , g. kllersj , j. weimer , k. hansen , h. hartman , and h. cederquist , rev . instrum . * 84 * , 055115 ( 2013 ) r. d. thomas , h. t. schmidt , g. andler , m. bjrkhage , m. blom , l. brnnholm , e. bckstrm , h. danared , s. das , n. haag , p. halldn , f. hellberg , a. i. s. holm , h. a. b. johansson , a. kllberg , g. kllersj , m. larsson , s. leontein , l. liljeby , p. lfgren , b. malm , s. mannervik , m. masuda , d. misra , a. orbn , a. pal , p. reinhed , k .- rensfelt , s. rosn , k. schmidt , f. seitz , a. simonsson , j. weimer , h. zettergren , and h. cederquist , rev . instrum . * 82 * , 065112 ( 2011 ) y. nakano , w. morimoto , t. majima , j. matsumoto , h. tanuma , h. shiromaru and t. azuma , j.phys . : conf . ser . * 388 * , 142027 ( 2012 ) m. lange , m. froese , s. menk , j. varju , r. bastert , k. blaum , j. r. crespo lpez - urrutia , f. fellenberger , m. grieser , r. von hahn , o. heber , k .- u . khnel , f. laux , d. a. orlov , m. l. rappaport , r. repnow , c. d. schrter , d. schwalm , a. shornikov , t. sieber , y. toker , j. ullrich , a. wolf , and d. zajfman , rev . instrum . * 81 * , 055105 ( 2010 ) a. wolf , g. gwinner , j. linkemann , a. a. saghiri , m. schmitt , d. schwalm , m. grieser , m. beutelspacher , t. bartsch , c. brandau , a. hoffknecht , a. mller , s. schippers , o. uwira , and d. w. savin , nucl . instrum . methods phys . a * 441 * , 183 ( 2000 ) j. f. ziegler , nucl . instrum . methods phys . res . , sect . b * 219220 * , 1027 ( 2004 ) d. zajfman , a. wolf , d. schwalm , d. a. orlov , m. grieser , r. von hahn , c. p. welsch , j. r. crespo lopz - urrutia , c. d. schrter , x. urbain , and j. ullrich , j. phys . ser . * 4 * , 296 ( 2005 ) k. rinn , a. mller , h. eichenauer , and e. salzborn , rev . instrum . * 53 * , 829 ( 1982 ) s. rosn , h. t. schmidt , p. reinhed , d. fischer , r. d. thomas , and h. cederquist , rev . instrum . * 78 * , 113301 ( 2007 ) k. u. kuehnel , c. d. schrter , and j. ullrich , _ proceedings of the 11th european particle accelerator conference _ ( 2008 ) tupc055 p. roth and g. w. fraser , nucl . instrum . methods phys . res . , sect . a * 439 * , 134 ( 2000 ) j. a. schecker , m. m. schauer , k. holzscheiter , and m. h. holzscheiter , nucl . instrum . methods phys . res . , sect . a * 320 * , 556 ( 1992 ) r. a. baragiola , e. v. alonso , j. ferron , and a. oliva - florio , surf . sci . * 90 * , 240 ( 1979 ) e. m. baroody , phys . rev . * 78 * , 780 ( 1950 ) d. a. dahl , int . j. mass spectrom . * 200 * , 3 ( 2000 ) j. g. weisend ii ( editor ) , _ handbook of cryogenic engeneering _ , ( taylor & francis , philadelphia , 1998 ) pp . 163166 l. a. dietz and j. sheffield , rev . instrum . * 44 * , 183 ( 1973 ) j. r. prescott , nucl . instrum . methods * 39 * , 173 ( 1966 ) g. lakits , f. aumayr , and h. winter , rev . instrum . * 60 * , 3151 ( 1989 ) d. a. orlov , h. fadil , m. grieser , c. krantz , j. hoffmann , o. novotn , s. novotny , and a. wolf , _ proceedings of the cool 2007 workshop on beam cooling and related topics _ ( 2007 ) frm1c03 m. galanti , r. gott , and j. f. renaud , rev . instrum . * 42 * , 1818 ( 1971 ) j. l. wiza , nucl . instrum . methods * 162 * , 587 ( 1979 ) g. w. fraser , nucl . instrum . methods * 206 * , 445 ( 1983 ) r. moshammer and r. matthus , j. phys . colloques * 50 * , 111 ( 1989 ) j. fricke , a. mller , and e. salzborn , nucl . instrum . methods * 175 * , 379 ( 1980 ) p. schackert , z. physik * 197 * , 32 ( 1966 ) l. e. collins and p. t. stroud , brit . j. appl . phys . * 18 * , 1121 ( 1967 ) g. d. magnuson and c. e. carlston , phys . rev . * 129 * , 2403 ( 1963 ) m. grieser , a. becker , k. blaum , s. george , r. von hahn , c. krantz , s. vogel , and a. wolf , _ proceedings of the 5th international particle accelerator conference _ ( 2014 ) thpme121 |
recent cognitive experiments have shown that the negative impact of an edge crossing on the human understanding of a graph drawing , tends to be eliminated in the case where the crossing angles are greater than degrees . this motivated the study of _ rac drawings _ , in which every pair of crossing edges intersects at right angle . in this work , we demonstrate a class of graphs with unique rac combinatorial embedding and we employ members of this class in order to show that it is-hard to decide whether a graph admits a straight - line rac drawing . + + date : _ september 27 , 2010 . _ | 1009.5227 | in the graph drawing literature , the problem of finding aesthetically pleasant drawings of graphs has been extensively studied . the graph drawing community has introduced and studied several criteria that judge the quality of a graph drawing , such as the number of crossings among pairs of edges , the number of edge bends , the maximum edge length , the total area occupied by the drawing and so on ( see the books @xcite ) . motivated by the fact that the edge crossings have negative impact on the human understanding of a graph drawing @xcite , a great amount of research effort has been devoted on the problem of finding drawings with minimum number of edge crossings . unfortunately , this problem is @xmath1-complete in general @xcite . however , recent eye - tracking experiments by huang et al . @xcite indicate that the negative impact of an edge crossing is eliminated in the case where the crossing angle is greater than @xmath0 degrees . these results motivated the study of a new class of drawings , called _ right - angle drawings _ or _ rac drawings _ for short @xcite . a rac drawing of a graph is a polyline drawing in which every pair of crossing edges intersects at right angle . didimo , eades and liota @xcite proved that it is always feasible to construct a rac drawing of a given graph with at most three bends per edge . in this work , we prove that the problem of determining whether an input graph admits a straight - line rac drawing is @xmath1-hard . didimo et al . @xcite initiated the study of rac drawings and showed that any straight - line rac drawing with @xmath2 vertices has at most @xmath3 edges and that any graph admits a rac drawing with at most three bends per edge . a slightly weaker bound on the number of edges of an @xmath2-vertices rac drawing was given by arikushi et al . @xcite , who proved that any straight - line rac drawing with @xmath2 vertices may have @xmath4 edges . angelini et al.@xcite showed that the problem of determining whether an acyclic planar digraph admits a straight - line upward rac drawing is @xmath1-hard . furthermore , they constructed digraphs admitting straight - line upward rac drawings , that require exponential area . di giacomo et al . @xcite studied the interplay between the crossing resolution , the maximum number of bends per edges and the required area . didimo et al . @xcite presented a characterization of complete bipartite graphs that admit a straight - line rac drawing . arikushi et al . @xcite studied polyline rac drawings in which each edge has at most one or two bends and proved that the number of edges is at most @xmath5 and @xmath6 , respectively . dujmovic et al . @xcite studied _ @xmath7 angle crossing _ ( or _ @xmath7ac _ for short ) drawings , i.e. , drawings in which the smallest angle formed by an edge crossing is at least @xmath7 . in their work , they presented upper and lower bounds on the number of edges . van kreveld @xcite studied how much better ( in terms of area required , edge - length and angular resolution ) a rac drawing of a planar graph can be than any planar drawing of the same graph . closely related to the rac drawing problem , is the angular resolution maximization problem , i.e. , the problem of maximizing the smallest angle formed by any two adjacent edges incident to a common vertex . note that both problems correlate the resolution of a graph with the visual distinctiveness of the edges in a graph drawing . formann et al . @xcite introduced the notion of the angular resolution of straight - line drawings . in their work , they proved that determining whether a graph of maximum degree @xmath8 admits a drawing of angular resolution @xmath9 ( i.e. , the obvious upper bound ) is @xmath1-hard . they also presented upper and lower bounds on the angular resolution for several types of graphs of maximum degree @xmath8 . malitz and papakostas @xcite proved that for any planar graph of maximum degree @xmath8 , it is possible to construct a planar straight - line drawing with angular resolution @xmath10 . garg and tamassia @xcite presented a continuous tradeoff between the area and the angular resolution of planar straight - line drawings . for the case of connected planar graphs with @xmath2 vertices and maximum degree @xmath8 , gutwenger and mutzel @xcite presented a linear time algorithm that constructs planar polyline grid drawings on a @xmath11 grid with at most @xmath12 bends and minimum angle greater than @xmath13 . bodlaender and tel @xcite showed that planar graphs with angular resolution at least @xmath14 are rectilinear . recently , lin and yen @xcite presented a force - directed algorithm based on edge - edge repulsion that constructs drawings with high angular resolution . argyriou et al . @xcite studied a generalization of the crossing and angular resolution maximization problems , in which the minimum of these quantities is maximized and presented optimal algorithms for complete graphs and a force - directed algorithm for general graphs . the rest of this paper is structured as follows : in section [ sec : preliminaries ] , we introduce preliminary properties and notation . in section [ sec : racgraphs ] , we present a class of graphs with unique rac combinatorial embedding . in section [ sec : np ] , we show that the straight - line rac drawing problem is @xmath1-hard . we conclude in section [ sec : conclusions ] with open problems . let @xmath15 be a simple , undirected graph drawn in the plane . we denote by @xmath16 the drawing of @xmath17 . given a drawing @xmath16 of a graph @xmath17 , we denote by @xmath18 the line passing through vertices @xmath19 and @xmath20 . by @xmath21 , we refer to the semi - line that emanates from vertex @xmath19 , towards vertex @xmath20 . similarly , we denote by @xmath22 ( @xmath23 ) the line ( semi - line ) that coincides ( emanates from ) vertex @xmath19 and is perpendicular to edge @xmath24 . the following properties are used in the rest of this paper . in a straight - line rac drawing there can not be three mutually crossing edges . [ prp : three - crossing - edges ] in a straight - line rac drawing there can not be a triangle @xmath25 and two edges @xmath26 and @xmath27 , such that @xmath28 lies outside @xmath25 and @xmath29 , @xmath30 lie inside @xmath25 . [ prp : triangle - edges ] the @xmath1-hardness proof employs a reduction from the well - known @xmath31-sat problem @xcite . however , before we proceed with the reduction details , we first provide a graph , referred to as _ augmented square antiprism graph _ , which has the following property : all rac drawings of this graph have two `` symmetric '' combinatorial embeddings . figures [ fig : basic - gadget-2 ] and [ fig : basic - gadget-2_2 ] illustrate this property . observe that the augmented square antiprism graph consists of a `` central '' vertex @xmath32 , which is incident to all vertices of the graph , and two quadrilaterals ( refer to the dashed and bold drawn squares in figure [ fig : basic - gadget-2_2 ] ) , that are denoted by @xmath33 and @xmath34 in the remainder of this paper . removing the central vertex , the remaining graph corresponds to the skeleton of a square antiprism , and , it is commonly referred to as _ square antiprism graph_. if we replace the two quadrilaterals with two triangles , then the implied graph is the _ augmented triangular antiprism graph_. didimo et al . @xcite , who showed that any @xmath2-vertex graph which admits a rac - drawing can have at most @xmath3 edges , used the augmented triangular antiprism graph , as an example of a graph that achieves the bound of @xmath3 edges ( see figure 1.c in @xcite ) . in contrast to the augmented triangular antiprism graph , the augmented square antiprism graph does not achieve this upper bound . in general , the class of _ the augmented @xmath35-gon antiprism graphs , @xmath36 _ , is a class of non - planar graphs , that all admit rac drawings . recall that any planar @xmath2-vertices graph , should have @xmath37 edges , and since an augmented @xmath35-gon antiprism graph has @xmath38 vertices and @xmath39 edges , it is not planar for the entire class of these graphs . there does not exist a rac drawing of the augmented square antiprism graph in which the central vertex @xmath32 lies on the exterior of quadrilateral @xmath40 , @xmath41 , and an edge connecting @xmath32 with a vertex of @xmath40 crosses an edge of @xmath40 . [ lem : v0nocross ] let @xmath42 be one of quadrilaterals @xmath40 , @xmath41 and let @xmath43 , @xmath44 , @xmath45 and @xmath46 be its vertices , consecutive along quadrilateral @xmath42 . assume to the contrary that vertex @xmath32 lies on the exterior of quadrilateral @xmath42 and there exists an edge , say @xmath47 , that emanates from vertex @xmath32 towards a vertex of quadrilateral @xmath42 , such that it crosses an edge , say @xmath48 is symmetric . ] , of quadrilateral @xmath42 . vertices @xmath44 and @xmath45 have the following properties : ( a ) they are both connected to vertex @xmath32 , and , ( b ) have a common neighbor @xmath49 , which is incident to vertex @xmath32 and @xmath50 ( see figure [ fig : basic - gadget-1 ] ) . observe that if vertex @xmath49 lies in the non - colored regions of figure [ fig : v0nocross - regions ] , then at least one of the edges incident to @xmath49 crosses either @xmath47 or @xmath48 , which are already involved in a right - angle crossing . this leads to a situation where three edges mutually cross , which , by property [ prp : three - crossing - edges ] is not permitted . hence , vertex @xmath49 should lie in the interior of the dark - gray colored regions @xmath51 , @xmath52 or @xmath53 of figure [ fig : v0nocross - regions ] . we consider each of these cases separately in the following . note that , there exist cases where @xmath54 ( i.e. , vertex @xmath32 is close to the intersection point of @xmath47 and @xmath48 ) , or @xmath55 ( i.e. , vertex @xmath45 is close to the intersection point of @xmath47 and @xmath48 ) , or @xmath56 ( i.e. , vertex @xmath44 is close to the intersection point of @xmath47 and @xmath48 ) . should lie in the interior of @xmath51 or @xmath52 or @xmath53 . ] case i : : : _ vertex @xmath49 is in the interior of @xmath51_. this case is depicted in figure [ fig : v0nocross - a-1 ] . let @xmath57 be the region formed by vertices @xmath49 , @xmath44 and @xmath45 ( i.e. , the dark - gray colored region of figure [ fig : v0nocross - a-1 ] ) . vertex @xmath46 , which has to be connected to vertices @xmath43 and @xmath45 , and , the central vertex @xmath32 , can not lie within @xmath57 , since edge @xmath58 would have to cross edge @xmath48 , which is already involved in a right - angle crossing . since vertex @xmath46 has to be connected to vertex @xmath32 , has to coincide with semi - line @xmath59 , as illustrated in figure [ fig : v0nocross - a-1 ] . however , under this restriction , the common neighbor @xmath60 of vertices @xmath45 and @xmath46 can not be connected to vertex @xmath32 , since edge @xmath61 should be perpendicular to one of the edges of @xmath57 , which can not be accomplished without introducing an edge overlap with edge @xmath62 . case ii : : : _ vertex @xmath49 is in the interior of either @xmath52 or @xmath53 . _ say without loss of generality that vertex @xmath49 is in the interior of @xmath53 . this case is depicted in figure [ fig : v0nocross - b ] . let @xmath19 be a vertex of the augmented antiprism graph ( except @xmath43 ) and assume that @xmath19 lies in the interior of the triangular face , say @xmath57 , formed by vertices @xmath44 , @xmath45 and @xmath49 . vertex @xmath19 has to be connected to the central vertex @xmath32 . edge @xmath63 should not be involved in crossings with edges @xmath47 and @xmath48 , since they are already involved in a right - angle crossing . if edge @xmath63 crosses edge @xmath64 , then the three vertices @xmath44 , @xmath45 and @xmath49 that define triangle @xmath57 must be collinear , which leads to a contradiction . therefore , triangle @xmath57 can not accommodate any other vertex ( except @xmath43 ) . now observe that each vertex of quadrilateral @xmath42 has degree five and there do not exist three vertices of quadrilateral @xmath42 , that have a common neighbor ( see figure [ fig : basic - gadget-1 ] ) . these properties trivially hold for vertex @xmath43 , since @xmath65 . based on the above properties , each neighbor of vertex @xmath43 can lie either in the interior of the dark - gray region of figure [ fig : v0nocross - b ] , or , on the external face of the already constructed drawing ( along the dashed semi - lines @xmath66 and @xmath67 of figure [ fig : v0nocross - b ] , respectively ) . this implies that we can route only four vertices out of those incident to vertex @xmath43 , i.e. , one of them should lie in the light - gray colored region of figure [ fig : v0nocross - b ] and thus , it can not be connected to vertex @xmath32 . in any rac drawing of the augmented square antiprism graph , quadrilateral @xmath40 , @xmath41 is drawn planar . [ lem : p6planar ] let @xmath42 be one of quadrilaterals @xmath40 , @xmath41 , and let , as in the previous lemma , @xmath43 , @xmath44 , @xmath45 and @xmath46 be its vertices , consecutive along quadrilateral @xmath42 . assume to the contrary that in a rac drawing of the augmented square antiprism graph , quadrilateral @xmath42 is not drawn planar , and say that edges @xmath68 and @xmath69 form a right - angle crossing . this case is illustrated in figure [ fig : q4planar ] . in the following , we will lead to a contradiction the cases , where central vertex @xmath32 lies ( i ) in the interior of a triangular face of quadrilateral @xmath42 , and , ( ii ) on the external face of quadrilateral @xmath42 . note that it is not feasible a non - planar rac drawing of a quadrilateral to contain more than one ( right - angle ) crossing . hence , its bounded faces are triangular . * * case i : * _ vertex @xmath32 lies in the interior of a triangular face of quadrilateral @xmath42 . _ assume without loss of generality that vertex @xmath32 ( which is incident to all vertices of quadrilateral @xmath42 ) lies in the interior of the triangular face formed by vertices @xmath44 , @xmath45 and the intersection point of edges @xmath68 and @xmath69 , as in figure [ fig : q4planar-2 ] . in this case , edges @xmath47 , @xmath68 and @xmath69 mutually cross , which leads to a contradiction , due to property [ prp : three - crossing - edges ] . * * case ii : * _ vertex @xmath32 lies on the external face of quadrilateral @xmath42 . _ this case is illustrated in figure [ fig : q4planar-3 ] . recall that by lemma [ lem : v0nocross ] , vertex @xmath32 can not introduce additional crossings on quadrilateral @xmath42 . we will first show that the common neighbor @xmath70 of vertices @xmath43 and @xmath44 can not lie in the region `` above '' line @xmath71 . in the case , where vertex @xmath70 lies in the region `` above '' @xmath71 and to the `` left '' of both edge @xmath69 and semi - line @xmath72 , edge @xmath73 would cross edge @xmath69 , which is not permitted by property [ prp : three - crossing - edges ] . symmetrically , vertex @xmath70 can not lie in the region `` above '' @xmath71 and to the `` right '' of both edge @xmath68 and semi - line @xmath74 . if vertex @xmath70 lies within the left gray - colored unbounded region of figure [ fig : q4planar-3 ] ( that is formed by semi - lines @xmath74 , @xmath75 ) , then , edge @xmath76 crosses two non - parallel edges @xmath58 and @xmath58 . in the case where , @xmath70 lies in the right gray - colored unbounded region of figure [ fig : q4planar-3 ] ( that is formed by semi - lines @xmath72 , @xmath77 ) , then @xmath78 either crosses both @xmath58 and @xmath69 which are non - parallel , or crosses edge @xmath62 forming a non - right angle crossing . in the case , where vertex @xmath70 lies in the interior of the triangle formed by vertices @xmath44 , @xmath45 and the intersection point of edges @xmath68 and @xmath69 , edge @xmath78 would cross edge @xmath69 , which leads to a violation property [ prp : three - crossing - edges ] . therefore , vertex @xmath70 should be `` below '' @xmath71 . + we continue our reasoning on vertex @xmath70 . vertex @xmath70 can not lie to the `` left '' of edge @xmath47 , since edge @xmath73 or @xmath78 would cross more than one ( non - parallel ) edges incident to vertex @xmath32 . if vertex @xmath70 lies to the `` right '' of edge @xmath79 , then edge @xmath78 either crosses edge @xmath69 , that it is not permitted by property [ prp : three - crossing - edges ] , or , both edges @xmath79 and @xmath80 , that are non - parallel . we complete our reasoning on vertex @xmath70 by the triangle formed by vertices @xmath32 , @xmath44 and @xmath45 . in this case , @xmath78 should be perpendicular to edge @xmath81 . this suggests that the angle formed by edges @xmath69 and @xmath80 is greater that @xmath82 and therefore , edge @xmath62 either crosses @xmath68 , or another edge of quadrilateral @xmath42 , which trivially leads to a contradiction , due to property [ prp : three - crossing - edges ] , or due to lemma [ lem : v0nocross ] , respectively . based on the above , vertex @xmath70 should be within the left gray - colored region of figure [ fig : q4planar-3 ] , along semi - line @xmath83 . following a similar reasoning scheme , as for vertex @xmath70 , we can prove that the common neighbor @xmath60 of vertices @xmath45 and @xmath46 should lie within the right light - gray colored region of figure [ fig : q4planar-3 ] , along semi - line @xmath84 . however , in this case , a common neighbor of vertices @xmath70 and @xmath60 , say @xmath49 , should lie on the intersection of semi - lines @xmath85 and @xmath84 , which leads to edge overlaps . thus , there exists no feasible placement for vertex @xmath49 . in any rac drawing of the augmented square antiprism graph , the central vertex @xmath32 lies in the interior of quadrilateral @xmath40 , @xmath41 . [ lem : v0internal ] from lemma [ lem : p6planar ] , it follows that quadrilateral @xmath40 should be drawn planar , for each @xmath41 . in order to prove this lemma , we assume to the contrary that central vertex @xmath32 lies on the exterior of one of the two quadrilaterals . say , w.l.o.g . , on the exterior of quadrilateral @xmath33 . let @xmath43 , @xmath44 , @xmath45 and @xmath46 be @xmath33 s vertices , consecutive along quadrilateral @xmath33 . then , by lemma [ lem : v0nocross ] , vertex @xmath32 can not contribute additional crossings on quadrilateral @xmath33 . this suggests that the drawing of the graph induced by quadrilateral @xmath33 and vertex @xmath32 will resemble the one depicted in figure [ fig : v0internal - gen ] . we denote by @xmath86 the triangle formed by vertex @xmath32 and the two vertices , which are on the convex hall of @xmath33@xmath87 ( refer to the gray - colored triangle of figure [ fig : v0internal - gen ] ) . and @xmath32 has to resemble to this one . ] before we proceed with the detailed proof of this lemma , we recall some properties of the augmented square antiprism graph . two consecutive vertices of @xmath33 ( @xmath34 ) share a common vertex of quadrilateral @xmath34 ( @xmath33 ) . each vertex of quadrilateral @xmath33 ( @xmath34 ) should be connected to two consecutive vertices of quadrilateral @xmath34 ( @xmath33 ) . we will prove that ( i ) no vertex of @xmath34 lies outside @xmath86 , ( ii ) @xmath34 can not entirely lie in the interior of @xmath33 , ( iii ) @xmath34 can not entirely lie in the interior of a triangular face of @xmath86 , ( iv ) @xmath34 can not entirely lie within two adjacent triangular faces of @xmath86 , ( v ) @xmath34 can not cross @xmath33 , such that some of the vertices of @xmath34 reside within a triangular face of @xmath86 , whereas the remaining ones within @xmath33 . note that quadrilateral @xmath34 can not entirely lie within three triangular faces of @xmath86 incident to vertex @xmath32 . case i : : : we prove that no vertex of quadrilateral @xmath34 lies on the external face of the graph induced by quadrilateral @xmath33 and vertex @xmath32 , i.e. , outside @xmath86 . for the sake of contradiction , assume that there exists a vertex of quadrilateral @xmath34 , say @xmath70 , that lies on the external face of the graph induced by quadrilateral @xmath33 and vertex @xmath32 ( see figure [ fig : q4external ] ) . vertex @xmath70 should be connected to vertices @xmath43 and @xmath44 of quadrilateral @xmath33 , and to the central vertex @xmath32 . if both vertices @xmath43 and @xmath44 are inside triangle @xmath86 , then vertex @xmath70 , which is assumed to lie on the external face of this graph , would violate property [ prp : triangle - edges ] , since vertices @xmath43 and @xmath44 would lie in the interior of @xmath86 , whereas vertex @xmath70 outside . therefore , at least one of vertices @xmath43 and @xmath44 should be a corner of @xmath86 . then , vertex @xmath70 contributes either none ( see figure [ fig : q4external-2 ] ) , or a single right - angle crossing ( see figure [ fig : q4external-1 ] ) . + + let now @xmath49 be a vertex of quadrilateral @xmath34 , which is incident to vertex @xmath70 . vertex @xmath49 is also incident to two consecutive vertices of quadrilateral @xmath33 , i.e. , @xmath44 and @xmath45 . we first turn our attention in the case where @xmath70 contributes a single right - angle crossing on the graph induced by quadrilateral @xmath33 and vertex @xmath32 ( see figure [ fig : q4external-1 ] ) . then , by property [ prp : triangle - edges ] , vertex @xmath49 should lie in the interior of triangle @xmath86 . this immediately leads to a contradiction , since edge @xmath88 should cross edge @xmath47 , which is already involved in a right - angle crossing ( see figure [ fig : q4external-1 ] ) . + consider now the case where vertex @xmath70 contributes no crossing on @xmath86 . observe that vertex @xmath49 , which is adjacent to vertex @xmath70 , vertices @xmath44 and @xmath45 of @xmath33 , and the central vertex @xmath32 can not lie in the dark - gray region of figure [ fig : q4external-2 ] , since in this case , edge @xmath68 would be crossed by more than one ( non - parallel ) edges , incident to @xmath49 . the case where vertex @xmath49 lies within the light - gray colored region of figure [ fig : q4external-2 ] , leads to a situation similar to the one depicted in figure [ fig : q4external-1 ] . therefore , vertex @xmath49 should lie `` somewhere '' in the interior of @xmath86 . let @xmath89 be the common neighbor of vertices @xmath43 and @xmath46 of @xmath33 , and vertex @xmath70 . this vertex can not lie within the dark - gray region of figure [ fig : q4external-2 ] , for the same reason that vertex @xmath49 could nt . in addition , vertex @xmath89 can not lie in the interior of @xmath86 , since in this case , both vertices @xmath49 and @xmath89 ( that are in @xmath86 ) , should be connected to @xmath70 ( that is not in @xmath86 ) , which trivially violates property [ prp : triangle - edges ] . therefore , vertex @xmath89 should be on the external face of the graph induced by @xmath33 and vertices @xmath32 and @xmath89 , along semi - line @xmath90 . however , in this case , we are also led to a situation similar to the one depicted in figure [ fig : q4external-1 ] , and subsequently , to a contradiction . case ii : : : say that quadrilateral @xmath34 entirely lies within quadrilateral @xmath33 ( see figure [ fig : q4internal-1 ] ) . in this case , its vertices should be connected to vertex @xmath32 . for three vertices of quadrilateral @xmath34 , this can be accomplished using the three available edges of quadrilateral @xmath33 ( refer to the dotted edges of figure [ fig : q4internal-1 ] ) , such that the right - angle crossings occur along them . however , the fourth vertex can not be connected to vertex @xmath32 , since only three edges of quadrilateral @xmath33 can be used to realize connections with vertex @xmath32 ( see the topmost edge of figure [ fig : q4internal-1 ] ) . case iii : : : assume now that quadrilateral @xmath34 entirely lies within a triangular face , say @xmath91 , incident to vertex @xmath32 ( see figure [ fig : q4internal-2 ] ) . then , there exists at least one vertex of quadrilateral @xmath33 , say vertex @xmath92 , which is incident to two vertices of quadrilateral @xmath34 , and is not identified with a vertex at the corners of @xmath91 ( see figure [ fig : q4internal-2 ] ) . vertex @xmath92 has to be connected to two vertices of quadrilateral @xmath34 . however , vertex @xmath92 is external to triangle @xmath91 , whereas its two incident vertices in the interior of this triangle , which leads to a violation of property [ prp : triangle - edges ] . case iv : : : say that quadrilateral @xmath34 entirely lies within two adjacent triangular faces , say @xmath93 and @xmath94 , incident to vertex @xmath32 ( see figure [ fig : q4internal-3 ] ) . then , quadrilateral @xmath34 should be `` perpendicular '' to the common edge of @xmath93 and @xmath94 . recall that two consecutive vertices of quadrilateral @xmath34 share a common vertex of quadrilateral @xmath33 . hence , we can find a vertex @xmath92 of quadrilateral @xmath33 , which is not identified with the common vertex of @xmath93 and @xmath94 , and is incident to a pair of vertices of quadrilateral @xmath34 , that do not lie in the same triangular face ( i.e. , the topmost vertices of quadrilateral @xmath34 or the bottommost vertices of quadrilateral @xmath34 in figure [ fig : q4internal-3 ] ) . this leads to a contradiction , since the common edge of @xmath93 and @xmath95 can not be crossed , as it is already involved in a right - angle crossing ( refer to the dotted - edges of figure [ fig : q4internal-3 ] ) . case v : : : we consider the case where quadrilateral @xmath34 crosses quadrilateral @xmath33 , such that some of the vertices of quadrilateral @xmath34 reside within a triangular face of @xmath86 , whereas the remaining ones within quadrilateral @xmath33 . we will lead to a contradiction the cases where : ( i ) two vertices of @xmath34 lie in the interior of a single triangular face incident to @xmath32 , ( ii ) two vertices of @xmath34 lie in the interior of two adjacent triangular faces , ( iii ) three vertices of @xmath34 lie in the interior of two adjacent triangular faces and two of them lie in the same triangular face of @xmath86 , ( iv ) three vertices of @xmath34 lie in the interior of three pairwise - adjacent triangular faces incident to vertex @xmath32 . recall that none of the vertices of @xmath34 lies in the external face of the graph induced by quadrilateral @xmath33 and vertex @xmath32 . let , with a slight abuse of notation , @xmath96 , @xmath97 , @xmath98 and @xmath99 be the vertices of quadrilateral @xmath34 . assume first that vertices @xmath96 and @xmath97 are in the interior of a single triangular face , whereas vertices @xmath98 and @xmath99 in the interior of quadrilateral @xmath33 ( see figure [ fig : q4internal-4 ] ) . in this case , edges @xmath100 and @xmath101 should perpendicularly cross quadrilateral @xmath33 . the connections between vertices @xmath98 and @xmath99 with vertex @xmath32 can be accomplished using two of the available edges of quadrilateral @xmath33 , such that the right - angle crossings occur along them ( refer to dotted edges of figure [ fig : q4internal-4 ] ) . thus , the triangular faces that are adjacent to the one that accommodates vertices @xmath96 and @xmath97 ( refer to the light - gray faces of figure [ fig : q4internal-4 ] ) can not be further used to connect vertices of quadrilateral @xmath33 to vertices of quadrilateral @xmath34 . then , there exists a vertex of quadrilateral @xmath33 , say @xmath92 , that it is not identified with any of the vertices of the face that accommodates vertices @xmath96 and @xmath97 , and either @xmath96 or @xmath97 has to be connected to vertex @xmath92 . however , this can not be accomplished , since the edge from either @xmath96 or @xmath97 to vertex @xmath92 would cross more than one non - parallel edges ( refer to the dashed edge of figure [ fig : q4internal-4 ] ) . + and @xmath97 are in the interior of a single triangular face incident to @xmath32 . ] + say now that vertices @xmath96 and @xmath97 are in the interior of two adjacent triangular faces incident to vertex @xmath32 , whereas vertices @xmath98 and @xmath97 within quadrilateral @xmath33 . this case is illustrated in figure [ fig : q4internal-5 ] . then , one of the vertices that lie in the interior of quadrilateral @xmath33 , say vertex @xmath99 , can be connected to vertex @xmath32 using one of the available edges of quadrilateral @xmath33 ( refer to the dotted edge of figure [ fig : q4internal-5 ] ) . however , vertex @xmath98 can not be connected to vertex @xmath32 , since only three of the edges of quadrilateral @xmath33 can be used to realize connections from the vertices that lie within quadrilateral @xmath33 , to vertex @xmath32 . + consider now the case where three vertices , say @xmath96 , @xmath97 and @xmath98 of quadrilateral @xmath34 are in the interior of two adjacent triangular faces incident to vertex @xmath32 , and two of vertices @xmath96 , @xmath97 and @xmath98 , say w.l.o.g . , @xmath97 and @xmath98 , lie in the same triangular face ( see figure [ fig : q4internal-7 ] ) . then , vertex @xmath99 , as in the previous case , can be connected to vertex @xmath32 using the `` last '' available edge of quadrilateral @xmath33 ( refer to the dotted edge of figure [ fig : q4internal-7 ] ) . however , in this case , there exists a vertex of quadrilateral @xmath33 , say @xmath92 , that it is not identified with any of the vertices of the face that accommodates vertices @xmath96 , @xmath97 and @xmath98 , which has to be connected to one of the vertices @xmath96 , @xmath97 or @xmath98 . however , this can not be accomplished , since an edge from either vertex @xmath96 , or @xmath97 , or @xmath98 , to vertex @xmath92 would cross more than one non - parallel edges ( refer to the dashed edge of figure [ fig : q4internal-7 ] ) . + the last case we have to consider is the one where three vertices , say @xmath96 , @xmath97 and @xmath98 of quadrilateral @xmath34 are in the interior of three pairwise - adjacent triangular faces incident to vertex @xmath32 , whereas the fourth vertex @xmath99 resides within quadrilateral @xmath34 ( see figure [ fig : q4internal-6 ] ) . in this case , vertex @xmath99 has to use the fourth edge of quadrilateral @xmath33 to reach vertex @xmath32 , which leads to a contradiction , since only three of the edges of quadrilateral @xmath33 can be used to realize connections from the vertices that lie within quadrilateral @xmath33 , to vertex @xmath32 . thus , we have considered all possible placements of @xmath34 , with vertex @xmath32 outside of @xmath33 , and are all led to a contradiction . we conclude that vertex @xmath32 is in the interior of quadrilateral @xmath33 ( and symmetrically in the interior of @xmath34 , too ) . there does not exist a rac drawing of the augmented square antiprism graph where an edge emanating from vertex @xmath32 towards a vertex of quadrilateral @xmath40 , @xmath41 , crosses quadrilateral @xmath40 . [ lem : v0internalnocross ] by lemma [ lem : v0internal ] , vertex @xmath32 should lie in the interior of quadrilateral @xmath40 , @xmath41 , which is drawn planar due to lemma [ lem : p6planar ] . assume to the contrary that in a rac drawing of the augmented square antiprism graph , an edge emanating from vertex @xmath32 towards a vertex of quadrilateral @xmath33 , say @xmath43 , crosses an edge , say @xmath69 , of quadrilateral @xmath33 ( see figure [ fig : planar - convex ] ) . consider vertex @xmath60 , which is incident to vertices @xmath45 and @xmath46 of quadrilateral @xmath34 . vertex @xmath60 can not lie `` above '' line @xmath102 and to the `` left '' of semi - line @xmath103 , since it can not be connected to vertex @xmath32 . in addition , it can not lie `` above '' line @xmath102 and to the `` right '' of semi - line @xmath103 , since in this case it can not be connected to either vertex @xmath45 or @xmath32 . furthermore , vertex @xmath60 can not be in the interior of the triangle formed by vertices @xmath32 , @xmath45 and @xmath46 , as it would not be feasible to be connected either to vertex @xmath45 or @xmath46 , since in either case , it crosses edge @xmath47 . also , @xmath60 can not be in the region formed by line @xmath102 and edges @xmath62 and @xmath79 , as it could not be connected to vertex @xmath45 . thus , vertex @xmath60 should lie in the light - gray triangular face of figure [ fig : planar - convex ] , along semi - line @xmath104 . following a similar reasoning scheme , we can prove that vertex @xmath89 , which is incident to vertices @xmath43 , @xmath46 of quadrilateral @xmath33 and vertex @xmath70 of quadrilateral @xmath34 , due to its adjacency with @xmath43 , @xmath46 , can lie in the face formed by vertices @xmath43 , @xmath44 , @xmath46 and the intersection point of edges @xmath105 and @xmath79 . however , under this restriction , vertex @xmath89 can not be connected to vertex @xmath60 , without crossing edge @xmath79 , which is already involved in a right - angle crossing ( refer to the dashed edge of figure [ fig : planar - convex ] ) . towards a vertex of @xmath33 , can not cross @xmath33 . ] there does not exist a rac drawing of the augmented square antiprism graph in which quadrilateral @xmath33 intersects @xmath34 . [ lem : quahexposition ] from lemmata [ lem : p6planar ] , [ lem : v0internal ] and [ lem : v0internalnocross ] , it follows that the graph induced by quadrilateral @xmath33 and vertex @xmath32 is drawn planar with vertex @xmath32 in the interior of both quadrilaterals @xmath33 and @xmath34 . therefore , it should resemble the one illustrated in figure [ fig : v0internalq4external-1 ] . in order to prove this lemma , we will contradict the following cases : ( i ) @xmath33 and @xmath34 cross and none of the vertices of @xmath34 is in the interior of quadrilateral @xmath33 , ( ii ) two vertices of @xmath34 lie in the interior of @xmath33 and @xmath34 crosses either a single edge of @xmath33 , or two edges of @xmath33 , ( iii ) three vertices of @xmath34 lie in the interior of @xmath33 , ( iv ) only one vertex of @xmath34 lies in the interior of @xmath33 . we first assume that quadrilateral @xmath33 and quadrilateral @xmath34 cross and none of the vertices of quadrilateral @xmath34 is in the interior of quadrilateral @xmath33 ( see figure [ fig : v0internalq4external-6 ] ) . in this case , an edge of quadrilateral @xmath34 , say @xmath106 , which is involved in the crossing , divides quadrilateral @xmath33 into two regions , say @xmath107 and @xmath108 . obviously , edge @xmath106 should cross parallel edges of quadrilateral @xmath33 . then , vertex @xmath32 , which lies in the interior of quadrilateral @xmath33 and is incident to all vertices of quadrilateral @xmath33 can not reside to none of @xmath107 and @xmath108 , without introducing a non - right angle crossing with edge @xmath106 . we proceed to consider the case where quadrilaterals @xmath33 and @xmath34 cross and some of the vertices of quadrilateral @xmath34 are in the interior of quadrilateral @xmath33 , whereas the remaining ones on its exterior . let @xmath96 , @xmath97 , @xmath98 and @xmath99 be the vertices of quadrilateral @xmath34 . assume that @xmath96 and @xmath97 lie within quadrilateral @xmath33 , whereas @xmath98 and @xmath99 on its external face , such that edges @xmath100 and @xmath101 are perpendicular either to one edge of quadrilateral @xmath33 ( see figure [ fig : v0internalq4external-2 ] ) , or to two edges of @xmath33 ( see figure [ fig : v0internalq4external-3 ] ) . note that edges @xmath100 and @xmath101 can not be crossed by any other edge incident to both quadrilaterals , since they are already involved in right - angle crossings . however , all vertices of quadrilateral @xmath33 have to be connected to vertex @xmath32 . assuming that one vertex of quadrilateral @xmath33 can utilize the `` last '' available edge of quadrilateral @xmath34 ( i.e. , edge @xmath109 ) to reach vertex @xmath32 ( refer to the dotted edges of figures [ fig : v0internalq4external-2 ] and [ fig : v0internalq4external-3 ] ) , there exists at least one vertex of @xmath33 , say vertex @xmath92 , that can not be connected to @xmath32 , without introducing non right - angle crossing ( refer to the dashed edges of figures [ fig : v0internalq4external-2 ] and [ fig : v0internalq4external-3 ] ) . following a similar reasoning scheme as for the previous cases , we can prove that the cases where ( i ) three vertices of @xmath34 , say w.l.o.g . , @xmath96 , @xmath97 and @xmath98 , lie in the interior of @xmath33 ( see figure [ fig : v0internalq4external-8 ] ) , and ( ii ) only one vertex of @xmath34 , say w.l.o.g , vertex @xmath97 , lies in the interior of @xmath33 ( see figure [ fig : v0internalq4external-9 ] ) , are led to a contradiction . [ thm : uniqueness ] any straight - line rac drawing of the augmented square antiprism graph has two combinatorial embeddings . so far , we have managed to prove that both quadrilaterals @xmath33 and @xmath34 are drawn planar , do not cross , and have central vertex @xmath32 to their interior . this suggests that either quadrilateral @xmath33 is in the interior of @xmath34 , or quadrilateral @xmath34 is in the interior of @xmath33 . however , in both cases , vertex @xmath32 , which has to be connected to the four vertices of the `` external '' quadrilateral , should inevitably perpendicularly cross the four edges of the `` internal '' quadrilateral , and this trivially implies only two feasible combinatorial embeddings . we extend the augmented square antiprism graph , by appropriately `` glueing '' multiple instances of it , the one next to the other , either horizontally or vertically . figure [ fig : basic - gadget - extension ] demonstrates how a horizontal extension of two instances , say @xmath17 and @xmath110 , is realized , i.e. , by identifying two `` external '' vertices , say @xmath20 and @xmath111 , of @xmath17 with two `` external '' vertices of @xmath110 ( refer to the gray - colored vertices of figure [ fig : basic - gadget - extension ] ) , and by employing an additional edge ( refer to the dashed drawn edge of figure [ fig : basic - gadget - extension ] ) , which connects an `` internal '' vertex , say @xmath19 , of @xmath17 with the corresponding `` internal '' vertex , say @xmath112 , of @xmath110 . let @xmath113 be the graph produced by a horizontal or vertical extension of @xmath17 and @xmath110 . since each of @xmath17 and @xmath110 has two rac combinatorial embeddings each , one would expect that @xmath113 would have four possible rac combinatorial embeddings . we will show that this is not true and , more precisely , that there only exists a single rac combinatorial embedding . let @xmath17 and @xmath110 be two instances of the augmented square antiprism graph . then , @xmath113 has a unique rac combinatorial embedding . assume first that in a rac drawing of @xmath113 , vertices @xmath20 and @xmath111 are on the external quadrilateral of @xmath17 and graph @xmath110 is drawn completely in the interior of @xmath17 ( see figure [ fig : basic - gadget - extension - no - cross ] ; since @xmath20 and @xmath111 are on the external face of @xmath110 , vertices @xmath7 and @xmath114 in figure [ fig : basic - gadget - extension - no - cross ] should also be on the external face of @xmath110 ) . first observe that vertex @xmath112 of @xmath110 , which is incident to vertices @xmath20 and @xmath111 , can not reside to the `` left '' of both edges @xmath115 and @xmath116 ( refer to the bold drawn edges of figure [ fig : basic - gadget - extension - no - cross ] ) , since this would lead to a situation where three edges mutually cross and , subsequently , to a violation of property [ prp : three - crossing - edges ] ( see the gray - colored square vertex of figure [ fig : basic - gadget - extension - no - cross ] ) . therefore , vertex @xmath112 should lie within the triangular face of @xmath17 formed by vertices @xmath19 , @xmath20 and @xmath111 . the same similarly holds for the central vertex of @xmath110 , which is also incident to vertices @xmath20 and @xmath111 . by property [ prp : triangle - edges ] , any common neighbor of vertices @xmath112 and @xmath20 should also lie within the same triangular face of @xmath17 , which progressively implies that entire graph @xmath110 should reside within this face , as in figure [ fig : basic - gadget - extension - no - cross ] . however , in this case and since @xmath112 is incident to @xmath20 and @xmath111 , edge @xmath117 , which is used on a horizontal or a vertical extension , crosses the interior of @xmath110 , which is not permitted . this suggests that graph @xmath110 should be on the exterior of @xmath17 . now assume that vertices @xmath20 and @xmath111 , which are identified , during a horizontal or vertical extension , are along the internal quadrilateral of @xmath17 in a rac drawing of @xmath113 . this is illustrated in figure [ fig : basic - gadget - extension - internal ] . then , the edge , say @xmath118 , which perpendicularly crosses edge @xmath119 and emanates from the external quadrilateral towards the central vertex of @xmath17 ( refer to the bold solid edge of figure [ fig : basic - gadget - extension - internal ] ) will be involved in crossings with @xmath110 . more precisely , we focus on vertex @xmath112 of @xmath110 , which is incident to vertices @xmath20 and @xmath111 . these edges will inevitably introduce non - right angle crossings , since one of them should cross edge @xmath118 . therefore , the vertices that are identified , during a horizontal or vertical extension , should always be on the external face of each augmented square antiprism graph and , subsequently , the drawing of the graph produced by a horizontal or vertical extension will resemble the one of figure [ fig : basic - gadget - extension ] , i.e. , it has a unique embedding . note that the extension which is given in figure [ fig : basic - gadget - extension ] , is ideal . in the general case , at each extension step the new instance of the augmented square antiprism graph may introduce a `` turn '' , as in figure [ fig : basic - gadget - turned ] . we observe that by `` glueing '' a new instance of the augmented square antiprism graph on @xmath113 either by a horizontal or a vertical extension , we obtain another graph of unique rac combinatorial embedding . in this way , we can define an infinite class of graphs of unique rac combinatorial embedding . this is summarized in the following theorem . there exists a class of graphs of unique rac combinatorial embedding . it is @xmath1-hard to decide whether an input graph admits a straight - line rac drawing . we will reduce the well - known @xmath31-sat problem @xcite to the straight - line rac drawing problem . in a @xmath31-sat instance , we are given a formula @xmath120 in conjunctive normal form with variables @xmath121 and clauses @xmath122 , each with three literals . we show how to construct a graph @xmath123 that admits a straight - line rac drawing @xmath124 if and only if formula @xmath120 is satisfiable . figure [ fig : gadgets ] illustrates the gadgets of our construction . each gray - colored square in these drawings corresponds to an augmented square antiprism graph . adjacent gray squares form an extension ( refer , for example , to the topmost gray squares of figure [ fig : gadgets]a , which form a `` horizontal '' extension ) . there also exist gray squares that are not adjacent , but connected with edges . the legend in figure [ fig : gadgets ] describes how the connections are realized . the gadget that encodes variable @xmath125 of formula @xmath120 is given in figure [ fig : gadgets]a . the gadget of variable @xmath125 consists of a combination of augmented square antiprism graphs , and , `` horizontal '' and `` vertical '' edges , which form a tower , whose rac drawing has unique combinatorial embedding . one side of the tower accommodates multiple vertices that correspond to literal @xmath125 , whereas its opposite side accommodates vertices that correspond to literal @xmath126 ( refer to vertices @xmath127 and @xmath128 in figure [ fig : gadgets]a ) . these vertices are called _ variable endpoints_. then , based on whether on the final drawing the negated vertices will appear to the `` left '' or to the `` right '' side of the tower , we will assign a true or a false value to variable @xmath125 , respectively . pairs of consecutive endpoints @xmath129 and @xmath130 are separated by a _ corridor _ ( see figure [ fig : gadgets]a ) , which allows perpendicular edges to pass through it ( see the bottommost dashed arrow of figure [ fig : gadgets]a ) . note that this is not possible through a `` corridor '' formed on a variable endpoint , since there exist four non - parallel edges that `` block '' any other edge passing through them ( see the topmost dashed arrow of figure [ fig : gadgets]a ) . the corridors can have variable height . in the variable gadget of variable @xmath125 , there are also two vertices ( they are drawn as gray circles in figure [ fig : gadgets]a ) , which have degree four . these vertices serve as `` _ _ connectors _ _ '' among consecutive variable gadgets , i.e. , these vertices should be connected to their corresponding vertices on the variable gadgets of variables @xmath131 and @xmath132 . note that the connector vertices of the variable gadgets associated with variables @xmath133 and @xmath134 are connected to connectors of the variable gadgets that correspond to variables @xmath135 and @xmath136 , respectively , and to connectors of _ dummy variable gadgets_. figure [ fig : gadgets]b illustrates a dummy variable gadget , which ( similarly to the variable gadget ) consists of a combination of augmented square antiprism graphs , and , `` horizontal '' and `` vertical '' edges , which form a tower . any rac drawing of this gadget has also unique combinatorial embedding . a dummy variable gadget does not support vertices that correspond to literals . however , it contains connector vertices ( they are drawn as gray circles in figure [ fig : gadgets]b ) . in our construction , we use exactly two dummy variable gadgets . the connector vertices of each dummy variable gadget should be connected to their corresponding connector vertices on the variable gadgets associated with variables @xmath133 and @xmath134 , respectively . the gadget that encodes the clauses of formula @xmath120 is illustrated in figure [ fig : gadgets]c and resembles to a valve . let @xmath137 be a clause of @xmath120 . as illustrated in figure [ fig : gadgets]c , the gadget which corresponds to clause @xmath138 contains three vertices and vertices of @xmath123 . ] , say @xmath139 , @xmath140 , and @xmath141 , such that : @xmath139 has to be connected to @xmath142 , @xmath140 to @xmath143 and @xmath141 to @xmath144 . these vertices , referred to as the _ clause endpoints _ , encode the literals of each clause . obviously , if a clause contains a negated literal , it should be connected to the negated endpoint of the corresponding variable gadget . the clause endpoints are incident to a vertex `` trapped '' within two parallel edges ( refer to the bold drawn edges of figure [ fig : gadgets]c ) . therefore , in a rac drawing of @xmath145 , only two of them can perpendicularly cross these edges , one from top ( _ top endpoint _ ) and one from bottom ( _ bottom endpoint _ ) . the other one ( _ right endpoint _ ) should remain in the interior of the two parallel edges . the one that will remain `` trapped '' on the final drawing will correspond to the true literal of this clause . the gadgets , which correspond to variables and clauses of @xmath120 , are connected together by the skeleton of graph @xmath145 , which is depicted in figure [ fig : skeleton]a . the skeleton consists of two main parts , i.e. , one `` horizontal '' and one `` vertical '' . the vertical part accommodates the clause gadgets ( see figure [ fig : skeleton]a ) . the horizontal part will be used in order to `` plug '' the variable gadgets . the long edges that perpendicularly cross ( refer to the crossing edges slightly above the horizontal part in figure [ fig : skeleton]a ) , imply that the vertical part should be perpendicular to the horizontal part . the horizontal part of the skeleton is separately illustrated in figure [ fig : skeleton]b . observe that it contains one set of horizontal lines . figure [ fig : example ] shows how the variable gadgets are attached to the skeleton . more precisely , this is accomplished by a single edge , which should perpendicularly cross the set of the horizontal edges of the horizontal part . therefore , each variable gadget is perpendicularly attached to the skeleton , as in figure [ fig : example ] . note that each variable gadget should be drawn completely above of these horizontal edges , since otherwise the connections among variable endpoints and clause endpoints would not be feasible . the connector vertices of the dummy variable gadgets , the variable gadgets and the vertical part of the construction , ensure that the variable gadgets will be parallel to each other ( i.e. , they are not allowed to bend ) and parallel to the vertical part of the construction . -sat to the straight - line rac drawing problem . the input formula is @xmath146 . the drawing corresponds to the truth assignment @xmath133=@xmath147=true , @xmath135=false . ] we now proceed to investigate some properties of our construction . any path of length two that emanates from a top- or bottom - clause endpoint can reach a variable endpoint either on the left or on the right side of its associated variable gadget . the first edge of this path should perpendicularly cross the vertical edges of the vertical part of the construction and pass through some corridors , the corridors are the gray - colored regions that reside at each variable gadget . ] , whereas the second edge will be used to realize the `` final '' connection with the variable gadget endpoint ( see figure [ fig : example ] ) . however , the same does nt hold for the paths that emanate from a right - clause endpoint . these paths can only reach variable endpoints on the right side of their associated variable gadgets . more precisely , the first edge of the @xmath148-length path should cross one of the two parallel edges ( refer to the bold drawn edges of figure [ fig : gadgets]c ) that `` trap '' it , whereas the other one should be used to reach ( passing through variable corridors ) its variable endpoint ( see figure [ fig : example ] ) . our construction ensures that up to translations , rotations and stretchings any rac drawing of @xmath123 resembles the one of figure [ fig : skeleton ] . it is clear that the construction can be completed in @xmath149 time . assume now that there is a rac drawing @xmath124 of @xmath123 . if the negated vertices of the variable gadget that corresponds to @xmath125 , @xmath150 , lie to the `` left '' side in @xmath124 , then variable @xmath125 is set to true , otherwise @xmath125 is set to false . we argue that this assignment satisfies @xmath120 . to realize this , observe that there exist three paths that emanate from each clause gadget . the one that emanates from the right endpoint of each clause gadget can never reach a false value . therefore , each clause of @xmath120 must contain at least one true literal , which implies that @xmath120 is satisfiable . conversely , suppose that there is a truth assignment that satisfies @xmath120 . we proceed to construct a rac drawing @xmath124 of @xmath123 , as follows : in the case where , in the truth assignment , variable @xmath125 , @xmath150 is set to true , we place the negated vertices of the variable gadget that corresponds to @xmath125 , to its left side in @xmath124 , otherwise to its right side . since each clause of @xmath120 contains at least one true literal , we choose this as the right endpoint of its corresponding clause gadget . as mentioned above , it is always feasible to be connected to its variable gadgets by paths of length two . this completes our proof . in this paper , we proved that it is @xmath1-hard to decide whether a graph admits a straight - line rac drawing . didimo et al.@xcite proved that it is always feasible to construct a rac drawing of a given graph with at most three bends per edge . if we permit two bends per edge , does the problem remain @xmath1-hard ? it is also interesting to continue the study on the interplay between the number of edges and the required area in order to fill the gaps between the known upper and lower bounds . 10 angelini , p. , cittadini , l. , di battista , g. , didimo , w. , frati , f. , kaufmann , m. , symvonis , a. : on the perspectives opened by right angle crossing drawings . in : proc . of 17th international sympsioum on graph drawing ( gd09 ) . lncs , vol . 5849 , pp . 2132 ( 2009 ) arikushi , k. , fulek , r. , keszegh , b. , moric , f. , toth , c. : drawing graphs with orthogonal crossings . in : proc . 36th international workshop on graph theoretic concepts in computer science ( wg 2010 ) ( 2010 ) , to appear , e. , didimo , w. , liotta , g. , meijer , h. : area , curve complexity , and crossing resolution of non - planar graph drawings . in : proc . of 17th international sympsioum on graph drawing ( gd09 ) . lncs , vol . 5849 , pp . 1520 ( 2009 ) formann , m. , hagerup , t. , haralambides , j. , kaufmann , m. , leighton , f. , symvonis , a. , welzl , e. , woeginger , g. : drawing graphs in the plane with high resolution . siam journal of computing 22(5 ) , 10351052 ( 1993 ) |
we describe two dimensional models with a metallic fermi surface which display quantum phase transitions controlled by strongly interacting critical field theories below their upper critical dimension . the primary examples involve transitions with a topological order parameter associated with dislocations in collinear spin density wave ( `` stripe '' ) correlations : the suppression of dislocations leads to a fractionalization of spin and charge collective modes , and this transition has been proposed as a candidate for the cuprates near optimal doping . the coupling between the order parameter and long - wavelength volume and shape deformations of the fermi surface is analyzed by the renormalization group , and a runaway flow to a non - perturbative regime is found in most cases . a phenomenological scaling analysis of simple observable properties of possible second order quantum critical points is presented , with results quite similar to those near quantum spin glass transitions and to phenomenological forms proposed by schrder _ et al . _ ( nature * 407 * , 351 ( 2000 ) ) . | cond-mat0207167 | an important property of most quantum phase transitions studied to date in systems with a metallic fermi surface in spatial dimensions @xmath0 is that the critical field theory for the order parameter is a free gaussian theory @xcite . this result has its origin in the fact that the order parameters considered can be expressed as a fermion bilinears , and consequently the order parameter fluctuations are efficiently overdamped and suppressed by fermionic particle and hole excitations near the fermi surface . the temperature dependencies of physical observables near the quantum critical point have been perturbatively computed @xcite , and these can be understood as corrections to scaling@xcite at the gaussian critical point@xcite . the quantum and thermal fluctuations near quantum phase transitions in fermi systems have been used to interpret experiments in a variety of correlated electron materials . in the cuprate superconductors , the anomalous properties in the normal state have been described in terms of the proximity to a quantum phase transition associated with the onset of spin density wave ( sdw ) or charge density wave ( cdw ) order in a fermi liquid @xcite . however , the observed anomalous properties extend to quite high temperatures , and it would be preferable to explain these in theories with stronger non - linearities among the order parameter modes . further , many of the anomalous properties extend to the optimal doping regime where there is no strong evidence of a large correlation length of the sdw and cdw orders . this paper describes an alternative class of quantum critical points whose critical theories are non - gaussian , and which remain strongly coupled even in the presence of a fermi surface because they are below their upper critical dimension . only in such theories can the order parameter relaxation rate , and possibly the quasiparticle energy width , be generically equal to @xmath1 times a universal numerical constant@xcite ( @xmath2 is the absolute temperature ) . examples of such quantum phase transitions abound in insulators and superconductors@xcite , and some of these have been used to explain low temperature properties of the cuprate superconductors @xcite . this paper will discuss transition in metals with a fermi surface , and so the possible regime of applicability is restricted to higher temperatures where superconductivity is absent . our primary focus will be on a transition with a non - local `` topological '' order associated with certain defects in the sdw / cdw order . however , long - range sdw or cdw order will not be present on either side of the critical point . this transition ( shown in fig . [ phase ] below ) is being offered as a candidate@xcite for a possible optimal doping quantum critical point @xcite in the cuprate superconductors . this proposal for the optimal doping quantum critical point must be distinguished from a conventional sdw ordering transition in a _ superconductor _ discussed recently by one of us@xcite ; the latter transition occurs at lower doping concentrations ( near 1/8 ) and was used to predict and explain a number of recent _ low temperature _ neutron scattering and stm experiments . here we are interested in higher dopings and temperatures , where , as we have already noted , the correlation length of conventional sdw / cdw order is surely quite small , but , as we shall discuss below , correlations in a topological order may be far more robust . we also hope for applications of our theory in other correlated materials , like the heavy fermion systems , where many experiments are in disagreement with the conventional weakly coupled theories@xcite . in particular , we will show that the strongly - coupled critical points predict a scaling structure for physical observables which is quite similar to those proposed by schrder _ @xcite related scaling structures are also known to appear near strongly - coupled spin glass quantum critical points , and this is reviewed in appendix [ app : qsg ] . there is one simple route to obtaining a strongly coupled field theory in the presence of a fermi surface that has been discussed earlier @xcite : use restrictions from momentum conservation to prevent the order parameter from coupling linearly to low energy fermionic excitations . for example , a transition with the onset of sdw order at the wavevector @xmath3 , but the geometry of the fermi surface such that no two points on the fermi surface are separated by @xmath3 . in such a situation the order parameter excitations remain undamped , and the fermions appear to be innocuous bystanders to the transition . however , it turns out that the fermi surface can still be relevant for the structure of the critical theory in certain situations ; these effects are related to those that arise in the transitions we do discuss below , and so we defer discussion of this point until section [ sec : sdw ] . turning to transitions characterized by a non - local , ` topological ' order parameter , a situation that has been much discussed in the recent literature@xcite has the electron , @xmath4 , ( @xmath5 is a spin index ) fractionalize into a spinless charge @xmath6 boson , @xmath7 , and a neutral spin 1/2 fermion @xmath8 : @xmath9 this fractionalization transition is described by a @xmath10 gauge theory @xcite , and both @xmath7 and @xmath8 carry a unit gauge charge ; the physical fermion , @xmath4 , is , of course , gauge neutral . if we denote the @xmath10 gauge field by @xmath11 ( @xmath12 are site labels ) , then terms like @xmath13 will generically appear in the effective hamiltonian @xcite . in the fractionalized phase , where the fluctuations of @xmath11 are effectively quenched , this term is a bcs - like pairing of the @xmath8 fermions , and it implies that the fermi surface will generically be gapped out ( except possibly at special points if the pairing is anisotropic ) . consequently , this fractionalization transition does not provide us with a candidate quantum phase transition in the presence of a fermi surface , and we will not consider it further in this paper . the main focus of this paper shall be on a fractionalization transition that does less damage to the integrity of the electron : the electron ( and hence its fermi surface ) remains intact on both sides of the transition , but charge neutral collective modes ( in the particle - hole sector ) do fractionalize ; only at the quantum critical point , and in the associated non - zero temperature quantum critical region , is the electronic quasiparticle ill - defined . _ @xcite have recently given an appealing pictorial description of such a transition . a transition with closely related physical content was introduced in ref . ; we will follow the latter approach here , and will describe the physical content and experimental motivation below in section [ sec : op ] . there have also been earlier discussions of the fractionalization of order parameters in other physical contexts @xcite . after introducing the order parameters in section [ sec : op ] , we write down and classify quantum field theories for the critical points in section [ sec : qft ] . we pause briefly in section [ sec : sdw ] to consider the conventional transitions mentioned above , in which the fermion bilinear order parameter is not damped by fermi surface excitations because of restrictions arising from momentum conservation , and show that it is described by field theories similar to those in section [ sec : qft ] . the latter are subjected to a renormalization group analysis in section [ sec : rg ] , and the physical implications of the results are noted in section [ sec : phys ] . consider a correlated electronic system in two dimensions with enhanced sdw correlations at the wavevectors @xmath14 and @xmath15 ; these are vectors of equal length pointing along the @xmath16 and @xmath17 axes respectively . similar considerations apply to other choices for the ordering wavevectors , but we will focus on this case because it directly applies to the doped cuprates @xcite at doping concentrations above @xmath18 . we can write for the spin operator @xmath19 ( @xmath20 is a spatial co - ordinate , @xmath21 labels spin components , and @xmath22 is imaginary time ) : @xmath23 ; \label{e1}\ ] ] @xmath24 and @xmath25 are then the sdw order parameters . except of the case of two sublattice order ( @xmath26 or @xmath27 ) , these order parameters are complex numbers . concommitant with this sdw correlations , symmetry demands @xcite that there are also `` bond order '' correlations at half the wavelength : @xmath28 + \ldots \label{e2}\end{aligned}\ ] ] there is an implied summation over the repeated spin index @xmath29 . the vector @xmath30 represents a bond ( say the nearest neighbor vector ) , we have assumed that @xmath24 does not vary significantly over the spatial distance @xmath30 , and the ellipses denote numerous other similar terms which can be deduced from ( [ e1 ] ) . for @xmath30 a nearest - neighbor vector , @xmath31 measures the modulations in the exchange energy , while for @xmath32 , @xmath33 measures the local charge density ( for the @xmath34-@xmath35 model the charge density is linearly related to the on - site @xmath36 ) . so , quite generally , the bond order parameters , @xmath37 , measure modulations in all observables invariant under spin rotations and time reversal ( such the electron kinetic , pairing , or exchange energies ) . the modulation in the total charge density may be strongly suppressed by the long - range coulomb interactions , but this suppression is not expected to apply to observables with @xmath38 . we will loosely refer to such bond order modulations as cdw order in this paper , but the caveats discussed in this paragraph must be kept in mind . now examine the structure of the effective potential , @xmath39 controlling fluctuations of @xmath40 ( parallel considerations apply implicitly in the remainder of this subsection to @xmath41 ) . on general symmetry grounds @xcite , this potential has the form @xmath42 where @xmath43 , @xmath44 , @xmath45 are phenomenological landau parameters , and there is again an implied summation over repeated spin indices @xmath46 . in the usual landau theory , the transition to the onset long - range sdw order happens when @xmath43 first becomes negative , and the minimum of the potential moves from @xmath47 to a non - zero value . the theory of the critical point then accounts for fluctuations of @xmath40 about the value of @xmath48 , and these allow for strong fluctuations in both the amplitudes and the phases of the three complex fields @xmath49 . in much of the discussion on stripe physics in the literature ( see _ e.g. _ ref . ) , the physical picture advocated for the onset of long range sdw / cdw ( or `` stripe '' ) order is quite different . as one decreases the doping from a fermi liquid towards a mott insulator , initial local stripe - like correlations form , the the orientation and phase of these regions changes from point to point ; only when these locally ordered regions align with each other , is there long range sdw / cdw order . in terms of our order parameter formulation , this means that locally the amplitude @xmath50 first becomes large , and that long range order develops subsequently upon alignment of the orientation and phases of @xmath40 . in the familiar @xmath51 field theory of a _ real _ vector order parameter , this change in perspective from a `` soft - spin '' landau theory ( with strong amplitude fluctuations ) to a `` hard - spin '' perspective ( with amplitude fluctuations quenched ) does not make a fundamental difference in the physics being considered . the orientation fluctuations in the hard - spin approach are described by a non - linear @xmath52 model , whose critical properties can be analyzed by expansion away the lower critical dimension . conversely , the soft - spin approach leads to an expansion away from the upper critical dimension . however , these are merely complementary approaches to the same critical point , and the quantum numbers of the excitations in the phases flanking the critical point are the same in both approaches ; indeed a @xmath53 expansion ( @xmath54 is the number of field components ) can interpolate between the soft - spin and hard - spin theories , and also between the upper and lower critical dimensions @xcite . for the present situation with a _ complex _ vector order parameter the situation is dramatically different , and there is a fundamental difference between the soft - spin and hard - spin approaches to fluctuations implied by the effective potential ( [ vphi ] ) . the key difference is that not all order parameter configurations with a fixed @xmath55 are physically equivalent , and they can not all be rotated into each other by a symmetry transformation . a minimization of ( [ vphi ] ) at fixed @xmath56 shows two distinct classes of minima , chosen by the sign of @xmath45 ; their most general form is : @xmath57 by inserting into ( [ e1 ] ) , it is easy to see that ( i ) describes a spiral sdw for which the charge order in ( [ e2 ] ) vanishes , while ( ii ) is a collinear sdw with a modulation in the length of the spin as a function of @xmath58 , and a concommitant charge order . all experimental indications @xcite are that the spin fluctuations in the cuprates are always of the form ( ii ) . it therefore seems physically reasonable to impose the amplitude constraint implied by ( ii ) at an early stage , and to examine the theory associated with @xmath59 in the limit of large and negative @xmath45 ( stability requires @xmath60 ) . this last constraint is the more formal statement of the physical requirement that the spin and charge ordering is `` stripe - like '' . as has been discussed in ref . , the imposition of the limit @xmath61 at the outset leads to a theory for phase and orientation fluctuations with a non - trivial structure that is very naturally described by a @xmath10 gauge theory . with the parameterization ( ii ) in ( [ ab ] ) , and for fixed @xmath62 , the order parameter @xmath49 takes values in the space @xmath63 , where the @xmath64 pertains to the orientation of @xmath65 , the @xmath66 is the phase factor @xcite @xmath67 , while the all - important @xmath10 quotient accounts for the invariance of @xmath40 under @xmath68 and @xmath69 . we can define global actions @xcite for single - valued fields @xmath65 and @xmath67 only at the cost of introducing a @xmath10 gauge field @xmath70 . a simple phenomenological form for the spatial terms in such an action is @xcite : @xmath71 where @xmath12 are lattice sites , and @xmath72 , @xmath73 , and @xmath74 are coupling constants that vary with doping . the @xmath10 gauge invariance ensures that all physical properties are invariant under the @xmath10 gauge transformation @xmath75 , @xmath76 and @xmath77 . the point defects associated with order parameter space @xmath63 were analyzed in ref . , and a fundamental role is played by the `` stripe dislocation '' , sketched in fig [ dislocation ] , which is a half - vortex in the angular field @xmath78 . ( see ( [ e2 ] ) and ( [ ab ] ) ) with @xmath79 containing a half vortex . as noted below ( [ e2 ] ) @xmath80 should not be literally interpreted as the electron charge density , and the modulation about could be associated with _ e.g. _ the exchange per link . the quantum phase transitions discussed in this paper involve a transition from a fermi liquid state in which these dislocations proliferate to one in which they are suppressed ; there is no long - range spin or charge density wave order in either state ( see also fig . [ phase]).,width=240 ] this vortex also carries a @xmath10 gauge flux of @xmath81 , as is easily seen from ( [ z2flux ] ) . we are now in a position to introduce our candidate transition for strongly coupled quantum criticality in the presence of a fermi surface . begin in a fermi liquid state , with only short range spin and charge order correlations : @xmath82 , @xmath83 ( @xmath43 in ( [ vphi ] ) is large and positive , @xmath84 , @xmath73 in ( [ z2flux ] ) are not large ) and only a small core energy for the stripe dislocations , which therefore proliferate . the dislocation core energy is controlled by the maxwell term for the @xmath10 gauge field , @xmath74 , and a small value of @xmath74 implies that the @xmath10 gauge field is in its confining state . in this state , the strong fluctuations of the @xmath10 gauge field `` bind '' the @xmath85 and @xmath86 fluctuations together , and the appropriate collective excitation is the conventional field @xmath40 itself , as it appears in ( [ e1 ] ) . the single @xmath40 collective mode therefore describes both the sdw fluctautions ( via [ e1 ] ) ) and the bond order / cdw fluctuations ( via ( [ e2 ] ) ) . now increase the dislocation core energy by increasing @xmath74 , while keeping @xmath43 reasonably large ( or @xmath84 , @xmath73 small ) : this will decrease the number of stripe dislocations but still maintain @xmath87 , @xmath88 . in the cuprates we imagine that the increase in @xmath74 is associated with the decrease in carrier concentration towards the mott insulator . at a critical value of @xmath89 , the @xmath10 gauge theory undergoes a transition into a deconfined state in which the stripe dislocations are suppressed . the transition at @xmath90 is the focus of interest in this paper . in this deconfined state , the @xmath65 and @xmath67 fluctuations become independent collective excitations : _ i.e. _ the single spin / charge collective mode , @xmath40 , has _ fractionalized _ into two independent modes . the bond order , ( [ e2 ] ) , fluctuations are now given by the ` square ' of the @xmath91 collective mode , while the spin fluctuations , ( [ e1 ] ) , are the product of the @xmath65 and @xmath92 collective modes . we reiterate that there is no long range spin or charge density wave order on either side of this fractionalization transition , and it is entirely associated with the suppression of the defects in fig [ dislocation ] ; see fig [ phase ] . is the one studied in this paper . the phases on both sides of @xmath89 have well defined electronic quasiparticles near a fermi surface and no long range spin or charge order ( @xmath93 , @xmath94 ) . the order parameters @xmath95 are defined in section [ sec : qft ] : these are ising order parameters dual to the @xmath10 gauge theory . the phase on the right is a conventional fermi liquid ; that on the left is a fermi liquid with local stripe correlations with random phases and spin orientations , and a ` topological rigidity ' associated with the suppression of stripe dislocations . in the application to the cuprates , the horizontal axis represents increasing doping and the critical point is near optimal doping . states with long range spin or charge density wave order do appear at smaller doping but these have not been shown . the phase transition at @xmath90 should be distinguished from the conventional sdw ordering transition in a superconductor ; this occurs at lower doping and was used recently by one of us @xcite to predict and explain the results of neutron scattering and stm experiments.,width=288 ] at dopings lower than those shown in fig [ phase ] , the cuprates possess phases with a variety of possible long range orders ( _ e.g. _ sdw or cdw ) which were reviewed in ref . . these transitions are associated with conventional order parameters , and we will not consider them in this paper , as theories are already available in the literature . we are interested in describing the universal properties of the transition at @xmath90 . these turn out to be controlled by two distinct characteristics of the electronic system . ( _ i _ ) tetragonal or orthorhombic crystal symmetry : with tetragonal symmetry we have to simultaneously consider the separate @xmath10 gauge theories associated with fractionalization transitions in both @xmath40 and @xmath41 . with orthorhombic symmetry , only one of these will be selected ( the `` stripes '' have a preferred direction ) . ( _ ii _ ) presence or absence of long - range coulomb interactions : as we argue below , an important coupling of the critical @xmath10 gauge degrees of freedom to the fermions arises from long - wavelength deformations of the shape of the fermi surface . as is well known , in the presence of long - range coulomb interactions the `` dilatational '' mode , involving local changes in the area of the fermi surface , is strongly suppressed . for completeness we will also consider the case of fermions with only short - range interactions ( perhaps the coulomb interactions have been screened by a metallic gate ) , where the coupling to the dilatational mode has to be included . with these two criteria , four distinct categories emerge , which we label a , b , c , d , as summarized in table [ table ] . .[table ] classification of quantum field theories . [ cols="^,^,^,^",options="header " , ] we will describe these cases in the subsections below . we begin with case a , for which the critical theory has the simplest form . as we noted in section [ sec : op ] , the transition at @xmath90 is described by a @xmath10 gauge theory . we know that the pure 2 + 1 dimensional @xmath10 gauge theory is dual a 2 + 1 dimensional ising model@xcite ; we represent this ising order parameter by a real scalar field @xmath96 . for now , we neglect the independent @xmath10 gauge theory associated with @xmath41 , because orthorhombic symmetry has oriented all the spin and charge density waves along @xmath49 . the action for @xmath97 in the vicinity of the transition is the familiar field theory for the ising critical point : @xmath98 ; \label{sx}\ ] ] we have chosen the spatial units so that the velocity of @xmath96 excitations is unity in both directions , @xmath99 is a quartic non - linearity , and the quadratic coupling @xmath100 is the tuning parameter across the transition . the field @xmath96 can be viewed as the creation / annihilation operator for the dislocations : @xmath96 is condensed in the confining phase where the dislocations proliferate , and @xmath96 has short - range correlations in the deconfined phase where dislocations are suppressed . in principle , we also need to account for the `` matter '' fields @xmath78 , @xmath85 in ( [ z2flux ] ) near this confinement transition . closely related @xmath10 gauge models have been considered in the literature@xcite , and it has been established that their fluctuations do not change the universality class of the transition . their presence does modify the effective couplings in ( [ sx ] ) , but the confinement transition continues to be described by the ising field theory ( [ sx ] ) all along the phase boundary . the main question we need to address here is the coupling between @xmath101 and the electronic excitations near the fermi surface ; the gapless excitations near the fermi surface have the potential of inducing long - range interactions which are more destructive than the matter fields in ( [ z2flux ] ) . the electrons _ do not _ carry any ising gauge charges , and couple to degrees of freedom of interest here only via the bilinears in ( [ e1 ] ) and ( [ e2 ] ) : @xmath102 where @xmath103 are the spin pauli matrices . we now imagine integrating out @xmath40 and generating terms which couple the electrons to the @xmath10 gauge degrees of freedom . the electrons are gauge singlets , and so can only couple to the physical @xmath10 gauge flux : the simplest such coupling is one in which the coupling @xmath74 in ( [ z2flux ] ) becomes @xmath104 dependent via its dependence on the local fermion bilinears @xmath105 where the ellipses represent terms with higher derivatives ; we have to include all such terms because the typical electron momentum is of order the fermi momentum , and this is not small . upon performing the duality transition to the ising field @xmath96 , the mapping applies to @xmath106 , the co - efficient of @xmath107 , via its dependence upon @xmath74 , and this is also noted in ( [ kmap ] ) . we can now integrate over the fermionic degrees of freedom and examine the structure of the terms which modify @xmath108 : it is not difficult to see that apart from the modification of the numerical values of the couplings in ( [ sx ] ) , all such terms are formally irrelevant . the leading non - analytic terms induced by the low energy fermionic excitations arise from long wavelength deformations of the fermi surface : in the presence of long - range coulomb interactions , all such excitations allowed by symmetry to couple to @xmath107 are suppressed . so we reach the important conclusion that the critical theory for case a is specified by ( [ sx ] ) . there is one important caveat to this conclusion that deserves highlighting . our analysis has neglected topological berry phase terms that could be present in the coupling between @xmath96 , the fermions , and the @xmath10 gauge field @xmath11 . similar phase factors do arise in other analyses of @xmath10 gauge theories @xcite . we leave the study of this issue to future work , and restrict our attention here to couplings associated with ( [ kmap ] ) . the higher tetragonal symmetry has two important consequences : we have two consider fractionalization in both @xmath40 and @xmath41 simultaneously , and distortions of the fermi surface with @xmath109 symmetry do couple to the critical degrees of the freedom even in the presence of long - range coulomb interactions . there are now two dual ising fields , @xmath96 and @xmath110 , associated with the independent @xmath10 gauge fields in the @xmath16 and @xmath17 directions ; simple symmetry considerations show that @xmath108 in ( [ sx ] ) is replaced by @xmath111 ; \label{sxy}\end{aligned}\ ] ] for simplicity , we have neglected a possible difference in the velocity of the @xmath97 modes in the @xmath16 and @xmath17 directions . the critical properties of the theory @xmath112 are well understood @xcite : they are described by a stable fixed point with @xmath113 and a global o(2 ) ( or xy ) symmetry of rotations between @xmath95 . the fundamental new phenomenon here is that low energy deformations of the fermi surface destabilize the critical point just noted . although long - range coulomb interactions do suppress a local uniform dilatational change in the fermi surface , they do not suppress shape deformations which preserve the local volume . in particular , the tetragonal symmetry allows an oscillation of the fermi surface with @xmath109 ( or @xmath114 ) symmetry which has singular contributions at low energy , as has been emphasized recently by oganesyan _ _ @xcite ; moreover , by symmetry , this deformation will couple linearly to @xmath115 . integrating the fermi surface deformations out by a standard procedure @xcite , we obtain the following non - analytic term in the effective action : @xmath116 the @xmath117 factor is the standard density of states of long - wavelength fermi surface oscillations in dimensions @xmath118 . a simple scaling argument now gives the scaling dimension of the coupling @xmath119 at the @xmath113 fixed point of @xmath112 . the @xmath117 factor implies that ( [ l5 ] ) couples @xmath120 to itself at distinct spacetime points , which precludes any additional operator renormalization ; consequently @xmath121 = d+1 - 2 \mbox{dim } [ \phi_x^2 - \phi_y^2 ] ; \label{scale1}\ ] ] here and henceforth we will state such scaling dimensions for general spatial dimensions @xmath122 , although our interest is in @xmath123 . the scaling dimension of @xmath124 can in turn be related to the anisotropy crossover exponent @xcite , @xmath125 @xmath126 = d+1 - \frac{\phi_t}{\nu } \label{scale2}\ ] ] where @xmath127 is the correlation length exponent of the @xmath128 fixed point of @xmath112 . using the known numerical values@xcite of these exponents in @xmath123 , @xmath129 and @xmath130 , we obtain @xmath131 = 0.52 > 0 $ ] . so we have established our earlier claim that @xmath109 deformations of the fermi surface constitute a relevant perturbation . it is now necessary to extend the renormalization group to include contributions higher order in @xmath119 . however , a simple one - loop analysis shows that this is not a stable procedure @xcite : at order @xmath132 we generate additional quartic terms which have a different @xmath133 , @xmath134 dependence than that in ( [ l5 ] ) , and new relevant terms are generated at each successive order . the solution to this conundrum is to introduce a new real scalar field , @xmath135 , as a hubbard - stratonovich decoupling of ( [ l5 ] ) : this field will be a measure of the amplitude of @xmath109 deformations of the fermi surface . note that we expect @xmath136 , and so to make the hubbard - stratonovich transformation positive definite , it is necessary to subtract a @xmath137 term before decoupling , and to compensate for this by corresponding changes to @xmath99,@xmath138 . in this manner we get the action @xmath139 our final effective action for case b is @xmath140 . in obtaining @xmath141 we have expanded the coefficient of @xmath142 for small @xmath143 , and rescaled @xmath135 to make the co - efficient of the analytic @xmath142 term unity . we now have two new coupling constants , @xmath144 and @xmath145 , and it can be checked that @xmath136 implies @xmath146 . further , at the fixed points of @xmath112 , we will show in section [ rgb ] that @xmath147= \mbox{dim}[\lambda_5]$ ] . the most important property of @xmath148 is that this action is stable under renormalization , and no new relevant terms are generated at any order . consequently , it can be subjected to a standard field - theoretic renormalization group analysis , and this will be presented in section [ sec : rg ] . the analysis of the remaining cases is a straightforward generalization of the discussion in section [ b ] . as in case a , the orthorhombic symmetry implies that we need only one ising field @xmath96 . however , because of the absence of long - range coulomb interactions , dilatational changes in the local fermi surface volume with full lattice symmetry do carry large low energy spectral weight . we measure these changes by the real scalar field @xmath149 , and obtain its effective action @xmath150 the effective action for case c is now @xmath151 . an argument similar to that in section [ b ] can be used to compute the scaling dimension of @xmath152 : as in ( [ scale1],[scale2 ] ) we find that the fixed point describing the transition of @xmath108 @xmath153 = \frac{2}{\nu } - d -1 = \frac{\alpha}{\nu } \label{spsia}\ ] ] where @xmath29 , @xmath127 are the standard exponents of the 2 + 1 dimensional ising model . using their known values @xcite , we obtain @xmath154 = 0.174 > 0 $ ] . so the fixed point of case a is unstable to a fermi surface with short - range interactions , and there is new universality class of critical behavior . we note that the discussion in this subsection , with zero sound modes of a fermi surface coupling to the square of an order parameter , bears a passing resemblance to a discussion of the influence of phonon modes in the superfluid state of bosons presented by frey and balents @xcite with tetragonal symmetry and no long - range coulomb interactions , both dilatational and @xmath109 modifications of the fermi surface have to be included . so the full effective action is @xmath155 where @xmath156 at the @xmath113 fixed point of @xmath112 , we can show , as in section [ c ] , that @xmath154 = \alpha/\nu$ ] , where @xmath29 , @xmath127 are now critical exponents of the 2 + 1 dimensional xy model . numerically@xcite we have @xmath154 = -0.022 < 0 $ ] , and so @xmath152 is irrelevant at this fixed point . however , the perturbations in @xmath141 remain relevant , and we can not a priori conclude that the terms in @xmath157 will also be irrelevant at any possible stable fixed point of @xmath158 ; so it is necessary to include ( [ spsip ] ) in the analysis of case d. this section makes a brief detour to a separate class of quantum phase transitions mentioned in section [ sec : intro ] : those with conventional order parameters which do not couple linearly to low energy excitations on the fermi surface . we will consider here a simple example of such a transition , although the results discussed here are far more general : an sdw ordering transition with the order parameters defined as in ( [ e1 ] ) , from a fermi liquid state with @xmath159 to an sdw state with @xmath160 . further , we assume that the values of @xmath161 are such that they can not connect any two points on the fermi surface ; consequently the @xmath162 fluctuations are undamped and the quantum critical modes are strongly coupled . the effective action for @xmath162 will have a structure similar to that of @xmath108 , @xmath112 in ( [ sx ] ) , ( [ sxy ] ) , apart from changes in the quartic terms which are similar to those in ( [ vphi ] ) . it should now be clear that the zero momentum composite fields @xmath163 can couple to long - wavelength deformations of the fermi surface in much the same way as the composite fields @xmath164 did in section [ sec : qft ] . for orthorhombic symmetry and short - range interactions ( as in case c ) , only @xmath24 is critical ( say ) , and its coupling to the long - wavelength particle - hole continuum leads to the term @xmath165 which is the analog of ( [ l5 ] ) . just as in ( [ spsi],[spsia ] ) @xmath166 = \frac{\alpha}{\nu } , \label{l6d}\ ] ] and so the coupling in ( [ l6 ] ) is relevant if and only if @xmath167 . most conventional order parameters have @xmath168 in 2 + 1 dimensions , in which case the fermi surface has no influence on the critical theory . the main exception is the ising case , found _ e.g. _ at a two sublattice cdw ordering transition , in which case the theory will identical to that in section [ c ] . for a system with tetragonal symmetry , there is stronger coupling to quadrupolar distortions of the fermi surface , just as discussed in section [ b ] . the analog of ( [ l5 ] ) is the term @xmath169 and the scaling dimension of @xmath170 is related to an anisotropy crossover exponent ; this always such that @xmath170 is strongly relevant . the analysis of the resulting theory will be similar to case b above , although we will not present it explicitly . this section returns to the collective mode fractionalization transitions considered in sections [ sec : op ] and [ sec : qft ] ; closely related analyses will apply to the cases mentioned in section [ sec : sdw ] . case a is described by the purely bosonic theory ( [ sx ] ) , whose properties are very well known : its critical point is in the universality class of the 2 + 1 dimensional ising model . so we move directly to case b in the first subsection , and mention the straightforward generalization to the other cases later . renormalization group ( rg ) analyses of models with a structure similar to @xmath171 have appeared elsewhere in the literature ( see _ e.g. _ ref . ) , and so we can omit most of the technical details . we consider the rescaling transformation @xmath172 under which the fields @xmath95 , @xmath135 transform as @xmath173 here @xmath174 , @xmath175 , and @xmath176 are @xmath177-dependent parameters to be determined by loop corrections to the effective action under the rg transformation the values of these parameters at any fixed point of the rg equations will specify the corresponding critical exponents . the one - loop graphs have the same topology as those considered in ref . , and we defer details on the evaluation of the loop integrals to appendix [ oneloop ] ; the results are @xmath178 where @xmath179 ( @xmath180 integer ) , @xmath181 , and @xmath182 are scheme dependent functions which are specified in appendix [ oneloop ] . we have assumed that the strongly relevant `` mass '' term @xmath106 has been tuned to its critical value , and are examining above only the flow within this critical manifold . the first of the equations in ( [ rg1 ] ) is the most important : it is an exact equation ( although the values of @xmath174 and @xmath175 do have corrections at higher order ) which follows from the fact that there are no loop corrections to the non - analytic term with coefficient @xmath145 in the effective action . first , let us examine ( [ rg1 ] ) in the subspace with @xmath183 , where the long - range forces associated with the fermi surface are absent . the results in appendix [ rg1 ] ) show that @xmath184 independent of @xmath180 , and @xmath185 ; here @xmath186 is a scheme - dependent constant whose value does not effect the physical properties . it is then easy to search for fixed points of ( [ rg1 ] ) : there are 8 fixed points , and all of them obey @xmath187 ( this is exact to all orders for @xmath183 ) and @xmath188 ( to this order ) . the fixed points are : @xmath189 where @xmath190 . the physical meaning of these fixed points becomes clear after integrating out the @xmath135 field in @xmath112 at @xmath183 : for this value of @xmath145 this yields an effective action with the same structure as @xmath112 in ( [ sxy ] ) but with modified coupling constants @xmath191 the reader can now observe that the fixed points in ( [ fp2 ] ) all separate into pairs with equal values of @xmath192 and @xmath193 , and our notation for the fixed points has been chosen to make this evident : @xmath194 and @xmath195 form one such pair , and similarly for the remainder . at this stage , we can compare results with the analysis of aharony@xcite of the consequences of a cubic anisotropy on fixed points with o(@xmath196 ) symmetry : he analyzed the analog of the model @xmath112 for @xmath196-component fields , but with a somewhat different notation . the fixed points ( [ fp2 ] ) are in complete agreement with his results : @xmath197 are the gaussian fixed points , @xmath198 are the `` cubic '' fixed points , @xmath199 are the xy fixed points with o(2 ) symmetry and no cubic anisotropy , and @xmath200 are the ising fixed points at which the fields @xmath201 are decoupled . as shown by aharony , an xy fixed point is stable for the case of a two - component field being considered here . here , we have to perform some further analysis to distinguish between the two possible @xmath202 fixed points : these two have the same values of @xmath192 , @xmath193 , but only one has @xmath203 . a standard computation shows that fixed point @xmath204 with @xmath203 is the only stable fixed point among all those in ( [ fp2 ] ) . the theory prefers the fixed point with @xmath203 because the fluctuations of @xmath205 are characterized by a divergent susceptibility at the critical point @xcite . we are now in a position to consider non - zero values of @xmath145 . for small @xmath145 , the above analysis shows that near the fixed point @xmath204 @xmath206 in other words , @xmath145 is a _ relevant _ perturbation with @xmath207 = 3 \epsilon /5 $ ] . we argued in section [ b ] that @xmath207 = 2 \phi_t /\nu - d -1 $ ] , and it can be checked that this is consistent with our present result and the @xmath208 expansion result@xcite for @xmath209 . for general @xmath145 , we have to consider the complete flow equations in ( [ rg1 ] ) . while these are quite complicated , it is clear from the first flow equation that a finite and non - zero @xmath145 fixed point , @xmath210 can only exist if and only if there is a solution to @xmath211 . from ( [ rg1 ] ) we observe that at the one loop order , this condition translates into an equation for @xmath212 alone : @xmath213 using the results in appendix [ oneloop ] , we plot half the left hand side of this equation in fig [ plotai ] for two different renormalization schemes . as determined by ( [ rg1 ] ) . there is no zero crossing , and hence no finite , positive , non - zero @xmath145 fixed point . the full line is the minimal subtraction scheme , while the dashed line is the fixed dimension scheme.,width=288 ] no solution is found in either case , or in the additional cutoff schemes discussed in appendix [ oneloop ] . the value of @xmath214 does initially decrease from its value at @xmath183 which led to ( [ dai ] ) , but this decrease is not strong enough to lead to a zero crossing . so we have not found a finite @xmath145 fixed point , but we can not rule out the possibility that one will appear upon including terms higher order in @xmath144 , @xmath99 , @xmath138 . it is clear now that any non - zero , positive @xmath145 runs away to @xmath215 , and we have to consider the theory in this limit . an examination of the limiting values of the results in appendix [ oneloop ] showed that there were no finite fixed point values for @xmath99 , @xmath138 in the limit of large @xmath145 either . in principle , it was possible that both @xmath216 and @xmath144 became large in a manner that permitted @xmath99 and @xmath138 to approach a finite fixed point value . however simple analytic and numerical analyses of ( [ rg1 ] ) showed that this was not the case , and we always had flows to strong coupling for @xmath99 and @xmath138 also . starting close to the stable fixed point at @xmath217 , we found that the flows sent @xmath218 and @xmath219 . this is possibly suggestive of a first - order transition , but no definitive conclusion can be reached because the rg equations themselves breakdown once @xmath99 or @xmath138 become of order unity . the analysis and results for case c are similar to those for case b above , and so we will be brief . the rg flow equations ( [ rg1 ] ) are replaced by @xmath220 where the functions @xmath221 , @xmath222 , @xmath223 remain the ones specified in appendix [ oneloop ] . these equations do have fixed points at @xmath217 , which have been specified completely in earlier work@xcite . however , the stable fixed point for @xmath217 is unstable to a non - zero @xmath152 , and the condition for a finite and non - zero @xmath152 fixed point is ( replacing ( [ eqai ] ) ) : @xmath224 the left hand side of this equation is plotted in fig . [ plota ] , and again there is no solution . , but for ( [ eqa]).,width=288 ] we find instead flows to strong coupling , as in case b. finally , we note that the flow equations and results for case d are a combination of those presented already for cases c and d , and we will not present them explicitly . this section considers some simple observable physical properties of the fractionalization transition described in section [ sec : qft ] . we will find that the properties of many observables are quite similar to another case with an order parameter which is a composite of the electron spin operator : the onset of spin glass order . to help make this comparison , we review the physical properties of the quantum spin glass transition in appendix [ app : qsg ] . our analysis of collective mode fractionalization theories has found only one stable fixed point : that for case @xmath225 with orthorhombic symmetry and long - range coulomb interactions . the other cases exhibited flows to strong coupling . this section will study the general scaling properties of various physical observables for case a , and also extrapolate the results to possible strongly coupling fixed points for the other cases . we focus on the spin susceptibility near the wavevectors @xmath226 ; other susceptibilities of conventional order parameters can be treated in a similar way . the order parameters associated with these susceptibilities , @xmath162 , are controlled by an effective action of the form ( @xmath227 is a matsubara frequency at a temperature @xmath2 ) @xmath228 along with a similar term with @xmath229 . unlike the situation in section [ sec : sdw ] , we are considering the case where @xmath161 do connect two points on the fermi surface , and hence there is over - damping of @xmath162 fluctuations represented by the term proportional to @xmath230 . also , the @xmath162 fluctuations remain non - critical with @xmath231 at the fractionalization transition , and hence the parameter @xmath232 . we now have to consider the coupling of @xmath162 to the critical modes @xmath95 . as noted at the end of section [ a ] , we ignore possible non - local topological terms , and focus on the simplest local couplings allowed by symmetry ; for the cases with tetragonal symmetry ( cases b and d ) , these are @xmath233 while for orthorhombic symmetry we merely drop the non - critical @xmath110 field above . we will consider the influence of the critical @xmath95 modes on @xmath162 in a perturbation theory in @xmath234 . for case a , where a stable fixed point was found , such a procedure appears reasonably safe . the validity for the other cases , with flows to strong coupling , has a greater uncertainty and this should be kept in mind below . the influence of the critical fluctuations on the non - critical @xmath162 susceptibility will be determined by two ( and higher ) point correlators of the operators @xmath124 and @xmath235 for tetragonal symmetry , and @xmath107 alone for orthorhombic symmetry . let us represent these operators generically by @xmath236 and its scaling dimension by @xmath237 = d_{\mathcal{o}}. \label{do}\ ] ] for case a , we have only @xmath238 and @xmath239 , where @xmath127 is the exponent of 2 + 1 dimensional ising model ; for other cases , the value of @xmath240 will be determined by the hypothetical strong - coupling fixed point . in perturbation theory in @xmath241 we can write for the dynamic spin susceptibility @xmath242 with the `` self energy '' @xmath243 where @xmath244 is related to @xmath241 , and @xmath245 is the two - point correlator of @xmath236 . we are interested in the behavior of @xmath246 at small @xmath134 and @xmath133 ; after inserting ( [ d1 ] ) into ( [ d2 ] ) , we use the fact that @xmath232 to expand the momentum and frequency dependence of @xmath247 in powers of @xmath248 . for the @xmath134 dependence of @xmath246 , a simple power - counting analysis shows that all terms in this series are free of infrared divergences as long as @xmath249 , in which case the @xmath134 dependence of @xmath246 is analytic and free of critical singularities . case a easily satisfies @xmath250 , and we expect this is also true for the other cases , because a negative @xmath251 would imply that correlations of @xmath236 increase with increasing spacetime distance . in contrast to its @xmath134 dependence , the @xmath133 dependence of @xmath246 is singular even for @xmath249 . this is most easily computed by analytically continuing ( [ d2 ] ) to real frequencies and taking its imaginary part . a standard analysis then shows that @xmath252 has a @xmath134 _ independent _ singular contribution in the quantum critical region which obeys the scaling form @xmath253 ( we are using @xmath254 ) , where @xmath255 is a universal scaling function determined by the non - trivial interacting fixed point of the strongly - coupled quantum field theories considered in section [ sec : qft ] ; we expect @xmath256 and latexmath:[$f_1 ( y \rightarrow \infty ) \sim \mbox{sgn}(y ) related to the local correlator of @xmath236 , which has a singular contribution obeying the scaling form @xmath258 ( the scaling function @xmath259 is a property of the strongly - interacting fixed point of the quantum field theories of section [ sec : qft ] ) by the integral relation @xmath260 \nonumber \\ & = & \int_{0}^{y } \frac{d\omega}{\pi } \omega \mbox{im } \left [ f_2(y - \omega ) \right ] \left ( 1 + n(\omega ) + n(y - \omega ) \right ) \nonumber \\ & + & \int_{0}^{\infty } \frac{d\omega}{\pi } \omega \mbox{im } \left [ f_2(\omega+y ) \right ] \left ( n(\omega ) - n(\omega+y ) \right ) \nonumber \\&+ & \int_{y}^{\infty } \frac{d\omega}{\pi } \omega \mbox{im } \left [ f_2(\omega - y ) \right ] \left(n(\omega - y ) - n ( \omega ) \right ) , \label{d5}\end{aligned}\ ] ] where @xmath261 is the bose function at unit temperature . note that the leading factor of @xmath262 in ( [ d5 ] ) arises from the linearly density of states of low frequency excitations implied by ( [ d1 ] ) , upon ignoring the influence of @xmath246 in @xmath247 on the right hand side of ( [ d2 ] ) . this is permissible as long as the contribution of @xmath246 here is subdominant : from ( [ d3 ] ) we see that this is so as long as @xmath249 . if this inequality is violated , then the perturbative expansion in @xmath244 fails , and a self consistent approach appears necessary . it is interesting to note that the results ( [ d1],[d3 ] ) imply a structure for the dynamic spin susceptibility which is closely related to phenomenological forms that have been proposed recently by schrder _ _ @xcite to describe neutron scattering measurements on cecu@xmath263au@xmath264 . elegant theoretical models of a `` local quantum criticality '' based on the self - consistent treatment of quantum impurity ( ` kondo ' ) models have also been advanced @xcite and discussed @xcite as explanations for the observations of schrder _ et al._. some important differences between the theories of si _ et al . _ @xcite and our work here should be noted : underlying our results is a bona fide @xmath265 dimensional quantum field theory ( presented in section [ sec : qft ] ) , and not a quantum impurity model . our theory leads to a @xmath134-independent singular contribution to the dynamic spin susceptibility because @xmath232 , as shown above . the condition @xmath232 also implies the static spin susceptibility is not divergent , in contrast to impurity model analyses in which it does diverge at the impurity critical point @xcite . an associated fact is that the exponent in ( [ d3 ] ) is likely to satisfy @xmath266 , although this need not be true for the spin glass transitions discussed in appendix [ app : qsg ] . it has also been noted @xcite that the divergence of the local susceptibility could preempt a local treatment of the impurity model critical points @xcite . it should be clear that the above mechanism for ` local ' quantum criticality in the dynamic spin susceptibility is quite a general property of a situation in which the primary order parameter , which exhibits strongly coupled bulk criticality , is coupled to the spin operator only with non - linear terms like those in ( [ zeta ] ) . as an illustration we recall in appendix [ app : qsg ] that related considerations apply to the quantum spin glass transition . we follow the same perturbative approach discussed in section [ sec : sus ] . singular fluctuations of @xmath236 will couple to the electrons via couplings like @xmath267 the ellipses represent additional terms with gradients of the electron operators ( these are not small because the typical electron momentum resides on the fermi surface ) . we should note at the outset that the status of a perturbative computation of the electron self energy , @xmath268 , remains open . for the case of overdamped fermion critical points below their upper critical dimension , the corresponding issue was addressed by altshuler _ _ @xcite and castellani _ et al . _ @xcite who concluded that the leading perturbative correction also gave the ultimate critical singularity . the extension of this result to the strongly coupled critical points being discussed here is less clear . indeed , for the case of a strongly coupled spin glass quantum critical point , the perturbative computation of the electron self energy is known to break down @xcite ( see also appendix [ app : qsg ] ) . we identify two distinct contributions to the critical electrical transport properties : the electronic quasiparticle and the bosonic collective modes . however , the cautions about a perturbative treatment of the coupling between these modes mentioned in section [ sec : self ] apply also to our discussion here . the electronic contribution is characterized by the transport relaxation time @xmath280 . the electron scattering discussed in section [ sec : self ] all occurs at small momenta @xmath281 and , as is well known , the transport rate then acquires an additional factor of @xmath282 over the quasiparticle scattering rate in the integral over scattered momenta . extending the scaling arguments of section [ sec : self ] we see easily that @xmath283 the collective mode contribution is the analog of the `` aslamazov - larkin '' term in the theory of fluctuating superconductivity : fluctuations of the order parameters @xmath95 may themselves contribute to the electrical current @xmath284 . the relationship between the dual fields @xmath95 and the microscopic degrees of freedom is extremely complex , and so we will restrict ourselves here to deducing the expressions for the current operator using symmetry arguments alone . the current @xmath284 is odd under time reversal and spatial inversion , while the fields @xmath95 are even under both . furthermore , the current should be invariant under the dual ising symmetries involving the change in sign of @xmath97 and/or @xmath285 . this indicates that the current operator has contributions from operators like @xmath286 and similarly for @xmath287 . the scaling dimensions of these operators can be determined by standard methods for case a ( some combination is related to components of the stress - energy tensor with scaling dimension @xmath265 , but the others require a @xmath288 expansion ) , while for the other cases the scaling dimensions remain unknown quantities to be determined by the hypothetical strong - coupling fixed point . if we generically denote the scaling dimension of these operators by @xmath289 , the kubo formula implies that they will yield a singular contribution to the conductivity , @xmath52 , which scales as @xmath290 note that because the current is carried by both the electronic and bosonic degrees of freedom , we expect that we do not have to worry about constraints imposed by overall current conservation , as is the case in theories of the superfluid - insulator transition in bosonic systems@xcite . this paper has described strongly interacting quantum phase transitions in the presence of a fermi surface . the primary examples considered involve the phase transition described in fig [ phase ] , from a conventional fermi liquid state in which the dislocation in fig [ dislocation ] proliferates , to a ` topologically ordered ' fermi liquid state in which these dislocations have been suppressed . note that there is no long - range sdw / cdw order in either state , although the amplitude of these orders must be locally large to allow identification of the dislocations . the conventional sdw / cdw orders are suppressed in both phases by ` spin - wave ' fluctuations and by the proliferation of other defects , such as the hedgehogs in @xmath291 and integer vortices in @xmath78 . we discussed some simple observable properties of such a topological phase transition in a fermi liquid , and found results which were quite similar to those near quantum spin glass transitions , and to the phenomenological ` local ' scaling forms proposed by schrder _ _ @xcite . despite their ` local ' nature , the critical singularities are controlled by a bulk quantum field theory , and not a quantum impurity model . the dislocation proliferation transition discussed here has been proposed as a candidate for a quantum critical point near optimal doping in the cuprate superconductors @xcite . in this context , it is interesting to note that recent scanning tunnelling microscopy images @xcite of the surface of bi@xmath292sr@xmath292cacu@xmath292o@xmath293 clearly show dislocations with the geometry of fig [ dislocation ] ( see fig 2c of ref . ) . the main shortcoming of our analysis here has been runaway flow of the renormalization group flow to a non - perturbative regime for the interesting cases . it is possible that this signals a more fundamental change in the nature of excitations near the fermi surface than is accounted for by our formalism . we integrate out the low energy fermions at an early stage , and the strong coupling situation may benefit from an analysis in which quasiparticle degrees of freedom are retained more explicitly . future work in this direction , along the lines of refs . , appears desirable . we thank a. aharony , a. chubukov , e. demler , m. p. a. fisher , m. grilli , w. metzner , a. millis , v. oganesyan , t. senthil , q. si , m. vojta , and j. zaanen for useful discussions . this research was supported by us nsf grant dmr 0098226 . tm was supported by a grant - in - aid from the ministry of education , culture , sports , science and technology of japan . the quantum spin glass transition in metallic systems has been studied in a number of recent works @xcite . here we review some physical properties whose scaling structure resembles the results in section [ sec : phys ] . the central actor in the theory of the spin glass transition is the order parameter functional @xmath294 where @xmath295 are replica indices . in the quantum critical region , this functional has an expectation value of the form @xmath296 if the quantum critical point obeys hyperscaling properties , then fourier transform of @xmath297 will obey `` @xmath298 scaling '' : @xmath299 it should be noted that the mean - field solution of the metallic spin glass does not obey ( [ qab3 ] ) , and instead has a characteristic frequency scale @xmath300 . however , other mean - field spin glass models which account for proximity to a mott insulator do obey @xmath298 scaling@xcite ; these latter solutions have the value @xmath301 , and @xmath302 has the form @xmath303 where @xmath304 was defined below ( [ fse5])for this value of @xmath305 there is an additive real contribution dependent upon the cutoff @xmath306 , but the remainder is a universal function of @xmath298 , and the function @xmath302 is analytic as @xmath307 at any nonzero @xmath2 . we will assume below that the low - dimensional spin - glass quantum critical point does obey the hyperscaling property ( [ qab3 ] ) , even if the mean - field theory for its particular situation does not . we now consider the computations of the physical quantities discussed in section [ sec : phys ] . for the dynamic spin susceptibility near the wavevector @xmath3 , note that symmetry allows a coupling like @xmath308,\label{qab5}\ ] ] along with the corresponding @xmath229 . using the expectation value ( [ qab2 ] ) in ( [ qab5 ] ) , we immediately obtain a dynamic spin susceptibility of the form ( [ d1 ] ) with @xmath309 the singular part of the inverse susceptibility is @xmath134 independent and a function of @xmath298 , as in the local quantum criticality scenario . again , however , the scaling function in ( [ qab3 ] ) is determined by a bulk @xmath122-dimensional quantum field theory . this appendix will present some further details of the derivation of the renormalization group equations for case b , described by the action @xmath316 . we noted in section [ rgb ] that the topological structure of the diagrams is identical to those considered in ref , and so we will not reproduce the diagrams here . the exponents @xmath176 and @xmath174 arise from the momentum and frequency dependence of the self energy @xmath317 of the @xmath318 field . to leading order , the only such contribution to @xmath268 is @xmath319 where @xmath320 . by an expansion of @xmath317 for small @xmath321 and @xmath133 , we find @xmath322 here the subscript @xmath306 is meant to indicate that the integrals in ( [ sen ] ) are to be carried out as part of a renormalization group procedure . the precise nature of the integral then depends upon the scheme being used : ( _ i _ ) in the dimensional regularization / minimal subtraction scheme , we perform the integrals in arbitrary dimension @xmath122 , and pick out online the poles in @xmath190 in ( [ etaz ] ) ; ( _ ii _ ) in the fixed dimension approach @xcite , we perform the integral in @xmath123 for a finite @xmath106 , take the derivative of the result with respect to @xmath323 , and set @xmath324 . ( _ iii _ ) in the soft - cutoff approach @xcite , we perform the integrals with a smooth cutoff on the scale @xmath306 , take the @xmath306 derivative , and then set @xmath325 . each of these approaches yields closely related expressions for @xmath326 and @xmath327 . the remaining terms in ( [ rg1 ] ) arise from simple one - loop integrals which can be evaluated at zero external momenta and frequency ; these yield the expression @xmath328 again the @xmath306 subscript indicates that the integrals are to be evaluated in one of the renormalization group schemes discussed below ( [ etaz ] ) . we should note here a possible exception to this canonical weak coupling scenario : chubukov and co - workers @xcite have studied spin and charge density wave ( sdw and cdw ) transitions in two dimensions , and have pointed out strong coupling effects in the electron self energy and possibly other observables . c. castellani , c. di castro , and m. grilli , phys . * 75 * , 4650 ( 1995 ) ; s. caprara , m. sulpizi , a. bianconi , c. di castro , and m. grilli , phys . rev . b * 59 * , 14980 ( 1999 ) ; s. andergassen , s. caprara , c. di castro , and m. grilli , phys lett . * 87 * , 056401 ( 2001 ) ; s. caprara , c. di castro , p. werner , and w. zwerger , phys . * 88 * , 066403 ( 2002 ) . y. zhang , e. demler , and s. sachdev , phys . b * 66 * , 094501 ( 2002 ) . this paper discussed mainly the physics of superconductors , and refered to the transition as `` exciton fractionalization '' . in a superconductor , the spin and charge collective modes appear as sharp excitations ( because there is no fermi surface to damp them ) , and so it is appropriate to refer to the collective modes as quanta of `` excitons '' . in the present paper , we are always dealing with a situation where the spin and charge collective modes are overdamped , and this terminology did not seem appropriate . otherwise , there is no change in the physical content of the fractionalizaion . c. pfleiderer , g. j. mcmullan , s. r. julian , and g. g. lonzarich phys . b * 55 * , 8330 ( 1997 ) ; s. a. grigera , r. s. perry , a. j. schofield , m. chiao , s. r. julian , g. g. lonzarich , s. i. ikeda , y. maeno , a. j. millis , and a. p. mackenzie , science * 294 * , 329 ( 2001 ) ; h. v. lhneysen , c. pfleiderer , t. pietrus , o. stockert , and b. will , phys . b * 63 * , 134411 ( 2001 ) . s. wakimoto , g. shirane , y. endoh , k. hirota , s. ueki , k. yamada , r. j. birgeneau , m. a. kastner , y. s. lee , p. m. gehring , and s. h. lee , phys . b * 60 * , r769 ( 1999 ) ; s. wakimoto , r. j. birgeneau , y. s. lee , and g. shirane , phys . b * 63 * , 172501 ( 2001 ) . a renormalization group argument on the implications of a divergent local susceptibility in _ impurity _ models runs as follows . consider an impurity model with a local spin @xmath336 with a local susceptibility @xmath337 with @xmath338 . then power counting implies @xmath339 = ( 1-\gamma)z/2 $ ] . now bring in a second identical impurity model , with spin @xmath340 initially decoupled from @xmath341 , which also has @xmath342 = ( 1-\gamma)z/2 $ ] . finally , turn on an infinitesimal coupling @xmath343 between the two impurities ; scaling implies @xmath344 = z \gamma > 0 $ ] , and so @xmath74 is a relevant perturbation . a possible escape from this instability is that @xmath74 is already accounted for in the ` baths ' of the two impurity models , but this has to be shown beyond mean - field theory . an analogy with the sliding luttinger liquid ( c. s. ohern , t. c. lubensky , and j. toner , phys . lett . * 83 * , 2745 ( 1999 ) , a. vishwanath and d. carpentier , phys . lett . * 86 * , 676 ( 2001 ) , and v. j. emery , e. fradkin , s. a. kivelson , and t. c. lubensky phys . lett . * 85 * , 2160 ( 2000 ) ) , another situation with apparently lower dimensional criticality embedded in a higher dimensional space , might be helpful : in this case there is only a marginal coupling between the luttinger liquids , but this does change all the exponents in the full theory . |
a quantum chain model of many molecule motors is proposed as a mathematical physics theory on the microscopic modeling of classical force - velocity relation and tension transients of muscle fibre . we proposed quantum many - particle hamiltonian to predict the force - velocity relation for the slow release of muscle fibre which has no empirical relation yet , it is much more complicate than hyperbolic relation . using the same hamiltonian , we predicted the mathematical force - velocity relation when the muscle is stimulated by alternative electric current . the discrepancy between input electric frequency and the muscle oscillation frequency has a physical understanding by doppler effect in this quantum chain model . further more , we apply quantum physics phenomena to explore the tension time course of cardiac muscle and insect flight muscle . most of the experimental tension transients curves found their correspondence in the theoretical output of quantum two - level and three - level model . mathematically modeling electric stimulus as photons exciting a quantum three - level particle reproduced most tension transient curves of water bug . | 1004.3120 | muscle is a bundle of cylindrical muscle fibres . one muscle fibre enclose hundreds of thinner cylindrical fibre called fibril . a fibril is a train of sarcommeres connected head to tail . the sarcomere , as seen by light microscopy , is a composite structure of actin filament and myosin filament(fig . [ muscle ] ) . the two filaments are bridged by long myosin protein called myosin molecule motor(see ref . @xcite@xcite or biology textbook for a more detail review ) . the longitudinal section of filament array in muscle fibre . myosin molecules are long polymers as cross - bridges between the actin filament and myosin filaments.,scaledwidth=48.0% ] the length of a sarcomere is observed to decrease during muscle contraction . huxley modeled the muscle contraction as mutual movement between actin filament and myosin filament@xcite . when the myosin molecule in the myosin filament is excited up to attach on the actin filament , myosin molecule would combine with adenosine triphosphate molecules , and convert energy from adenosine triphosphate hydrolysis into mechanical force , this is the attached state . in the detached state , myosin molecule is at rest and doing nothing . the arrangement of myosin molecules along the length of filament is incommensurate . a stochastic model was introduced to describe the cooperative behavior of molecules motors which has disordered arrangement along a backbone@xcite . experiment research is beyond the description of classical mechanics two - state cross - bridge model@xcite which assumed the probability of the two states detached state and attached state satisfy a pair of coupled differential equations . myosin molecules looks like a long arm ended by a head domain . the myosin in muscle cell is usually called myosin - ii , for it has two heavy chains in the head domain . a second kind of myosin molecule with single heavy chain was found in non - muscle cells , they are termed as myosin - i . the average length of myosin - ii is about 160 nm@xcite . visible light corresponds to a wavelength range of 400 - 700 nm . the electromagnetic wavelength comparable to the length of myosin - ii falls in ultraviolet region . a muscle fibre surrounded by membrane is under control of nerve cell . membrane is a filter system which is permeable for certain ions but is impermeable for other ions . the imbalance distribution of ions across the membrane results in an electric potential difference of -60 to -90 mv . the electric signals from nerve cell can modify the permeability of membrane . an active muscle would generate electric signal . scientist use this electric signal to represent the tension inside the muscle . if we insert a needle with two fine - wire electrodes into the muscle , the electric activity of muscle can be detected and recorded by an oscilloscope . this electromyogram is in widespread use for medical examination of muscle . electromagnetic wave is termed as photon in quantum mechanics . the electric signal generated by active muscle is physically equivalent to a wave packet of photons . quantum scattering between photons and molecules within biological system is a common phenomena , so does in muscle . molecule only absorb photons at resonance frequency . we assume there exist a one - to - one correspondence between conformational change of motor molecule and the hopping between quantum states of molecule . we define the excited quantum level as the attached states in which the motor molecule is extended to a longe shape . the detached state is assumed to be equivalent to a quantum states with lower energy . the motor molecule becomes shorter in detached states . besides the two quantum states of molecules , another quantum state of photon is introduced to induce the hopping between the two quantum level . photon can propagate among different molecules . it is photon that couples different motor molecules to work together . quantum physics has many techniques to control quantum states , thus it is possible to control the conformation of molecule motors by electromagnetic wave . muscle fibre can also be stimulated by chemical solutions which involves many complicate biochemistry@xcite . there was an argument on chemical reactions in muscle using quantum mechanics@xcite , it is not related to my mathematical physics model . the time scale in my modeling is much larger than the chemical reaction . we only focus on the electrical stimulation of muscle fibre without adding any external chemical solutions . to make this mathematical physics model work for a real muscle fibre system , the following assumptions are necessary : ( 1 ) the strength of electric field should be so strong that it dominates the conformation of molecules over thermodynamic fluctuations . ( 2 ) for the muscle fibre without any external input of stimulus chemical ions , the conformational change of motor molecules governed by electromagnetic field is much larger than that induced by chemistry . this is because electric field is distributed across the whole space , it exerts a global force on the motor molecules . while chemical ion only interact with molecule locally , it does not come into action unless it finds the correct binding site . ( 3 ) the spatial conformation of motor molecule can be modified by concentration of some chemical ions . we simplified the muscle fibre as a one dimensional chain of many giant quantum particles which represents molecule motors . the velocity of sliding actin filament acts as environmental parameter . these giant quantum particles are excited by absorbing photons , and relaxed by emitting photons . we arrange these particles regularly along a one dimensional chain . the wavelength of the stimulus electromagnetic pulse is assumed to be much larger than a sarcomere . the configuration of spatial arrangement within a sarcomere is supposed to has no influence on the output physics . the article is organized as follows : in section ii , we present the quantum hamiltonian for deriving analogous force - velocity relation with respect to the coresponding macroscopic quantity . the empirical hill s relation is consistent with the steady solution of heisenberg equations . we predict the force velocity relation for the slow release case and the non - steady state solution . in section iii , we study how the force decays after the stimulus of electric pulse . exact solution of quantum two level model reproduced similar tension activation curves of cardiac muscle . the analytical solution of quantum three - level model can generate most similar curves of tension transients of insect flight muscle . in section iv , the tension transients curve of skeletal muscle is mathematically reproduced from the point view of quantum coherent state . the experiment curve coincides with the third order projection of coherent state . in section v , we derived the self - coupled quantum hamiltonian for solving strongly coupled differential equations . for this case , the sliding velocity depends on the internal states of the motor molecule . the last section is a brief summary and outlook . in muscle release experiment , the two ends of the muscle is fixed at its resting length . electric pulses with different frequency is superimposed to develop tension in the muscle . in a quick release experiment , one end of the muscle is set free quickly . for a slow release , the length of muscle is shortening with one end of the muscle oscillating or moving under control with in a relatively longer time interval . the tension - velocity relation is the relation between two macroscopic quantity . in the following , we shall express the force by microscopic density operator , while the velocity is taken as macroscopic parameter . we first assume the velocity of contraction is controlled by the density of some ions in the solution . we only model the excitation and decay process of muscle . a one dimensional chain of quantum two - level particles . each particle represents one myosin molecule motor.,scaledwidth=48.0% ] we summarize a myosin molecule motor as a multi - level quantum particle . one eigenstate of the particle with respect to each level represents one conformation . in the excited state @xmath1 , myosin molecule is attached on actin filament . in the ground state @xmath2 , the quantum particle is at rest in myosin filament . one photon may excite the quantum particle from ground state to excited state . the particle hops from state @xmath1 to state @xmath2 by emitting one photon . the quantum representation of the mutual sliding movement between actin filament and myosin filament is to annihilate one quantum state on one lattice site and generate it again at the next nearest neighboring lattice site . as the sliding motion has only one direction , switching two neighboring lattice sites would contributes a minus sign in front of the hopping operator . the quantum hamiltonian of this one dimensional chain(fig . [ pair ] ) is unitary , @xmath3.\end{aligned}\ ] ] where @xmath4 is the unit lattice spacing , and @xmath5 is the absolute sliding velocity . the hopping coefficient @xmath6 is electric field . usually we choose symmetric coupling coefficients , @xmath7 . this hamiltonian is neither single - particle hamiltonian nor tight - binding model . for the particle at lattice site @xmath8 , the probability of being in the excited level @xmath9 is measured by the density operator , @xmath10 . @xmath11 defines the probability of being in state @xmath12 . @xmath13 is proportional to the complex dipole moment . the probability density function are not independent , they must satisfy the constrain , @xmath14 . the time evolution of the density operators are governed by the equation of motion , @xmath15 $ ] . for the given hamiltonian ( [ hamiltonian1 ] ) , the equations are @xmath16 @xmath17 where @xmath18 is the energy difference between the two states . at time @xmath19 , and position @xmath20 , the initial number of molecule motors in state @xmath21 and @xmath22 are denoted as @xmath23 and @xmath24 respectively , both @xmath23 and @xmath24 are functions of velocity @xmath5 . in the steady states , the probability in both two sates are constants . put the steady state constrain , @xmath25 and @xmath26 , on the equations of motion eq . ( [ aa ] ) and eq . ( [ ab ] ) , one would arrive at an equality of probability distribution and velocity(detail calculation is presented in appendix [ quickrelease ] ) , @xmath27}{4},\ ] ] where @xmath28 is complicate function of velocity , @xmath29,\nonumber\\ g_{\pm}&=&\frac{4{v}^2/{s}^2}{4{v}^2/{s}^2+(\omega - f_{n}{\pm}kv)^2}.\end{aligned}\ ] ] where @xmath30 , @xmath31 is the speed of light . @xmath32 is the frequency of the stimulus electric pulses . @xmath33 is the total number of particles . we define the tension of the one dimensional chain as the total number of particles in excited states , @xmath34 where @xmath35 is a renormalization coefficient , we set it as @xmath36 for convenience . the myosin is attached on the actin filament in excited states . the more myosin molecules are excited , the stronger physical bonds there would exist between actin filament and myosin filament . the maximal tension is determined by the total number of particles , @xmath37 , i.e. , all particles are excited to connect the two filaments . the existence of a maximal tension value is consistent with experiment measurement . if one apply a single electric shock across membrane , the tension of the muscle fibre first increase from zero to a transient maximal point and then decays to zero . the electric shock is an energy packet with finite number of photons . suppose the energy packet contains @xmath38 photons , and @xmath39 . it can maximally excite @xmath38 molecule motors . in the meantime , the motor is decaying and emitting photons . if @xmath33 photons are sent into the muscle at one time before any decay process , all the molecules will be excited , then it will reach the maximal tension state . it always take certain time for the muscle to eat all the photons , more photons must be added to fill the loss of photon by decay . the numerical plot of tension - velocity relation according to eq . ( [ aa - v ] ) . , scaledwidth=30.0% ] the relation between total probability density and velocity obtained for steady states , eq . ( [ aa - v ] ) , is quite similar to hill s empirical force - velocity relation@xcite . here we know the physical origin of every parameter . although there is a complex function @xmath28 involved in eq . ( [ aa - v ] ) , the tension - velocity plots a hyperbolic curve , tension decreases with increased velocity(fig . the equations of motion , eq . ( [ aa ] ) and eq . ( [ ab ] ) , suggest that the velocity is actually a decay coefficient . larger velocity leads to faster decay . thus the number of physical bonds connecting the two filament decreases as velocity increases . this give us a natural explanation to experimental results . besides the decay induced by mutual sliding movement , there exist spontaneous decay due to thermal fluctuation and vacuum fluctuation . when we take the spontaneous decay into account , the equation of motion for the density operator is given by @xmath40-\frac{1}{2}\left\lbrace \gamma,\rho\right\rbrace,\end{aligned}\ ] ] where the @xmath41 is friction tensor , @xmath42 the equation of motion has similar form as that for quick release , @xmath43 @xmath44 where @xmath18 , @xmath45 . the decay coefficients of velocity now becomes the composition of velocity and spontaneous decay coefficients , @xmath46 for the steady state , @xmath25 , @xmath26 , the relation equation between the excited state density operator and the composite decay coefficients is @xmath47 = \frac{{\delta}n(z)}{2}\frac{n_{t}\gamma'_{a}\gamma'_{b}}{\gamma'_{a}+\gamma'_{b}},\end{aligned}\ ] ] where @xmath48 is initial probability difference . @xmath28 is a complex function of velocity , @xmath49,\nonumber\\ g_{\pm}&=&\frac{\gamma'^{2}_{ab}}{\gamma'^{2}_{ab}+(\omega - f_{n}{\pm}kv)^2}.\end{aligned}\ ] ] if the velocity term is much larger than spontaneous decay term , i.e. , @xmath50 , one neglects spontaneous decay , then eq . ( [ fv2 ] ) reduced to former tension - velocity relation eq . ( [ aa - v ] ) . therefore , eq . ( [ aa - v ] ) is the force - velocity relation for the quick release of muscle . if the muscle is released slowly , the spontaneous decay may become equally important as velocity term , @xmath51 . in that case , the tension - velocity relation should obey eq . ( [ fv2 ] ) . ( [ fv2 ] ) is the general force - velocity relation for steady state . as for the non - steady state , we have to solve the equations of motion to find how the excited state density operator evolves as a function of velocity . insect flight muscle oscillate far more rapidly than the frequency of the input nervous impulse . a muscle is unable to expand , it only contracts . the muscle expansion is achieved by contracting an opposing muscle . the oscillation frequency of muscle is the frequency of the oscillation between excited state and relaxed state . the @xmath52 terms in hamiltonian eq . ( [ hamiltonian1 ] ) describes two free harmonic oscillator which are oscillating with frequency @xmath53 and @xmath54 respectively . the ultimate output frequency of hamiltonian eq . ( [ hamiltonian1 ] ) is regulated by the hopping term between the two levels and the mutual sliding term . we perform fourier transformation on the density operator , @xmath55 and substitute eq . ( [ al - ak ] ) into hamiltonian eq . ( [ hamiltonian1 ] ) . the hamiltonian now becomes more brief , @xmath56 where @xmath57 is the exactly solvable part for regulated free oscillator , @xmath58|a_{k}\rangle\langle{a}_{k}|\nonumber\\ & + & [ \omega^{b}-\frac{2v}{s}\sin(ks)]|b_{k}\rangle\langle{b}_{k}|\}.\end{aligned}\ ] ] the regulated oscillation frequency depends on velocity and momentum vector @xmath59 . the free oscillation modes has a maximal frequency and a minimum frequency , @xmath60 the band width for existed free oscillation modes is determined by the sliding velocity . an oscillating electric field , @xmath61 would induced the hopping between two levels . the state vector for the schr@xmath62dinger wave equation , @xmath63 , is written as @xmath64\arrowvert{b}_{k}\rangle,\end{aligned}\ ] ] the density operator are given by @xmath65 . within rotating wave approximation , the terms that do not satisfy energy conservation are dropped , such as @xmath66 and @xmath67 . the eigenstate density operator under the first order perturbation of the hopping terms @xmath68 is @xmath69 } { [ \delta\omega-({2v}/{s})\sin(ks)-{f}]^2},\ ] ] where @xmath70 the non - steady state tension decrease following @xmath71 regulated by an oscillation factor . the ultimate oscillation frequency of tension includes three sources , @xmath72 when the frequency difference of the two level vanishes , @xmath73 , the output frequency is utterly dependent on the sum of perturbation frequency and sliding velocity . the fast oscillating ion flow of @xmath74 , @xmath75 across the membrane cause the potential to rise to + 40 mv and restore its original value within a few milliseconds@xcite . this oscillation frequency is coincide with some muscle oscillation . some insect flight muscle contract 1000 times a second . my conjecture is that the oscillation frequency of muscle has some possible relation with the oscillating of ion flow . certain insect flight muscle can oscillate more rapidly than the frequency of the input nervous impulse . ( [ dispersion ] ) told us the output frequency not only depends on the frequency of input nervous impulse , but also depends on the sliding velocity and the frequency difference between the two levels . the sliding velocity may increase the output oscillation frequency . the output tension measured in experiment is the superposition of different oscillator modes , @xmath76 the two ends of the one dimensional muscle fibre is always attached to some tissues . if there are @xmath77 particles along the chain , the beginning and the ending particle are fixed . as @xmath78 in dispersion eq . ( [ dispersion ] ) is a periodical function , independent standing wave modes is confined in the domain @xmath79 . the allowed wave modes are @xmath80 where @xmath81 is the length of one sarcomere . all particles are oscillating in a standing wave pattern with frequency @xmath82 . for a selected modes @xmath83 , the phase difference between nearest neighboring particles is @xmath84 . if the sliding velocity is zero , @xmath85 , the standing wave modes vanished . a stretch activation is to lengthen the muscle fibre by external force , or to release the fibre to certain length with in a extremely short time . the fibres were glued between two glass rods . one rod is attached to the anode pin of a force transducer . the length and tension signal are displayed on a double beam oscilloscope@xcite . both the myosin filament and actin filament are long helical chains . the elementary structure to build myosin filament looks like golf club . they entwined around each other to produce a helix with diameter of about 15 nm . the actin filament is a double helix structure composed of globular protein . the local electric charge distribution along the myosin chain and actin chain are not homogeneous . a mutual sliding between the myosin and actin chain would inevitable modify the local potential configuration . the raising or lowering of local potential barrier induce the hopping of electrons among different energy level . many photons are generated within very short time . the stretch activation by mutual sliding movement is equivalent to stimulating the muscle fibre with an electric shock pulse . the time evolution of tension induced by a stretch activation includes two periods . in the first period , there is a rapid decay due to the extremely large sliding velocity . a large number of photons are generated but are not absorbed by molecule motors . in the second period , the two filament stopped mutual sliding . the accumulated photon began coming into action . this period is dominated by the spontaneous emission and absorption of photon . the ideal definition for contraction velocity is @xmath86 @xmath81 is the length of sarcomere . in fact , what we measure in experiment is the average velocity within @xmath87 . at the moment of quick release , the length experienced a sudden change , @xmath88 the sliding velocity at @xmath19 is almost infinity , @xmath89 . while after the stretch , the length does not change anymore , the sliding velocity becomes zero . the heaviside function provides us a good mathematical description to the length variation during a stretch . in the microscopic phenomena , a stretch activation would induce a sudden change of the particle distribution on different energy levels . there exist an optimal distribution which acts as the minimal point of a potential configuration . any deviation from this optimal point has a tendency to go back . the optimal particle distribution is different for different types of muscle fibre . the external load applied upon the muscle is actually a high potential barrier . the particles are trapped in a local minimal point to counterbalance the external force . as a muscle is only able to contract , the intrinsic potential without external load is asymmetric . the stretch activation of cardiac muscle has been extensively studied in experiment@xcite . various different species of vertebrates were put into use , including humming bird , frog , guinea pig , rat , rabbit . the tension transients following length perturbation were recorded by force transducer . at the instant lengthening the muscle , the force jumps up to a maximal force . then the force drops to a temporary minimum . after several period of small oscillation , it finally reach a new maximal . the length of the muscle were kept invariant for a few seconds , and was suddenly release to its original length . a sharp drop of force appears immediately following by a fast recovery . the force transients for stretch and release looks more or less like mirror - image(fig . 5 . in ref . @xcite ) . the mechanical oscillator as an analogy of the quantum two level model.,scaledwidth=46.0% ] a quantum two - level system provide a good model for the physics of force transients of cardiac muscle . a mechanical analogy of this quantum model is an oscillator in thick liquid . as shown in fig . [ twolevel ] , a heavy ball is attached to one end of a elastic rod . the sudden lengthening of muscle is to hit the ball with a pulse . the ball pops up to hit the upper level . since the length of muscle after the sudden stretch keep invariant , a spring from the upper level will attach the ball to prevent it from going back . however the elastic rod will pull the ball down , the competition between the elastic rod and the spring from the upper level induced a local oscillation . when the kinetic energy is finally consumed by the friction of the thick liquid , the ball stopped oscillating . for the sudden release of muscle , the pule hit ball in the opposite direction of the lengthening . as the molecule motor runs only in one direction , it is reasonable to introduce a bias potential . this bias potential makes the ball a little bit harder to restore from the bottom level . here the bias potential is gravitational field . we assume the relaxed state of a muscle fibre is reached at an optimal ratio between attached states and detached states . one straightforward ratio is to set the number of particles in excited state and ground states as equal . this relaxed state is the vacuum state of the quantum two - level model . the vacuum state does not means there is no particle , but means the number of particle is exactly the number of antiparticle . in the mechanical analogy , this relaxed state corresponds to the equilibrium position of the ball without external pulses and gravitation field . if there exist bias field , the mechanical equilibrium is obtained by the balance between gravitational field and elastic force of the rod , @xmath90 . we assume the optimal ratio is determined by the chemical potential of the two states , @xmath91:\exp[-\frac{\hbar\mu_b}{k_b{t}}].\ ] ] where @xmath92 and @xmath93 are the chemical potential of excited state and ground state respectively . in the following , the density matrix is actually the difference between the usual density operator and the vacuum density operator , we simply denote @xmath94 for convenience . the vacuum density distribution determines the resting stiffness of a muscle fibre . a lengthening activation is equivalent to a positive pulse . all the particle in state @xmath95 are driven to state @xmath96 . @xmath97 the sudden release at @xmath98 is negative pulse , all particles shift to state @xmath95 . @xmath99 thus the length perturbation provide the initial condition for time evolution of density operator . since the length is constant after the stretch activation , we ignore the velocity term in hamiltonian eq . ( [ hamiltonian1 ] ) . the equation of motion including spontaneous decay reads , @xmath100 where @xmath101 . the coupling coefficient are chosen as the symmetric , @xmath102 we focus on the case of two degenerated states , i.e. , @xmath103 in the beginning , all particles are in state @xmath95 , @xmath104 this corresponds to a quick release . applying the standard ansatz @xmath105,\;\;\;\;\alpha,\beta = a , b,\ ] ] one reduces the differential equation to algebra equation@xcite . it finally yields the time evolution of probability in state @xmath96@xcite , @xmath106.\ ] ] this is the conventional solution in the textbook of quantum optics@xcite . the plot of this solution for three different cases were also shown in the book of ref.@xcite . i will reproduce some figures of ref.@xcite in the following to show what the tension transients curve looks like . i will compare different cases with experimental curves which i have no copyright to use here . @xmath107 governs the tension transient . here @xmath108 after the stretch activation , there is no more external stimulus . the molecules undergoes radiative damping through spontaneous emission . the time evolution of @xmath107 relies on the ratio between driven field @xmath109 and damping rate @xmath110 , here we have set @xmath111 @xmath112 . if the driven field is much larger than damping field , there is an oscillation upon the exponential increase of tension(the upper curve in fig . [ high - low ] ( a ) ) . in the opposite case , the damping dominates the dynamic evolution , it is an exponential increasing without oscillation(the upper curve in fig . [ high - low ] ( b ) ) . as the density operator must satisfy @xmath113 , the time evolution of @xmath114 just follow the image curve of @xmath115 . for the intermediate case , the pumping strength is comparable with damping rate , both the lifetime and amplitude of oscillation are reduced , it approaches to an exponential increase for @xmath115(the upper curve in fig . [ midfriction ] ) or an exponential decay for @xmath116(the lower curve in fig . [ midfriction ] ) . comparing these tension transients curve with the experiment records in ref . @xcite , we find @xmath117 in fig . [ high - low ] ( a ) ) fits well with the muscle tension transient under stretch activation , including the tension - time course of rabbit papillary muscle in fig . 2 of ref . @xcite , the tension - time course of cardiac papillary muscle and insect flight muscle in fig . 26 of ref . @xcite . as for the release activation , the tension - time course of rabbit papillary muscle of fig . a of ref . @xcite and fig . 2 of ref . @xcite fits very well with @xmath117 in the theoretical fig . [ midfriction ] . the tension - time course of humming bird muscle under stretch activation and release activation looks like mirror image , however the life time and amplitude of tension oscillation under stretch activation is larger than that for the release case(fig . 5 of ref . the stretch activation curve is well fitted by @xmath118 in fig . [ high - low ] ( a ) ) . the release activation curve fits better with @xmath117 in fig . [ midfriction ] . the release activation is the tension recovery from state @xmath95 , the stretch activation is the recovery from @xmath96 . the asymmetric tension - time course indicates the damping rate for @xmath95 is higher than the damping for state @xmath96 . in the mechanical analogical phenomena , the damping may comes from friction of liquid or the bias external field . in polymer physics , the extension process of a long polymer is not always exactly the reverse process of contraction . i guess the motor molecule has similar behavior , this leads to the the asymmetric tension time course . we may call this a hysteresis phenomena . @xmath119{lowfriction.eps}&\includegraphics[width=0.24\textwidth]{highfriction.eps}\\ \mbox{(a ) } & \mbox{(b)}\\ \end{array } $ ] the time evolution of @xmath117 for @xmath120=1.5(the upper curve ) . the lower curve is @xmath121 for @xmath120=1.5.,scaledwidth=30.0% ] the tensions time course for a weak pumping field and high damping rate(fig . [ high - low ] ( b ) ) , @xmath120=0.3 , reproduced the high tension state of rabbit psoas muscle . the tension transient of stretch and release are exactly mirror - image of exponential decay(fig . 15 a of ref . @xcite ) . the tension transient eq . ( [ taa ] ) has two extremal case , one is for extremely high friction @xmath122 , the tension first decreases , and then exponentially rise to a stable value . the other case is for very strong pumping field , @xmath123 . in this case , the tension shows swift oscillation , the envelop of the oscillation is an exponential decay curve . these two cases have no resemblance among the tension course of the experiment in ref . there might exist some mechanism to prevent the cardiac muscle falling into these two extremal cases . @xmath119{weakpulse.eps}&\includegraphics[width=0.23\textwidth]{cos20.eps}\\ \mbox{(a ) } & \mbox{(b)}\\ \end{array } $ ] the tension transients of insect flight muscle under length perturbation is different from that of vertebrate including rabbit , humming birds , frog , pig , and so on . the experiment in ref . @xcite reported the muscle tension course of water bug @xmath0 . if the muscle is prepared in a regular solution without @xmath124 , the muscle tension almost has no response to length perturbation . in the presence of @xmath124 and @xmath125 , significant tension transient of @xmath0 muscle were observed in reaction to length perturbation@xcite . the tension course curve under both stretch and release activation showed a fast rise to a transient maximal and then followed by a slow decay@xcite . the shape of the decaying tail is beyond the theoretical prediction of the quantum two level model . while a quantum three - level model fits the experiment data very well this suggests that the molecule motors in insect flight muscle may be different from the motors in cardiac muscle . we still model the insect flight muscle as one dimensional chain of quantum molecule motors . now the molecule motor has three states : the attached state @xmath96 , the metastable state @xmath95 and the detached state @xmath126 . we assume the hopping between different states is induced by vacuum fluctuation . since muscle works at room temperature , the stochastic environmental fluctuation may destroy the coherent correlation between different motors . the motors are almost independent from each other , thus the one particle model is a good approximation . the usual hamiltonian for a three - level particle interacting with vacuum modes reads , @xmath127\nonumber\\ & & -i\sum_{k}[g^{k}_{_{1}}e^{-i(\omega_{ac}-\omega_{k})t}d^{\dag}_{k}\arrowvert{c}\rangle\langle{a}\arrowvert + g^{k}_{_{2}}e^{-i(\omega_{bc}-\omega_{k})t}d^{\dag}_{k}\arrowvert{c}\rangle\langle{b}\arrowvert],\nonumber\\\end{aligned}\ ] ] where @xmath128 is the frequency difference between state @xmath129 and @xmath130 , @xmath131 . @xmath132 and @xmath133 are the annihilation and creation operator of the @xmath59th vacuum mode with frequency @xmath134 . @xmath135 and @xmath136 denotes the coupling strength between vacuum modes and the three - level particle . the mechanical oscillator analogy of the quantum three level model.,scaledwidth=47.0% ] the quantum three - level system has a similar mechanical analogy with the damping oscillator for the two - level model . the key difference is that the oscillator has a metastable partner suspended above the equilibrium position(fig . [ threelevel ] ) . to perform a lengthening perturbation , we sent a shock pulse to hit the oscillator in the opposite direction of gravitation . the oscillator will rise to attach the partner ball and becomes a pair . the pair first rise to the attached state@xmath96 , and then decayed to the detached state @xmath126 . for a sudden release perturbation , the shock pulse hits the ball along the direction of gravitation field . the oscillator first drops to the detached state @xmath126 , and then is pulled up by the elastic rod . it continues to rise high enough to attach the metastable ball , and oscillates as a pair . this damping oscillator model gives us a rough phenomena of tension time course for insect flight muscle . we assume the myosin molecule motors as an equivalent three - level system . the tension of the one dimensional chain is proportional to the total number of particles in the attached state , @xmath137 . another equivalent way of modeling this system is to take the whole chain as one giant three - level particle . the tension of the particle is proportional to the probability in the attached state , in this case , we take muscle as single particle system which is relatively easy to handle with . an analytical investigation of hamiltonian eq . ( [ h3 ] ) was well performed for the spontaneous emission of a three - level atom with quantum interference@xcite . the dynamics of a quantum particle is governed by the schr@xmath62dinger equation , @xmath63 . the general state vector is written as @xmath138 the molecule is initially in the attached state @xmath96 and metastable state @xmath95 . when the frequency difference between the attached state and metastable state is much smaller than the frequency difference between the attached state and detached state , i.e. , @xmath139 , the schr@xmath62dinger equation admits an analytical solution@xcite , @xmath140 where the coefficients @xmath141 and @xmath142 are @xmath143 the decay rate of the two upper level are @xmath144 , @xmath145 . @xmath146 is the mode density . @xmath147 determines the amplitude of the decaying metastable state , @xmath148 where @xmath149 is the difference of decay rates for the two upper level . the probability of being in the attached state is given by the density operator @xmath117 . here the weights of state vector , @xmath150 and @xmath151 , define @xmath117 and @xmath152 , @xmath153 this density operator is actually a reduced density operator by tracing out all vacuum modes , @xmath154 to get a clear comparison with the two - level model , we map the schr@xmath62dinger equations of motion for @xmath155 @xcite into the equations of density operator , @xmath156 the sum of the probability in the three levels satisfy the constrain , @xmath157 . for the two - level model , the sum of probability in the two level is conserved , @xmath158=0 $ ] . while for the three - level model , @xmath159 , the sum of the probability in the upper two level depends on the probability in the detached state @xmath126 . the equations of motion ( [ rho3 ] ) have already incorporated the evolution of @xmath160 . one may verify @xmath158\neq0 $ ] in eq . ( [ rho3 ] ) . the two excited states are coupled to the detached state by the same vacuum modes . the photon emitted by the transition from @xmath95 to @xmath126 may induce the hopping from @xmath126 to @xmath96 . the photon may comes from one molecule , but absorbed by another molecule . thus there exist strong quantum coherence between the two decay channel . a sudden lengthening or release of muscle leads to the mutual sliding between the charged filaments . a packet of photons are created immediately but are bot absorbed . the stretch activation and release activation set up the initial value of population at different level . the deviation from equilibrium position results in a target potential barrier , which keeps a certain number of excited molecule from tunneling into lower level . the number of survived attached states determine the final tension value after decay . @xmath161{08 - 02 - 7 - 4 - 4.eps}&\includegraphics[width=0.23\textwidth]{05 - 0 - 7 - 7 - 4.eps}\\ \mbox{(a ) } & \mbox{(b)}\\ \end{array } $ ] @xmath119{05 - 02 - 7 - 14 - 4.eps}&\includegraphics[width=0.24\textwidth]{08 - 02 - 7 - 4 - 1.eps}\\ \mbox{(a ) } & \mbox{(b)}\\ \end{array } $ ] one qualitative conjecture of ref . @xcite is that @xmath125 is able to vary the @xmath124 sensitivity of the contractile system . when the @xmath0 muscle is in a regular solution with only @xmath124 ions , the tension reaches the delayed maximum within in 70 msec@xcite . when @xmath125 is added , the time to reach the maximum tension becomes much longer , it is about @xmath162 msec@xcite . 18 b in ref . @xcite showed the tension time course in @xmath124 and @xmath125 poor contraction solution . the experiment is performed at temperature t=@xmath163 with 5 mm atp . the ion concentration is estimated as pga=8@xmath1649 . pmg=7@xmath1648 , ph=6.5 . another fig . 23 in ref . @xcite showed the case that the concentration of @xmath124 and @xmath125 is even much higher than that for fig . 18 b in ref . the tension time course of fig . 23 in ref . @xcite is recorded at temperature t=@xmath165 with 5 mm atp , ph=6.5 , pmg@xmath1645 , [email protected] . my analogue theoretical conjecture is the damping rate of exited states is related to the relative ratio of @xmath125 to @xmath124 and their absolute value of concentration . increasing @xmath125 maybe reduce the decay rate . higher concentration induces lower decay rate . this theoretical conjecture is based on the solution of three - level model in eq . ( [ cacb ] ) . we found two very similar tension time course to the experimental curve of fig . 18 b in ref . @xcite . the first fig . [ 7 - 4 - 4 ] ( a ) simulates the stretch activation , for which the initial probability in the attached state is @xmath166 . the second fig . [ 7 - 4 - 4 ] ( b ) is for the release activation , no particle exists initially in the attached state . to get the similar tension transient curve with respect to fig . 23 in ref . @xcite from the three - level model , we raised the value of damping rate @xmath167 from @xmath168 to @xmath169 in the numerical calculation . the theoretical output of fig . [ 7 - 4 - 1 ] can be identified as the experimental curve of fig . 23 in ref . this theoretical curve in fig . [ 7 - 4 - 1 ] shows a slower decay process compared with the two decay curves lower concentration of @xmath125 and @xmath124(fig . [ 7 - 4 - 4 ] ) . @xmath161{08 - 02 - 7 - 4 - 14.eps}&\includegraphics[width=0.23\textwidth]{08 - 02 - 7 - 4 - 40.eps}\\ \mbox{(a ) } & \mbox{(b)}\\ \includegraphics[width=0.23\textwidth]{05 - 02 - 7 - 7 - 200.eps}&\includegraphics[width=0.23\textwidth]{05 - 05 - 7 - 7 - 900.eps}\\ \mbox{(c ) } & \mbox{(d)}\\ \end{array } $ ] the frequency difference @xmath170 indicates the oscillation between the two excited states . the exponential decay of tension is strongly regulated by this oscillation . for small value of @xmath171 , the tension has a dominant peak followed by small waves ( fig . [ 7 - 4 - 200 ] ( a ) ) . if we magnify the tension course of experiment fig . 19 in ref . @xcite , the theoretical fig . [ 7 - 4 - 200 ] ( a ) is coincide with experimental curve . this curve of fig . 19 in ref . @xcite is obtained at temperature = @xmath172 with 15 mm atp , and pga=8@xmath1649 . pmg=7@xmath1648 , ph=6.5 . comparing the experimental condition with that for fig . 18 in ref . @xcite , one may see the temperature raise about @xmath173 , the concentration of atp grows from 5 mm atp to 15 mm atp . my conjecture is that increasing atp and temperature enhanced the oscillation between two excited states . if we make the @xmath170 larger , i.e. , @xmath174 , the decay process shows more periods of oscillation(fig . [ 7 - 4 - 200 ] ( b ) ) . when @xmath175 , the tension - time course becomes a wavy curve of exponential decay(fig . [ 7 - 4 - 200 ] ( c ) ) . this wavy curve is quit similar to the tension transient of rabbit psoas muscle(fig . 15 b in ref . @xcite ) . for an infinite large frequency difference , @xmath176 , the oscillation is completely suppressed . the tension transients approaches to a smooth exponential decay(fig . [ 7 - 4 - 200 ] ( d ) ) , which looks very similar to the experimental curve in the high tension states(fig . 23 . c , in ref . the high tension states of @xmath0 muscle is reached by increasing pga from 8@xmath1649 to 6.8 , and keeping pmg=7@xmath1648 . from the theoretical point of view , a large frequency difference @xmath170 produced a large energy barrier . the quantum interference between the two excited states vanished . in mind of the theoretical approximation we used , @xmath177 and @xmath178 , a large @xmath170 indicates that the excited states are almost decoupled from ground state . the dynamics of the three - level system is dominated by decay of excited state . since no pumping field can pass an infinite barrier , the evolution of excited states is an exponential decay . when the frequency difference @xmath170 vanishes , the two excited states becomes degenerated . both the two states obey an exponential decay , @xmath179 the tension now is the sum of the probability of the two degenerated states , @xmath180 , @xmath181 the tension evolution is an exponential decay plus a constant term . the constant tension depends on the initial value of the density operator . for this special case , three - level model with two degenerated excited states gives similar result as the two - level model . however , there is always an induced decay rate by emitting photons in three - level model due to the metastable state . in two - level model , the emitting photons can be brought back to excited state without violating selection rules , thus the output of two - level model does not produce all of the decay curves as the three - level model did . comparing with cardiac muscle and insect flight muscle , skeletal muscle has two special characters : one is low resting stiffness , the other is its ability of extending much beyond its optimal length . when we summarize these two special characters into theoretical modeling , low stiffness indicates there are less number of molecules motors in the attached state when the muscle is not excited . two filaments connected by less bonding is easy to extend . the electrical signal representing muscle tension is physically explained as polarized photon . the strength of electric signal is proportional to the density of photon . quantum coherent photons exist in many living biological system@xcite . even though an isolated muscle fibre is not alive , we can still shed some new light on the twitch tension evolution in reaction to the stimulus of electric signal from the point of view of coherent states . external stimulus field will generate tension in the muscle . if the two ends of a muscle fibre is fixed , the tension developed by external stimulus is called isometric tension . the isometric tension of skeletal muscle was measured many decades ago@xcite . a superamaximal rectangular pulses is used to stimulate the muscle . both the second and third derivatives of twitch tension are recorded by some software@xcite . we take the same quantum hamiltonian eq . ( [ hkk ] ) of one dimensional chain of quantum two - level particles for modeling skeletal muscle , @xmath182 the exactly solvable hamiltonian @xmath57 is given by eq . ( [ h0 ] ) . a quantized electric field motivated the hopping between the two levels , @xmath183,\ ] ] where @xmath184 is annihilation operator of photons . applying the conventional rotating wave approximation , the total hamiltonian reads , @xmath185 where @xmath186 hamiltonian eq . ( [ hdd ] ) shares the very similar formulation with the general hamiltonian for a coherent state@xcite . if there is no special projection to certain number of photon , we would always have a coherent state for photons . however , one must project the coherent state to a fixed number of photons to get similar curve to the tension - time of skeletal muscle . we denote the off - diagonal density operator @xmath187 by a brief symbol @xmath188 . the schr@xmath62dinger equation , @xmath189 may be expressed by fock states , @xmath190 eq . ( [ n(t ) ] ) is equivalent to a radioactive decay equation from which one can get the standard poissonian distribution@xcite . physicist define coherent state as the eigenstate of annihilation operator @xmath184 . the coherent state with respect to hamiltonian eq . ( [ hdd ] ) is @xmath191 , @xmath192|0\rangle , \;\;\;\;\;\ ; d|\rho{t}\rangle=\rho{t}|\rho{t}\rangle.\ ] ] the general expression of wave function as solution of schr@xmath62dinger equation is @xmath193|\rho{t}\rangle.$ ] projecting the coherent state @xmath191 to fock space gives us possonian distribution , @xmath194\sum_{n=0}^{\infty}\frac{(\rho{t})^{n}}{\sqrt{n!}}|n\rangle.\ ] ] this coherent state is the superposition of many photon states . a photon is the electromagnetic radiation of quantum particles . usually the electromagnetic radiation of human muscle is in the infrared region at a wavelength of tens of microns . the detector of electric signal in the muscle is metallic needle . we assume the detector of photon has an optimal window which only allow finite number of photon pass , it can not capture all the photons . then the projection of coherent state to this optimal photon number state is what the experiment measured . this electric signal is proportional to the amplitude of the optimal photon state , @xmath195.\ ] ] the experiment of ref . @xcite only reported one complete tension - time course(fig . 1 e in ref . @xcite ) . the tension transient of fig . 1 e in ref . @xcite can be theoretically fitted by @xmath196 , \;\;\;\alpha\tau = t^2,\ ] ] where @xmath197 is an adjustable parameter to rescale the unit time scale . one can choose different @xmath197 to get different amplitude for the curve . both the second and the third derivative of @xmath198 looks very similar to experiment curve(fig . [ tva ] ) , except the minor oscillating tail of the curve . this oscillating tail is most likely due to fluctuating experimental environment or the poor quality of equipment in 1968 . because the experimental curves in 1977 of ref . @xcite rarely showed similar wavy tails . the tension transients @xmath199 . the first derivative of tension is @xmath200 . the second derivative of tension is @xmath201 . all the three curves are quit similar to those in fig . 1 e of ref . there exist relative sliding motion between two filaments . the total number of particles described by the quantum hamiltonian is sandwiched between the two filament . for an almost infinitely large system , the sliding motion of filament acts as environmental parameters . so we summarize the sliding velocity as a constant parameter in the model . in the quantum hamiltonian of the one dimensional chain model , the total probability of being in the two quantum states is conserved . the quantum density operator measures the probability for a particle being in certain state . for the two - level model , the density operator follow @xmath202 . for the three - level model , the density operators of three levels obey @xmath203 . when the one dimensional quantum chain has finite length , the sliding velocity of filament may depend on the internal states of motor molecule that drags the filament . the total number of particles sandwiched between the two filaments is proportional to the length of the overlapping region . when the myosin molecules between the two filaments are excited , they becomes active to drag the filament . however if the molecules falls out of the overlap region , even if they are excited , there is no active bond connecting the two filaments . we call the motor molecules out of the overlap region as unemployed molecules . the molecules within the overlap region are named after employed molecules . the mutual sliding movement make the quantum chain an open system . the total number of employed particles are increasing during contraction . a faster sliding velocity means the total number of employed particle increase or decrease faster , so we define the mutual sliding velocity as the speed of increasing employed molecules plus a initial velocity @xmath204 , @xmath205 the absolute value of @xmath206 is equivalent to the decrease of unemployed molecules . we assume the initial velocity @xmath204 is constant . the initial velocity maybe induced by chemical or biological effect . for the simplest case , one may take @xmath207 . the time dependent sliding velocity for the quantum chain of two - level particles can be expressed in terms of density operator , @xmath208 according to heisenberg equation , @xmath209 $ ] , this velocity is given by the commutator equation of hamiltonian operator , @xmath210-i\sum_{j}[h,{\rho}_{bb}^j]+v_{c}.\ ] ] usually a quantum hamiltonian is expressed by a functional of many operators which does not include the hamiltonian operator itself . here the expanding sequence of operators for hamiltonian eq . ( [ hamiltonian1 ] ) includes a velocity which is a functional of hamiltonian operator itself . the explicit formulation of hamiltonian eq . ( [ hamiltonian1 ] ) including the heisenberg equation of velocity reads , @xmath211.\nonumber\end{aligned}\ ] ] we may call this hamiltonian a self - coupled hamiltonian . transforming the self - coupled hamiltonian into a conventional hamiltonian must follow the algebra rules of operating matrix . we denote the projection of hamiltonian to local eigen states as , @xmath212 @xmath213 and @xmath214 is the hopping rate between two neighboring sites . @xmath215 and @xmath216 is the local eigen energy of the particle at the @xmath8th site . after some algebra , we obtained the conventional expansion of the self - coupled hamiltonian , @xmath217|a\rangle_{i , i}\langle{a}|\nonumber\\ & + & [ \omega_{i}^{b}+{t_{i , i+1}^{b}}{s}^{-1}]|b\rangle_{i , i}\langle{b}|\nonumber\\ & + & g_{ab}|a\rangle_{i , i}\langle{b}|+g_{ba}|b\rangle_{i , i}\langle{a}|+[\frac{iv_{c}}{s}-\frac{\omega_{i}^{b}}{s}]|b\rangle_{i}\langle{b}|_{i+1}\nonumber\\ & -&\frac{iv_{c}}{s}[|a\rangle_{i+1}\langle{a}|_{i}+|b\rangle_{i+1}\langle{b}|_{i}]+[\frac{iv_{c}}{s}-\frac{\omega_{i}^{a}}{s}]|a\rangle_{i}\langle{a}|_{i+1},\nonumber\end{aligned}\ ] ] where @xmath218 is the renormalization operator , @xmath219}.\nonumber\end{aligned}\ ] ] we can map the self - decoupled hamiltonian expressed by density operator into a more familiar equation for quantum field theory . each of the two quantum states is represented by a boson operator , @xmath220 notice here we study an open finite system , the total particle number is not conserved , i.e. , @xmath221 . the decoupled hamiltonian is simplified into a two - component boson system , @xmath222a^{+}_{i}a_{i } + [ \omega_{i}^{b}+{t_{i , i+1}^{b}}{s}^{-1}]|b^{+}_{i}b_{i}\nonumber\\ & + & [ \frac{iv_{c}}{s}-\frac{\omega_{i}^{a}}{s}]a^{+}_{i}a_{i+1 } + [ \frac{iv_{c}}{s}-\frac{\omega_{i}^{b}}{s}]b^{+}_{i}b_{i+1},\nonumber\\ & + & g_{ab}a^{+}_{i}b_{i}+g_{ba}b^{+}_{i}a_{i}-\frac{iv_{c}}{s}[a^{+}_{i+1}a_{i}+b^{+}_{i+1}b_{i}],\nonumber\\ g&=&\frac{1}{1-\frac{1}{s}\sum_{i}[a^{+}_{i}a_{i+1}+b^{+}_{i}b_{i+1 } -a^{+}_{i+1}a_{i}-b^{+}_{i+1}b_{i}]}.\end{aligned}\ ] ] this nonlinear hamiltonian has no resemblance in conventional quantum field theory due to the renormalization operator @xmath218 . if we assume @xmath223 , then the operator @xmath218 is approximated by a constant @xmath224 . such a approximation is not always appropriate , as the hopping step size @xmath4 is usually very small . for the other extreme case of @xmath225 , the constant term @xmath226 in the nominator of @xmath218 can be neglected , @xmath227 the key rules for handling this inverse quantum operator @xmath218 is the elementary commutator relation between a matrix operator and the inverse of another operator , @xmath228=\hat{c}_{ij}.\ ] ] we can assume if @xmath229 , @xmath230 , this is a physical assumption since two elementary bosonic operators are sitting at different lattice sites . while for @xmath231 , @xmath232 depends on the specific formulation of the mattix @xmath233 and @xmath234 . we know only one universal relation , @xmath235 , i.e. , @xmath236=0 $ ] . physically speaking , @xmath233 annihilate one particle at the @xmath8th site . its dual operator @xmath237 is not the conjugate operator @xmath238 . more over , it is unknown how to decompose the commutator between one elementary operator and the inverse of many composite operators , such as , @xmath239=\hat{c}(a , b).\ ] ] maybe we should first find out how to map the inverse operator into a series expansion of the conventional polynomial operators . in fact , quantum operator provides a natural description of directed molecule motors . we can define a walking operator @xmath240 , when it operates on a spatial coordinate , it walks one step forward or backward , @xmath241 @xmath242 is the unit step size . a directed molecule motor is characterized by only annihilation operator or creation operator , @xmath243 the velocity of the motor is denoted as the number of step @xmath244 within one unit time interval . for the sliding filaments of muscle fibre , @xmath244 has a upper bound and lower bound . this is an open system in which the hamiltonian is only consists of either annihilation operator or creation operator . this open physical system is strongly coupled with another chemical system or biological system . when single muscle fibre is teased away from the whole muscle , it is no longer a living organism . in the eye of a physicist , one muscle fibre is a long cylindrical material with complex internal structure . the polymers , chiral proteins and ions inside the fibre are properly organized both in space and functionality . an input electric signal induced the deformation of molecule motors , which in turn shortens the fibre . generally i believe the electromagnetic interaction(or in other words chemical bonding ) determines the equilibrium conformation of myosin molecules . the stochastic fluctuation from thermal environment only results in some minor modifications of local structure , even though sometimes this minor modification plays a crucial role . one myosin motor molecule can be simplified as a giant quantum particle with finite quantum states . each state corresponds to one stable molecule configuration . the input electric field induce the hopping among different states . we mathematically model a muscle fibre as a one dimensional chain of quantum particles representing myosin motor molecules . besides the two quantum states , exited state and ground state in analogy of attached state and detached state of classical model , an external quantum state of photon is introduced to couple the dynamics of one quantum particle with another . the mutual sliding between two filaments is included in the directed hopping operator of a quantum hamiltonian . quantum dynamics of this quantum hamiltonian system gives us similar force - velocity relation to the classical model for the quick release case . the force - velocity relation for the case of slow release and unstable states is beyond the capable description of classical mechanics model . however using this quantum hamiltonian , we can derived a rather brief force - velocity relation for the case of slow release and unstable states . the quantum equation of motion suggests that the mutual sliding is a source of the decay of excited states . another source of decay is vacuum fluctuation or thermal fluctuation , we summarize them into a spontaneous decay process . the mutual sliding movement means breaking bonds and establishing new bonds between two filaments . the ratio between the life time of the sliding induced decay and spontaneous decay determines the behavior of force - velocity relation . for a quick release , the sliding decay is much larger than spontaneous decay . for a slow release , the spontaneous decay dominates . both the experimental measurement of quick release and slow release only records the behavior of steady states solution . for the non - steady state , the force - velocity relation is an oscillating function of time . there is a difference between the frequency of force oscillation and the input frequency of electric stimulus . this theoretical conclusion is consistent with biological observation some insect flight muscle indeed can oscillate more rapidly than the frequency of input nervous impulse . we proposed the mathematical definition of the tension of a muscle fibre as the total number of particles in the excited states . thus tension transients is described as the evolution of the density operator of the excited states . based on this hypothesis , we apply the conventional quantum two - level model and three - level model to check the tension transients course of muscle fibre . the quantum two - level model reproduced most of tension - time courses of cardiac muscle under stretch activation . the three - level model can theoretically generate most curves of the tension transient of insect flight muscle . the quantum three - level model also produced some tension - time course of rabbit psoas muscle for which the two - level model can not account . the key difference between three - level model and two - level model is that three - level model takes into account of the induced decay due to metastable state while the two - level model does not . the quantum three - level model can also generate the tension transients of a skeletal muscle fibre under stimulus of electric pulse . however we checked the tension transients of skeletal muscle from a different point of view of quantum coherent state . the tension , the first derivative of tension and the second derivative of tension are all coincide with the third order projection of coherent states . it is not fully understood in physics why it projects to the third order , maybe there are better ways to model skeletal muscle . in the muscle stimulus experiment almost 35 years ago@xcite , the tension - time course varies according to the concentration of @xmath245 and @xmath246 . in the quantum chain model , those mathematical parameters can varies to produce similar curves to the experiment data . if there are too many theoretical parameters , it is big challenge to understand the physical or chemical relation between the concentration of @xmath245 and @xmath246 and those parameters . fortunately there are only three effective parameters in the quantum model related to the muscle tension transient : the two decay rates of excited states and the frequency difference between the two excited states . the decay rates controls the long time behavior of tension transient . the frequency difference introduces wavy oscillations upon the conventional exponential decay or exponential increase . it needs a series of systematic experiments to find out how ions control the three parameters , the existed experiment data is not enough . one experiment fact is that if solution does not contain @xmath245 ions , delayed tension changed disappeared , if the @xmath245 and @xmath246 are added , distinctive tension transients were observed@xcite . it seems as if @xmath245 ion drive the myosin motor molecule into metastable conformation and transform the molecule into a three - level quantum particle . while @xmath246 seems like to fix the conformation of myosin molecules so that they are invariant under stretch activation . maybe the ratio between the concentration of @xmath245 and @xmath246 is a possible controller of the two decay rate of the two excited states . both theoretical and experimental works are required to solve the puzzle . i did not find the contact information of the author of ref @xcite to ask for the permission of using the experimental curves in my article . if the reader want to see the experimental curve but can not find the old ref . @xcite in 1977 , please sent me an e - mail . i will ask the electronic library in max - planck - institute for the physics of complex systems for help . i present the detail calculations of force - velocity equation here . the assumptions and approximations used in calculation are explained at every step . i first show how i derived the differential equations from the quantum chain model . then i add the general procedure for solving similar differential equations in ref . @xcite for the benefit of readership . the assumption used in solving the very similar differential equations have more detail explanation in laser physics theory@xcite . for a quantum particle , the probability of being in a quantum state is given by the square of the norm of @xmath247 , i.e. , @xmath248 . this probability is well defined by density operator , @xmath249 . for a two level system , there are four density operators , @xmath250 the diagonal density operator satisfy the constrain , @xmath251,\ ] ] the equation of motion for these density operator is @xmath252=-\frac{i}{\hbar}[h\rho-\rho{h}].\ ] ] the hamiltonian of the one dimensional chain model is @xmath253.\nonumber\end{aligned}\ ] ] substituting this hamiltonian and density operator into eq . ( [ dtrho ] ) , one may get the equation of motion of the probability density in different states . here we take @xmath254 as an example , @xmath255\langle{a}|_{i}+|a\rangle_{i}[\langle{a}|_{i+1}-\langle{a}|_{i}]\}\nonumber\\ & + & g_{ba}|b\rangle_{i , i}\langle{a}|-g_{ab}|a\rangle_{i , i}\langle{b}|-2\frac{iv}{s}|a\rangle_{i}\langle{a}|_{i}.\end{aligned}\ ] ] if the lattice spacing @xmath4 is very small , we summarize the first term of eq . ( [ raa ] ) into a continuum limit , @xmath256\langle{a}|_{i}+|a\rangle_{i}[\langle{a}|_{i+1}-\langle{a}|_{i}]\ } = \partial_{x}{\rho}_{aa}\ ] ] here @xmath257 is set to @xmath258 for convenience . ( [ raa ] ) becomes more compact , @xmath259 following a similar calculation , we get the equation of @xmath260 and @xmath261 , @xmath262 @xmath263 where @xmath264 . since @xmath265 . the equation of motion for @xmath266 is just the conjugate of @xmath267 . mathematical theory told us , if the equation bear the form of @xmath268 the solution is formulated as @xmath269 the above equations of motion for density operator has exactly this kind of form . the solutions can be obtained by analogy of the calculation in quantum optics@xcite . within the rotating wave approximation , the coupling coefficient are @xmath270 integration of eq . ( [ rab ] ) gives out @xmath271\nonumber\\ & & { \times}g_{ab}(z',t')({\rho}_{bb}^i(z',v , t')-{\rho}_{aa}^i(z',v , t')).\end{aligned}\ ] ] following the assumptions of laser theory@xcite , the population difference @xmath272 , the electric field mode amplitude @xmath273 and the phase @xmath274 vary little in the time @xmath275 , then @xmath267 is rewritten as , @xmath276\nonumber\\ & & \left\lbrace \int_{-\infty}^{t}dt'u_{n}(z')\exp[-[i(\omega - f_{n})+\frac{2v}{s}](t - t')]\right\rbrace \nonumber\\ & & { \times}[{\rho}_{bb}^i(z',v , t')-{\rho}_{aa}^i(z',v , t')].\end{aligned}\ ] ] @xmath277 is the normal - mode function in space . for the one dimensional chain model , @xmath277 is the distribution of electric field along the length direction of the chain . as the electric field is oscillating in space , one usually take sinusoidal @xmath278 dependence : @xmath279 $ ] . as all know , @xmath280=\cos[k_n{z}]{\pm}i\sin[k_n{z}]$ ] , @xmath277 is decomposed into the sum of two opposite waves . for a symmetric velocity distribution @xmath281 , the quick @xmath282 $ ] terms are canceled in the integration , it finally yields @xmath283 . the density operator @xmath267 now reads , @xmath284 where @xmath285 is the propagator function , @xmath286 where @xmath30 , @xmath31 is the speed of light . substituting eq . ( [ rabdd ] ) into the equation of motion of @xmath115 and @xmath116 , one obtain the rate equation , @xmath287 where @xmath288,\end{aligned}\ ] ] @xmath289 has the similar form of lorentzian function @xmath290 for the steady state , @xmath25 , @xmath26 , the pair of rate equation leads to @xmath291 here @xmath292 is the initial population difference in the absence of field oscillation , @xmath293.\ ] ] eq . ( [ fv0 ] ) gives the relation between population difference and decay coefficient . in our one dimensional chain model , the decay coefficient is just the linear function of velocity . so eq . ( [ fv0 ] ) is actually the force - velocity relation , @xmath294}{2}.\ ] ] the rate function @xmath28 is a complex function of velocity . at the resonance points @xmath295 , @xmath28 gained maximal value . g. j. steiger , stretch activation and tension transients in cardiac , skeletal and insect flight muscle , page 221 , in _ insect flight muscle _ edited by r. t. tregear , north - holland publishing company , ( 1977 ) |
we take a new look at the neutron radiative capture by a deuteron at thermal energy with the pionless effective field theory ( eft( ) ) approach . we present in detail the calculation of amplitudes for incoming doublet and quartet channels leading to the formation of a triton fully in the projection method based on the cluster - configuration space approach . in the present work , we consider all possible one - body and two - body photon interaction diagrams . in fact , additional diagrams that make significant changes in the results of the calculation of the total cross section in the process are included in this study . the properly normalized triton wave function is calculated and taken into consideration . we compare the cross section of the dominant magnetic m1-transition of up to next - to - next - to - leading order with the results of the previous model - dependent theoretical calculations and experimental data . the more acceptable results for cross section\:\textrm{mb}$ ] show order by order convergence and cutoff independence . no three - body currents are needed to renormalize observables up to in this process . o # 1#2_#1 -2.5pt # 1#2#3_#1 | 1406.6530 | studies of the radiative capture reactions on numerous light atomic nuclei have been continued at thermal and astrophysical energies with the model - independent pionless effective field theory ( eft(@xmath0 ) ) approach in the recent years @xcite-@xcite . the calculation of radiative capture amplitude and cross section of @xmath1 @xmath2 and @xmath5 @xmath6 are an essential input in the calculation of the parity - violating radiative capture of the above processes at the thermal energy @xcite . in the present paper , we study the @xmath1 @xmath2 process fully with the projection operator method based on the cluster - configuration space which is introduced by @xcite . we also consider the calculation of observables with m1 transition up to next - to - next - to - leading order ( @xmath3 ) with the following significant changes in comparison with the previous eft(@xmath0 ) calculation @xcite : a ) including the diagrams with radiation from external nucleon leg , external deuteron leg , and on - shell two - body bubble ( see the diagrams `` @xmath7 '' , `` @xmath8 '' , and `` @xmath9 '' in fig.[fig : nd capture ] of page ) , b ) considering both contributions corresponding to two nucleon poles before and after photon creation in the first diagram of the second row in fig.[fig : nd capture ] , c ) inserting the diagram with the radiation directly from the exchanged nucleon , d ) adding the contribution of the @xmath10 m1 transition and e ) introducing .and using the properly normalized triton wave function . the triton or helium - three wave functions consist of two parts , one is the nucleon and dibaryon cluster wave function and the other is the two nucleon structure of the dibaryon cluster . we follow the bethe - salpeter ( bs ) equation in @xcite and we use the normalization condition of the relativistic two - body vertex function and work out the nonrelativistic one which is suitable for neutron - deuteron ( @xmath11 ) scattering leading to the formation of a triton . the theoretical calculations of the observables in the @xmath1 @xmath2 process were previously performed based on model - dependent approaches @xcite . the cross section and polarization observables were studied theoretically for radiative capture reactions @xmath12@xmath13@xmath14 and @xmath12@xmath15@xmath16 at low energies @xcite . the cross section for thermal neutron radiative capture on the deuteron was measured to be @xmath17 mb @xcite , in agreement with the results of earlier experiments @xcite . in the present work , the calculation of all m1 diagrams are calculated for the incoming doublet and quartet @xmath11 channels fully in the cluster - configuration space up to @xmath3 in sec.[sec : amplitude ] . the calculation of the cross section for @xmath1 @xmath2 is presented in sec.[sec : cross section ] . in sec.[sec : numerical ] numerical aspects of the calculation of m1 amplitudes are discussed . the results and comparison with other theoretical and experimental works are explained in sec.[sec : results ] . finally , we summarize the paper and discuss future investigations in sec.[sec : conclusion ] . in this section , we focus on the introduction of the eft(@xmath0 ) amplitude for the @xmath1 @xmath2 process up to @xmath3 . we concentrate on the zero - energy regime and try to calculate the amplitude of the neutron radiative capture by deuteron at thermal energy ( @xmath18 mev ) in the center - of - mass ( c.m . ) frame . in the eft(@xmath0 ) method , the electromagnetic ( em ) interactions in the three - body systems can be inserted principally by considering the one- , two- , and three - body currents . however , we show that the cutoff independence is achieved up to @xmath3 with one- and two - body currents and therefore there is no need for additional three - body currents up to @xmath3 calculations . in the very - low - energy regime the m1 transition has a dominant piece in the amplitude of @xmath1 @xmath2 . e2 transition also contribute to the @xmath1 @xmath2 reaction but comparing with the m1 interaction , it has a negligible contribution . in the following , we evaluate the eft(@xmath0 ) amplitude of the neutron radiative capture by deuteron reaction by considering the dominant m1 transitions using one- and two - body currents up to @xmath3 . note that the convection current of the proton ( e1 transition ) has odd parity ( due to one power of nucleon momentum ) , so this mixes an incoming p - wave state to the final s - wave triton . capture from the p wave introduces the factor of the external nucleon momentum forcing the amplitude to vanish at threshold . the lagrangian of the s - wave strong interactions using a dibaryon auxiliary field are given by @xcite @xmath19d^{a}_{s } \qquad\qquad\qquad\qquad\quad\nonumber \\ + d^{i^\dag}_{t}\big[\delta_t - c_{0t}\big(id_0+\frac{\vec{d}^2}{4m_n}+\frac{\gamma^2_t}{m_n}\big)\big]d^{i}_{t } -y\big(d^{a^\dag}_{s}(n^\dag p^a n)+d^{i^\dag}_{t}(n^\dag p^i n)+h.c.\big)\quad\!\!\!\nonumber\\ + \frac{m_n y^2 \mathcal{h}(e,\lambda)}{6}\;n^\dag\bigg(\big(d^{i}_{t}\sigma_i\big)^\dag\big(d^{j}_{t}\sigma_j\big)-\big[\big(d^{i}_{t}\sigma_i\big)^\dag\big(d^{a}_{s}\sigma_a\big ) + h.c.\big]+\big(d^{a}_{s}\tau_a\big)^\dag\big(d^{b}_{s}\sigma_b\big)\bigg)n\!\!\!\!\!\!\!\!\nonumber\\+\cdot\cdot\cdot\,,\;\quad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\:\;\end{aligned}\ ] ] where @xmath20 is the covariant derivative which acts on the nucleon and dibaryon fields with @xmath21 and @xmath22 relations , respectively . @xmath23 is the external field and @xmath24 and @xmath25 for proton - proton , neutron - neutron and neutron - neutron dibaryons . the center dots in the last line denotes the other suppressed terms . in eq.([eq:1 ] ) , @xmath26 is the nucleon iso - doublet field . the dibaryon auxiliary fields for deuteron and iso di - nucleon systems are introduced by @xmath27 and @xmath28 , respectively . the operators @xmath29 and @xmath30 with @xmath31 ( @xmath32 ) as isospin ( spin ) pauli matrices are the projection operators of nucleon - nucleon ( @xmath33 ) @xmath34 and @xmath35 states , respectively . @xmath36 represents the nucleon mass and the three - nucleon force is introduced by @xmath37 , where @xmath38 and @xmath39 are the total energy and cutoff momentum . the @xmath37 , which absorbs all dependence on the cutoff as @xmath40 , is given by @xcite @xmath41 where the interactions proportional to @xmath42 enter at @xmath43 @xcite . in our calculation , we consider generally @xmath44 . the parameters @xmath45 and @xmath46 are given by matching the eft(@xmath0 ) @xmath33 scattering amplitude to the effective range expansion ( ere ) of the scattering amplitude of two non - relativistic nucleons around the @xmath47 @xcite . @xmath48 mev is the binding momentum of the deuteron and @xmath49 with @xmath50 fm as the scattering length in the @xmath35 state . the lagrangian of the m1 interaction is constructed by considering the nucleon and dibaryon operators coupling to the magnetic field @xmath51 , @xmath52 in the above equation , @xmath53 and @xmath54 with @xmath55 ( @xmath56 ) as the proton ( neutron ) magnetic moment are the isoscalar and isovector nucleon magnetic moments , respectively . @xmath57 is the electric charge and @xmath58 fm ( @xmath59 fm ) denotes the effective range of the triplet ( singlet ) @xmath33 state . the coefficients @xmath60 fm and @xmath61 fm , which enter at next - to - leading order ( nlo ) , have been fixed from the cross section of @xmath62 @xmath63 at thermal energy , @xmath64 mb and the deuteron magnetic moment @xmath65 , respectively @xcite . the m1 @xmath1 @xmath2 diagrams at @xmath66 ( @xmath67 ) . the `` @xmath68 '' superscript denotes the contribution up to @xmath66 . all possible diagrams in the m1 transition of the @xmath1 @xmath2 process up to @xmath66 are shown in the first and second lines . the diagrams in the third line are the expanded version of the first diagram in the second line . the solid , wavy , and double lines represent a nucleon , a photon , and a dibaryon field , respectively . the nucleon - nucleon - photon ( @xmath69 ) and dibaryon - dibaryon - photon ( @xmath70 ) vertices show the one- and two - body m1 interactions . @xmath71 is the @xmath72 propagator matrix of the dibaryon auxiliary fields and the three - body force is indicated by @xmath73 . the dashed oval and dashed half oval denote the @xmath74 scattering amplitude and the normalized triton wave function up to @xmath66 , respectively.,width=623 ] the diagrams of the m1 transition in the @xmath1 @xmath2 process up to @xmath66 ( @xmath67 ) are schematically shown in fig.[fig : nd capture ] . note that in the entire paper , the superscript `` ( n ) '' denotes the contribution from the sum of all pieces up to , and including , order n. @xmath71 indicates the propagator of the dibaryon fields up to @xmath66 which is given in the cluster - configuration space as @xmath75 where @xmath76 we emphasize that the above propagators can be applied up to @xmath77 and should be corrected for the higher orders . in fig.[fig : nd capture ] , the dashed oval denotes the nucleon - deuteron ( @xmath74 ) scattering amplitudes which are presented by @xmath78 and @xmath79 for the doublet and quartet channels up to @xmath66 , respectively . the faddeev equations of @xmath80 are introduced in appendix [ appendix a ] . the dashed half oval indicates the normalized triton wave function up to @xmath66 which is introduced by @xmath81 in the following . the procedure of making the triton wave function and its normalization condition are briefly presented in appendix [ appendix b ] . we consider the contribution of all diagrams shown in fig.[fig : nd capture ] in the amplitude of neutron radiative capture by a deuteron reaction . the third diagram of the second line and all diagrams of the first line in fig.[fig : nd capture ] have not been considered in the previous eft(@xmath0 ) calculations of the @xmath1 @xmath2 amplitude @xcite . we have also added the contribution of the @xmath82@xmath83 m1 transition to the amplitude of the @xmath1 @xmath2 process which was previously not considered in @xcite . this two - body m1 transition is indicated in the lagrangian of eq.([eq:6 ] ) by the @xmath84 coefficient which enters first at nlo as @xmath85 . however the contribution of the @xmath82@xmath83 m1 transition is small at nlo but its effect is significant at @xmath3 . before we evaluate the contribution of the diagrams in fig.[fig : nd capture ] , let us make a comment about the computational process of the amplitude at nlo and @xmath3 . introducing the @xmath66 diagrams as in fig.[fig : nd capture ] , includes some diagrams of higher order , for example the nlo calculation includes @xmath3 , @xmath77 and @xmath86 terms . so , this calculation includes higher - order terms , but it is not - or at least , not immediately - a full higher - order correction , and so does not achieve that precision . on the other hand , the additional diagrams are small in a well behaved expansion , so the precision is not compromised . this procedure is made only for convenience in the computational process . now , we make a comment about the evaluation of the first diagram in the second line of fig.[fig : nd capture ] . this diagram is different somewhat from other ones because it has two contributions corresponding to the poles in the nucleon propagators before and after the photon creation . therefore , these two poles are corresponding to two contributions , one is that the photon is emitted during the exchange of a nucleon and the other is that the photon is emitted after exchanging the nucleon . if we add the half - offshell @xmath74 scattering amplitude from left to the first diagram in the first line of fig.[fig : nd capture ] , we miss the contribution of the case that the photon emission occurs during the nucleon exchange . so , we have to replace the half - offshell @xmath74 scattering amplitude by four diagrams which are introduced in fig.[fig : nd scattering ] . thus , we must substitute the first diagrams of the second line by four diagrams introduced in the third line of fig.[fig : nd capture ] . the effect of the photon emitted during the exchange of another nucleon can also be applied when the triton formation precedes the photon - nucleon interaction . but the evaluation of this effect makes no significant changes in the final results . by working in the coulomb gauge , the m1 amplitude of the @xmath1 @xmath2 can be written as two orthogonal terms , @xmath87 with @xmath88 , @xmath89 , @xmath90 , and @xmath91 are the final @xmath14 ( or @xmath16 ) field , the three - vector polarization of the produced photon , the three - vector polarization of the deuteron and the unit vector along the 3-momentum of the photons , respectively . in the @xmath1 @xmath2 process , two initial doublet ( @xmath92 ) and quartet ( @xmath93 ) channels can make the final triton state using the m1 transition . if we evaluate the contributions of all diagrams in fig.[fig : nd capture ] , we can generally write the @xmath66 ( @xmath67 ) amplitude of the @xmath1 @xmath2 process as @xmath94n\end{aligned}\ ] ] where @xmath95 the @xmath96 diagram in fig.[fig : nd capture ] . @xmath97 and @xmath98 denote the incoming c.m . momentum and the total energy of the initial @xmath11 system , respectively . all notation are the same as in fig.[fig : nd capture].,width=226 ] for example , we concentrate on the detailed evaluation of the diagram @xmath96 contribution . the energy and momentum of the incoming particles are shown in fig.[fig : a4 ] . we start by writing the amplitude of the diagram in fig.[fig : a4 ] using the lagrangians in eqs.([eq:1 ] ) and ( [ eq:6 ] ) . generally , before applying the projection operators , we can write the contribution of the diagrams in fig.[fig : a4 ] in the cluster - configuration space up to @xmath66 ( @xmath67 ) as @xmath99 where @xmath97 is the incoming momentum and @xmath100 denotes the energy of the initial @xmath11 system . @xmath101 represents the final state energy which is given by @xmath102 with @xmath103 mev as the binding energy of the triton . the @xmath104 ( @xmath105 ) and @xmath106 ( @xmath107 ) indices are the spin ( isospin ) components of the incoming and outgoing dibaryons . to solve the energy integration , we introduce the poles of eq.([eq:22 ] ) in the complex plane . it is obvious that we have the three following pole : @xmath108 where they result from the denominator of the nucleon propagators . with respect to the poles in eq.([eq:022 ] ) and doing the integration over energy and the solid angle , the @xmath109 is @xmath110\nonumber\\ \times\left ( \begin{array}{cc } ( k_0+k_1\tau_3)\sigma_k\sigma_s\sigma_rb_k & ( k_0+k_1\tau_3)\sigma_k\tau_a\sigma_rb_k \\ ( k_0+k_1\tau_3)\sigma_k\sigma_s\tau_b b_k & ( k_0+k_1\tau_3)\sigma_k\tau_a\tau_bb_k \\ \end{array } \right).\qquad\qquad\qquad\qquad\qquad\end{aligned}\ ] ] with @xmath111 as the zeroth legendre polynomial of the second kind . in order to obtain the contribution of the diagram in fig.[fig : a4 ] for the @xmath1 @xmath2 reaction , we have to project the initial @xmath74 system to the doublet and quartet cases ( corresponding to two possible m1 transitions ) while the final state should be @xmath92 because of the triton . the contribution of the m1 transition with the initial quartet channel ( @xmath93 ) is calculated by applying the projection operators @xmath112 and @xmath113 with @xmath114 as the spin component of the deuteron in the quartet channel , from left and right in eq.([eq:0022 ] ) , respectively . so , we gain @xmath115 \nonumber\\ \times \left ( \begin{array}{cc } k_0+k_1\tau_3 & 0 \\ 0 & 0 \\ \end{array } \right ) t^\dag\big[2\,i\vec{\varepsilon}_d\cdot\vec{\varepsilon}^\ast_\gamma \times\vec{\tilde{q}}-\vec{\sigma}\times\vec{\varepsilon}_d\cdot\vec{\varepsilon}^\ast_\gamma\times\vec{\tilde{q}}\big]n.\qquad\qquad\qquad\qquad\;\,\end{aligned}\ ] ] taking into account the projection operators @xmath116 and @xmath117 for the incoming and outgoing channels , respectively , the contribution of the diagram in fig.[fig : a4 ] for the initial doublet channel is given by @xmath118 \nonumber\\ \times \left ( \begin{array}{cc } k_0+k_1\tau_3 & -3(k_0+k_1\tau_3 ) \\ 3(3k_0-k_1\tau_3 ) & -(3k_0-k_1\tau_3 ) \\ \end{array } \right ) t^\dag\big[i\vec{\varepsilon}_d\cdot\vec{\varepsilon}^\ast_\gamma \times\vec{\tilde{q}}+\vec{\sigma}\times\vec{\varepsilon}_d\cdot\vec{\varepsilon}^\ast_\gamma\times\vec{\tilde{q}}\big]n.\qquad\!\!\end{aligned}\ ] ] the results of eqs.([eq:26 ] ) and ( [ eq:25 ] ) are calculated using @xmath119 and the sum over the repeated indices . finally , the total contribution of the diagram in fig.[fig : a4 ] can be written as @xmath120n.\end{aligned}\ ] ] where @xmath121 with @xmath122 . by ignoring the normalization factor of the incoming deuteron , eq.([eq:27 ] ) is as we expected . one can evaluate the contribution of all diagrams in fig.[fig : nd capture ] using the same procedure as for the @xmath96 diagram . after applying the integration over energy and solid angle , the contribution of all m1 diagrams in the @xmath123 function ( eq.([eq:8 ] ) ) , before multiplying the deuteron wave function normalization factor , is given by @xmath124 where @xmath125 in the above , @xmath126 can be @xmath127 or @xmath128 for doublet and quartet channels , respectively . the @xmath129 matrix function @xmath130 with @xmath131 represents the contribution of the @xmath132 diagram in fig.[fig : nd capture ] for the initial @xmath126 channel up to @xmath66 ( @xmath67 ) . for the initial doublet ( @xmath133 ) state , in the cluster - configuration space , we obtain @xmath134 @xmath135\left ( \begin{array}{cc } 2k_0 & k_1\tau_3 \\ k_1\tau_3 & 0 \\ \end{array } \right ) , \end{aligned}\ ] ] @xmath136\left ( \begin{array}{cc } -5k_0 + 5k_1\tau_3 & 3k_0+k_1\tau_3 \\ 3k_0+k_1\tau_3 & 3k_0 + 5k_1\tau_3 \\ \end{array } \right ) , \end{aligned}\ ] ] @xmath137 @xmath138 \left ( \begin{array}{cc } k_0+k_1\tau_3 & -3(k_0+k_1\tau_3 ) \\ 3(3k_0-k_1\tau_3 ) & -(3k_0-k_1\tau_3 ) \\ \end{array } \right ) , \end{aligned}\ ] ] @xmath139 \left ( \begin{array}{cc } k_0+k_1\tau_3 & -(k_0+k_1\tau_3 ) \\ 3k_0-k_1\tau_3 & -(3k_0-k_1\tau_3 ) \\ \end{array } \right).\end{aligned}\ ] ] also , in the incoming quartet channel ( @xmath140 ) , we have @xmath141 @xmath142\left ( \begin{array}{cc } 2k_0 & 0 \\ k_1\tau_3 & 0 \\ \end{array } \right ) , \end{aligned}\ ] ] @xmath143\left ( \begin{array}{cc } -k_0+k_1\tau_3 & 0 \\ -3k_0-k_1\tau_3 & 0 \\ \end{array } \right ) , \end{aligned}\ ] ] @xmath144 @xmath145 \left ( \begin{array}{cc } k_0+k_1\tau_3 & 0 \\ 0 & 0 \\ \end{array } \right ) , \end{aligned}\ ] ] @xmath146 the results of @xmath147 are obtained after applying the appropriate projection operators for initial and final states . we note that the @xmath148 must be zero since in the quartet ( @xmath149=@xmath150 ) channel all spins are aligned and there is no three - body interaction in this channel because the pauli principle forbids the three nucleons to be at the same point in space . low - energy observables of the @xmath1 @xmath2 process are cutoff - independent by the introduction of @xmath151 and @xmath152 up to @xmath3 ( see table [ tab : cutoff var ] ) . namely , they are renormalized and therefore no new three - body forces are needed up to @xmath3 . the same argument can be applied equally with three - body currents @xcite , so no three - body currents are included in the present calculation . we stress that the @xmath153 amplitude is a @xmath129 matrix which is written in the cluster - configuration space and so the contributions of both initial @xmath154 and @xmath155 systems are taken into account . thus , the physical amplitude of the @xmath1 @xmath2 process is given by @xmath156 where @xmath157 indicates the normalization factor of the incoming deuteron wave function at @xmath66 , @xmath158 must be applied for the @xmath1 @xmath2 process . in the following , we use the @xmath159 amplitude for calculating the total cross section of the @xmath1 @xmath2 . in order to proceed to calculate the cross section , we use the following spin sums : @xmath160 where the factor @xmath161 comes from the average over initial state polarizations . the above calculations are done in the coulomb gauge @xmath162 and the results are given using @xmath163 , @xmath164 and @xmath165 , where the upper and lower signs denote the photon with the right and left helicity , respectively . from eq.([eq:15 ] ) , the total cross section of the neutron radiative capture by a deuteron can be written as @xmath166 where @xmath167 superscript denotes @xmath66 results and @xmath168 is the incident neutron velocity in the c.m . frame . in the computation of the m1 amplitude of the diagrams in fig.[fig : nd capture ] , we need to obtain the triton wave function and the half - off - shell @xmath74 scattering amplitude at leading ( lo ) , next - to - leading , and next - to - next - to - leading orders . the half - off - shell neutron - deuteron scattering is obtained order by order by solving numerically the faddeev equations which are introduced in appendix a for both initial doublet and quartet channels . we solve them by the hetherington - schick method @xcite in a @xmath169 code with a specific cutoff momentum @xmath39 . we also obtain the triton wave function at each order by solving the homogenous part of the faddeev equations of @xmath74 scattering in the doublet channel with the same cutoff and then normalize it by the method which is introduced in appendix [ appendix b ] . using the order - by - order results of @xmath170 and @xmath171 ( @xmath172 and @xmath173 ) , we can be able to solve the integrations in eqs.([eq:10 ] ) , ( [ eq:12 ] ) and ( [ eq:13 ] ) to obtain the m1 amplitude of @xmath1 @xmath2 . we solve these integrations numerically using the gaussian quadrature weights and also the same cutoff momentum @xmath39 as before . as we see from eq.([eq:05 ] ) , the parameter @xmath174 and @xmath175 must be determined order by order for the cutoff @xmath39 . at each order , we obtain the value of @xmath151 by constructing the exact triton scattering length , @xmath176 fm . the @xmath152 parameter which enters at @xmath3 is determined for an arbitrary cutoff value by matching the triton binding energy to the experimental value , @xmath177 mev . in this work , we have concentrated on the evaluation of the cross section of the @xmath1 @xmath2 process up to @xmath3 . our eft(@xmath0 ) results for the amplitudes and cross sections of the @xmath1 @xmath2 process at thermal energy , @xmath178 mev are shown in table [ tab : results ] . we compare schematically our eft(@xmath0 ) results at thermal energy for the cross section with the previous model - dependent theoretical calculations and the experimental data in fig.[fig : plot ] . cccccc @xmath179 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & @xmath184 + @xmath25 & @xmath185 & @xmath186 & 0.232 & 0.065 & 0.297 @xmath187 0.196 + @xmath188 & @xmath189 & @xmath190 & 0.264 & 0.157 & 0.421 @xmath187 0.093 + @xmath191 & @xmath192 & @xmath193 & 0.273 & 0.196 & 0.469 @xmath187 0.033 + we use the power counting introduced by bedaque @xmath194 in @xcite . the eft(@xmath0 ) expansion parameter is @xmath195 , where @xmath196 and @xmath197 are the small and large parameters , so the nlo and @xmath3 diagrams enter 33@xmath198 and 11@xmath198 corrections to the leading- and next - to - leading - order amplitudes , respectively . also , with respect to our power counting , the error of @xmath3 amplitude must be less than 3.7@xmath198 of the exact value . the cross section is proportional to the square of the amplitude . it is obvious that if the systematic eft(@xmath0 ) error in the amplitude is @xmath199 , as an example , the cross section has a maximum error @xmath200 in the eft(@xmath0 ) approach . so , we expect to have a maximum error of 7@xmath198 at @xmath3 for the cross section . comparison between different theoretical results for the total cross section of the @xmath1 @xmath2 process . the points from left to right denote the results computed by experiment @xcite , our lo eft(@xmath0 ) , our nlo eft(@xmath0 ) , our @xmath3 eft(@xmath0 ) , earlier @xmath3 eft(@xmath0 ) @xcite , av14/viii(ia+mi+md ) @xcite , av18/ix(ia+mi+md ) @xcite , av14/viii(ia+mi+md+@xmath201 ) @xcite , av18/ix(ia+mi+md+@xmath201 ) @xcite , av14/viii(ia+mi+md+@xmath202 ) @xcite , av18/ix(ia+mi+md+@xmath202 ) @xcite , av18/ix(gauge inv . ) @xcite , and av18/ix(gauge inv.+@xmath203-current ) @xcite methods , respectively . the thin band indicates the error band of the experimental result of the cross section . two horizontal dashed lines determine the upper and lower limits due to our systematic eft(@xmath0 ) error at @xmath3.,width=604 ] our results in table [ tab : results ] show the convergence in our power counting from lo to @xmath3 . at nlo , 0.124 mb adds to the leading - order value and at @xmath3 0.048 mb to the next - to - leading order . our eft result for the total cross section of neutron radiative capture by a deuteron at @xmath3 , @xmath204 mb , has an error of 7@xmath198 compare with the experimental value , @xmath17 mb . we stress that the contribution of the e2 transition has not been included in our calculation for the amplitude of the @xmath1 @xmath2 reaction . the e2 transition is suppressed by two powers of the initial nucleon momentum or photon energy compared to the dominant m1 transition . therefore , this effect numerically has a contribution of @xmath205 correction in the quartet - initial - channel amplitude of @xmath1 @xmath2 and so @xmath206 in total cross section at threshold regime . also , with respect to the power counting as discussed above , we expect a maximum error @xmath207 in @xmath3 eft(@xmath0 ) results of the cross section . thus the 7@xmath198 error in our @xmath3 results is acceptable . we believe that the higher - order corrections make this discrepancy narrow . cc @xmath179 & abs[@xmath208 + @xmath25 & 0.098756 + @xmath188 & 0.045714 + @xmath191 & 0.004006 + according to table [ tab : cutoff var ] , we have computed the cutoff variation of our eft(@xmath0 ) results for the total cross section within a natural range of @xmath209 to @xmath210 mev at lo , nlo , and @xmath3 . the range of cutoff variation should be a few times the pion mass because , here , the existence of a definite @xmath40 limit in an eft calculation does not guarantee that the results found in that limit are rigorous consequences of the eft @xcite . our results in table [ tab : cutoff var ] indicate that the m1 amplitudes and the cross section of @xmath1 @xmath2 are cutoff independent and properly renormalized . the differences of our results and the previous eft(@xmath0 ) calculation of total cross section at thermal energy @xcite are due to the ignored diagrams and the @xmath82@xmath83 m1 transition effects . @xmath211 & @xmath212 & @xmath213 difference@xmath213 + @xmath25 & @xmath214 & @xmath214 & 0 + @xmath188 & @xmath215 & @xmath216 & @xmath217 + @xmath191 & @xmath218 & @xmath219 & @xmath220 + the @xmath84 coefficient corresponding to the contribution of the @xmath82@xmath83 m1 transition is small compared with @xmath85 which comes from the @xmath221@xmath83 m1 transition @xcite . so , we expect that the @xmath82@xmath83 m1 transition has a small ( and negligible ) effect at nlo results but at @xmath3 the @xmath82@xmath83 m1 transition could have a significant effect . our results for the total cross section with and without the @xmath84 coefficient effect which are summarized in table [ tab : l2 effect ] are as we expected . .the investigation of the ignored contributions in the previous eft(@xmath0 ) calculation @xcite at each order . @xmath222 indicates the total @xmath66 amplitude of m1 @xmath1 @xmath2 without the contribution of the time ordering that corresponds to the photon emitted during the nucleon exchange in the first diagram of the second line in fig.[fig : nd capture ] . @xmath223 denotes the sum of the amplitudes of the diagrams `` @xmath7 '' , `` @xmath8 '' , and `` @xmath9 '' for the incoming @xmath126 channel . also , @xmath224 is only the contribution of the `` @xmath225 '' diagram in the first line of fig.[fig : nd capture ] for the initial @xmath126 channel , respectively . @xmath226 and @xmath227 in the first column indicate the initial quartet ( @xmath228@xmath229 ) and doublet ( @xmath230@xmath231 ) channels , respectively . the results of the amplitudes are presented in units of @xmath232 . [ cols="^,^,^,^,^,^",options="header " , ] the effects of the diagrams in fig.[fig : nd capture ] which have been neglected in the previous eft(@xmath0 ) calculation @xcite have been investigated in table [ tab : ignored diagrams effects ] . the results in the third column of table [ tab : ignored diagrams effects ] are the total doublet and quartet amplitudes of the m1 @xmath1 @xmath2 transition at lo , nlo and @xmath66 . in the fourth column , we present the computed values of the contribution which is only corresponding to the nucleon pole before photon creation in the first diagram in the second line of fig.[fig : nd capture ] at each order . the fifth and sixth columns of table [ tab : ignored diagrams effects ] represent only the evaluated values for the amplitudes of the `` @xmath233 '' and `` @xmath225 '' diagrams in the first line of fig.[fig : nd capture ] , respectively , for both doublet and quartet channels . the lack of a correct calculation of the first diagram of the second line in fig.[fig : nd capture ] creates the significant errors as indicated in the fourth column of table iv at each order . the results shown in the fifth column of table [ tab : ignored diagrams effects ] indicate that the diagrams with radiation from external nucleon leg , external deuteron leg , and on - shell two - body bubble in the first line of fig.[fig : nd capture ] are lo effects and so , one expects that these diagrams have a very important effect in the final results of the amplitude of the m1 @xmath1 @xmath2 transition and could not be ignored . but the last column depicts that the diagram `` @xmath225 '' in the first line of fig.[fig : nd capture ] has a small effect at lo especially in the quartet channel as estimated in the previous eft(@xmath0 ) calculation @xcite . finally , we emphasize that our results have been evaluated using the properly normalized triton wave function . in the present parity - conserving eft(@xmath0 ) calculation , we have calculated the amplitudes and cross section for @xmath1 @xmath2 fully in the cluster - configuration space up to @xmath3 . we have considered one- and two - body currents . no three - body currents are needed to renormalize the observables in this work up to @xmath3 . the m1 is the dominant transition at the low energies . we have included the contribution of the possible diagrams which have not been included in the previous eft(@xmath0 ) calculations and used the properly normalized triton wave function . we have also considered the effects of the @xmath82@xmath83 m1 transition ( @xmath84 coefficient in eq.([eq:6 ] ) ) in the @xmath1 @xmath2 amplitudes together with other transitions which are included in the previous eft(@xmath0 ) calculations . the @xmath3 eft(@xmath0 ) total cross section is determined to be @xmath234 mb . the reliable calculation of the doublet and quartet amplitudes can be used in the calculation of parity - violating observables in the @xmath1 @xmath2 process . the @xmath3 eft(@xmath0 ) total cross section @xmath235 is within 7@xmath198 of the measured values . the remaining discrepancies between theory and experiment indicate that inclusion of 1 ) higher order corrections and 2 ) higher - order multipoles contributions , may refine the differences . this work was supported by the research council of the university of tehran . the diagrams of the @xmath74 scattering amplitude up to @xmath66 are shown in fig.[fig : nd scattering ] . the faddeev equation of the diagrams in fig.[fig : nd scattering ] for the quartet channel in the cluster - configuration space is given by @xmath236 and for the @xmath74 scattering in the doublet ( @xmath149=@xmath237 ) channel , we have @xmath238 \;\nonumber\\-\frac{1}{\pi}\int^\lambda_0dqq^2\left[\mathcal{k}(e;q , p)\left ( \begin{array}{cc } 1 & -3 \\ -3 & 1 \\ \end{array } \right)+\mathcal{h}(e,\lambda)\left ( \begin{array}{cc } 1 & -1 \\ -1 & 1 \\ \end{array } \right)\right ] \nonumber\\ \times\mathcal{d}^ { ( n)}(e , q)\left ( \begin{array}{cc } t^{(n)}_{d_{d_tn\rightarrow d_tn } } & t^{(n)}_{d_{d_sn\rightarrow d_tn } } \\ t^{(n)}_{d_{d_tn\rightarrow d_sn } } & t^{(n)}_{d_{d_sn\rightarrow d_sn } } \\ \end{array } \right)(e;k , q),\qquad\qquad\;\:\end{aligned}\ ] ] where @xmath239 , @xmath97 and @xmath240 are the total energy of @xmath74 system , the incoming and outgoing momentums , respectively . the @xmath74 scattering diagrams up to @xmath66 ( @xmath241 ) . all notations are the same as in fig.[fig : nd capture].,width=491 ] in eq.([eq:21 ] ) , @xmath242 denotes the @xmath243 transition amplitude ( @xmath244 or @xmath88 ) in the doublet channel . the propagator of the exchanged nucleon , @xmath245 , is @xmath246 where @xmath247 indicates the angle between @xmath248 and @xmath249 vectors . other variables in the above equation are similar to the text . the results of eqs.([eq:20 ] ) and ( [ eq:21 ] ) are evaluated by considering the operators for projecting the @xmath74 system to @xmath92 and @xmath93 channels . for the doublet and quartet channels the projection operators @xmath250 and @xmath251 are used , respectively , with iso - spin index @xmath105 and spin indices @xmath252 and @xmath253 @xcite . the normalized triton wave function is obtained by solving the homogeneous part of eq.([eq:21 ] ) with the application of @xmath254 , where @xmath255 is the binding energy of the triton . so , the homogeneous part of eq.([eq:21 ] ) for the calculation of the triton wave function up to @xmath66 can be written as @xmath256 \mathcal{d}^ { ( n)}(-b_t , q)\,t^{(n)}_{^3h}(q),\end{aligned}\ ] ] where @xmath257 . generally , @xmath258 denotes the contribution of the @xmath259 transition ( @xmath244 or @xmath88 ) for making the triton . one can be able to normalize the solution of eq.([eq:0005 ] ) for the incoming deuteron channel by @xcite @xmath260\bigg|_{e =- b_t}\!\!\!t^{(n)}_{^3h}(q')\bigg(\begin{array}{c } 1 \\ 0 \end{array } \bigg),\end{aligned}\ ] ] where @xmath261 is given by @xmath262.\end{aligned}\ ] ] |
we point out that when a d - brane is placed in an ns - ns field background with non - vanishing field strength ( ) along the d - brane worldvolume , the coordinate of one end of the open string does not commute with that of the other in the low energy limit . the degrees of the freedom associated with both ends are not decoupled and accordingly , the effective action must be quite different from that of the ordinary noncommutative gauge theory for a constant background . we construct an associative and noncommutative product which operates on the coordinates of both ends of the string and propose a new type of noncommutative gauge action for the low energy effective theory of a d-brane . this effective theory is bi - local and lives in twice as large dimensions ( ) as in the case . when viewed as a theory in the-dimensional space , this theory is non - local and we must force the two ends of the string to coincide . we will then propose a prescription for reducing this bi - local effective action to that in dimensions and obtaining a local effective action . maketitle 2em title 1.5em .5em [ cols="^ " , ] 1.5em = -1 cm = 1.5 cm = -.0 cm = 15.5 cm | hep-th0109125 | noncommutative field theories@xcite@xcite are obtained in an @xmath7 limit of the open string theory . this noncommutativity is inherent in the open string theory@xcite . by considering an open string in an ns - ns @xmath0 field background and taking a specific field theory limit the effects of the @xmath0 field are encoded into a special multiplication rule of functions , the @xmath8 product or moyal product . by studying such field theories we expect that open string physics may be understood in the framework of field theory . until recently such investigations are mostly restricted to constant @xmath0 backgrounds@xcite@xcite@xcite@xcite . the algebra of the coordinates of the end points of the open string was studied , the connection of the commutative and noncommutative descriptions of the gauge theory was elucidated , and so on . when the @xmath0 field is constant , the background metric @xmath9 is flat . attempts to extend the analysis to the curved background , _ i.e. _ , the nonvanishing @xmath0 field strength @xmath1 , appeared in the context of open string theory in wzw models@xcite . in this approach it was found that the algebra of functions on the d - brane worldvolume is given by the @xmath10-deformation of the lie algebra . the general framework for general backgrounds was , however , not yet obtained . more recently a noncommutative field theory for a weak field strength @xmath11 and a weakly curved background was investigated@xcite . term were also considered in @xcite and the related underlying geometry was called h - poisson or twisted poisson one . ] several correlation functions of the string theory are obtained in the @xmath12 limit and it was concluded that the algebra of functions on the d - brane worldvolume is non - associative and noncommutative . various correlation functions are expressed in terms of a non - associative product @xmath13 . in @xcite the commutation relations of the coordinate @xmath14 of the end point of the open string were studied by using the approximation of a very short and slowly moving string . it was found that the commutators of the coordinate @xmath14 contain a momentum @xmath3 and it was argued that the space on which the d - brane lives is not the usual noncommutative space which we are getting used to . in @xcite a new product @xmath15 was defined in order to make the product @xmath13 associative on the functions of @xmath14 and @xmath3 and an attempt to define a gauge transformation in this ` @xmath14-@xmath3 space ' was presented . an explicit construction of the gauge theory action , however , was not obtained yet . the purpose of the present paper is to understand how the low energy effective theory can be formulated as an associative and noncommutative gauge theory , when the background space is curved and the field strength @xmath1 along the direction of the d - brane worldvolume is nonvanishing . our analysis will be restricted to the bosonic string . we assume that the field strength is weak and perform analysis in perturbation series in @xmath11 up to @xmath16 . we will first obtain the commutation relations of the coordinates @xmath17 , @xmath18 of the two ends ( at @xmath19 and @xmath20 ) of the open string in the low energy limit.(sec 2 ) this will be performed by discarding the oscillators of the string variable @xmath21 , but without further approximation . the result is striking . @xmath17 and @xmath18 do not commute with each other in contrast to the @xmath5 case . this means that when we take the @xmath12 limit , the gauge and scalar degrees of freedom at both ends of the string are not decoupled in the low energy effective theory . we are forced to take into account the degrees of freedom at both ends in constructing the field theory description and the numbers of gauge and scalar fields are doubled . furthermore the noncommutative product @xmath2 constructed according to the algebra of @xmath14 and @xmath22 is _ associative _ , but operates on both coordinates.(sec 3 ) we are thus lead to consider a bi - local field theory . all fields are functions of @xmath14 and @xmath22 . the gauge theory must be formulated in the 2@xmath6 dimensional @xmath23 space instead of the ordinary @xmath6 dimensional @xmath14 space . this bi - local field theory will , however , be acausal and must be regarded as an intermediate step toward construction of the local effective field theory , which will be discussed later . construction of this bi - local gauge theory can be performed in a standard way . the derivatives are generalized in such a way that they satisfy leibnitz rule with respect to the product @xmath2.(sec 4 ) the commutator of the gauge covariant derivatives defined in terms of these new derivatives gives the gauge field strength.(sec 5 ) the trace or the integral is defined in such a way that the cyclic property is respected . an obstruction in this prescription is the curved d - brane worldvolume . the metric of the effective gauge theory is the open string metric@xcite , which is curved already at @xmath24@xcite . because the gauge symmetry of the noncommutative gauge theory is realized by differential operators on the fields , if the action contains such a non - constant metric , one anticipates that the gauge symmetry may be broken . remarkably , we found that there exists a coordinate transformation in the @xmath25 space ( not in the @xmath17 space ) which effectively makes @xmath5 . in other words in a suitably chosen frame @xmath26 the coordinates satisfy the commutation relations of the @xmath5 theory . therefore it is natural to further assume that the metric is flat in this frame . moreover the derivatives with respect to @xmath27 and @xmath28 coincide with the new derivatives defined above to satisfy leibnitz rule . by writing down the action for the noncommutative gauge theory with @xmath29 in the flat @xmath30 frame and by going to the @xmath23 frame by means of the inverse coordinate transformation , we obtain the action for a bi - local , noncommutative gauge theory in curved backgrounds . as mentioned above , the above bi - local theory is non - local and will be acausal . next we will present a prescription to reduce our bi - local theory to a ` local theory'.(sec 6 ) actually , in order to obtain a massless string we must consider a limit @xmath31 . this is not automatically achieved in the @xmath12 limit . we will set @xmath32 by hand in the integrand of the effective action and integrate this over @xmath14 with a suitable weight function @xmath33 . we will show cyclic property of this reduced effective action which relates various correlation functions of the fields at @xmath14 and @xmath22 , _ i.e. _ , at both ends of the string . this reduced effective action is still invariant under a part of the gauge transformation . we propose this as the low energy effective action of the open string in the @xmath34 background . this reduced effective action , however , contains @xmath2 in the integrand and the manipulation such as differentiation with respect to @xmath14 and/or @xmath22 must be performed before setting @xmath32 . in this sense this reduced effective action is not the ordinary action in the @xmath14 space . finally in sec 7 we will speculate on the relation of our noncommutative gauge theory to the matrix model in curved backgrounds . we will quantize an open string theory with a d@xmath3-brane in the presence of an ns - ns @xmath0 field along the d@xmath3-brane , whose worldvolume spans a subspace @xmath35 . the field strength of the @xmath0 field , @xmath1 , is assumed to be small but non - vanishing in the direction of the d@xmath3-brane worldvolume . in this case we can perform a perturbative analysis in @xmath11 . such an analysis was first attempted in @xcite . we will simplify the problem by taking @xmath36 constant . by the consistency of the background due to weyl invarinace the metric tensor @xmath9 must be a constant up to @xmath24 @xcite . we will choose a gauge @xmath37 where @xmath38 is a constant . we assume that @xmath36 is nonvanishing for @xmath39 . when @xmath40 , @xmath38 is also in general non - vanishing because we can always shift @xmath41 by a constant . we also assume that @xmath0 is block diagonal , _ i.e. _ , @xmath42 for @xmath43 and @xmath6 is even and @xmath44 . in what follows we will restrict our attention only to this @xmath6-dimensional subspace . the worldsheet action of the open string in the conformal gauge is given by @xmath45 the coordinate @xmath41 must satisfy the following boundary conditions at @xmath46 . @xmath47 inside the worldsheet it must satisfy the equations of motion . @xmath48 for small @xmath11 the metric @xmath49 is constant up to @xmath24 and eq of motion ( [ eom1 ] ) reads @xmath50 where the dot and prime stand for @xmath51 and @xmath52 , respectively . the boundary condition ( [ bc1 ] ) is now expressed as @xmath53 where @xmath54 . we will solve ( [ eom2 ] ) under the condition ( [ bc2 ] ) up to @xmath24 . we set @xmath55 . the leading order solution is given by @xmath56 where @xmath57 stands for @xmath58 and @xmath59 @xmath17 , @xmath60 are ` zero modes ' and @xmath61 oscillators . we will next plug this in ( [ eom2 ] ) and derive an equation for the @xmath24 correction @xmath62 . this can easily be solved . @xmath63 here @xmath64 and @xmath65 are some functions to be determined by ( [ bc2 ] ) . the energies of the oscillation modes are proportional to @xmath66 and in the @xmath12 limit they can be ignored . in this paper we will restrict attention only to the ` zero modes ' and simply set @xmath67 . one may speculate that @xmath68 and @xmath69 may mix up due to the interaction and the commutators of @xmath68 may contain @xmath69 . we have explicitly checked that this is not the case up to @xmath24 . the full analysis including the oscillators is now in progress and will be reported elsewhere@xcite . eq ( [ bc2 ] ) forces @xmath64 , @xmath65 to satisfy an equation @xmath70 at @xmath71 . the primes on @xmath64 , @xmath65 stand for the derivatives with respect to the arguments . this equation can be solved by subtracting ( [ bc3 ] ) with @xmath19 from ( [ bc3 ] ) with @xmath72 and @xmath73 . this determines @xmath74 and then ( [ bc3 ] ) gives @xmath75 . @xmath76 the next step is to quantize this system . we will use the phase - space integral invariant of poincar @xmath77 to derive poisson brackets@xcite . it is defined by @xmath78 where @xmath79 is the canonical momentum conjugate to @xmath41 , @xmath80 @xmath81 and @xmath82 represent differential 1-forms . the quantity ( [ invariant ] ) is indepedent of @xmath83 and defines the poisson structure . substitution of ( [ x0 ] ) , ( [ x1 ] ) , ( [ p ] ) into ( [ invariant ] ) yields @xmath77 as a 2-form in the basis @xmath84 , @xmath85 , @xmath86 . the coefficients are a 2@xmath6 by 2@xmath6 matrix and its inverse gives the poisson brackets . they are then converted to the commutation relations . @xmath87 & = & 2\pi \alpha ' \ { i \theta^{\mu \nu } + \frac{i}{3 } ( \theta^{\mu \lambda } \theta^{\nu \rho } - g^{\mu \lambda } g^{\nu \rho } ) \rho \sigma } \ x^{\sigma } \nonumber \\ & & + \frac{\pi i}{9 } ( 3\theta^{\mu \lambda } \theta^{\nu \rho } - g^{\mu \lambda } g^{\nu \rho } ) \ \rho \sigma } \ ( jp)^{\sigma } - \frac{\pi i}{9}(\theta^{\mu \lambda } g^{\nu \rho}+g^{\mu \lambda } \theta^{\nu \rho } ) \ \rho \sigma } \ p^{\sigma } \ } , \nonumber \\ \ [ x^{\mu } , p^{\nu } ] & = & 2\pi \alpha ' \ { -\frac{i}{\pi } g^{\mu \nu}+\frac{i } { 3\pi } \theta^{\mu \lambda } g^{\nu \rho } \ h_{\lambda \rho \sigma } \ x^{\sigma } \nonumber \\ & & + \frac{i}{6}(3\theta^{\mu \lambda } \theta^{\nu \rho } - g^{\mu \lambda } g^{\nu \rho } ) \ \rho \sigma } \ p^{\sigma } -\frac{i}{2}(g^{\mu \lambda } \theta^{\nu \rho}-\theta^{\mu \lambda } g^{\nu \rho } ) \ \rho \sigma } \ ( jp)^{\sigma } \ } , \nonumber \\ \ [ p^{\mu } , p^{\nu } ] & = & \frac{4i}{3}\alpha ' \ g^{\mu \lambda } g^{\nu \rho } \ \rho \sigma } ( jp)^{\sigma}. \label{crpp}\end{aligned}\ ] ] here @xmath88 and @xmath89 are defined by @xmath90 these are the noncommutative parameter is different from that used in the literature by a factor @xmath91 . when multiplied by this factor , @xmath88 remains finite in the @xmath12 limit . we used the definition ( [ thetag ] ) in order to avoid messy expression for ( [ crpp ] ) . ] and the open string metric for the @xmath5 background@xcite . similar commutation relations in the @xmath12 limit are obtained in @xcite by using some approximation of short and slowly moving string . this @xmath12 limit is a specific one taken with fine tuning@xcite . @xmath9 is adjusted to be @xmath92 , while @xmath93 , @xmath94 . hence @xmath95 , @xmath96 and @xmath97 . defining the total momentum at @xmath98 by @xmath99 , we get in the @xmath12 limit @xmath100 & = & 2\pi \alpha ' \left(i \hat{\theta}^{\mu \nu}(x ) + \frac{2\pi i}{3 } \ \alpha'\theta^{\mu \lambda}\theta^{\nu \rho } \ h_{\lambda \rho \sigma } \ \theta^{\sigma \alpha } \ \tilde{p}_{\alpha } \right ) , \nonumber \\ \ [ x^{\mu } , \tilde{p}_{\nu } ] & = & i \delta^{\mu}_{\nu } + \frac{\pi i}{3 } \alpha ' \theta^{\mu \lambda } \theta^{\rho \alpha } \ h_{\nu \lambda \rho } \ \tilde{p}_{\alpha } , \qquad [ \tilde{p}_{\mu } , \tilde{p}_{\nu } ] = 0 , \end{aligned}\ ] ] which agrees with the result of @xcite , @xcite . here @xmath101 is @xmath88 in ( [ thetag ] ) with @xmath102 replaced by @xmath103 . in @xcite it is argued that the appearence of @xmath60 on the righthand sides of ( [ crpp ] ) makes the construction of the low energy effective action difficult . in @xcite some proposal for a construction of gauge transformations is presented . an explicit form of the action integral in the @xmath14-@xmath3 space is , however , not proposed . in this paper we will observe the relations ( [ crpp ] ) from a different view point and construct a noncommutative gauge theory based on this algebra . crucial point is that the coordinates of the end points of the open string ( at @xmath98 ) are given by @xmath104 throughout this paper we will use @xmath14 and @xmath22 to stand for the coordinates of the two ends of the open string . the difference of the two end point coordinates is thus proportional to @xmath3 . @xmath105 note that this is @xmath106 and does not automatically go to zero but is fixed in the @xmath12 limit . for large momenta the length of the open string grows.@xcite if @xmath38 is invertible , which we assume in this paper , so are @xmath107 and @xmath108 , and we can invert ( [ yxp ] ) . @xmath109 now the algebra ( [ crpp ] ) can be reexpressed in terms of @xmath17 and @xmath18 .. ] @xmath100 & = & i \theta^{\mu \nu } + \frac{i}{3 } ( \theta^{\mu \lambda}\theta^{\nu \rho } - g^{\mu \lambda}g^{\nu \rho } ) h_{\lambda \rho \alpha } x^{\alpha } + \frac{i}{9 } ( 3\theta^{\mu \lambda}\theta^{\nu \rho } - g^{\mu \lambda}g^{\nu \rho } ) h_{\lambda \rho \alpha}(y^{\alpha } - x^{\alpha } ) \nonumber \\ & & \qquad - \frac{i}{9 } ( \theta^{\mu \lambda}g^{\nu \rho } + g^{\mu \lambda}\theta^{\nu \rho } ) h_{\lambda \rho \alpha } { ( j^{-1})^{\alpha}}_{\beta}(y^{\beta } - x^{\beta } ) , \label{cr1}\end{aligned}\ ] ] @xmath110 & = & -\frac{i}{18 } ( 3\theta^{\mu \lambda}\theta^{\nu \rho } - g^{\mu \lambda}g^{\nu \rho } ) h_{\lambda \rho \alpha}(y^{\alpha } - x^{\alpha } ) \nonumber \\ & & + \frac{i}{18 } ( \theta^{\mu \lambda}g^{\nu \rho } + g^{\mu \lambda}\theta^{\nu \rho } ) h_{\lambda \rho \alpha}{(j^{-1})^{\alpha}}_{\beta}(y^{\beta } - x^{\beta } ) , \label{cr2}\end{aligned}\ ] ] @xmath111 & = & - i \theta^{\mu \nu } - \frac{i}{3 } ( \theta^{\mu \lambda}\theta^{\nu \rho } - g^{\mu \lambda}g^{\nu \rho } ) h_{\lambda \rho \alpha } y^{\alpha } + \frac{i}{9 } ( 3\theta^{\mu \lambda}\theta^{\nu \rho } - g^{\mu \lambda}g^{\nu \rho } ) h_{\lambda \rho \alpha}(y^{\alpha } - x^{\alpha } ) \nonumber \\ & & \qquad - \frac{i}{9 } ( \theta^{\mu \lambda}g^{\nu \rho } + g^{\mu \lambda}\theta^{\nu \rho } ) h_{\lambda \rho \alpha } { ( j^{-1})^{\alpha}}_{\beta}(y^{\beta } - x^{\beta } ) \label{cr3}\end{aligned}\ ] ] ( [ cr1 ] ) and ( [ cr3 ] ) shows that the coordinates of the ends of the open string @xmath17 , @xmath18 do not commute . the commutator of @xmath22 s is obtained from that of @xmath14 s by interchanging @xmath112 and reversing the overall sign . furthermore ( [ cr2 ] ) shows that the coordinate of one end @xmath14 does not commute with that of the other @xmath22 . this is in contrast to the @xmath5 case@xcite@xcite . the gauge and scalar degrees of freedom at the @xmath19 end and those at @xmath72 are not decoupled . to construct a low energy effective theory we need to consider both degrees of freedom . this means that if the two ends of the string are separated , the low energy effective field theory will become acausal . to avoid this problem we will take a limit @xmath31 in sec 5 . neverthelss these two ends must be separated for a while in order to construct an associative product . we expect that the full string theory including the oscillators is causal . the reason for the non - commutativity of @xmath14 and @xmath22 may lie in the fact that we discarded oscillators . it will be interesting to study the commutators of the full string variable @xmath113 . it can be shown that the commutation relations ( [ cr1])-([cr3 ] ) satisfy jacobi identities . we will then construct a product @xmath2 which realizes this algebra . the parameter @xmath36 is treated as a small purturbation but @xmath108 must be taken into account to the full order . we report only the results here . @xmath114 here @xmath8 is the ordinary noncommutative product for @xmath14 and @xmath22 in the @xmath5 case . @xmath115 the noncommutative parameter is @xmath108 for @xmath14 and @xmath116 for @xmath22 . the product ( [ newstar ] ) is associative . the proof of the associativity ( up to @xmath24 ) @xmath117 is straightforward . because the product @xmath2 contains both @xmath14 and @xmath22 , the product of the functions of @xmath14 only , @xmath118 will also depend on @xmath22 . hence we must regard all fields to be functions of both @xmath14 and @xmath22 , _ i.e. _ , bi - local fields . the low energy effective theory on the d - brane is expected to be some kind of noncommutative gauge theory . to construct such a theory we need to define proper derivatives . because the @xmath2 product ( [ newstar ] ) depends on the coordinates explicitly , this will not satify leibnitz rule . @xmath119 to circumvent this problem we must modify the derivatives . we define @xmath120 here @xmath121 , @xmath122 , @xmath123 and @xmath124 are linear functions of @xmath14 , @xmath22 and are @xmath24 . @xmath125 is the anti - commutator with respect to the product ( [ oldstar ] ) . we note that because @xmath69 , @xmath126 , @xmath127 , @xmath128 are linear functions , @xmath129 can be replaced by @xmath130 in ( [ derx ] ) , ( [ dery ] ) . these functions must be determined in such a way that the new derivatives ( [ derx ] ) , ( [ dery ] ) satisfy leibnitz rule . by ( [ derx ] ) we obtain @xmath131 _ * * \frac{\partial f_2}{\partial x^{\nu } } - \frac{\partial f_1}{\partial x^{\nu } } * \left [ { a_{\mu}}^{\nu } , f_2\right ] _ * \nonumber \\ & & + \left [ { b_{\mu}}^{\nu } , f_1\right ] _ * * \frac{\partial f_2}{\partial y^{\nu } } - \frac{\partial f_1}{\partial y^{\nu } } * \left [ { b_{\mu}}^{\nu } , f_2\right ] _ * + \delta \label{leibnitz}\end{aligned}\ ] ] here @xmath132_*$ ] is the commutator with respect to the product ( [ oldstar ] ) . @xmath133 can be computed from the defintions ( [ newstar ] ) , ( [ vleibnitz ] ) . by requiring ( [ leibnitz ] ) to vanish we obtain the equations . @xmath134 @xmath135 @xmath136 general solution to these equations is given by @xmath137 @xmath138 here @xmath139 , @xmath140 and @xmath141 are some constants to be discussed later . similarly @xmath123 , @xmath124 can be determined . by using the @xmath112 symmetry we find @xmath142 the field strength of the gauge fields will be defined as the commutator of the gauge covariant derivatives . for this procedure to be consistent the derivatives ( [ derx ] ) , ( [ dery ] ) must commute . otherwise the commutator becomes a differential operator . for instance from ( [ derx ] ) we obtain @xmath143 f = 2\left ( \frac{\partial { a_{\nu}}^{\lambda}}{\partial x^{\mu}}- \frac{\partial { a_{\mu}}^{\lambda}}{\partial x^{\nu } } \right ) \ \frac{\partial f}{\partial x^{\lambda } } + 2\left ( \frac{\partial { b_{\nu}}^{\lambda}}{\partial x^{\mu}}-\frac{\partial { b_{\mu}}^{\lambda } } { \partial x^{\nu } } \right ) \ \frac{\partial f}{\partial y^{\lambda}}\ ] ] and the two coefficients of the derivatives of @xmath144 must vanish . we obtain 4 more conditions from the other commutators . it turns out the following three out of the six conditions are independent . @xmath145 these equations impose the following conditions on @xmath146 , @xmath147 , @xmath148 . @xmath149 @xmath150 @xmath151 if we recall the symmetry @xmath152 , ( [ 2 ] ) comes out as a result of ( [ 5 ] ) and hence is not an independent condition . the solution to ( [ 1 ] ) is given by @xmath153 @xmath154 is a constant . @xmath155 and @xmath156 are arbitrary as long as ( [ 5 ] ) is satisfied . unfortunately , further conditions which may determine these values are not yet available to the authors . the fact that the derivatives ( [ derx ] ) , ( [ dery ] ) commute suggests the existence of a coordinate transformation @xmath157 by which the derivatives ( [ derx ] ) , ( [ dery ] ) are reexpressed as ordinary derivatives . @xmath158 ( and a similar equation for @xmath159 . ) here we used the property , @xmath160 which holds for a linear function @xmath69 . in this transformation @xmath14 , @xmath22 are treated as commuting variables . indeed eq ( [ integrability ] ) guarantees the existence of such a transformation and it is given by @xmath161 and a similar equation with the replacement @xmath162 . by using ( [ cr1])-([cr3 ] ) it is possible to show that in the primed coordinate system @xmath27 and @xmath28 commute with each other . @xmath163=i \theta^{\mu \nu } , \qquad [ y^{\prime \mu } , y^{\prime \nu } ] = -i \theta^{\mu \nu } , \qquad [ x^{\prime \mu } , y^{\prime \nu } ] = 0\ ] ] here the quadratic terms in ( [ transfx ] ) are assumed to be symmetrized . in other words the effect of the field strength @xmath11 can be compensated by a ` coordinate transformation ' . this fact suggests that the metric tensor in the ` @xmath30 space ' may be also flat and given by @xmath164 . we should note that this is not the ordinary space but a dimensionally - doubled space combining the coordinates of the two ends of the string and that the transformation ( [ transfx ] ) mixes up @xmath14 and @xmath22 as a whole . it is tempting to assume that the existence of such a transformation persists to all orders of @xmath11 . we do not have a proof at present . let us construct a trace operation @xmath165 . this must satisfy the cyclic property . @xmath166 we assume the form @xmath167 and determine the function @xmath168 by the condition @xmath169 for @xmath5 we must also have @xmath170 . note that on the righthand side of ( [ trace ] ) the product is the ordinary one . @xmath89 is @xmath171 . it is not hard to show that this function is given by @xmath172 it may be natural to expect that this extra factor @xmath173 in the trace has an origin in the jacobian from @xmath174 to @xmath30 . we indeed obtain @xmath175 and we can show from ( [ 1])-([5 ] ) that this coincides with @xmath173 . we are now ready to define gauge transformations and write down the gauge theory action . this action will be an integral over @xmath17 and @xmath18 and defines a bi - local theory . we will construct this bi - local theory as an intermediate step for obtaining a ` local gauge theory ' in the next section . let us define the gauge covariant derivatives by @xmath176 here @xmath177 , @xmath178 are the gauge fields in the @xmath30 frame and transformed to those in the @xmath23 frame like the derivatives ( [ derivative ] ) , @xmath179 we will use a ` bar ' to denote functions associated with the coordinate @xmath22 . in ( [ covder ] ) , ( [ a ] ) the signs in front of @xmath177 and @xmath178 are opposite , because the two ends of the open string have opposite charges . the gauge transformation is defined by @xmath180 where @xmath181 is a gauge function . @xmath182 is an exponential defined in terms of @xmath2 . @xmath183 is a scalar field in the adjoint representation . for an infinitesimal @xmath184 these read @xmath185_{\star } - \frac{\nabla \lambda } { \nabla x^{\mu } } , \qquad \delta \overline{a}_{\mu } ' = i [ \lambda , \overline{a}_{\mu}']_{\star } + \frac{\nabla \lambda}{\nabla y^{\mu } } , \qquad \delta \phi = i [ \lambda , \phi]_{\star}. \label{infgauge}\end{aligned}\ ] ] here @xmath132_{\star}$ ] is the commutator with respect to the product ( [ newstar ] ) . an infinitesimal transformation of @xmath186 , @xmath187 is consistently obtained . @xmath188_{\star } - \frac{\partial \lambda } { \partial x^{\mu } } + i \ { [ \lambda , \ { a_{\mu}}^{\lambda } ] _ { \star } , \ a_{\lambda } \}_{\star } - i \ { [ \lambda , \ { b_{\mu}}^{\lambda } ] _ { \star } , \ \overline{a}_{\lambda } \}_{\star } , \end{aligned}\ ] ] @xmath189_{\star } + \frac{\partial \lambda } { \partial y^{\mu } } + i \ { [ \lambda , \ \overline{a}_{\mu}{}^{\lambda } ] _ { \star } , \ \overline{a}_{\lambda } \}_{\star } - i \ { [ \lambda , \ \overline{b}_{\mu}{}^{\lambda } ] _ { \star } , \ a_{\lambda } \}_{\star}\end{aligned}\ ] ] now by using the standard prescription we can write down the action integral for the bi - local noncommutative gauge theory @xmath190_{\star } \ { \mbox{\boldmath $ \star$ } } \ [ d_{\nu } , \ d_{\rho } ] _ { \star } - 2 g^{\mu \nu } \ [ d_{\mu } , \ x^m ] _ { \star } { \mbox{\boldmath $ \star$ } } [ d_{\nu } , \ x^m ] _ { \star } \right . \nonumber \\ & & \qquad \qquad \qquad + \ [ x^m , \ x^n]_{\star}^2 \ + \ g^{\mu \nu } g^{\lambda \rho } \ [ \overline{d}_{\mu } , \ \overline{d}_{\lambda } ] _ { \star } \ { \mbox{\boldmath $ \star$ } } \ [ \overline{d}_{\nu } , \ \overline{d}_{\rho } ] _ { \star } \nonumber \\ & & \qquad \qquad \qquad \left . - 2 \ g^{\mu \nu } \ [ \overline{d}_{\mu } , \ \overline{x}^m ] _ { \star } { \mbox{\boldmath $ \star$ } } [ \overline{d}_{\nu } , \ \overline{x}^m ] _ { \star } + \ [ \overline{x}^m , \ \overline{x}^n]_{\star}^2 \ \right\ } \nonumber \\ & \equiv & \int d^dx d^dy \ g v^{-1 } \ \omega \ { \cal l}(x , y ) , \label{gauge}\end{aligned}\ ] ] where @xmath191 is the volume @xmath192 and @xmath193 the yang - mills coupling . we explicitly write only the integration measure relevant to our discussions . @xmath194 and @xmath195 are the scalar fields describing the fluctuations of the d - brane . the scalar fields are doubled , because there are two such independent fluctuations at both ends of the open string . this action is invariant under the gauge transformation ( [ gaugetr ] ) due to the cyclic property ( [ cyclicproperty ] ) of the trace ( [ trace ] ) . because @xmath196 is constant , this does not break the gauge symmetry . there is no constraint on the gauge function @xmath197 . by substituting ( [ derx ] ) , ( [ dery ] ) , ( [ covder ] ) , ( [ a ] ) into ( [ gauge ] ) we will find a non - constant metric in the action . for example we obtain for the commutator of the covariant derivatives @xmath198_{\star } & = & i \ \ { { \cal f}_{\mu \nu}^{(x , x ) } + 2{a_{\mu}}^{\lambda } \ { \cal f}_{\lambda \nu}^{(x , x ) } + 2{a_{\nu}}^{\lambda } \ { \cal f}_{\mu \lambda}^{(x , x ) } + 2{b_{\mu}}^{\lambda } \ { \cal f}_{\lambda \nu}^{(y , x ) } \nonumber \\ & & \qquad + 2{b_{\nu}}^{\lambda } \ { \cal f}_{\mu \lambda}^{(x , y ) } + { \cal r}_{\mu \nu } \},\end{aligned}\ ] ] where @xmath199 s are the field strength in terms of the ordinary derivatives . @xmath200_{\star } , \nonumber \\ { \cal f}_{\mu \nu}^{(y , x ) } & = & -\frac{\partial}{\partial y^{\mu}}a_{\nu } -\frac{\partial}{\partial x^{\nu}}\overline{a}_{\mu } + i \ [ \ \overline{a}_{\mu } , \ a_{\nu } \ ] _ { \star } = { \cal f}_{\nu \mu}^{(x , y ) } , \nonumber \\ { \cal f}_{\mu \nu}^{(y , y ) } & = & \frac{\partial}{\partial y^{\mu } } \overline{a}_{\nu}-\frac{\partial}{\partial y^{\nu}}\overline{a}_{\mu } - i \ [ \ \overline{a}_{\mu } , \ \overline{a}_{\nu } \ ] _ { \star}\end{aligned}\ ] ] @xmath201 is a sum of terms like @xmath202 here we omit its explicit form . by defining a new curved metric @xmath203 we obtain the part of the action for the gauge field . @xmath204 here @xmath205 is the determinant of the 2@xmath6 @xmath206 2@xmath6 matrix ( @xmath207 , @xmath208 / @xmath209 , @xmath210 ) and coincides with @xmath211 . @xmath212 is a sum of terms similar to those of @xmath201 . in the above ` @xmath213 ' is an ordinary product . in this way the noncommutative gauge theory in the curved space is obtained . although the metric @xmath214 depends on the coordinates , the action is gauge invariant . similarly , an action for the scalars @xmath215 , @xmath216 can be obtained . in the special case @xmath217 , @xmath14 and @xmath22 commute and the product of functions of @xmath14 only is still a function of @xmath14 only . thus it is natural to set @xmath218 , @xmath219 , @xmath220 , @xmath221 . the product @xmath2 reduces to @xmath8 . then the action @xmath222 becomes the sum of two decoupled noncommutative gauge theories with noncommutative parameters @xmath88 and @xmath223 , respectively . in @xmath5 noncommutative gauge theory , one obtains unitary and causal s - matrix as long as the noncommutative parameter has only space components.@xcite if we regard ( [ gauge ] ) as the action integral in 2@xmath6-dimensional space , we may be able to expect that this theory is also causal , even if there are terms containing the field strength @xmath11 . if we regard @xmath14 and @xmath22 as two distinct points in the same @xmath6-dimensional space , however , this theory will be non - local and we can not expect to obtain causal s - matrix . we must reduce @xmath222 to a single @xmath14-integral and make the theory ` local ' . we need to find a prescription to connect this action to the amplitudes of the string theory in the low energy limit . let us recall that the energy of the open string is proportional to the length of the string divided by @xmath225 and given by @xmath226 . in the low energy limit this goes to a finite value . a massive string will be unstable and eventually decay into a massless string . to obtain a massless string we must take the @xmath227 limit by hand . if one sends @xmath22 to @xmath14 directly in the product @xmath2 , ( [ newstar ] ) , this product becomes non - associative . in this sense the separation @xmath228 works as a kind of regularization to realize the associative product @xmath2 . at the final step we need to reduce the @xmath14 , @xmath22 integral to a single @xmath14 integral . the integration measure is not just @xmath229 but must be modified . as a simple example let us first consider the following integral in the @xmath23 space . @xmath230 @xmath231 s are scalar fields . we propose that the following reduced integral in the @xmath14 space should correspond to this integral in the @xmath31 limit . this prescription is similar to the one ( @xmath232 ) adopted in @xcite due to the relation ( [ pyx ] ) . ] @xmath233|_{y = x } \label{reduced}\ ] ] in ( [ reduced ] ) we set @xmath32 only after all the algebra associated with @xmath2 is finished . here @xmath33 is a function which is determined by the requirement of cyclic property . actually if we set @xmath234 the two point function of the functions of @xmath14 only , @xmath235 \equiv \int d^dx \ \omega(x ) \left . ( \phi_1(x ) { \mbox{\boldmath $ \star$ } } \phi_2(x ) ) \right|_{y = x}\ ] ] is given by @xmath236 = \int d^dx \phi_1 ( x ) \cdot \phi_2(x ) \cdot ( 1 + \frac{1}{3 } x^{\lambda } \theta^{\mu \nu } \ \nu \lambda } ) . \label{i2}\ ] ] this is cyclic invariant . @xmath236 = i_2[\phi_2,\phi_1 ] \label{cyc2}\ ] ] contrary to our expectation the function @xmath33 is not proportional to @xmath237 but to @xmath238 , where @xmath239 is the open string metric in the presence of @xmath11 . furthermore we can check that this correlation function enjoys the following interesting property . @xmath240 this property reveals the string theory origin of the @xmath2 product . to see this it is helpful to visualize the @xmath2 product as an operation adding one rung to a ladder . let us consider a ladder and name the two legs as @xmath14 and @xmath22 , respectively . the @xmath241 product adds one unit on top of the ladder . @xmath242 adds one rung with the @xmath14-leg marked , while @xmath243 one with the @xmath22-leg marked . this operation must be performed step by step and is noncommutative . finally setting @xmath32 and integrating over @xmath14 join the two legs at the top and bottom , separately . after this joining the legs are merged into a circle and we expect to have the cyclic property as in the string amplitude . this is the case here . eq ( [ cycl2 ] ) is illustrated in fig.1 . 1.3 mm ccccccc ( 3,0)(0,0 ) ( 1,-1)(0,0)@xmath244 ( 1,3)(0,0)@xmath245 ( 0,0)(0,0 ) ( 1,-5)(0,0)@xmath14 ( 10,-5)(0,0)@xmath22 = & = & ( 3,0)(0,0 ) ( 1.5,-1)(0,0)@xmath244 ( 0,0)(0,0 ) ( 1,-5)(0,0)@xmath14 ( 10,-5)(0,0)@xmath22 = ( 3,0)(0,0 ) ( 2.5,3)(0,0)@xmath246 & = & ( 3,0)(0,0 ) ( 1.5,3)(0,0)@xmath244 ( 0,0)(0,0 ) ( 1,-5)(0,0)@xmath14 ( 10,-5)(0,0)@xmath22 = ( 3,0)(0,0 ) ( 2.5,-1)(0,0)@xmath246 & = & ( 0,0)(0,0 ) ( 1,-5)(0,0)@xmath14 ( 10,-5)(0,0)@xmath22 = ( 3,0)(0,0 ) ( 2.5,3)(0,0)@xmath247 ( 2.5,-1)(0,0)@xmath246 + @xmath248 & & @xmath249 & & @xmath250 & & @xmath251 + + similarly the three point function @xmath252 \equiv \int d^dx \ \omega(x ) \left . ( \phi_1(x ) { \mbox{\boldmath $ \star$ } } \phi_2(x ) { \mbox{\boldmath $ \star$ } } \phi_3(x ) ) \right|_{y = x}\ ] ] is given by @xmath253 & = & \int d^dx \ { \ \omega(x ) \cdot \phi_1(x ) * \phi_2(x ) * \phi_3(x ) -\frac{i}{6 } ( g^{\mu \lambda}g^{\nu \rho}-\theta^{\mu \lambda } \theta^{\nu \rho } ) h_{\lambda \rho \sigma } \ x^{\sigma } \nonumber \\ & & \qquad \times ( \partial_{\mu}\phi_1 * \partial_{\nu}\phi_2 * \phi_3 + \partial_{\mu}\phi_1 * \phi_2 * \partial_{\nu}\phi_3 + \phi_1 * \partial_{\mu}\phi_2 * \partial_{\nu}\phi_3 ) \nonumber \\ & & \qquad + \frac{1}{18 } ( g^{\mu \lambda}\theta^{\nu \alpha}- \theta^{\mu \lambda}g^{\nu \alpha } ) \ g^{\sigma \rho } \ \rho \alpha } \ \partial_{\mu}\phi_1 * \partial_{\nu}\phi_2 * \partial_{\sigma}\phi_3 \ \}. \label{cycl3}\end{aligned}\ ] ] in the @xmath12 limit those terms in ( [ i2 ] ) and ( [ cycl3 ] ) which contain the metric @xmath196 drop and agree with the results of correlation functions obtained in @xcite . it is not difficult to show that the three point function ( [ cycl3 ] ) satisfies @xmath252 = i_3[\phi_3,\phi_1,\phi_2]= i_3[\phi_2,\phi_3 , \phi_1 ] . \label{cyc3}\ ] ] we can also show the following identities . @xmath254 the locations of the fields can be shifted from one end @xmath14 of the string to the other @xmath22 , and _ vice versa _ , as long as the ordering on a circle is unchanged . this is natural from the string theory point of view . higher point functions @xmath255 $ ] do not have the cyclic property . in @xcite it was shown that in order to obtain the four point function in terms of the non - associative product @xmath13 one must take a specific linear combination of the ordered products of the functions with various positionings of the parenthesis . in our approach the correlation functions will be obtained by functional derivatives of the effective action and various orderings will be automatically taken into account due to the bose symmetry . the cyclic property ( [ cyc2 ] ) , ( [ cycl2 ] ) , ( [ cyc3 ] ) , ( [ cy3 ] ) , and the agreement of ( [ i2 ] ) , ( [ cycl3 ] ) with the results of @xcite in the @xmath12 limit justifies the prescription ( [ reduced ] ) . let us now turn to the gauge theory action ( [ gauge ] ) . the reduced effective action which corresponds to this is given by @xmath256 \ = \ \int d^dx \sqrt{g } \ \omega(x)\ [ \ { \cal l}(x , y ) \ ] |_{y = x } \nonumber \\ \qquad & & = -\frac{1}{4g^2_{\rm{ym}}}\int d^dx \ \sqrt{g } \ \omega(x ) \ \cdot \ { \ g^{\mu \nu } g^{\lambda \rho } \ [ d_{\mu } , \ d_{\lambda } ] _ { \star } \ { \mbox{\boldmath $ \star$ } } \ [ d_{\nu } , \ d_{\rho } ] _ { \star } \nonumber \\ & & \qquad \qquad \qquad - 2 g^{\mu \nu } \ [ d_{\mu } , \ x^m ] _ { \star } { \mbox{\boldmath $ \star$ } } [ d_{\nu } , \ x^m ] _ { \star } + [ x^m , \ x^n]_{\star}^2 \nonumber \\ & & \qquad \qquad \qquad + \ g^{\mu \nu } g^{\lambda \rho } \ [ \overline{d}_{\mu } , \ \overline{d}_{\lambda } ] _ { \star } \ { \mbox{\boldmath $ \star$ } } \ [ \overline{d}_{\nu } , \ \overline{d}_{\rho } ] _ { \star } \nonumber \\ & & \qquad \qquad \qquad \left . - 2 g^{\mu \nu } \ [ \overline{d}_{\mu } , \ \overline{x}^m ] _ { \star } { \mbox{\boldmath $ \star$ } } [ \overline{d}_{\nu } , \ \overline{x}^m ] _ { \star } + [ \overline{x}^m , \ \overline{x}^n]_{\star}^2 \ } \right|_{y = x } , \label{red}\end{aligned}\ ] ] where @xmath257 is the lagrangian density defined in ( [ gauge ] ) . the restriction @xmath32 here breaks some part of the gauge symmetry . the reduced effective action ( [ red ] ) , however , turned out to be still invariant under the gauge transformation with the gauge function @xmath258 this follows from the identity @xmath259_{\star } \right\ } \right|_{y = x } = 0 , \label{identity}\ ] ] which can be proved by expanding @xmath260 as @xmath261 and using the identities ( [ cy3 ] ) . because @xmath262 contains derivatives , the integrand of ( [ identity ] ) does not trivially vanish , even if we set @xmath32 . let us note that in ( [ red ] ) the product inside the bracket is @xmath262 . hence the reduced action is not an ordinary action in the @xmath6-dimensional space , because the product in the integrand contains both derivatives @xmath263 , @xmath264 . the gauge transformation is also defined in the @xmath23 space . to write down the gauge invariant action in the ordinary sense we must introduce a set of coordinates @xmath23 and consider 2@xmath6 dimensional action integral , as we did in the previous section . finally , in order to derive the correlation functions we must further set @xmath265 , @xmath266 , @xmath267 and @xmath268 before taking the functional derivatives of ( [ red ] ) , because these describe the gauge particles emitted and/or absorbed from the @xmath19 and @xmath20 ends of the open string and the fluctuations at the two ends , respectively . the same functions at both ends because of the bose symmetry . then this reduced action becomes a functional of @xmath269 , @xmath270 . we must mention some ambiguity in our prescription . when we perform a partial integration or use the cyclic property ( [ cyclicproperty ] ) in ( [ gauge ] ) , we will obtain different integrands which will , however , lead to the same result . the corresponding reduced integral does not have this property . this is unavoidable , because when we reduce the integral , we insert into the integrand a function proportional to a delta function . the result will depend on where this delta function is inserted and we must specify this . our proporsal is to reduce the integral in the standard form of the action ( [ gauge ] ) . because we have not determined @xmath271 , @xmath155 , @xmath156 completely , we can not decide whether ( [ red ] ) reproduces the open string amplitudes in the low energy limit . moreover such an analysis will meet difficulties , because one can not generally define free states in an asymptotically non - flat space . we , however , expect that once @xmath271 , @xmath155 , @xmath156 are determined , the gauge invariance is strong enough to restrict the form of the effective action . in this paper we considered an open string theory in an ns - ns @xmath0 field background with a nonzero constant field strength @xmath1 . the background space is curved although if @xmath11 is small , the metric @xmath9 is constant up to @xmath24 . we performed a perturbative analysis of the commutation relations of the string coordinates without oscillators up to @xmath24 and found that the coordinate @xmath14 of one end of the string does not commute with the coordinate @xmath22 of the other . we then constructed an associative and noncommutative product @xmath2 which realizes the commutation relations of the coordinates . this product is an operation for functions of both the coordinates @xmath14 , @xmath22 and even if two functions of @xmath14 only are multiplied , the result is a function of both @xmath14 and @xmath22 . in this way we are lead to consider bi - local fields which depends on the coordinates of both ends of the string . we then found that derivatives @xmath272 , @xmath273 do not satisfy leibnitz rule with the @xmath2 product and we are forced to modify the derivatives . at present the modified derivatives are not uniquely determined , but remarkably , the new derivatives @xmath274 , @xmath275 turned out to be rewritten as @xmath276 , @xmath277 , respectively , by some ` coordinate transformation ' @xmath278 . the commutation relations of the primed variables coincide with those of the unprimed ones with @xmath5 . hence if we combine the @xmath14 and @xmath22 coordinates and consider 2@xmath6 dimensional space @xmath23 , the primed coordinate system appears to be flat . this provides us with a clue to write down the gauge theory action . the open string metric @xmath279 is curved and it may seem difficult to write down the action without spoiling the gauge invariance due to the coordinate dependent metric . our proporsal is to write down the action in the flat @xmath30 frame and then transform it to the @xmath23 frame . in this way we obtained an action integral for the noncommutative gauge theory in curved backgrounds this gauge theory lives in a double - dimensional space , _ i.e. _ , the space with coordinates @xmath280 . this action is invariant under gauge transformations with @xmath14 , @xmath22 dependent gauge parameters . we then took a limit @xmath31 and proposed a prescription to reduce the action integral to that in @xmath6-dimensional space with a coordinate @xmath14 . this reduced action still has a part of the gauge symmetry . this will provide a ` local ' low energy effective action for string amplitudes . let us now turn to the application of our results to the matrix model . the formulation of the matrix model for curved backgrounds has been investigated @xcite but is not yet well established . as mentioned above , the backgound space is curved for nonvanishing @xmath11 . the closed string metric @xmath9 is flat only up to @xmath24 and the open string metric @xmath281 is curved already at @xmath24@xcite . on the other hand in the action ( [ gauge ] ) the metric @xmath164 is flat and has the ordinary form of the gauge theory action in the flat space . this is based on the fact that we can deform the algebra of the coordinates @xmath14 , @xmath22 to that of the flat background by some coordinate transformation @xmath282 . although we do not have a proof that this persists to all orders in @xmath11 , let us assume this here . then this observation suggests the following general formulation of the matrix model for curved backgrounds . * double the number of the matrices of the ordinary matrix model . denote these as @xmath41 and @xmath283 . this is natural because the open string has two ends . in this paper we considered only the bosonic string . in the case of the superstring the fermion fields on the d - brane must be also doubled . * the action of the matrix model is given by @xmath284 ^ 2 \ + [ \overline{x}^{m } , \ \overline{x}^{n}]^2 \nonumber \\ & & \qquad \qquad \qquad \qquad + \ \mbox{fermions } \ \}. \label{matrix}\end{aligned}\ ] ] * the action ( [ matrix ] ) is background independent . the space - time metric is minkowskian . we can introduce backgrounds as follows . by a separate linear coordinate transformation of @xmath41 , @xmath283 the constant metric @xmath164 can be put in . the matrix multiplication is associative and noncommutative and there should be an isomorphism between matrices @xmath41 , @xmath283 and the bi - local fields whose multiplication rule obeys the product @xmath2 . by the replacement @xmath285 , @xmath286 @xmath287 and @xmath288 , we will obtain the action ( [ gauge ] ) . the product @xmath2 and the gauge covariant derivatives @xmath289 , @xmath290 will introduce the background dependence . for the above formulation to be valid we must obtain the classical solution for the matrix model action ( [ matrix ] ) and then find a way to systematically determine the star product @xmath2 and the covariant derivatives @xmath6 , @xmath291 and demonstrate that the gauge theory action ( [ gauge ] ) can be obtained . we hope to report on the analysis in the future . 99 a. conne , m. r. douglas and a. schwartz , _ noncommutative geometry and matrix theory : compactification on tori _ , jhep * 9802 * ( 1998 ) 003 , hep - th/9711162 . m. r. douglas and c. hull , _ d - branes and noncommutative torus _ , jhep * 9802 * ( 1998 ) 008 , hep - th/9711165 . e. witten , _ noncommutative geometry and string field theory _ , nucl . phys . * b268 * ( 1986 ) 253 . v. schomerus , _ d - branes and deformation quantization _ , jhep * 9906 * , ( 1999 ) 030 , hep - th/9903205 . c .- s . chu and p .- _ noncommutative open string and d - brane _ , nucl . phys . * b550 * ( 1999 ) 151 , hep - th/9812219 ; + c .- s . chu and p .- m . ho , _ constrained quantization of open string in background @xmath0 field and noncommutative d - brane _ , nucl . b568 * ( 2000 ) 447 , hep - th/9906192 . f. ardalan , h. arfai and m. m. sheikh - jabbari , _ dirac quantization of open strings and noncommutativity in branes _ , b576 * ( 2000 ) 578 , hep - th/9906161 . n. seiberg and e. witten , _ string theory and noncommutative geometry _ , jhep * 9909 * ( 1999 ) 032 , hep - th/9908142 . a.y . alekseev , a. recknagel and v. schomerus , _ non - commutative world - volume geometries : branes on su(2 ) and fuzzy spheres _ , jhep * 9909 * ( 1999 ) 023 , hep - th/9908040 ; + a.y . alekseev , a. recknagel and v. schomerus , _ brane dynamics in background fluxes and non - commutative geometry _ , jhep * 0005 * ( 2999 ) 010 , hep - th/0003187 . l. cornalba and r. schiappa , _ nonassociative star product deformations for d - brane worldvolumes in curved backgrounds _ , hep - th/0101219 . m . ho and y .- t . yeh , _ noncommutative d - brane in non - constant ns - ns b field background _ lett . * 85 * ( 2000 ) 5523 , hep - th/0005159 . ho , _ making non - associative algebra associative _ , hep - th/0103024 . k. hayasaka , r. nakayama and y. shimono , in preparation . j. gomis and t. mehen , _ space - time noncommutative field theories and unitarity _ , nucl . phys . * b591 * ( 2000 ) 265 , hep - th/0005129 ; o. aharony , j. gomis and t. mehen , _ on theories with light - like noncommutativity _ , jhep * 0009 * ( 2000 ) 023 , hep - th/0006236 ; + l. alvarez - gaum and j. l. f.barbn , _ non - linear vacuum phenomena in non - commutative qed _ , a16 * ( 2001 ) 1123 , hep - th/0006209 n. seiberg , l. susskind and n. toumbas , _ space / time non - commutativity and causality _ , jhep * 0006 * ( 2000 ) 044 , hep - th/0005015 ; _ strings in background electric field , space / time noncommutativity and a new noncritical string theory _ , jhep * 0006 * ( 2000 ) 021 , hep - th/0005040 . see for example , + m. douglas , _ d - branes in curved space _ , adv . theor . math . * 1 * ( 1998 ) 198 , hep - th/9703056 . c. klimck and t. strobl , _ wzw - poisson manifolds _ , math.sg/0104189 ; p. severa and a. weinstein , _ poisson geometry with a 3-form background _ , math.sg/0107133 ; j. s. park , _ topological open p - branes _ , hep - th/0012141 . |
the importance of the-gaussian family of distributions lies in its power - law nature , and its close association with gaussian , cauchy and uniform distributions . this class of distributions arises from maximization of a generalized information measure . we use the power - law nature of the-gaussian distribution to improve upon the smoothing properties of gaussian and cauchy distributions . based on this , we propose a smoothed functional ( sf ) scheme for gradient estimation using-gaussian distribution . our work extends the class of distributions that can be used in sf algorithms by including the-gaussian distributions , which encompass the above three distributions as special cases . using the derived gradient estimates , we propose two - timescale algorithms for optimization of a stochastic objective function with gradient descent method . we prove that the proposed algorithms converge to a local optimum . performance of the algorithms is shown by simulation results on a queuing model . | 1206.4832 | @xcite provided the concept of entropy as a measure of information , or more precisely , a measure of uncertainty given by probability distributions . @xcite was the first to introduce the concept of generalized measures of information when he proposed the first well - known generalization of shannon entropy , known as @xmath1-entropy or rnyi entropy , based on kolmogorov - nagumo averages . several other generalizations have also been studied in the literature @xcite , and have been extensively used in physics , communication theory and other disciplines . one of the most recently studied generalized information measure is the nonextensive entropy , due to @xcite , defined as @xmath2 where @xmath3 is the probability mass function of the discrete random variable on the set @xmath4 , and the @xmath0-logarithm is defined as @xmath5 , @xmath6 , @xmath7 . it is called nonextensive because of its pseudo - additive nature @xcite . although this generalization had been introduced earlier in @xcite , tsallis provided interpretations in the context of statistical mechanics . @xcite generalized the shannon - khinchin axioms to the nonextensive case . the most important characteristic of these generalized information measures is that , on maximization , they give rise to power - law distributions , while maximization of shannon entropy gives rise to exponential distributions . the concept of maximization of information measure can be attributed to kullback s minimum discrimination theorem @xcite , which establishes important connections between statistics and information theory . this theorem shows that exponential distributions can be obtained by minimizing kullback - leibler divergence under given moment constraints . one can consider maximum entropy @xcite as a special case , where maximization of shannon entropy under moment constraints leads to exponential distributions . for example , given mean and variance of a random variable , maximum entropy gives rise to gaussian distribution . while exponential distributions have been extensively studied and used in statistical modeling , the power - law behavior has been observed in most of the real - world data , e.g. , @xcite . the importance of tsallis entropy can be attributed to its connections with power - law distributions , which has been studied in different contexts like finance , earthquakes and network traffic @xcite . compared to the exponential family , the tsallis distributions , _ i.e. _ , the family of distributions resulting from maximization of tsallis entropy , have an additional shape parameter @xmath0 , similar to that in that controls the nature of the power - law tails . one of the most studied tsallis entropy maximizer is the @xmath0-gaussian distribution @xcite , which is a power - law generalization of gaussian distribution . in this paper , we study the @xmath0-gaussian distribution in the context of smoothed functional algorithms for stochastic optimization . stochastic techniques play a key role in optimization problems , where the objective function does not have an analytic expression . such problems are often encountered in discrete event systems , which are quite common in engineering and financial world . most often , the data , obtained via statistical survey or simulation , contains only noisy estimates of the objective function to be optimized . one of the most commonly used solution methodologies involves stochastic approximation algorithms , originally due to @xcite , which is used to find the zeros of a given function . based on this approach , gradient descent algorithms have been developed , in which the parameters controlling the system track the zeros of the gradient of the objective . however , these algorithms require an estimate of the cost gradient . @xcite provide such a gradient estimate using several parallel simulations of the system . more efficient techniques for gradient estimation , have been developed based on the smoothed functional approach @xcite , simultaneous perturbation stochastic approximation @xcite etc . a stochastic variation of newton - based optimization methods , also known as adaptive newton - based schemes has also been studied in literature @xcite . when the above schemes for gradient estimation are employed in optimization methods involving long - run average cost objective , the time complexity of the algorithms increase as the long - run average cost needs to be estimated after each parameter update . a more efficient approach is to simultaneously perform the long - run averaging and parameter updates using different step - size schedules . these classes of algorithms constitute the multi - timescale stochastic approximation algorithms @xcite . two - timescale optimization algorithms have been developed using simultaneous perturbations @xcite and smoothed functional @xcite schemes . the main issue with such algorithms is that , although convergence of the algorithm to a local optimum is guaranteed , the global optimum is often not achieved in practice . @xcite proves that the gradient sf schemes based on gaussian perturbations converge to a local minimum , and provides comparison of the performance of various multi - timescale algorithms for stochastic optimization on a queuing system . the results presented there indicate that the performance of the sf algorithms depends considerably on several tuning parameters , such as the variance of the gaussian distribution , and also the step - sizes . the smoothed functional schemes for simulation based optimization have become popular due to their smoothing effects on local fluctuations . we derive for the first time , smoothed functional algorithms with power - law ( @xmath0-gaussian ) perturbations . prior work @xcite indicated that the class of distributions that can be used for the perturbation random variables in sf algorithms includes gaussian , cauchy and uniform distributions . our main contribution is to show that the @xmath0-gaussian family of distributions belong to this class , and encompasses the above three distributions as special cases . this allows us to work with a larger class of distributions in sf algorithms , in an unified way , where the `` shape parameter '' of the @xmath0-gaussian controls its power - law behavior . this parameter also controls the smoothness of the convolution , thereby providing additional tuning . we show that the multivariate @xmath0-gaussian distribution satisfies all the conditions for smoothing kernels discussed in @xcite . we then present estimators for gradient of a function using the @xmath0-gaussian smoothing kernel . we also present multi - timescale algorithms for stochastic optimization using @xmath0-gaussian based sf that incorporate gradient based search procedures , and prove the convergence of the proposed algorithms to the neighborhood of a local optimum . the convergence analysis presented in this paper differs from the approaches that have been studied earlier @xcite . here , we provide a more straightforward technique using standard results from @xcite . further , we perform simulations on a queuing network to illustrate the benefits of the @xmath0-gaussian based sf algorithms compared to their gaussian counterparts . a shorter version of this paper containing only the one - simulation @xmath0-gaussian sf algorithm , and without the convergence proof , has been presented in ieee 2012 international symposium on information theory @xcite . the rest of the paper is organized as follows . the framework for the optimization problem and some preliminaries on sf and @xmath0-gaussians are presented in section [ preliminaries ] . section [ qgrad ] validates the use of @xmath0-gaussian as smoothing kernel , and presents gradient descent algorithms using @xmath0-gaussian sf . the convergence analysis of the proposed algorithms is discussed in section [ gqsf_convergence ] . section [ sim_results ] presents simulations based on a numerical setting . finally , section [ conclusion ] provides the concluding remarks . in the appendix , we discuss a sampling technique for multivariate @xmath0-gaussians that is used in the proposed algorithms . let @xmath8 be a parameterized markov process , depending on a tunable parameter @xmath9 , where @xmath10 is a compact and convex subset of @xmath11 . let @xmath12 denote the transition kernel of @xmath13 when the operative parameter is @xmath9 . let @xmath14 be a lipschitz continuous cost function associated with the process . [ ergodic ] the process @xmath13 is ergodic for any given @xmath15 as the operative parameter , _ i.e. _ , as @xmath16 , @xmath17,\ ] ] where @xmath18 is the stationary distribution of @xmath13 . our objective is to minimize the long - run average cost @xmath19 by choosing an appropriate @xmath9 . the existence of the above limit is assured by assumption [ ergodic ] and the fact that @xmath20 is continuous , hence measurable . in addition , we assume that the average cost @xmath21 satisfies the following requirement . [ differentiable ] the function @xmath22 is continuously differentiable for all @xmath9 . a random sequence of parameter vectors , @xmath23 , controlling a process @xmath24 , is said to be non - anticipative if the conditional probability @xmath25 almost surely for @xmath26 and all borel sets @xmath27 , where @xmath28 , @xmath26 are associated @xmath29-fields . one can verify that under a non - anticipative parameter sequence @xmath30 , the sequence @xmath31 is markov . we assume the existence of a stochastic lyapunov function . [ lyapunov ] let @xmath30 be a non - anticipative sequence of random parameters controlling the process @xmath13 , and @xmath32 @xmath33 , @xmath26 be a sequence of associated @xmath29-fields . there exists @xmath34 , a compact set @xmath35 , and a continuous function @xmath36 , with @xmath37 , such that a. @xmath38 < \infty$ ] , and b. @xmath39 \leqslant v(y_n ) - \epsilon_0 $ ] , whenever @xmath40 , @xmath26 . while assumption [ differentiable ] is a technical requirement , assumption [ lyapunov ] ensures that the process under a tunable parameter remains stable . assumption [ lyapunov ] will not be required , for instance , if , in addition , the single - stage cost function @xmath20 is bounded . it can be seen that the sequence of parameters obtained using any of our algorithms below form a non - anticipative sequence . here , we present the idea behind the smoothed functional approach proposed by @xcite . we consider a real - valued function @xmath41 , defined over a compact set @xmath10 . its smoothed functional is defined as @xmath42 = \int\limits_{-\infty}^{\infty}g_{\beta}(\eta)f(\theta-\eta)\:\mathrm{d}\eta = \int\limits_{-\infty}^{\infty}g_{\beta}(\theta-\eta)f(\eta)\:\mathrm{d}\eta,\ ] ] where @xmath43 is a kernel function , with a parameter @xmath44 taking values from @xmath45 . the idea behind using smoothed functionals is that if @xmath46 is not well - behaved , _ i.e. _ , it has a fluctuating character , then @xmath47 $ ] is `` better - behaved '' . this can ensure that any optimization algorithm with objective function @xmath46 does not get stuck at a local minimum , but converges to a global minimum . the parameter @xmath44 controls the degree of smoothness . @xcite established that the sf algorithm achieves these properties if the kernel function satisfies the following sufficient conditions : 1 . @xmath48 , where @xmath49 corresponds to @xmath50 with @xmath51 , _ i.e. _ , @xmath52 2 . @xmath53 is piecewise differentiable in @xmath54 , 3 . @xmath53 is a probability distribution function , _ i.e. _ , @xmath47 = \mathsf{e}_{g_{\beta}(\eta)}[f(\theta-\eta)]$ ] , 4 . @xmath55 , where @xmath56 is the dirac delta function , and 5 . @xmath57 = f(\theta)$ ] . a two - sided form of sf is defined as @xmath58 & = \frac{1}{2}\int\limits_{-\infty}^{\infty}g_{\beta}(\eta)\big(f(\theta-\eta)+f(\theta+\eta)\big)\:\mathrm{d}\eta \nonumber \\&= \frac{1}{2}\int\limits_{-\infty}^{\infty}g_{\beta}(\theta-\eta)f(\eta)\:\mathrm{d}\eta + \frac{1}{2}\int\limits_{-\infty}^{\infty}g_{\beta}(\eta-\theta)f(\eta)\:\mathrm{d}\eta\;. \label{sf2}\end{aligned}\ ] ] the gaussian distribution satisfies the above conditions , and has been used as a smoothing kernel @xcite . the sf approach provides a method for estimating the gradient of any function , which satisfies assumptions [ ergodic][lyapunov ] @xcite . @xcite uses the gaussian smoothing , and derives a gradient estimator from as @xmath59 for large @xmath60 , @xmath61 and small @xmath44 . the stochastic process @xmath62 is governed by parameter @xmath63 , where @xmath64 is obtained through an iterative scheme , and @xmath65 is a @xmath66-dimensional vector of i.i.d . standard gaussian random variable . similarly , a two - simulation gradient estimator has been suggested using , which is of the following form @xmath67 for large @xmath60 , @xmath61 and small @xmath44 , where @xmath62 and @xmath68 are two processes governed by parameters @xmath63 and @xmath69 , respectively , @xmath70 and @xmath71 being as before . a continuous form of the shannon entropy , also known as differential entropy , has been extensively studied in statistical mechanics , probability and statistics . it is defined as @xmath72 where @xmath73 is a p.d.f . defined on the sample space @xmath4 . following the lines of discrete form generalization given in @xcite , @xcite provides a measure theoretic formulation of continuous form tsallis entropy functional , defined as @xmath74 this function results when the natural logarithm in is replaced by the @xmath0-logarithm defined earlier . the differential shannon entropy can be retrieved from as @xmath75 . the @xmath0-gaussian distribution was developed to describe the process of lvy super - diffusion @xcite , but has been later studied in other fields , such as finance @xcite and statistics @xcite . its importance lies in its power - law nature , due to which the tails of the @xmath0-gaussian decay at a slower rate than the gaussian distribution , depending on @xmath0 . it results from maximizing tsallis entropy under certain ` deformed ' moment constraints , known as normalized @xmath0-expectation defined by @xmath76 this form of an expectation considers an escort distribution @xmath77 , and has been shown to be compatible with the foundations of nonextensive statistics @xcite . @xcite maximized tsallis entropy under the constraints , @xmath78 and @xmath79 , which are known as @xmath0-mean and @xmath0-variance , respectively . these are generalizations of standard first and second moments , and tend to the usual mean and variance , respectively , as @xmath80 . this results in the @xmath0-gaussian distribution is used to maintain consistency with smoothed functional notations , which is used in section [ qgrad ] . ] that has the form @xmath81 where , @xmath82 is called the tsallis cut - off condition @xcite , which ensures that the above expression is defined , and @xmath83 is the normalizing constant , which is given by @xmath84 with @xmath85 being the gamma function , which exists over the specified intervals . the function defined in is not integrable for @xmath86 , and hence , @xmath0-gaussian is a probability density function only when @xmath87 . a multivariate form of the @xmath0-gaussian distribution has been discussed in @xcite . considering the @xmath0-mean and @xmath0-covariance matrix to be @xmath88 and @xmath89 respectively , the @xmath66-variate @xmath0-gaussian distribution can be expressed as @xmath90 for all @xmath91 , where the normalizing constant @xmath92 as in the one - dimensional case , the distribution is only defined for @xmath93 . can be mapped to the multivariate students-@xmath94 distribution . the support set of the above distribution is given by @xmath95 the multivariate normal distribution can be obtained as a special case when @xmath80 , while for @xmath96 , we obtain the uniform distribution with an infinitesimally small support around its mean . it is interesting to note that for @xmath97 , these distributions have a one - to - one correspondence with students-@xmath94 distribution , and in particular , for @xmath98 , we retrieve the cauchy distribution . a similar distribution can also be obtained by maximizing rnyi entropy @xcite . in this paper , we study the multivariate @xmath0-gaussian distribution as a smoothing kernel , and develop smoothed functional algorithms based on it . the first step in applying @xmath0-gaussians for sf algorithms is to ensure that the distribution satisfies the rubinstein conditions ( properties ( p1)(p5 ) in section [ preliminaries_sf ] ) . the rest of the paper uses the multivariate form of @xmath0-gaussian , with the @xmath0-mean @xmath99 , and @xmath0-covariance matrix @xmath100 , _ i.e. _ , the components are uncorrelated . @xmath66-dimensional @xmath0-gaussian distribution , with @xmath0-covariance @xmath101 satisfies the kernel properties ( p1)(p5 ) for all @xmath102 , @xmath7 . 1 . from , it is evident that @xmath103 . 2 . for @xmath104 , @xmath105 for all @xmath106 . thus , @xmath107 for @xmath108 , holds when @xmath109 . on the other hand , when @xmath110 , we have @xmath111 and hence , @xmath112 . thus , @xmath113 is differentiable for @xmath97 , and piecewise differentiable for @xmath108 . 3 . @xmath113 is a distribution for @xmath102 and hence , the corresponding sf @xmath114 , parameterized by both @xmath0 and @xmath44 , can be written as @xmath115 = \mathsf{e}_{g_{q,\beta}(x)}[f(\theta - x)]$ ] . 4 . @xmath116 is a distribution satisfying @xmath117 . so , @xmath118 . 5 . this property trivially holds due to convergence in mean as @xmath119 = \int\limits_{-\infty}^{\infty}{\lim_{\beta\to0}g_{q,\beta}(x)f(\theta - x)dx } = \int\limits_{-\infty}^{\infty}{\delta(x)f(\theta - x)\:\mathrm{d}x } = f(\theta).\ ] ] hence the claim . from the above result , it follows that @xmath0-gaussian can be used as a kernel function , and hence , given a particular value @xmath120 and some @xmath121 , the one - sided and two - sided sfs of any function @xmath122 are respectively given by @xmath123 & = \int\limits_{\omega_q } g_{q,\beta}(\theta - x ) f(x ) \:\mathrm{d}x , \\s'_{q,\beta}[f(\theta ) ] & = \frac{1}{2}\int\limits_{\omega_q}g_{q,\beta}(\theta - x)f(x)\:\mathrm{d}x + \frac{1}{2}\int\limits_{\omega_q}g_{q,\beta}(x-\theta)f(x)\:\mathrm{d}x , \label{qg_sf2_defn}\end{aligned}\ ] ] where the nature of the sfs are controlled by both @xmath0 and @xmath44 . the objective is to estimate the gradient of the average cost @xmath124 using the sf approach , where existence of @xmath124 follows from assumption [ differentiable ] . the gradient of smoothed functional ( smoothed gradient ) is defined as @xcite @xmath125 = \int\limits_{\omega_q } \nabla_{\theta}g_{q,\beta}(\theta-\eta ) j(\eta ) \:\mathrm{d}\eta\;,\ ] ] where @xmath126 is the support set defined as in . as there is no functional relationship between @xmath15 and @xmath54 over @xmath126 , _ i.e. _ , @xmath127 for all @xmath128 , @xmath129 where @xmath130 . hence , substituting @xmath131 , and using the symmetry of @xmath132 and @xmath133 , we can write @xmath134 & = \left(\frac{2}{\beta(n+2-nq)}\right)\int\limits_{\omega_q } \frac{\eta'}{\rho{(\eta ' ) } } g_{q}(\eta ' ) j(\theta+\beta\eta ' ) \:\mathrm{d}\eta ' \nonumber \\&= \left(\frac{2}{\beta(n+2-nq)}\right)\mathsf{e}_{g_{q}(\eta ' ) } \left[\left . \frac{\eta'}{\rho{(\eta ' ) } } j(\theta+\beta\eta ' ) \right| \theta \right]\;. \label{grad_sf1_formula}\end{aligned}\ ] ] in the sequel ( proposition [ grad_sf1_convergence ] ) , we show that @xmath135-\nabla_{\theta}j(\theta)\right\vert \to 0 $ ] as @xmath136 . hence , for large @xmath60 and small @xmath44 , the form of gradient estimate suggested by is @xmath137 where @xmath138 are uncorrelated identically distributed standard @xmath0-gaussian distributed random vectors . considering that in two - timescale algorithms ( discussed later ) , the value of @xmath15 is updated concurrently with the gradient estimation procedure , we estimate @xmath139 at each stage . by ergodicity assumption ( assumption [ ergodic ] ) , we can write as @xmath140 for large @xmath61 , where the process @xmath62 has the same transition kernel as defined in assumption [ ergodic ] , except that it is governed by parameter @xmath141 . in a similar manner , based on , the gradient of the two - sided sf can be written as @xmath142 = \frac{1}{2}\int\limits_{\omega_q } \nabla_{\theta}g_{\beta}(\theta-\eta)j(\eta)\:\mathrm{d}\eta + \frac{1}{2}\int\limits_{\omega_q } \nabla_{\theta}g_{\beta}(\eta-\theta)j(\eta)\:\mathrm{d}\eta.\ ] ] the first integral can be obtained as in . the second integral is evaluated as @xmath143 where @xmath144 . thus , we obtain the gradient as a conditional expectation @xmath142 = \left(\frac{1}{\beta(n+2-nq)}\right)\mathsf{e}_{g_{q}(\eta ) } \left[\left.\frac{\eta}{\rho{(\eta ) } } \big(j(\theta+\beta\eta)-j(\theta-\beta\eta)\big)\right|\theta\right ] . \label{grad_sf2_formula}\ ] ] in sequel ( proposition [ grad_sf2_convergence ] ) we show that @xmath145-\nabla_{\theta}j(\theta)\right\vert \to 0 $ ] as @xmath136 , which can be used to approximate , for large @xmath60 , @xmath61 and small @xmath44 , as @xmath146 where @xmath62 and @xmath68 are governed by @xmath141 and @xmath147 respectively . we propose two - timescale algorithms based on the estimates obtained in and . let @xmath148 and @xmath149 be two step - size sequences satisfying the following . [ stepsize ] @xmath148 and @xmath149 are two positive step - size sequences satisfying @xmath150 , @xmath151 , @xmath152 and @xmath153 , _ i.e. _ , @xmath154 as @xmath155 . it must be noted that in the algorithms , although @xmath60 is chosen to be a large quantity ( to ensure convergence ) , the quantity @xmath61 is arbitrarily picked and can be any finite positive number . the averaging of the inner summation in and is obtained in our algorithms using two - timescale stochastic approximation . in principle , one may select @xmath156 . however , it is generally observed that a value of @xmath61 typically between @xmath157 and @xmath158 results in better performance @xcite . further , the algorithms require generation of @xmath66-dimensional random vectors , consisting of uncorrelated @xmath0-gaussian distributed random variates . this method is described in appendix . for @xmath159 , let @xmath160 @xmath161 represent the projection of @xmath15 onto the set @xmath10 . for simulation , we need to project the perturbed random vectors @xmath63 onto @xmath10 using the above projection . the quantities @xmath162 are used to estimate @xmath124 in the recursions . fix @xmath60 , @xmath61 , @xmath0 and @xmath44 set @xmath163 fix the parameter vector @xmath164 output @xmath165 as the final parameter vector the g@xmath0-sf2 algorithm is similar to the g@xmath0-sf1 algorithm , except that we use two parallel simulations @xmath166 and @xmath167 governed with parameters @xmath63 and @xmath69 respectively , and update the gradient estimate , in step 9 , using the single - stage cost function of both simulations as in . fix @xmath60 , @xmath61 , @xmath0 and @xmath44 set @xmath163 fix the parameter vector @xmath164 output @xmath165 as the final parameter vector we provide here a more straightforward technique to prove that the algorithms converge to a local optimum as compared to @xcite . before presenting the details of convergence analysis , we present the following result on @xmath0-gaussians . it provides an expression for the moments of @xmath66-variate @xmath0-gaussian distributed random vector . this is a consequence of the results presented in @xcite . this result plays a key role in the proofs discussed below . [ moments ] suppose @xmath168 is a random vector , where the components are uncorrelated and identically distributed , each being distributed according to a @xmath0-gaussian distribution with zero @xmath0-mean and unit @xmath0-variance , with parameter @xmath169 . also , let @xmath170 . then , for any @xmath171 , we have @xmath172 = \left\{\begin{array}{l } \bar{k}\displaystyle\left(\frac{n+2-nq}{|1-q|}\right)^{\sum\limits_{i=1}^{n}\frac{b_i}{2 } } \left(\prod\limits_{i=1}^{n}\frac{b_i!}{2^{b_i}\left(\frac{b_i}{2}\right)!}\right ) \\ \hspace{10mm}\text{if } b_i \text { is even for all } i = 1,2,\ldots , n , \\\\ 0 \hspace{8mm}\text{otherwise , } \end{array}\right . \label{moments_eqn}\ ] ] where @xmath173 exists only if the arguments in the above gamma functions are positive , which holds for @xmath174 if @xmath108 , and @xmath175 if @xmath104 . since @xmath176 , and @xmath177 is non - negative over @xmath126 , we have @xmath178 \\&= \frac{1}{k_{q , n } } \int\limits_{\omega_q } \left(x^{(1)}\right)^{b_1}\left(x^{(2)}\right)^{b_2}\ldots\left(x^{(n)}\right)^{b_n } \left(1 - \frac{(1-q)}{\big(n+2-nq\big)}\vert{x}\vert^2\right)^{\frac{1}{1-q}-b}\:\mathrm{d}x . \end{aligned}\ ] ] the second equality in can be easily proved . if for some @xmath179 , @xmath180 is odd , then the above function is odd , and its integration is zero over @xmath126 , which is symmetric with respect to any axis by definition . for the other cases , since the function is even , the integral is same over every orthant . hence , we may consider the integration over the first orthant , _ i.e. _ , where each component is positive . for @xmath108 , we can reduce the above integral , using ( * ? ? ? * equation ( 4.635 ) ) , to obtain @xmath181 & = \frac{\prod\limits_{i=1}^{n } \gamma\left(\frac{b_i+1}{2}\right)}{k_{q , n}\gamma ( \bar{b } ) } \left(\frac{n+2-nq}{1-q}\right)^{\bar{b } } \int\limits_{0}^{1 } ( 1-y)^{\left(\frac{1}{1-q}-b\right)}y^{\left(\bar{b}-1\right)}\mathrm{d}x \label{integral } \ ] ] where we set @xmath182 . one can observe that the integral in is in the form of a beta function . since @xmath180 s are even , we can expand @xmath183 using the expansion of gamma function of half - integers to get @xmath184 . the claim can be obtained by substituting @xmath185 from and using the relation @xmath186 . it is easy to verify that all the gamma functions in the equality are positive provided @xmath174 . the result for the interval @xmath104 can be proved in a similar way ( see equations ( 4.635 ) and ( 4.636 ) of @xcite ) . however , in this case the gamma functions are positive if @xmath187 satisfy the mentioned condition . it may be noted here that this is always true for any dimension if @xmath188 since @xmath0-gaussians are defined only when @xmath189 . it is easy to verify the following result in the limiting case of @xmath80 . this would help to ensure that the convergence analysis done in this paper also holds in the case of gaussian sf . [ moments_coro ] in the limiting case , as @xmath80 , @xmath190 = \prod_{i=1}^{n } \mathsf{e}_{g(x ) } \left[\left(x^{(i)}\right)^{b_i}\right]\;.\ ] ] first , let us consider the update along the faster timescale , _ i.e. _ , step 9 of the g@xmath0-sf1 algorithm . we define @xmath191 , @xmath192 and @xmath193 for @xmath194 , @xmath26 . it follows from assumption [ stepsize ] that @xmath195 , @xmath196 and @xmath197 . we can rewrite step 9 of algorithm 1 as the following iteration for all @xmath198 @xmath199,\ ] ] where @xmath200 here , @xmath201 is defined as in , and @xmath202 is a markov process parameterized by @xmath203 . let @xmath204 denote the @xmath29-field generated by the mentioned quantities . we can observe that @xmath205 is a filtration , where @xmath206 is @xmath207-measurable for each @xmath198 . we summarize the results presented in ( * ? ? ? * chapter 6 , lemma 3 theorem 9 ) in the following theorem . this result leads to the stability and convergence of iteration , which runs on the faster timescale . [ borkar_cor_8 ] consider the iteration , @xmath208 $ ] . let the following conditions hold : 1 . @xmath202 is a markov process satisfying assumptions [ ergodic ] and [ lyapunov ] , 2 . for each @xmath106 and @xmath209 for all @xmath210 , @xmath211 has a unique invariant probability measure @xmath212 , 3 . @xmath213 are step - sizes satisfying @xmath214 and @xmath215 , 4 . @xmath216 is lipschitz continuous in its first argument uniformly w.r.t the second , 5 . @xmath217 is a martingale difference noise term with bounded variance , 6 . if @xmath218 $ ] , then the limit @xmath219 exists uniformly on compacts , and 7 . the ode @xmath220 is well - posed and has the origin as the unique globally asymptotically stable equilibrium . then the update @xmath221 satisfies @xmath222 , almost surely , and converges to the stable fixed points of the ordinary differential equation ( ode ) @xmath223 rewriting the update as @xmath224 - z(p ) + a_p\big ] , \label{faststep1}\ ] ] where @xmath225 $ ] is @xmath207-measurable . the following result shows that @xmath226 satisfies condition 5 in theorem [ borkar_cor_8 ] . [ a_n_convergence ] for all values of @xmath120 , @xmath227 is a martingale difference sequence with a bounded variance . it is easy to see that for all @xmath198 , @xmath228 = 0 $ ] . so @xmath227 is a martingale difference sequence . expanding the terms , we have @xmath229 \\&\leqslant \frac{8}{\beta^2(n+2-nq)^2 } \mathsf{e}\left[\left . \left(\frac{\vert\tilde{\eta}(p)\vert h(y_p)}{\rho{(\tilde{\eta}(p))}}\right)^2 + \left(\left.\mathsf{e}\left[\frac{\vert\tilde{\eta}(p)\vert h(y_p)}{\rho{(\tilde{\eta}(p))}}\right|\mathcal{g}_{p-1}\right]\right)^2 \right|\mathcal{g}_{p-1}\right].\end{aligned}\ ] ] applying conditional jensen s inequality on the second term , we obtain @xmath230 & \leqslant \frac{16}{\beta^2(n+2-nq)^2 } \mathsf{e}\left[\left.\frac{\vert\tilde{\eta}(p)\vert^2}{{\rho{(\tilde{\eta}(p))}^2}}{h^2(y_p)}\right| \mathcal{g}_{p-1}\right ] . \label{jensen_g1}\end{aligned}\ ] ] for @xmath231 , we use holder s inequality to write as @xmath230 & \leqslant \frac{16}{\beta^2(n+2-nq)^2 } \sup\limits_{\eta } \left(\frac{\vert\tilde{\eta}(p)\vert^2}{{\rho{(\tilde{\eta}(p))}^2}}\right ) \mathsf{e}\left[\left.{h^2(y_p)}\right| \mathcal{g}_{p-1}\right ] \\&\leqslant \frac{16}{\beta^2 ( 1-q)(n+2-nq ) } \mathsf{e}\left[\left.{h^2(y_p)}\right| \mathcal{g}_{p-1}\right].\end{aligned}\ ] ] since , @xmath232 and @xmath233 for all @xmath234 . by lipschitz continuity of @xmath20 , there exists @xmath235 such that @xmath236 for all @xmath3 , and hence , by assumption [ lyapunov ] , we can claim @xmath237 \leqslant 2\alpha_1 ^ 2\left ( 1 + \mathsf{e } \left[\vert{y_p}\vert^2|\mathcal{g}_{p-1}\right]\right ) < \infty \text { a.s . } \label{eq_jensen_g1}\end{aligned}\ ] ] on the other hand , for @xmath238 , we apply cauchy - schwartz inequality for each of the components in to obtain @xmath230 & \leqslant \frac{16}{\beta^2(n+2-nq)^2 } \sum_{j=1}^{n } \mathsf{e}\left[\left.\frac{\left(\tilde{\eta}^{(j)}(p)\right)^2}{{\rho{(\tilde{\eta}(p))}^2}}{h^2(y_p)}\right| \mathcal{g}_{p-1}\right ] \\&\leqslant \frac{16}{\beta^2(n+2-nq)^2 } \sum_{j=1}^{n } \mathsf{e}\left[\frac{\left(\tilde{\eta}^{(j)}(p)\right)^4}{{\rho{(\tilde{\eta}(p))}^4}}\right]^{1/2 } \mathsf{e}\left[\left.{h^4(y_p)}\right| \mathcal{g}_{p-1}\right]^{1/2}\;.\end{aligned}\ ] ] the second expectation can be shown to be finite as in , while we apply proposition [ moments ] to study the existence of @xmath239 $ ] . we can observe that in this case , @xmath240 and @xmath241 if @xmath242 , otherwise @xmath243 . and so @xmath188 . proposition [ moments ] ensures that the term is finite , and hence , the claim . we can write the parameter update ( step 13 of g@xmath0-sf1 ) as @xmath244 where @xmath245 since @xmath246 . thus , the parameter update recursion can be seen to track the ode @xmath247 hence , the recursion @xmath70 , @xmath26 appears quasi - static when viewed from the timescale of @xmath248 , and hence , in the update , one may let @xmath249 and @xmath250 for all @xmath210 . consider the following ode @xmath251 [ z_bounded ] the sequence @xmath252 is uniformly bounded with probability 1 . further , @xmath253 almost surely as @xmath254 . it can be easily verified that iteration satisfies all the conditions of theorem [ borkar_cor_8 ] . thus , by theorem [ borkar_cor_8 ] , @xmath252 converges to ode as @xmath255 = \frac{2\eta j(\theta+\beta\eta)}{\beta(n+2-nq)\rho{(\eta)}}\;.\ ] ] we can also see that @xmath256 all the conditions in theorem [ borkar_cor_8 ] are seen to be verified and the claim follows . from lemma [ z_bounded ] , steps 13 and 15 of g@xmath0-sf1 can be written as @xmath257\right ) \nonumber \\&= \mathcal{p}_c \bigg(\theta(n ) + a(n)\left[- \nabla_{\theta(n ) } j\big(\theta(n)\big ) + \delta \big(\theta(n)\big ) + \xi_n\right]\bigg ) , \label{slowstep}\end{aligned}\ ] ] where the error in the gradient estimate is given by @xmath258\ ] ] and the noise term is @xmath259 - \frac{2\eta(n)j\big(\theta(n)+\beta\eta(n)\big)}{\beta(n+2-nq)\rho{(\eta(n ) ) } } \nonumber \\&= \frac{2}{\beta(n+2-nq)}\bigg ( \mathsf{e}_{g_q(\eta)}\left[\left.\frac{\eta(n)}{\rho{(\eta(n))}}j\big(\theta(n)+\beta\eta(n)\big)\right| \theta(n)\right ] - \frac{\eta(n)}{\rho{(\eta(n))}}j\big(\theta(n)+\beta\eta(n)\big)\bigg ) , \label{noise_term}\end{aligned}\ ] ] which is a martingale difference term . let @xmath260 denote the @xmath29-field generated by the mentioned quantities . we can observe that @xmath261 is a filtration , where @xmath262 are @xmath263-measurable for each @xmath26 . we state the following result due ( * ? ? ? * theorem 5.3.1 , pp 189196 ) , adapted to our scenario , which leads to the convergence of the updates in . [ kushner_thm_5.3.1 ] given the iteration , @xmath264 , where 1 . @xmath265 represents a projection operator onto a closed and bounded constraint set @xmath10 , 2 . @xmath266 is a continuous function , 3 . @xmath267 is a positive sequence satisfying @xmath268 , @xmath269 , and 4 . @xmath270 converges a.s . under the above conditions , the update @xmath271 converges to the asymptotically stable fixed points of the ode @xmath272 where @xmath273 . the next result shows that the noise term @xmath274 satisfies the last condition in lemma [ kushner_thm_5.3.1 ] , while the subsequent result proves the error term @xmath275 is considerably small . [ xi_n_convergence ] let @xmath276 . then , for all values of @xmath120 , @xmath277 is an almost surely convergent martingale sequence . we can easily observe that for all @xmath278 , @xmath279 $ ] @xmath280 - \mathsf{e}\left[\left.\frac{\eta(k)j\big(\theta(k)+\beta\eta(k)\big)}{\beta\rho{(\eta(k))}}\right|\mathcal{f}_k\right ] \bigg ) . \end{aligned}\ ] ] so @xmath279 = 0 $ ] , since @xmath281 is @xmath282-measurable , whereas @xmath283 is independent of @xmath282 . it follows that @xmath284 is a martingale difference sequence , and hence @xmath277 is a martingale sequence . now , use of conditional jensen s inequality leads to @xmath285 & = \sum_{j=1}^n \mathsf{e}\left[\left.\left(\xi_{k}^{(j)}\right)^2\right|\mathcal{f}_{k}\right ] \\&\leqslant \frac{16}{\beta^2(n+2-nq)^2 } \sum_{j=1}^n \mathsf{e}\left[\left.\frac{\left(\eta(k)^{(j)}\right)^2}{{\rho{(\eta(k))}^2}}{j\big(\theta(k)+\beta\eta(k)\big)^2}\right|\theta(k)\right ] . \ ] ] for any @xmath286 , by definition @xmath287 $ ] , where the expectation is with respect to the stationary measure . by jensen s inequality , we can claim @xmath288 $ ] and @xmath289 $ ] for all @xmath286 . using these facts along with arguments similar to lemma [ a_n_convergence ] , it can be seen that @xmath290 < \infty$ ] for all @xmath291 , and hence , if @xmath292 , @xmath293 & = \sum_{n=0}^{\infty } a(n)^2\mathsf{e}\left[\vert\xi_{n}\vert^2\right ] \leqslant \sum_{n=0}^{\infty } a(n)^2 \sup_n \mathsf{e}\left[\vert\xi_{n}\vert^2\right ] < \infty \text { a.s.}\end{aligned}\ ] ] the claim follows from martingale convergence theorem @xcite . [ grad_sf1_convergence ] for a given @xmath294 , @xmath7 , and for all @xmath9 , the error term @xmath295-\nabla_{\theta}j(\theta)\big\vert = o(\beta).\ ] ] for small @xmath121 , using taylor series expansion of @xmath296 around @xmath9 , @xmath297 so we can write as @xmath298 \nonumber = \frac{2}{(n+2-nq)}\bigg ( \frac{j(\theta)}{\beta}\mathsf{e}_{g_q(\eta)}\left[\frac{\eta}{\rho{(\eta)}}\right ] + \mathsf{e}_{g_q(\eta)}\left[\frac{\eta\eta^t}{\rho{(\eta)}}\right]\nabla_{\theta}j(\theta ) \\&\hspace{60 mm } + \frac{\beta}{2}\mathsf{e}_{g_q(\eta)}\left[\left.\frac{\eta\eta^{t}\nabla_{\theta}^2j(\theta)\eta}{\rho{(\eta)}}\right|\theta\right ] + o(\beta)\bigg)\;. \label{gqsf1_expand}\end{aligned}\ ] ] we consider each term in . the @xmath299 component in the first term is @xmath300 = 0 $ ] by proposition [ moments ] for all @xmath301 . similarly , the @xmath299 component in the third term can be written as @xmath302^{(i ) } = \frac{\beta}{2}\sum\limits_{j=1}^{n}\sum\limits_{k=1}^{n}\left[\nabla_{\theta}^2j(\theta)\right]_{j , k } \mathsf{e}_{g_q(\eta)}\left[\frac{\eta^{(i)}\eta^{(j)}\eta^{(k)}}{\rho{(\eta)}}\right].\ ] ] it can be observed that in all cases , each term in the summation is an odd function , and so from proposition [ moments ] , we can show that the third term in is zero . using a similar argument , we claim that the off - diagonal terms in @xmath303 $ ] are zero , while the diagonal terms are of the form @xmath304 $ ] , which exists for all @xmath305 as the conditions in proposition [ moments ] are always satisfied on this interval . further , @xmath306 = \frac{(n+2-nq)}{2}\;. \label{grad_expand_term}\ ] ] the claim follows by substituting the above expression in . now , we consider the following ode for the slowest timescale recursion @xmath307 where @xmath308 . in accordance with lemma [ kushner_thm_5.3.1 ] , it can be observed that the stable points of lie in the set @xmath309 we have the following key result which shows that iteration tracks ode , and hence , the convergence of our algorithm is proved . [ thm_gqsf1 ] under assumptions [ ergodic ] [ stepsize ] , given @xmath310 and @xmath305 , there exists @xmath311 such that for all @xmath312 $ ] , the sequence @xmath30 obtained from g@xmath0-sf1 converges to a point in the @xmath313-neighborhood of the stable attractor of , defined as @xmath314 with probability 1 as @xmath155 . it immediately follows from lemmas [ kushner_thm_5.3.1 ] and [ xi_n_convergence ] that the update in converges to the stable fixed points of the ode @xmath315 now starting from the same initial condition , the trajectory of converges to that of uniformly over compacts , as @xmath316 . since from proposition [ grad_sf1_convergence ] , we have @xmath317 for all @xmath318 , the claim follows . it may be noted that we can arrive at the same claim more technically using hirsch s lemma @xcite . since the proof of convergence here is along the lines of g@xmath0-sf1 , we do not describe it explicitly . we just briefly describe the modifications that are required in this case . in the faster timescale , as @xmath155 , the updates given by @xmath319 track the function @xmath320 so we can rewrite the slower timescale update for g@xmath0-sf2 algorithm , in a similar manner as , where the noise term @xmath274 has two components , due to the two parallel simulations , each being bounded ( as in lemma [ xi_n_convergence ] ) . we have the following proposition for the error term @xmath321-\nabla_{\theta}j(\theta)\;.\ ] ] [ grad_sf2_convergence ] for a given @xmath294 , @xmath7 , and for all @xmath9 , @xmath322-\nabla_{\theta}j(\theta)\big\vert = o(\beta).\ ] ] using taylor s expansion , we have for small @xmath44 , @xmath323 one can use similar arguments as in proposition [ grad_sf1_convergence ] to rewrite as @xmath324 & = \frac{1}{(n+2-nq ) } \mathsf{e}_{g_q(\eta)}\left[\frac{2}{\rho{(\eta)}}\eta\eta^t\right]\nabla_{\theta}j(\theta ) + o(\beta),\end{aligned}\ ] ] which leads to the claim . finally , we have a similar result to prove the convergence of the g@xmath0-sf2 algorithm . [ thm_gqsf2 ] under assumptions [ ergodic ] [ stepsize ] , for @xmath310 and @xmath305 , there exists @xmath311 such that for all @xmath312 $ ] , the sequence @xmath30 obtained from g@xmath0-sf2 converges to a point in the @xmath313-neighborhood of the stable attractor of , with probability 1 as @xmath155 . theorems [ thm_gqsf1 ] and [ thm_gqsf2 ] give the existence of some @xmath325 for a given @xmath310 such that the gradient - descent algorithms converge to @xmath313-neighborhood of a local minimum . however , these results do not give the precise value of @xmath326 . further , they do not guarantee that this neighborhood lies within a close proximity of a global minimum . we make a note on the analysis for gaussian sf algorithms . though the above results exclude the case @xmath327 , it is easy to verify that all the claims hold as @xmath328 due to corollary [ moments_coro ] . hence , the above convergence analysis provides an alternative to the analysis presented in @xcite for gaussian sf algorithms . we consider a multi - node network of @xmath329 queues with feedback as shown in the figure below . there are @xmath330 nodes , which are fed with independent poisson external arrival processes with rates @xmath331 , respectively . after departing from the @xmath299 node , a customer either leaves the system with probability @xmath332 or enters the @xmath333 node with probability @xmath334 . once the service at the @xmath335 node is completed , the customer may rejoin the @xmath336 node with probability @xmath337 . the service time processes of each node , @xmath338 are defined as @xmath339 where for all @xmath340 , @xmath341 are constants and @xmath342 are independent samples drawn from the uniform distribution on @xmath343 . the service time of each node depends on the @xmath344-dimensional tunable parameter vector @xmath345 , whose individual components lie in a certain interval @xmath346 $ ] , @xmath347 , @xmath340 . @xmath348 represents the @xmath349 update of the parameter vector at the @xmath299 node , and @xmath350 represents the target parameter vector corresponding to the @xmath299 node . the cost function is chosen to be the sum of the total waiting times of all the customers in the system . for the cost to be minimum , @xmath351 should be minimum , and hence , we should have @xmath352 , @xmath353 . let us denote @xmath354 and @xmath355 it is evident that @xmath356 , where @xmath357 . in order to compare the performance of the various algorithms , we consider the performance measure to be the euclidean distance between @xmath70 and @xmath358 , @xmath359^{1/2}.\ ] ] the choice for such a performance measure is due to the fact that when the above distance is low , the queuing network provides globally optimal performance . hence , in the results presented below , a low value of the distance ( performance measure ) implies that the algorithm converges to a closer proximity of the global minimum . for the simulations , we first consider a two queue network with the arrival rates at the nodes being @xmath360 and @xmath361 respectively . we consider that all customers leaving node-1 enter node-2 , _ i.e. _ , @xmath362 , while customers serviced at node-2 may leave the system with probability @xmath363 . we also fix the constants in the service times at @xmath364 and @xmath365 , respectively , for the two nodes . the service time parameters for either node are two - dimensional vectors , @xmath366 , with components lying in the interval @xmath346 = [ 0.1,0.6]$ ] for @xmath367 , @xmath368 . thus the constrained space @xmath10 is given by @xmath369 ^ 4\subset\mathbb{r}^4 $ ] . we fix the target parameter at @xmath370 . the simulations were performed on an intel core @xmath371 machine with @xmath372 memory space and linux operating system . we run the algorithms by varying the values of @xmath0 and @xmath44 , while all the other parameters are held fixed at @xmath373 and @xmath374 . for all the cases , the initial parameter is assumed to be @xmath375 . for each set of parameters , 20 independent runs were performed with each run of @xmath376 iterations taking about 0.5 seconds . we compare the performance of the proposed algorithms with gradient based sf algorithms proposed in @xcite , which use gaussian smoothing . box - mller method has been used to sample standard gaussian vectors , while samples from multivariate @xmath0-gaussians are drawn using the method discussed in appendix . this method uses generation of a @xmath377-distributed random variable , which is implemented using standard methods described in ( * ? ? ? * algorithms 4.33 and 4.37 ) . figure [ plot_4dim ] shows the convergence behavior of the gaussian and proposed @xmath0-gaussian based algorithms with @xmath378 , where the smoothness parameter is @xmath379 . .performance of all algorithm for different values of @xmath0 with @xmath379 , and step - sizes @xmath380 and @xmath381 . [ cols="^,^,^,^,^,^,^",options="header " , ] we perform similar experiments in a higher dimensional case . for this , we consider a four node network with @xmath382 for all @xmath383 . the probability of leaving the system after service at each node is @xmath384 for all nodes . the service process of each node is controlled by a @xmath157-dimensional parameter vector , and a constant set at @xmath385 . thus , we have a @xmath386-dimensional constrained optimization problem , where each component can vary over the interval @xmath387 $ ] and the target is @xmath388 . the parameters of the algorithms are held fixed at @xmath373 , @xmath374 and @xmath389 . each component in the initial parameter vector is assumed to be @xmath390 for all @xmath391 . the step - sizes were taken to be @xmath392 and @xmath393 . for each @xmath394 tuple , 20 independent runs were performed . each run of @xmath376 iterations took about 1 second on an average . figure [ plot_50dim ] shows the convergence behavior of the algorithms in this case . the performance of one - simulation algorithm is relatively poor . table [ gq2_20 ] shows the performance of g@xmath0-sf2 algorithms for different @xmath394-pairs . we observe that the trend is similar in this case showing that the algorithms scale in a similar manner . the @xmath0-gaussian distribution is an important power - law distribution that has connections with generalized information measures . the power - law behavior of @xmath0-gaussians provide a better control over smoothing of functions as compared to the gaussian distribution . we have extended the gaussian smoothed functional approach for gradient estimation to the @xmath0-gaussian case by showing that the @xmath0-gaussian distribution satisfies the rubinstein conditions @xcite . further , we developed optimization algorithms that incorporate @xmath0-gaussian smoothing . this extension turns out to be more significant when we note that the @xmath0-gaussians encompass all the existing smoothing kernels - gaussian ( @xmath80 ) , cauchy ( @xmath395 ) and uniform ( @xmath96 ) . we proposed two @xmath0-gaussian sf algorithms for simulation optimization . we use a queuing network example to show that for certain values of @xmath0 , the results provided by the proposed algorithms are significantly better than gaussian sf algorithms . our simulation results even indicate that for some @xmath0-values , the performance is better as compared to all the above mentioned special cases . we also presented proof of convergence of the proposed algorithms to a local minimum of the objective function . it would be interesting to develop hessian estimators incorporating the @xmath0-gaussian smoothed functionals , and developing newton based algorithms along these lines . the algorithms discussed in the paper require generation of a multivariate @xmath0-gaussian distributed random vector , whose individual components are uncorrelated and identically distributed . this implies that the random variables are @xmath0-independent @xcite . for the limiting case of @xmath80 , @xmath0-independence is equivalent to independence of the random variables . hence , we can use standard algorithms to generate i.i.d . this is typically not possible for @xmath0-gaussians with @xmath7 . @xcite proposed an algorithm for generating one - dimensional @xmath0-gaussian distributed random variables using generalized box - mller transformation . but , there exists no standard algorithm for generating @xmath66-variate @xmath0-gaussian random vectors . a method can be obtained by making use of the one - to - one correspondence between @xmath0-gaussian and students-@xmath94 distributions for @xmath97 . further , a duality property of @xmath0-gaussians can be used to relate the distributions for @xmath396 and @xmath231 . this observation , first made by @xcite , is shown below . we denote the @xmath0-gaussian distribution with @xmath0-mean @xmath88 and @xmath0-covariance @xmath89 as @xmath397 . based on this , we formally present an algorithm , which can be used to generate multivariate @xmath0-gaussian distribution . theoretical justification behind the algorithm is provided in the following results @xcite . a. rnyi . on measures of entropy and information . in _ fourth berkeley symposium on mathematical statistics and probability , 1960 _ , volume 1 , pages 547561 , berkeley , california , 1961 . university of california press . a. perez . risk estimates in terms of generalized @xmath411-entropies . in _ proceedings of the colloquium on information theory , debrecen 1967 _ , pages 299315 , budapest , 1968 . journal bolyai mathematical society . s. bhatnagar and v. s. borkar . two timescale stochastic approximation scheme for simulation - based parametric optimization . _ probability in the engineering and informational sciences _ , 12:0 519531 , 1998 . s. bhatnagar , m. c. fu , s. i. marcus , and i. j. wang . two - timescale simultaneous perturbation stochastic approximation using deterministic perturbation sequences . _ acm transactions on modeling and computer simulation _ , 130 ( 2):0 180209 , 2003 . a. dukkipati , s. bhatnagar , and m. n. murty . on measure - theoretic aspects of nonextensive entropy functionals and corresponding maximum entropy prescriptions . _ physica a : statistical mechanics and its applications _ , 3840 ( 2):0 758774 , 2007 . j. costa , a. hero , and c. vignat . on solutions to multivariate maximum @xmath1-entropy problems . _ energy minimization methods in computer vision and pattern recognition , lecture notes in computer science _ , 2683:0 211226 , 2003 . w. j. thistleton , j. a. marsh , k. nelson , and c. tsallis . generalized box - muller method for generating @xmath0-gaussian random deviates . _ ieee transactions on information theory _ , 530 ( 12):0 48054810 , 2007 . |
in this paper we introduce the notion of _ parabolic - like mapping_. such an object is similar to a polynomial - like mapping , but it has a parabolic external class , _ i.e. _ an external map with a parabolic fixed point . we define the notion of parabolic - like mapping and we study the dynamical properties of parabolic - like mappings . we prove a straightening theorem for parabolic - like mappings which states that any parabolic - like mapping of degree is hybrid conjugate to a member of the family \,| \ , p_a(z)=z+ \frac{1}{z}+ a,\,\,a \in \c\right\},\ ] ] a unique such member if the filled julia set is connected . | 1111.7150 | a _ polynomial - like map _ of degree @xmath2 is a triple ( @xmath3 ) where @xmath4 and @xmath5 are open subsets of @xmath6 isomorphic to discs , @xmath4 is compactly contained in @xmath5 , and @xmath7 is a proper degree @xmath2 holomorphic map ( see @xcite ) . a degree @xmath2 polynomial - like map is determined up to holomorphic conjugacy by its internal and external classes , that is , the ( conjugacy classes of the ) maps which encode the dynamics of the polynomial - like map on the filled julia set and its complement . in particular the external class consists of degree @xmath2 real - analytic orientation preserving and strictly expanding self - coverings of the unit circle . the definition of a polynomial - like map captures the behaviour of a polynomial in a neighbourhood of its filled julia set . by changing the external class of a degree @xmath2 polynomial - like map with the external class of a degree @xmath2 polynomial ( see @xcite ) , a degree @xmath2 polynomial - like map can be straightened to a polynomial of the same degree . in this paper we introduce a new object , a _ parabolic - like mapping _ , similar to but different from a polynomial - like mapping . the similarity resides in the fact that a parabolic - like map is a local concept , it is characterized by a filled julia set and an external map , and the external map of a degree @xmath2 parabolic - like mapping is a degree @xmath2 real - analytic orientation preserving self - covering of the unit circle . the difference resides in the fact that a parabolic - like map has a parabolic fixed point with an attracting petal outside the filled julia set , and the external map of a parabolic - like mapping has a parabolic fixed point . the aim of this paper is to extend the theory of polynomial - like mappings ( in the dynamical plane ) to parabolic - like mappings . let us give an example which illustrates the class of maps we are considering . the map @xmath8 has a parabolic fixed point at @xmath9 . since the parabolic basin of attraction of the parabolic fixed point resides in the interior of the filled julia set , while the repelling direction resides on the julia set and outside of it , the external map of @xmath10 is hyperbolic . the map @xmath10 presents polynomial - like restrictions . on the other hand , let us interchange the roles of the filled julia set and the closure of the basin of attraction of infinity for @xmath11 . in other words , let us conjugate @xmath10 by @xmath12 and obtain the map @xmath13 , and let us define as filled julia set for @xmath14 the closure of the basin of attraction of the superattracting fixed point @xmath15 . the basin of attraction of the parabolic fixed point @xmath16 now resides outside the filled julia set , and gives rise to an the external class with a parabolic fixed point . appropriate restrictions of the map @xmath14 belong to the class of parabolic - like mappings . as polynomial - like mappings are straightened to polynomials , we straighten degree @xmath0 parabolic - like mappings to members of a model family of maps with a parabolic external class . we take as model family the family of quadratic rational maps with a parabolic fixed point of multiplier @xmath17 , normalized by fixing the parabolic fixed point to be infinity and the critical points to be @xmath17 and @xmath18 , this is @xmath19 \,| \ , p_a(z)=z+ 1/z+ a,\,\,a \in \c\}.\ ] ] [ ca ] [ m1 ] all the maps in @xmath20 have a completely invariant fatou component @xmath21 , namely the parabolic basin of attraction of infinity . we define the filled julia set for these maps as @xmath22 ( note that for every @xmath23 , @xmath24 has a unique completely invariant fatou component @xmath21 , hence @xmath25 is well defined , while for the map @xmath26 we need to make a choice , after which the filled julia set @xmath27 is well defined ) . the external class of this family is parabolic , and we prove in proposition [ extmap ] that it is given by the class of @xmath28 . in this paper we will first define parabolic - like maps and the filled julia set of a parabolic - like map . then we will construct and discuss the external class in this setting . finally , we will prove that we can straighten every degree @xmath0 parabolic - like map to a member of the family @xmath20 , by replacing the external map of the parabolic - like map by @xmath29 ( see fig . [ ca ] and [ m1 ] ) . + the author would like to thank her advisor , carsten lunde petersen , for suggesting the idea of parabolic - like mapping , and for his help , support and encouragement . this paper was written during the author s ph.d . hence the author would like to thank roskilde university and universit paul sabatier for their hospitality , and roskilde university , the anr-08-jcjc-0002 founded by the agence nationale de la recherche and the marie curie rtn 035651-cody for their financial support during her ph.d . in this paper we are studying restrictions of maps with a parabolic fixed point of multiplier @xmath17 . by a change of coordinates we can consider the parabolic fixed point to be at @xmath15 , hence we will consider maps of the form @xmath30 the integer @xmath31 is the _ degeneracy / parabolic multiplicity _ of the parabolic fixed point . in a neighborhood of a parabolic fixed point of parabolic multiplicity @xmath31 , there are @xmath31 attracting petals , which alternate with @xmath31 repelling petals ( for the definition of petal see @xcite or @xcite ) . we will denote the petals by @xmath32 . on each petal @xmath32 there exists a conformal map which conjugates the map @xmath33 to a translation ( see @xcite or @xcite ) . this map is called a _ fatou coordinate _ for the petal @xmath32 , and it is unique up to composition with a translation . we will denote fatou coordinates by @xmath34 . often it is convenient to consider the quotient of a petal @xmath32 under the equivalence relation identifying @xmath35 and @xmath36 if both @xmath35 and @xmath36 belong to @xmath32 . this quotient manifold is called the _ calle cilinder _ , and it is conformally isomorphic to the infinite cylinder @xmath37 ( see @xcite and @xcite ) . + the straightening theorem is obtained by surgery , applying the measurable riemann mapping theorem ( stated below ) . for a proof of the measurable riemann mapping theorem and the notion of almost complex structure , quasiconformal mappings and quasisymmetric mappings , the reader is referred to @xcite or , for a modern treatment , to @xcite . let @xmath38 be a bounded almost complex structure on a domain @xmath39 . then there exists a quasiconformal homeomorphism @xmath40 such that @xmath41 we will use the following notation : @xmath42 @xmath43 a parabolic - like map is an object introduced to extend the notion of polynomial - like maps to maps with a parabolic external map . the domain of a parabolic - like map is not contained in the range , and the set of points with infinite forward orbit is not contained in the intersection of the domain and the range . this calls for a partition of the set of points with infinite forward orbit into a filled julia set compactly contained in both domain and range and exterior attracting petals . [ definitionparlikemap ] * ( parabolic - like maps ) * a _ parabolic - like map _ of degree @xmath44 is a 4-tuple ( @xmath45 ) where * @xmath4 and @xmath5 are open subsets of @xmath6 , with @xmath46 and @xmath47 isomorphic to a disc , and @xmath4 not contained in @xmath5 , * @xmath7 is a proper holomorphic map of degree @xmath48 with a parabolic fixed point at @xmath49 of multiplier 1 , * @xmath50 \rightarrow \overline { u}$ ] is an arc with @xmath51 , forward invariant under @xmath33 , @xmath52 on @xmath53 $ ] and on @xmath54 $ ] , and such that @xmath55 @xmath56 ) \subseteq u \setminus u',\,\,\,\,\,\,\gamma(\pm 1 ) \in \partial u.\ ] ] it resides in repelling petal(s ) of @xmath57 and it divides @xmath4 and @xmath5 into @xmath58 and @xmath59 respectively , such that @xmath60 ( and @xmath61 ) , @xmath62 is an isomorphism ( see fig . [ aaa ] ) and @xmath63 contains at least one attracting fixed petal of @xmath57 . we call the arc @xmath64 a _ dividing arc_. [ gamma ] we can consider @xmath65 , where @xmath66\rightarrow \overline { u } , \,\,\,\gamma_{-}:[0,-1]\rightarrow \overline { u } , \,\,\ , \gamma_{\pm}(0)=z_0 $ ] . where it will be convenient ( e.g. in the examples ) we will refer to @xmath67 instead of @xmath64 . 1 . consider the function @xmath28 . this map has critical points at @xmath15 and at @xmath68 , and a parabolic fixed point at @xmath69 of multiplier @xmath17 and parabolic multiplicity @xmath0 . the attracting directions of the parabolic fixed point are along the real axis , while the repelling ones are perpendicular to the real axis . the repelling petals @xmath70 and @xmath71 intersect the unit circle and can be taken to be reflection symmetric around the unit circle , since @xmath29 is autoconjugate by the reflection @xmath72 . let @xmath73 be fatou coordinates with axis tangent to the unit circle at the parabolic fixed point . the image of the unit circle in the fatou coordinate planes are horizontal lines , which we can suppose coincide with @xmath74 , possibly changing the normalizations of @xmath75 . choose @xmath76 and define @xmath77 , and @xmath78 . let @xmath79 be intersection points of @xmath80 respectively and @xmath81 . thus @xmath82 with @xmath83 , and @xmath84 with @xmath85 . define the dividing arcs as : @xmath86 \rightarrow u\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\gamma_{- } : [ 0,-1 ] \rightarrow u\ ] ] @xmath87 then ( @xmath88 ) is a parabolic - like map of degree @xmath0 . 2 . let @xmath89 , for @xmath90 . this map has a superattracting fixed point @xmath91 at @xmath92 , a critical point @xmath93 at @xmath94 and a parabolic fixed point at @xmath15 with multiplier and parabolic multiplicity @xmath17 . call @xmath95 the immediate basin of attraction of the parabolic fixed point . then the critical point @xmath93 belongs to @xmath95 . let @xmath96 be the riemann map normalized by setting @xmath97 and @xmath98 , and let @xmath99 be its inverse . by the carathodory theorem the map @xmath100 extends continuously to @xmath101 . note that @xmath102 . let @xmath103 be an @xmath29 periodic point in the first quadrant , such that the hyperbolic geodesic @xmath104 connecting @xmath103 and @xmath105 separates the critical value @xmath106 from the parabolic fixed point @xmath69 . let @xmath5 be the jordan domain bounded by @xmath107 , union the arcs up to potential level @xmath17 of the external rays landing at @xmath108 and @xmath109 , together with the arc of the level @xmath17 equipotential connecting this two rays around @xmath91 ( see fig . [ cubicbcnnew ] ) . let @xmath4 be the connected component of @xmath110 containing @xmath111 and the dividing arcs @xmath67 be the fixed external rays landing at the parabolic fixed point @xmath111 and parametrized by potential . then ( @xmath45 ) is a parabolic - like map of degree @xmath0 ( see fig . [ cubicbcnnew ] ) . 3 . let @xmath112 , for @xmath113 ( fat rabbit ) . its third iterate @xmath114 has a parabolic fixed point at @xmath115 of multiplier @xmath17 and parabolic multiplicity @xmath116 . let @xmath117 be the component of the immediate basin of attraction of the parabolic fixed point containing @xmath15 . number the connected components of the immediate attracting basin in the dynamical order ( which here is the counterclockwise direction around @xmath118 ) . let @xmath119 be the riemann map , normalized by @xmath120 and @xmath121 , and let @xmath122 be its inverse . the map @xmath100 extends continuously to @xmath101 , and @xmath123 . as above let @xmath103 be a @xmath29 periodic point in the first quadrant such that the hyperbolic geodesic @xmath124 connecting @xmath103 and @xmath105 separates the critical value @xmath106 from the parabolic fixed point @xmath69 . define @xmath125 and @xmath126 let @xmath5 be the jordan domain bounded by @xmath127 union the arcs up to potential level 1 of the external rays landing at @xmath108 and @xmath128 union @xmath129 union the arcs up to potential level @xmath17 of the external rays landing at @xmath130 and @xmath131 , together with the two arcs of the level @xmath17 equipotential connecting this four rays around the parabolic fixed point . let @xmath4 be the connected component of @xmath132 containing @xmath133 and the dividing arcs @xmath134 and @xmath135 be the external rays for angles @xmath136 and @xmath137 respectively parametrized by potential . then ( @xmath138 ) is a parabolic - like map of degree @xmath0 ( see fig . [ fatrabbit ] ) . + more generally , define @xmath139 with @xmath140 and @xmath141 coprime , @xmath142 and consider @xmath143 . the map @xmath144 has a parabolic fixed point of multiplier @xmath145 at @xmath146 , therefore @xmath147 has a parabolic fixed point of multiplier @xmath17 and parabolic multiplicity @xmath141 . + repeating the construction done above one can see that the map @xmath147 restricts to a degree @xmath0 parabolic - like map . let @xmath148 be a parabolic - like map . we define the _ filled julia set @xmath149 _ of @xmath33 as the set of points in @xmath4 that never leave @xmath150 under iteration : @xmath151 an equivalent definition for the filled julia set of @xmath33 is @xmath152 the filled julia set is a compact subset of @xmath153 and it is full ( since it is the intersection of topological disks ) . as for polynomials , we define the julia set of @xmath33 as the boundary of the filled julia set : @xmath154 a parabolic - like map can be seen as the union of two different dynamical parts : a polynomial - like part ( on @xmath155 ) and a parabolic one ( on @xmath63 ) , which are connected by the dividing arc @xmath64 . the parabolic fixed point belongs to the interior of the domain of a parabolic - like map in order to insure that the filled julia set is compactly contained in the intesection of the domain and the range . the dividing arc separates the exterior attracting petals from the filled julia set of the parabolic - like mapping , and for this reason the dividing arc is part of the definition of parabolic - like mapping ( note that we could have constructed the dividing arc a posteriori by fatou coordinates ) . the definition of parabolic - like map also guarantees the existence of an annulus , @xmath156 , essential in defining the external class and to perform the surgery which will give the straightening theorem . there are many prospect definitions of a parabolic - like map . the one introduced here is flexible enough to capture many interesting examples , and rigid enough to allow for a viable theory . we say that @xmath157 is a _ parabolic - like restriction _ of @xmath158 if @xmath159 and @xmath160 are parabolic - like maps with the same degree and filled julia set . * ( conjugacy for parabolic - like mappings ) * we say that the parabolic - like mappings ( @xmath161 ) and ( @xmath162 ) are _ topologically conjugate _ if there exist parabolic - like restrictions ( @xmath163 ) and ( @xmath164 ) , and a homeomorphism @xmath165 such that @xmath166 and @xmath167 if moreover @xmath168 is quasiconformal ( and @xmath169 a.e . on @xmath149 ) , we say that @xmath33 and @xmath170 are _ quasiconformally _ ( _ hybrid _ ) conjugate . a topological conjugacy between parabolic - like maps is a homeomorphism defined on a neighborhood of the filled julia set , which conjugates dynamics just on @xmath171 . this definition allows flexibility regarding the parabolic multiplicity of the parabolic fixed point . in this paper we will prove the following : [ tst ] 1 . every degree @xmath0 parabolic - like mapping ( @xmath161 ) is hybrid equivalent to a member of the family @xmath20 . moreover , if @xmath149 is connected , this member is unique . part @xmath17 follows from proposition [ stex ] together with theorem [ thm ] , while part @xmath0 follows from proposition [ unic ] . two parabolic - like maps are _ equivalent _ , and we do not distinguish between them , if they have a common parabolic - like restriction . given a parabolic - like map @xmath157 , the arc @xmath172\rightarrow \overline u$ ] with @xmath173 is _ isotopic _ to @xmath174 if there exists a domain @xmath175 for which @xmath176 have a common parabolic - like restriction . [ is ] let ( @xmath45 ) be a parabolic - like map , and let @xmath177\rightarrow \overline u$ ] be an arc forward invariant under @xmath33 , with @xmath178 and @xmath52 on @xmath53 $ ] and @xmath179 $ ] . then @xmath180 and @xmath64 are isotopic if and only if their projections to calle cylinders are isotopic and the isotopies are disjoint from the projections of the filled julia set and the critical points . let us prove that , if the projections of @xmath64 and @xmath180 to calle cylinders are isotopic and the isotopies are disjoint from the projections of the filled julia set and the critical points , then @xmath180 and @xmath64 are isotopic . the vice versa is trivial . let @xmath181 and @xmath71 be repelling petals where @xmath134 and @xmath135 respectively reside ( note that @xmath181 and @xmath71 may coincide ) . then the quotient manifolds @xmath182 , @xmath183 are conformally isomorphic to the bi - infinite cylinder . call @xmath184 the isomorphism between @xmath185 and @xmath186 , and @xmath187 the isomorphism between @xmath188 and @xmath186 . let @xmath189 \times [ \tau , d\tau ] \rightarrow \c / \z\ ] ] @xmath190 @xmath191 \times [ d \hat \tau , \hat \tau ] \rightarrow \c / \z\ ] ] @xmath192 be isotopies , disjoint from the projections of the filled julia set and the critical points , such that for every fixed @xmath193,$ ] both @xmath194 are at least @xmath52 . set @xmath195=\beta^{-1}(h_+(s , \cdot))$ ] and @xmath196=\delta^{-1}(h_-(s , \cdot))$ ] define @xmath180 by extending @xmath197 and @xmath198 by the dynamics of @xmath33 to forward invariant curves in @xmath181 and @xmath71 respectively ( see picture [ isotopy ] ) , _ i.e. _ : 1 . @xmath199 @xmath200 ; 2 . @xmath201 @xmath202 ; 3 . @xmath203 and @xmath204 where @xmath205 is the branch which gives continuity . then @xmath180 divides @xmath5 and @xmath4 in @xmath206 and @xmath207 respectively , and by construction @xmath208 contains @xmath149 and all the critical points of @xmath209 . hence @xmath210 is a parabolic - like restriction of @xmath209 , and thus the arcs @xmath64 and @xmath180 are isotopic . note that , by construction , if ( @xmath45 ) is a parabolic - like map and @xmath180 is isotopic to @xmath64 , then the arc @xmath211 resides in the same petal as @xmath134 and the arc @xmath212 resides in the same petal as @xmath135 . in analogy with the polynomial - like setting , we want to associate to any parabolic - like map ( @xmath213 ) of degree @xmath2 a real - analytic map @xmath214 of the same degree @xmath2 and with a parabolic fixed point , unique up to conjugacy by a real - analytic diffeomorphism . we will call @xmath215 an _ external map _ of @xmath33 , and we will call @xmath216 $ ] ( its conjugacy class under real - analytic diffeomorphisms ) the _ external class _ of @xmath33 . the construction of an external map of a parabolic - like map with connected julia set follows the construction of an external map in @xcite , up to the differences given by the geometry of our setting . let ( @xmath45 ) be a parabolic - like map of degree @xmath2 with connected filled julia set @xmath149 . then @xmath149 contains all the critical points of @xmath33 and hence @xmath217 is a holomorphic degree @xmath2 covering map . let @xmath218 be the riemann map , normalized by @xmath219 and @xmath220 as @xmath221 . write @xmath222 and @xmath223 ( see fig . [ abb ] ) and define the map : @xmath224 then the map @xmath225 is a holomorphic degree @xmath2 covering . let @xmath226 denote the reflection with respect to the unit circle , and define @xmath227 , @xmath228 , @xmath229 and @xmath230 . applying the strong reflection principle with respect to @xmath101 we can extend analytically the map @xmath231 to @xmath232 . let @xmath215 be the restriction of @xmath233 to the unit circle , then the map @xmath234 is an _ external map _ of @xmath33 . an external map of a parabolic - like map is defined up to a real - analytic diffeomorphism . let ( @xmath45 ) be a parabolic - like map of degree @xmath2 . to deal with the case where the filled julia set is not connected , we will lean on the similar construction in the polynomial - like case . we construct annular riemann surfaces @xmath235 and @xmath236 that will play the role of @xmath237 and @xmath238 respectively , and an analytic map @xmath239 that will play the role of @xmath33 . let @xmath240 be a full relatively compact connected subset of @xmath241 containing @xmath242 , the critical values of @xmath33 and such that @xmath243 ( after riscaling @xmath64 ) is a parabolic - like restriction of ( @xmath244 ) . call @xmath245 and @xmath246 . define @xmath247 , @xmath248 , @xmath249 , @xmath250 , @xmath251 and @xmath252 . note that @xmath253 is an annular domain . let @xmath254 be a degree @xmath2 covering map for some riemann surface @xmath255 , and define @xmath256 . define @xmath257 . the map @xmath258 is proper holomorphic of degree d , and @xmath259 is a proper holomorphic map of degree d. therefore we can choose @xmath260 , a lift of @xmath258 to @xmath259 , and @xmath261 is an isomorphism . the subset @xmath262 has @xmath2 preimages under the map @xmath263 . let us call @xmath264 the preimage of @xmath262 under @xmath263 such that @xmath265 . since @xmath266 is an isomorphism , we can extend the map @xmath261 to @xmath63 . let us call @xmath267 . since @xmath268 , the extension @xmath269 is an isomorphism ( see fig [ gencase1 ] ) . let us call @xmath270 . define @xmath271 and @xmath272 . the map @xmath11 is proper , holomorphic and of degree @xmath2 ( see fig.[gencase2 ] ) . indeed @xmath273 is a degree @xmath2 covering by definition and @xmath274 is an isomorphism because it is a restriction of an isomorphism . define @xmath275 , then @xmath276 . let @xmath277 be a degree @xmath2 covering map for some riemann surface @xmath278 , and call @xmath279 . define @xmath280 as a lift of @xmath11 to @xmath281 . then @xmath282 is an isomorphism , since @xmath283 is a degree @xmath2 covering and @xmath284 is a degree @xmath2 covering as well . define @xmath285 , and @xmath286 . define @xmath287 and @xmath288 . the map @xmath14 is proper , holomorphic and of degree @xmath2 , indeed @xmath289 is a degree @xmath2 covering and @xmath290 is an isomorphism . define @xmath291 , then @xmath292 . define recursively @xmath293 for @xmath294 as a holomorphic degree @xmath2 covering for some riemann surface @xmath295 and call @xmath296 . define recursively @xmath297 as a lift of @xmath298 to @xmath299 . then @xmath300 is an isomorphism . define @xmath301 , and @xmath302 . define @xmath303 and @xmath304 . then all the @xmath305 are proper holomorphic maps of degree @xmath2 , indeed @xmath306 are degree @xmath2 coverings and @xmath307 are isomorphisms . define @xmath308 , then @xmath309 . we define @xmath310 and @xmath311 ( disjoint union ) . let @xmath236 be the quotient of @xmath312 by the equivalence relation identifying @xmath313 with @xmath314 , and @xmath235 be the quotient of @xmath315 by the same equivalence relation . then @xmath236 is an annulus , since it is constructed by identifying at each level an inner annulus @xmath316 with an outer annulus @xmath317 in the next level . similarly @xmath235 is an annulus , since it is constructed by identifying at each level an inner annulus @xmath318 with an outer annulus @xmath319 in the next level . hence ( since @xmath320 ) @xmath321 is an annulus , since @xmath322 is an annulus and @xmath261 identifies an inner annulus of @xmath322 ( which is @xmath323 ) with an outer annulus of @xmath255 ( which is @xmath324 ) , and @xmath235 is an annulus . the covering maps @xmath325 induce a degree @xmath2 holomorphic covering map @xmath326 . indeed , @xmath327 is well defined , since at each level @xmath328 by definition and @xmath329 is a lift of @xmath330 to @xmath331 . therefore @xmath332 , and the following diagram commutes @xmath333 finally , the map @xmath327 is proper of degree @xmath2 since by definition @xmath334 is a proper map ( and @xmath335 is proper onto its range , which is @xmath253 ) . now , let us construct an external map for @xmath33 . let @xmath336 be the modulus of the annulus @xmath337 . let @xmath338 be any annulus with inner boundary @xmath101 and modulus @xmath339 . then there exists an isomorphism @xmath340 with @xmath341 when @xmath342 and @xmath343 when @xmath344 within @xmath345 ( where @xmath346 ) . then we just have to repeat the construction done for the case @xmath149 connected , with @xmath235 and @xmath236 playing the role of @xmath237 and @xmath238 respectively , and @xmath327 playing the role of @xmath33 . two parabolic - like maps ( @xmath161 ) and ( @xmath162 ) are _ externally equivalent _ if their external maps are conjugate by a real - analytic diffeomorphism , _ i.e. _ if their external maps belong to the same external class . let ( @xmath161 ) and ( @xmath162 ) be two parabolic - like mappings with connected julia sets . by the construction of an external map we gave ( see section [ em ] ) , it is easy to see that ( @xmath161 ) and ( @xmath162 ) are externally equivalent if and only if there exist parabolic - like restrictions ( @xmath163 ) and ( @xmath164 ) , and a biholomorphic map @xmath347 such that @xmath348 and @xmath349 on @xmath350 . we call @xmath100 an _ external equivalence _ between @xmath33 and @xmath170 . the following lemma shows that the situation is analogous also in the case where the julia sets are not connected . [ lem ] let @xmath351 , @xmath352 be two parabolic - like mappings with disconnected julia sets . let @xmath353 be a full relatively compact connected subset of @xmath354 containing @xmath355 and the critical values of @xmath356 , and such that @xmath357 is a parabolic - like restriction of @xmath358 . define @xmath359 . suppose @xmath360 is a biholomorphic map such that @xmath361 on @xmath362 . then @xmath363 and @xmath364 are externally equivalent , and we say that @xmath365 is an external equivalence between them . let ( @xmath366 be as in the construction of an external map for a parabolic - like map with disconnected julia set . let us set @xmath367 and define recursively @xmath368 . then the following diagram commutes : @xmath369 ( for @xmath370 , @xmath371 ) . then every @xmath372 thus defined is an isomorphism and a conjugacy between @xmath373 and @xmath374 , and the following diagram commutes : @xmath375 thus the family of isomorphisms @xmath376 induces an isomorphism @xmath377 compatible with dynamics , and so the external maps of @xmath11 and @xmath14 are real - analytically conjugate . the filled julia set @xmath378 of a polynomial @xmath379 is defined as the complement of the basin of attraction of infinity , which is a completely invariant fatou component . for a degree @xmath2 rational map @xmath380 with a completely invariant fatou component @xmath21 we may define the filled julia set as @xmath381 note that a degree @xmath2 map can have up to @xmath0 completely invariant fatou components @xmath382 ( since a degree @xmath2 map defined on the riemann sphere has @xmath383 critical points , and a completely invariant fatou component contains at least @xmath384 critical points ) . in the case @xmath385 has precisely @xmath17 completely invariant fatou component @xmath21 , the filled julia set @xmath386 is well defined . in the case @xmath385 has @xmath0 such components @xmath382 , there are @xmath0 possibilities for the filled julia set , hence we need to make a choice . after choosing a completely invariant component @xmath387 , the filled julia set @xmath388 is well defined . every member of the family @xmath20 has a parabolic fixed point at @xmath68 with multiplier @xmath17 , and the basin of attraction of the parabolic fixed point is a completely invariant fatou component . for all the members of the family @xmath20 with @xmath23 the parabolic multiplicity of the parabolic fixed point is @xmath17 , hence all these maps have precisely one completely invariant fatou component @xmath21 . thus for all the members of the family @xmath20 with @xmath23 the filled julia set @xmath389 is well defined . on the other hand , since for the map @xmath390 the parabolic multiplicity of @xmath68 is @xmath0 , this map has 2 completely invariant fatou components , namely @xmath391 and @xmath392 . since @xmath26 is conformally conjugate to the map @xmath28 under the map @xmath393 , for consistency with the example 1 in section [ ex ] we consider @xmath394 . let @xmath395 be a rational map of degree @xmath2 . the map @xmath33 has a _ parabolic - like restriction _ if there exist open connected sets @xmath5 and @xmath4 and a dividing arc @xmath64 such that ( @xmath45 ) is a parabolic - like map of some degree @xmath396 . a parabolic - like restriction of a member @xmath24 of the family @xmath20 has degree @xmath0 , hence the filled julia set @xmath397 defined as above coincides with the filled julia set of the parabolic - like restriction of @xmath24 . therefore , we consider as external class of @xmath24 the external class of its parabolic - like restriction . [ extmap ] for every @xmath398 the _ external class of @xmath24 _ is given by the class of @xmath399 . since the maps @xmath26 and @xmath28 are conformally conjugate , in order to prove that @xmath29 is an external map of @xmath24 , it is sufficient to prove that @xmath400 is externally equivalent to @xmath24 , for @xmath398 . let @xmath401 be an attracting petal of @xmath400 containing the critical value @xmath16 , and let @xmath402 be the incoming fatou coordinates of @xmath400 normalized by @xmath403 . replacing @xmath404 by @xmath405 if necessary , we can assume that @xmath69 is the first critical point attracted by @xmath68 . let @xmath406 be an attracting petal of @xmath24 and let @xmath407 be the incoming fatou coordinates of @xmath24 with @xmath408 . let us contruct an external equivalence between @xmath400 and @xmath24 first in the case @xmath397 is connected . the map @xmath409 is a conformal conjugacy between @xmath400 and @xmath24 on @xmath401 . defining @xmath410 as the connected component of @xmath411 containing @xmath401 , and @xmath412 as the connected component of @xmath413 containing @xmath406 , we can lift the map @xmath414 to @xmath415 . since @xmath397 is connected by iterated lifting of @xmath414 we obtain a conformal conjugacy @xmath416 between @xmath400 and @xmath24 . in the case @xmath397 is not connected the map @xmath414 is a conformal conjugacy between @xmath400 and @xmath24 on the region delimited by the fatou equipotential passing through @xmath69 . we are now going to construct parabolic - like restrictions ( @xmath417 ) and ( @xmath418 ) of the maps @xmath400 and @xmath24 respectively and extend the map @xmath414 to an external equivalence between them . the critical point @xmath69 is the first attracted by infinity for both the maps @xmath400 and @xmath24 , so it can not belong to the domains @xmath419 of their parabolic - like restrictions but it may belong to the codomains @xmath420 , while the critical point @xmath421 belongs to @xmath422 and @xmath423 . let us denote by @xmath424 and @xmath425 the fatou coordinates of @xmath24 and @xmath400 respectively ( normalized by @xmath426 and @xmath427 ) , extended to the whole basin of attraction of @xmath68 by iterated lifting . the maps @xmath424 and @xmath425 have univalent inverse branches @xmath428\ } \rightarrow \widehat \xi_a\ ] ] and @xmath429 respectively , and the map @xmath430\ } ) \rightarrow \widehat \xi_a\ ] ] is a biholomorphic extension of @xmath414 conjugating dynamics . choose @xmath431 and @xmath432 such that @xmath433 . then for @xmath434 with @xmath435 sufficiently close to @xmath436 we have @xmath437 . let @xmath438 , @xmath439 be horizontal lines , symmetric with respect to the real axis , starting at @xmath440 and landing at @xmath441 , such that the point @xmath442 is contained in the strip between them ( see fig . [ exteqnuovo1 ] ) and they do not leave the disk @xmath443 ( where @xmath443 is the disk of radius @xmath436 and center @xmath444 ) after having entered to it . define @xmath445 , @xmath446 , @xmath447 , and @xmath448 . in the same way define @xmath449 , @xmath450 , @xmath451 and @xmath452 . then the parabolic - like restriction of @xmath400 we consider is ( @xmath453 ) , and the parabolic - like restriction of @xmath24 we consider is ( @xmath454 ) . note that , by construction , the map @xmath414 is a conformal conjugacy between @xmath400 and @xmath24 on @xmath455 . in order to obtain an external equivalence we need @xmath414 to be defined on a fundamental annulus . define @xmath456 , @xmath457 , @xmath458 , and @xmath459 ( see fig . [ exteqnuovo2 ] ) . since @xmath460 and @xmath461 belong to the regions delimited by the fatou equipotential passing through @xmath69 , the restriction @xmath462 is a holomorphic conjugacy between @xmath400 and @xmath24 . since @xmath463 and @xmath464 , the restrictions @xmath465 and @xmath466 are degree @xmath0 coverings . hence we can lift the map @xmath414 to @xmath467 . finally , we obtain a biholomorphic map @xmath468 which conjugates dynamics . define @xmath469 and @xmath470 , and consequently @xmath471 and @xmath472 . the sets @xmath473 and @xmath474 are compactly contained in @xmath475 and @xmath476 respectively , containing respectively @xmath477 ( which contains the critical value @xmath478 ) , and @xmath479 and the critical value @xmath480 , and such that @xmath481 and @xmath482 are parabolic - like restrictions of ( @xmath483 ) and ( @xmath484 ) respectively , and the map @xmath485 is a biholomorphic conjugacy between @xmath400 and @xmath24 . therefore the result follows by lemma [ lem ] . let ( @xmath45 ) be a parabolic - like map of degree @xmath2 , and let @xmath215 be a representative of its external class . the map @xmath234 is by construction real analytic , symmetric with respect to the unit circle , and it has a parabolic fixed point @xmath486 of multiplier @xmath17 and even parabolic multiplicity @xmath487 ( where @xmath31 is the number of petals of @xmath57 outside @xmath149 ) . let @xmath488 be an isomorphism which defines @xmath215 . hence @xmath215 inherits via @xmath488 dividing arcs @xmath489 and @xmath490 , which divide @xmath491 and @xmath492 into @xmath493 and @xmath494 respectively , such that @xmath495 is an isomorphism and @xmath496 contains at least one attracting fixed petal of @xmath486 . note that @xmath497 is not compactly contained in @xmath498 ( since they share the inner boundary ) , and that @xmath499 and @xmath500 form a positive angle ( since there is at least one attracting fixed petal of @xmath486 in @xmath501 . we prove in @xcite that this angle is @xmath502 ) . moreover , we prove in @xcite ( theorem 2.3.3 ) that there exists @xmath503 $ ] such that for all @xmath504 , @xmath505 , and the equality holds only at the parabolic fixed point . so far we have considered external maps only in relation to parabolic - like maps ( and members of the family @xmath20 ) . we now want to separate these two concepts , and then consider external maps as maps of the unit circle to itself with some specific properties , without refering to a particular parabolic - like map . in order to do so we need to give an abstract definition of external map , which endows it with all the properties it would have , if it would have been constructed from a parabolic - like map . * ( parabolic external map ) * let @xmath506 be a degree @xmath2 orientation preserving real - analytic and metrically expanding ( i.e. @xmath507 ) map . we say that @xmath233 is a _ parabolic external map , _ if there exists a unique @xmath508 such that @xmath509 and @xmath510 , and @xmath511 for all @xmath512 . the multiplicity of @xmath513 as parabolic fixed point of @xmath233 is even ( since the map @xmath233 is symmmetric with respect to the unit circle ) . as @xmath233 is metrically expanding , the repelling petals of @xmath513 intersect the unit circle . let @xmath514 be an extension which is a degree @xmath2 covering ( where @xmath515 for an @xmath76 , and @xmath516 ) . we define a _ dividing arc for @xmath233 _ to be an arc @xmath517 \rightarrow \overline{w } \setminus \d$ ] , forward invariant under @xmath233 , @xmath52 on @xmath53 $ ] and @xmath179 $ ] , residing in the union of the repelling petals which intersect the unit circle and such that @xmath518 @xmath519 ) \subseteq w \setminus w',\,\,\,\,\,\,\widetilde\gamma(\pm 1 ) \in \partial w.\ ] ] a dividing arc for a parabolic external map can be constructed by taking preimages of horizontal lines by repelling fatou coordinates with axis tangent to the unit circle at the parabolic fixed point . the dividing arc divides @xmath520 and @xmath521 into @xmath493 and @xmath522 respectively , such that @xmath523 is an isomorphism and @xmath496 contains at least an attracting fixed petal of @xmath513 . we prove in @xcite ( see lemma 2.3.9 ) that there exists a range @xmath524 for an extension @xmath525 degree @xmath2 covering such that @xmath526 is a topological quadrilateral . therefore external maps contructed from parabolic - like mappings and parabolic external maps are equivalent concepts . * for clarity of exposition we consider in this paper parabolic - like maps with external map having exactly one parabolic fixed point ( cfr . definition [ definitionparlikemap ] ) . this concept naturally generalizes to maps with external maps having several parabolic fixed points . a general parabolic - like map has as many pairs of dividing arcs @xmath67 ( which divide @xmath5 and @xmath4 in @xmath527, ... ,@xmath528 and @xmath529, ... ,@xmath530 respectively ) as the number of parabolic fixed points . * moreover this concept generalizes in a similar way to maps with external maps having several parabolic periodic orbits . an external map for such an object is an orientation preserving real - analytic and metrically expanding map @xmath506 with @xmath510 for every @xmath513 belonging to a parabolic cycle and @xmath511 for all the other points of the unit circle . a _ degree @xmath2 covering extension _ @xmath514 of a parabolic external map @xmath531 is an extension to some neighborhood @xmath532 for an @xmath76 , and @xmath516 such that the map @xmath533 is a degree @xmath2 covering and there exists a dividing arc @xmath534 which divides @xmath535 into @xmath493 and @xmath522 respectively , such that @xmath536 is a topological quadrilateral . the concept of _ parabolic - like restriction _ naturally applies to parabolic external maps ( @xmath537 is a parabolic - like restriction of @xmath514 if they are both degree @xmath2 covering extension of the same parabolic external map and @xmath538 ) . let @xmath64 be a dividing arc for some parabolic external map . we say that @xmath177 \rightarrow \c \setminus \d$ ] is _ isotopic _ to @xmath64 if their projections to calle cylinders are isotopic and the isotopies are disjoint from the projections of the unit circle . from the definitions of dividing arc and isotopy of arcs for parabolic external maps is easy to see that dividing arcs for the same external map are isotopic . [ arcprop ] let @xmath539 be parabolic external maps of the same degree @xmath2 , @xmath540 degree @xmath2 covering extensions , and @xmath541 dividing arcs . then the following statements hold . 1 . let @xmath177 \rightarrow \c \setminus \d$ ] be a foward invariant arc for @xmath542 , isotopic to @xmath174 , and @xmath52 on @xmath53 $ ] and @xmath54 $ ] . then ( possibly after rescaling ) @xmath180 is a dividing arc for @xmath542 . 2 . assume @xmath543 and @xmath544 are constructed by taking preimages of the same periodic curves by repelling fatou coordinates @xmath545 and @xmath546 ( with axis tangent to the unit circle at the parabolic fixed point ) of @xmath547 . then the map @xmath548 defined as : @xmath549 is a quasisymmetric conjugacy between @xmath550 and @xmath551 . property @xmath17 comes from the definition of isotopy in the parabolic external maps setting and a similar argument as in the proof of lemma [ is ] . let us prove property @xmath0 . in the following , @xmath352 . let @xmath552 and @xmath553 be repelling petals where @xmath543 and @xmath544 respectively reside , and let @xmath554 and @xmath555 be repelling fatou coordinates with axis tangent to the unit circle at the parabolic fixed point @xmath556 of @xmath547 . then there exist @xmath134 and @xmath135 , @xmath17-periodic curves in @xmath392 bounded from above and below , such that @xmath557 and @xmath558 . the map @xmath559 is clearly a conjugacy between @xmath550 and @xmath551 . let us prove that this map is quasisymmetric . to fix the notation let us assume the multiplicity of @xmath556 as parabolic fixed point of @xmath547 is @xmath560 . by an iterative local change of coordinates applied to eliminate lower order terms one by one , we obtain conformal diffeomorphisms @xmath561 which conjugate @xmath547 to the map @xmath562 on @xmath563 . since the forward invariant arcs @xmath564 reside in the repelling petals @xmath563 , it suffices to consider @xmath565 . the map @xmath566 conjugates @xmath547 to @xmath567 . shishikura proved in @xcite that fatou coordinates which conjugate the map @xmath568 to @xmath569 on @xmath570 take the form @xmath571 . therefore @xmath572 , and we can write : @xmath573 @xmath574 @xmath575 call @xmath576 , @xmath577 and @xmath578 . the map @xmath579 : @xmath580 is quasisymmetric on a neighborhood of @xmath111 . define @xmath581 , and the map @xmath582 as follows : @xmath583 the map @xmath584 is the restriction to @xmath585 of a conformal map . again by shishikura @xcite the maps @xmath586 have derivatives @xmath587 , hence the map @xmath582 is a diffeomorphism ( one may take @xmath588 as a chart ) . the map @xmath589 conjugates the map @xmath547 to the map @xmath590 on @xmath543 , and to the map @xmath591 on @xmath544 . hence @xmath592 . the map @xmath593 is a diffeomorphism because it has the same analytic expression as @xmath594 , and therefore the map @xmath595 is a diffeomorphism . since the map @xmath596 is quasisymmetric on a neighborhood of @xmath111 , the inverses are quasisymmetric on a neighborhood of @xmath68 . hence the composition @xmath597 is quasisymmetric . let ( @xmath161 ) and ( @xmath162 ) be two parabolic - like mappings . we say that @xmath33 and @xmath170 are _ holomorphically equivalent _ if there exist parabolic - like restrictions ( @xmath163 ) and ( @xmath164 ) , and a biholomorphic map @xmath598 such that @xmath166 and @xmath599 by proposition [ extmap ] , the external class of every member of the family @xmath20 is given by the class of @xmath29 , hence a parabolic - like map holomorphically conjugate to a member of the family @xmath20 has external map in the class of @xmath29 . let us prove that a degree @xmath0 parabolic - like map @xmath600 with external map @xmath29 is holomorphically conjugate to a member of the family @xmath20 . let @xmath601 be an external equivalence between the maps @xmath170 and @xmath29 . let @xmath602 be the riemann surface obtained by gluing @xmath603 and @xmath604 , by the equivalence relation identifying @xmath35 to @xmath605 , i.e. @xmath606 by the uniformization theorem , @xmath602 is isomorphic to the riemann sphere . consider the map @xmath607 since the map @xmath29 is an external map of @xmath170 , the map @xmath608 is holomorphic . let @xmath609 be an isomorphisim that sends the parabolic fixed point of @xmath610 to infinity , the critical point of @xmath610 to @xmath421 , and the preimage of the parabolic fixed point of @xmath610 to @xmath15 . define @xmath611 . the map @xmath612 is a degree @xmath0 holomorphic map defined on the riemann sphere , so it is a quadratic rational function . by construction it has a parabolic fixed point of multiplier @xmath17 at @xmath613 with preimage @xmath15 , and it has a critical point at @xmath421 . hence @xmath612 belongs to the family @xmath20 . [ thm ] let ( @xmath614 ) be a parabolic - like mapping of some degree @xmath615 , and @xmath616 be a parabolic external map of the same degree @xmath2 . then there exists a parabolic - like mapping ( @xmath617 ) which is hybrid equivalent to @xmath33 and whose external class is @xmath618 $ ] . throughout this proof we assume , in order to simplify the notation , @xmath5 and @xmath4 with @xmath52 boundaries ( if @xmath5 and @xmath4 do not have @xmath52 boundaries we consider a parabolic - like restriction of ( @xmath619 ) with @xmath52 boundaries ) . let @xmath616 be a parabolic external map of degree @xmath615 , @xmath513 be its parabolic fixed point and @xmath620 be a degree @xmath2 covering extension . define @xmath621 and @xmath622 . we are going to construct now a dividing arc @xmath517\rightarrow \overline b \setminus \d$ ] for @xmath233 , such that on @xmath623 the dynamics of @xmath233 is conjugate to the dynamics of @xmath33 . let @xmath215 be an external map of @xmath33 , @xmath486 its parabolic fixed point , @xmath624 a degree @xmath2 covering extension and @xmath488 an external equivalence between @xmath33 and @xmath215 . the dividing arcs @xmath625 are tangent to @xmath101 at the parabolic fixed point @xmath486 , and they divide @xmath626 and @xmath627 in @xmath628 and @xmath629 respectively ( see section [ propex ] ) . let @xmath630 be repelling petals for the parabolic fixed point @xmath486 which intersect the unit circle and @xmath631 be fatou coordinates with axis tangent to the unit circle at the parabolic fixed point @xmath486 . on the other hand , let @xmath632 be repelling petals for the parabolic fixed point @xmath513 of @xmath233 which intersect the unit circle and @xmath633 be fatou coordinates with axis tangent to the unit circle at the parabolic fixed point @xmath513 . define @xmath634 and @xmath635 the arc @xmath636 is ( possibly after rescaling ) a dividing arc for @xmath233 . it divides the set @xmath637 into @xmath638 and @xmath639 ( with @xmath640 ) and the set @xmath641 into @xmath642 and @xmath643 ( with @xmath644 ) . define the map @xmath645 as follows : @xmath646 by proposition [ arcprop](2 ) the map @xmath647 is a quasisymmetric conjugacy between @xmath648 and @xmath649 . let @xmath57 be the parabolic fixed point of @xmath33 , and define the map @xmath650 as follows : @xmath651 the map @xmath650 is an orientation preserving homeomorphism , real - analytic on @xmath652 , which conjugates the dynamics of @xmath33 and @xmath233 . let @xmath653 be an orientation preserving @xmath52-diffeomorphism coinciding with @xmath100 on @xmath654 ( it exists because both @xmath5 and @xmath637 have smooth boundaries ) . the set @xmath664 is a quasicircle , since it is a piecewise @xmath52 closed curve with non - zero interior angles . indeed , @xmath665 and @xmath500 form a positive angle since they are separated by at least one attracting petal , and we can assume the angles between @xmath660 and @xmath666 to be positive ( we may take parabolic - like restrictions ) . the same argument shows that @xmath667 is a quasicircle . let @xmath668 and @xmath669 be riemann maps , and let @xmath670 and @xmath671 be their inverse maps . by the carathodory theorem the maps @xmath672 and @xmath673 extend continuously to the boundaries , and since @xmath664 and @xmath667 are quasicircles , the restrictions @xmath674 and @xmath675 are quasisymmetric . define the map @xmath676 as follows : @xmath677 the map @xmath676 is quasisymmetric , because the extensions of @xmath678 and @xmath679 to the unit circle are quasisymmetric , @xmath488 is conformal , the map @xmath657 is a @xmath52-diffeomorphism and by proposition [ arcprop](2 ) the map @xmath680 is quasisymmetric . hence it extends by the douady - earle extension ( see @xcite ) to a quasiconformal map @xmath681 which is a real - analytic diffeomorphism on @xmath682 . thus @xmath683 is a quasiconformal map between @xmath684 and @xmath685 , which is a real - analytic diffeomorphism on @xmath501 , and which coincides with @xmath686 on @xmath660 and with @xmath661 on @xmath662 . let us define @xmath687 . on the other hand define @xmath688 , @xmath689 , @xmath690 . consider @xmath691 define @xmath692 , and @xmath693 . the map @xmath694 is a degree @xmath2 proper and quasiregular map which coincides with @xmath33 on @xmath695 . define @xmath696 , @xmath697 and @xmath698 . then @xmath699 is a parabolic - like restriction of @xmath700 , and @xmath701 . set @xmath702 , and @xmath703 . let @xmath704 be an orientation preserving @xmath52-diffeomorphism coinciding with @xmath657 on @xmath705 , and let @xmath706 be a lift of @xmath707 to @xmath233 . let us start by constructing a quasiconformal map @xmath712 between @xmath713 and @xmath714 which agrees with @xmath100 on @xmath715 , with @xmath657 on @xmath81 and with @xmath716 on @xmath717 . the sets @xmath718 and @xmath719 are quasicircles , since they are piecewise @xmath52 closed curves with non - zero interior angles . let @xmath720 and @xmath721 be riemann maps , and let @xmath722 and @xmath723 be their inverses . by the carathodory theorem @xmath724 and @xmath725 extend continuously to the boundaries , and since @xmath718 and @xmath719 are quasicircles , @xmath726 and @xmath727 are quasisymmetric . hence the map @xmath728 defined as : @xmath729 is quasisymmetric , and it extends ( see @xcite ) to a quasiconformal map @xmath730 which is a real - analytic diffeomorphism on @xmath682 . finally , the map @xmath731 is a quasiconformal map which coincides with @xmath100 on @xmath715 , with @xmath657 on @xmath81 and with @xmath716 on @xmath717 . moreover the map @xmath732 is a real - analytic diffeomorphism on @xmath733 . by the measurable mapping theorem , there exists a quasiconformal map @xmath736 such that @xmath737 . let @xmath738 let us call @xmath739 , @xmath740 , @xmath741 and @xmath742 . then ( @xmath617 ) is a parabolic - like map hybrid equivalent to @xmath33 . indeed , since @xmath743 , @xmath744 and @xmath699 is a parabolic - like restriction of @xmath745 , the map @xmath168 is a quasiconformal conjugacy between @xmath33 and @xmath170 , and @xmath746 on @xmath149 by construction . if @xmath149 is connected , define the quasiconformal map @xmath747 as follows : @xmath748 then the quasiconformal map @xmath749 is an external equivalence between @xmath170 and @xmath233 , since by construction on @xmath750 @xmath751 , and @xmath601 is holomorphic ( indeed @xmath752 ) . if @xmath149 is not connected , let @xmath753 be a full relatively compact connected subset of @xmath754 , containing @xmath755 , the critical values of @xmath711 and such that ( @xmath756 ) is a parabolic - like restriction of @xmath757 . call @xmath758 . define the map @xmath759 as follows : @xmath760 let @xmath761 be a full relatively compact connected subset of @xmath762 containing @xmath763 , the critical values of @xmath170 and such that @xmath764 is a parabolic - like restriction of @xmath765 . call @xmath766 . then the map @xmath767 is an external equivalence between @xmath170 and @xmath233 ( cfr . lemma [ lem ] ) . let @xmath770 be a hybrid equivalence between @xmath33 and @xmath170 , and @xmath771 an external equivalence between @xmath33 and @xmath170 . let @xmath772 be an external map of @xmath33 constructed from the riemann map @xmath773 . let @xmath774 be a topological disc compactly contained in @xmath775 and such that @xmath776 is compactly contained in @xmath777 , and @xmath778 is compactly contained in @xmath637 . set @xmath779 . the map @xmath780 restricts to an external equivalence between @xmath170 and @xmath233 . by construction the map @xmath784 conjugates the maps @xmath33 and @xmath170 conformally on @xmath774 and quasiconformally with @xmath785 on @xmath786 . we want to prove that the map @xmath787 is holomorphic . by rickmann lemma ( see below ) @xmath787 is holomorphic if @xmath787 is continuous . thus we just need to prove that it is continuous . define @xmath788 and @xmath789 . the restriction @xmath790 is proper holomorphic and of degree @xmath2 . the map @xmath791 is a quasi - conformal homeomorphism ( into its image ) which autoconjugates @xmath233 on @xmath792 . setting @xmath793 , @xmath794 , @xmath795 , and applying the strong reflection principle with respect to the unit circle , we obtain a quasiconformal homeomorphism ( into its image ) @xmath796 , which autoconjugates @xmath233 on @xmath797 . thus the restriction @xmath798 is a quasisymmetric autoconjugacy of @xmath233 on the unit circle . since the preimages of the parabolic fixed point @xmath69 are dense in @xmath101 , an autoconjugacy of @xmath233 on the unit circle is the identity . therefore @xmath799 . since the map @xmath800 is a quasiconformal homeomorphism which coincides with the identity on @xmath101 , the hyperbolic distance between a point near @xmath101 and its image is uniformly bounded , i.e. @xmath801 and @xmath802 such that : @xmath803 since @xmath488 and @xmath184 are isometries , we obtain @xmath804 then @xmath805 and @xmath806 converge to the same value as @xmath35 converges to @xmath807 , i.e. @xmath808 extends continuously to @xmath809 by @xmath810 . thus @xmath787 is continuous . the result follows by the rickmann lemma ( for a proof of the rickmann lemma we refer to @xcite , lemma @xmath0 pg . 303 ) : _ * rickmann * _ let @xmath39 be open , @xmath811 be compact , @xmath40 and @xmath812 be two maps which are homeomorphisms onto their images . suppose that @xmath168 is quasi - conformal , that @xmath787 is quasi - conformal on @xmath813 and that @xmath814 on @xmath815 . then @xmath787 is quasiconformal and @xmath816 almost everywhere on @xmath815 . since @xmath25 and @xmath820 are connected , the external conjugacies between @xmath24 and @xmath821 respectively and @xmath29 can be extended to the discs @xmath822 and @xmath823 ( see proposition [ extmap ] ) , i.e. there exist holomorphic conjugacies @xmath824 and @xmath825 between @xmath24 and @xmath821 respectively and @xmath29 . therefore @xmath826 is a holomorphic conjugacy between @xmath24 and @xmath821 . let @xmath745 and @xmath827 be parabolic - like restrictions of @xmath24 and @xmath821 respectively , and let @xmath770 be a hybrid equivalence between them . define the map @xmath828 as follows : @xmath829 the proof of proposition [ hol ] shows that the map @xmath830 is holomorphic , hence it is a mbius transformation . since @xmath787 conjugates @xmath24 and @xmath821 , it fixes the parabolic fixed point @xmath613 and its preimage @xmath15 , and it can fix or interchange the critical points @xmath69 and @xmath421 . hence either @xmath787 is the identity or @xmath831 is the map @xmath832 . therefore @xmath833= \ { p_a,\,\,p_{-a}\}$ ] , and finally @xmath819 . 13 l. ahlfors , lectures on quasiconformal mappings , second edition . ams university lecture series , vol . a. douady & c. j. earle , conformally natural extension of homeomorphisms of the circle , _ acta math._,169 ( @xmath834 ) , no . 3 - 4 , 229 - 325 . a. douady & j. h. hubbard , on the dynamics of polynomial - like mappings , _ ann . sci . cole norm . sup._,(4 ) , vol.@xmath835 ( @xmath836 ) , 287 - 343 . j. hubbard , _ teichmller theory , volume 1 : teichmller theory _ , matrix editions , ( 2006 ) . l. l. a. lomonaco , _ parabolic - like mappings . ph.d .- thesis _ , roskilde universitet , ( 2012 ) . j. milnor , _ dynamics in one complex variable _ , annals of mathematics studies , ( 2006 ) . m. shishikura , bifurcation of parabolic fixed points , _ the mandelbrot set , theme and variations , _ ( 325 - 363 ) , _ london math . lecture note ser . , 274 _ cambridge univ . press , ( 2000 ) . |
motivated by recent experiments , we consider a generic spin model in the basis for the hyperhoneycomb and harmonic - honeycomb iridates . based on microscopic considerations , the effect of an additional bond - dependent anisotropic spin exchange interaction ( ) beyond the heisenberg - kitaev model is investigated we obtain the magnetic phase diagrams of the hyperhoneycomb and harmonic - honeycomb ( ) lattices via a combination of the luttinger - tisza approximation , single- variational ansatz , and classical monte carlo simulated annealing . the resulting phase diagrams on both systems show the existence of incommensurate , non - coplanar spiral magnetic orders as well as other commensurate magnetic orders . the spiral orders show counter - propagating spiral patterns , which may be favorably compared to recent experimental results on both iridates . the parameter regime of various magnetic orders and ordering wavevectors are quite similar in both systems . we discuss the implications of our work to recent experiments and also compare our results to those of the two dimensional honeycomb iridate systems . | 1407.4125 | iridium - based compounds@xcite have spurred considerable interest because the interplay of strong atomic spin - orbit coupling ( soc ) , electronic correlations , crystal field effects , and sizable orbital overlaps in these compounds have been shown to give rise to exotic states of matter ranging from topological insulators to weyl semimetals and quantum spin liquids.@xcite drawing significant attention is the a@xmath4iro@xmath5 family of layered honeycomb iridates@xcite : these materials were first described by the kitaev model@xcite and later by the heisenberg - kitaev ( hk ) model@xcite , both of which hosts a @xmath6 quantum spin - liquid ground state first discovered in the context of kitaev s exactly solvable spin-1/2 model.@xcite subsequently , similar exactly solvable spin models on several other two - dimensional ( 2d ) and three - dimensional ( 3d ) lattices were investigated.@xcite experimentally , the sodium honeycomb iridate na@xmath4iro@xmath5has been shown to order magnetically in the zigzag phase@xcite , while the lithium counterpart li@xmath7iro@xmath8is believed to also order magnetically but in a different ground state.@xcite the quest to find an appropriate minimal model that can explain the observed orders , low energy excitations , as well as bulk thermodynamic properties has been actively pursued . to address these points , localized pseudospin models involving exchanges in addition to the heisenberg and kitaev couplings have been considered , including the symmetry - allowed off - diagonal exchange@xcite , further neighbour exchanges@xcite , and distortion - induced anisotropies@xcite . electronic band structure calculations@xcite have also been investigated to further our understanding of these compounds . indeed , a suitable minimal model may predict the behaviour of these layered honeycomb iridates under a variety of perturbations and may direct us to the discovery of exotic physics , including the illusive quantum spin liquid phase . recently , two three - dimensional polymorphs of the layered honeycombs with chemical formula li@xmath7iro@xmath8 have been discovered , fuelling intrigue over the family of honeycomb - based iridate compounds . curiously , these two materials the hyperhoneycomb ( @xmath9)@xcite and harmonic - honeycomb ( @xmath2)@xcite behave rather similarly : they have the same stoichiometry as the layered honeycomb with analogous three - dimensional crystal structures , they order magnetically with no net moment at similar temperatures ( @xmath10 ) , and their magnetization curves behave non - linearly at low fields . furthermore , thermodynamic measurements@xcite suggested that both 3d iridates might order unconventionally in a non - collinear fashion and recent single - crystal resonant magnetic x - ray diffraction experiments@xcite find a complex incommensurate structure with non - coplanar and counter - propagating moments in both iridates . indeed , with the discovery of these two 3d honeycomb compounds and to build on past work in both the intermediate - coupling regime@xcite and localized@xcite limits , the search for a relevant minimal model to describe the family of honeycomb iridates has now become more fascinating . can the same minimal model describe both the 2d and 3d honeycomb iridates while capturing the similarities and heterogeneity amongst these compounds ? in this work , driven by the fact that these compounds have similar local structures , we suggest that the nearest - neighbor @xmath11 model , which was recently used to describe the layered honeycomb iridates@xcite , is a good starting point in the discussion of this family of compounds , especially in regards to the magnetic order . the rest of the paper is organized as follows . we begin , in sec . [ sec : lattices ] , by discussing the hyperhoneycomb and @xmath2 lattices , drawing special attention to the similarities and differences between these three - dimensional analogues of the honeycomb lattice . in sec . [ sec : model ] , we describe the minimal effective model of @xmath0 pseudospins that describes the low - energy physics of the hyperhoneycomb and @xmath2 lattices in the mott insulating limit . the presence of the off - diagonal exchange term , @xmath1 , along with the familiar heisenberg and kitaev terms , is shown to be necessary from a strong - coupling expansion of the underlying , multi - band electronic model . following this , in sec . [ sec : pd ] we present the classical phase diagram of these models using a combination of luttinger - tisza approximation , single-@xmath3 ansatz minimization , and simulated annealing . we show that the hyperhoneycomb and @xmath2 lattices have similar phase diagrams , but differ from the phase diagram of the 2d honeycomb iridates in important ways . in addition to analogues of the ferromagnetic , antiferromagnetic , zigzag , and stripy phases of the hk model , we find a number of non - coplanar , counter - propagating spiral phases , multiple-@xmath3 states , as well as phases that have no direct analogues in the hk - limit . we expound these phases and their static structure factors in sec . [ sec : orders ] , thus providing results that can be compared with experiments . finally , in sec . [ sec : discussion ] , we discuss the relevance of our results to the newly discovered 3d - li@xmath7iro@xmath8 compounds . the ideal iridium ion ( ir ) network in the hyperhoneycomb ( fig . [ fig : hhc_lattice ] ) and @xmath2 ( fig . [ fig : har_lattice ] ) lattices share many common features but also have some important differences . in this section , we highlight these main points and introduce conventions relevant to the description of the magnetic phases explored in sec . [ sec : orders ] . the conventional unit cells of both hyperhoneycomb and @xmath2 lattices are orthorhombic with the former being face - centered@xcite and the latter being base - centered@xcite . the hyperhoneycomb primitive unit cell has four ir while the @xmath2 primitive unit cell has eight ; hence , the conventional unit cell of hyperhoneycomb contains four primitive units while in the @xmath2 , there are two . in our chosen coordinate system , the conventional lattice vectors of both hyperhoneycomb and @xmath2 lattices are given by the orthorhombic set @xmath12 , @xmath13 , and @xmath14 . all the ir sites reside in their respective oxygen octahedra and these octahedra share edges between nn ir sites . the ir sites are three - fold coordinated , with the three nn bonds labelled as @xmath15 , @xmath16 , and @xmath17 as illustrated in figs . [ fig : hhc_lattice ] and [ fig : har_lattice ] . in the coordinate system that we have chosen , the label represents the direction to which the bond is normal . for example , the sites joined by an @xmath15-bond form either the vector @xmath18 or @xmath19 . we will see that this labelling scheme is also used in describing the low - energy hamiltonian of these lattices in sec . [ sec : model ] . the hyperhoneycomb ( also known as the @xmath9 ) and @xmath2 lattices have been described as strips of layered honeycombs joined in an abab fashion along the @xmath20 direction . the a and b strips are mirrored copies of each other and are hence related by glide planes.@xcite in the @xmath21 notation , @xmath22 refers to the number of rows of hexagons in the each of the honeycomb strips . for the hyperhoneycomb lattice , @xmath23 indicates that no complete hexagons are present . for the @xmath2 lattice , each honeycomb strip is one row in width . despite the difference in the topology of the ir network , the projection along the conventional planes ( @xmath24 , @xmath25 , and @xmath26 planes ) are identical between the two lattices . we define a _ honeycomb - plane _ associated with every site : it is the plane that is spanned by the site s three nn bonds . alternatively , it can be defined as the plane that contains the honeycomb strip to which the site belongs . since the honeycomb strips are related by glide planes , there are only two orientations that honeycomb - planes can take . the honeycomb - planes are important when discussing the magnetic orders in sec . [ sec : orders ] , especially for the spiral phases . all ir are symmetry - equivalent in the hyperhoneycomb lattice , but in the @xmath2 lattice , there are two inequivalent sets . these two inequivalent sets of ir in the @xmath2 lattice can be classified by considering whether the site shares the same honeycomb - plane with all its nns . we denote the sites that do nt share the same honeycomb - plane with all its nns as _ bridging - sites _ since they connect the a and b honeycomb strips . consequently , we call the @xmath17-bonds that connect bridging - sites _ bridging-@xmath17-bonds _ , while the other symmetry - inequivalent @xmath17-bonds _ non - bridging-@xmath17-bonds_. this terminology is used in the discussion of the spin patterns in several of the magnetic orders . in compounds containing @xmath28 ions residing in oxygen octahedral crystal fields , the combination of large atomic spin - orbit coupling and strong electronic correlations can result in low - energy physics describable by highly anisotropic @xmath0 pseudospin models ( hereafter , we will use _ spin _ and _ pseudospin _ interchangeably for brevity).@xcite when the bonds are comprised of edge - shared octahedra , such as the nn bonds found in both the hyperhoneycomb and the @xmath2 lattices , a symmetric off - diagonal exchange term denoted by @xmath1is generically present in addition to the heisenberg and kitaev exchange terms.@xcite as such , a minimal hamiltonian that describes the low - energy physics of these systems can be written as @xmath29,\ ] ] where @xmath30 is the @xmath0 pseudospin at site @xmath31 , the summation is over nn bonds @xmath32 labelled by @xmath33 , and @xmath34 is shorthand for @xmath35 . the @xmath36 sign in front of @xmath1 is a reminder that , unlike the @xmath37 and @xmath38 terms , the @xmath1 term can have relative minus signs on different bonds . the sign structures of @xmath1 for both the hyperhoneycomb and @xmath2 lattices are shown in fig . [ fig : sign_lattice ] ( see appendix [ app : sign ] for details on the choice of the @xmath1 sign structure ) . the exchange interactions @xmath37 , @xmath38 , and @xmath1 can be obtained microscopically via a strong coupling expansion of a @xmath39 multi - band tight - binding model with on - site interactions and large atomic soc . the details can be found in appendix [ app : model ] and ref . . here we summarize the exchange interactions dependence on the parameters of the multi - band hubbard model , which include coulomb repulsion @xmath40 , hund s coupling @xmath41 , spin - orbit coupling @xmath42 , and three @xmath39 hopping amplitudes denoted by @xmath43 through @xmath44 @xmath45\nonumber \\ k&= \frac{32j_h}{9}\left [ \frac{(t_1-t_3)^2 - 3t^2_2}{(2u+3\lambda)^2}\right],\ \gamma= \frac{64j_h}{9 } \frac{t_2(t_1-t_3)}{(2u+3\lambda)^2}.\end{aligned}\ ] ] we note that all three exchange interactions are present at this order of perturbative expansion and that the presence of hund s coupling @xmath41 is needed for a finite @xmath38 and @xmath1 . furthermore , the signs of the exchange interactions sensitively depend on relative magnitudes and signs of @xmath43 through @xmath44 , which themselves are functions of slater - koster amplitudes@xcite that parametrize @xmath39 orbital overlaps and oxygen - iridium orbital overlaps ( details can be found in appendix [ app : model ] and ref . ) . generically , this sensitivity implies that all the phases we find are potentially experimentally relevant to the 3d - li@xmath7iro@xmath8 compounds . the nn @xmath0 pseudospin hamiltonian in eq . [ eq : ham ] describes the local , low - energy physics of the hyperhoneycomb and @xmath2 lattices in the strong soc and electron correlation limit . unlike the 2d honeycomb iridates , both the hyperhoneycomb and the @xmath2 lattices are three - dimensional , thus making numerical techniques for quantum spin models , such as exact diagonalization and density matrix renormalization group , challenging to implement this is further exacerbated by the large unit cells in these 3d lattices : compared to the two ir ions in the 2d - honeycomb lattice s primitive unit cell , there are four and eight ir ions in the primitive unit cell for the hyperhoneycomb and @xmath2 lattices , respectively . although these numerical issues in the quantum analysis are difficult to overcome , a classical understanding of these three - dimensional pseudospin models is both important and relevant insomuch that the hyperhoneycomb and @xmath2 iridates are believed to be magnetically - ordered.@xcite a classical analysis may be sufficient since , in the 2d - honeycomb pseudospin model , the classical phase diagram well - approximated the magnetic phases found via exact diagonalization.@xcite furthermore , quantum fluctuations the key ingredient in generating exotic , non - classical phases are generally believed to be suppressed in 3d systems compared to 2d systems . to this end , we analyze the @xmath0 pseudospin hamiltonians of both hyperhoneycomb and @xmath2 lattices in the classical limit using a combination of the luttinger - tisza approximation@xcite , single-@xmath3 ansatz minimization , and classical monte carlo simulated annealing . in the luttinger - tisza approximation ( lta),@xcite the local constraint of @xmath46 is relaxed to a global constraint @xmath47 . the lowest energy configuration satisfying the global constraint can be easily solved in momentum space . this configuration is then checked against the local constraint : if it is satisfied , then the lta succeeded in finding the exact classical ground state ; otherwise , the lta failed but has instead found a lower bound of the ground state energy . in the parameter regions where lta failed , we implemented a combination of single-@xmath3 ansatz minimization and classical monte carlo simulated annealing to find the classical ground state configuration . in the single-@xmath3 ansatz minimization , an upper bound to the classical energy is determined variationally by a trial spin configuration of the form @xmath48,\ ] ] where @xmath49 is the pseudospin on sublattice @xmath50 in the ( conventional ) unit cell @xmath51 , @xmath52 is an orthonormal set of vectors defining a local frame for sublattice @xmath50 , @xmath53 is the canting angle for sublattice @xmath50 , and @xmath3 is the @xmath3-vector common to all sublattices . the energy of this ansatz is minimized to find a variational bound to the classical ground state energy . since the wavevectors @xmath3 and @xmath54 ( where @xmath55 is a reciprocal lattice vector ) parametrize the same set of spin configurations , we can restrict the domain over which minimization of @xmath3 is performed to the first brillouin zone . classical simulated annealing using the single - spin metropolis algorithm and periodic boundary conditions was performed in conjunction with the single-@xmath3 ansatz minimization . simulated annealing has the advantage over single-@xmath3 ansatz minimization in that it can access states that are characterized by multiple @xmath3-vectors and is less prone to being trapped by local minima in energy . on the other hand , although simulated annealing can access single-@xmath3 states given by eq . [ eq : singleqansatz ] , the allowable @xmath3-vectors must be commensurate with the finite system sizes chosen in our simulations . a saw - tooth like temperature annealing profile was used to better traverse the spin - configuration space and a minimum of @xmath56 sweeps ( updates per spin ) were used for each simulation . over 8000 @xmath57 parameter points for each lattice were analyzed , with particular focus on spiral phases . for each parameter point , simulated annealing was performed with random initial conditions on systems with @xmath58 conventional unit cells where @xmath59 . for the spiral regions , we performed additional annealing on systems with @xmath60 conventional unit cells , where @xmath61 is the direction along the spiral wavevector . in addition , a minimum of @xmath62 single-@xmath3 ansatz minimization runs were also performed per parameter point . the minimum energy amongst these simulated annealing and single-@xmath3 ansatz minimization runs is deemed the variational bound of the ground state energy and the corresponding pseudospin configuration is used to characterize the magnetic order of the ground state at that parameter point . before delving into the details of the classical magnetic ground states of eq . [ eq : ham ] , we first examine some general features of the hyperhoneycomb and @xmath2 phase diagrams , which were obtained via the methods outlined in sec . [ subsec : methods ] . to fix the overall energy scale , we parametrize the @xmath57 parameter space using an angular representation @xmath63 such that @xmath64 . the phase diagrams are plotted as polar plots , where the angular component is given by @xmath65 and the radial component , @xmath66 , is given by @xmath67 . in figs . [ fig : pd_hhc_pos ] and [ fig : pd_har_pos ] , we show the phase diagrams for the hyperhoneycomb and @xmath2 lattices , respectively , when @xmath68 ( _ i.e. _ @xmath69 and @xmath70 $ ] ) ; the @xmath71 results can be obtained by applying time - reversal on the odd sublattice pseudospins , which transforms @xmath72 and can be seen in fig . [ fig : pds_neg ] . prior to sec . [ subsec : mgamma ] , we will concentrate on the @xmath73 case with the understanding that equivalent statements can be made for the @xmath74 case . important properties of the @xmath71 phases especially those relating to experimental results will be discussed in sec . [ subsec : mgamma ] . at first glance , we note the striking similarities between the two phase diagrams : despite the different topology of the hyperhoneycomb and @xmath2 lattices , the parameter regimes where we find the various magnetic orders and phase boundaries are similar in both systems . this notable result emphasizes the commonalities shared between the two systems as far as local physics is concerned . on the other hand , when compared to the classical phase diagram of the 2d honeycomb iridate from ref . , we notice that the zigzag region has reduced in size and the spiral phases have become more prevalent . in particular , the 120@xmath75 phase of the 2d model is now a spiral phase in the two 3d models . on the outer edges of the two phase diagrams , @xmath1 vanishes and the hamiltonians reach the isotropic hk limit . with increasing @xmath65 , we encounter the nel ( hk - af ) , skew - zigzag ( hk - sz ) , ferromagnet ( hk - fm ) , and skew - stripy phases ( hk - ss ) , which were discussed in the context of the 2d - honeycomb and 3d - honeycomb iridates.@xcite when @xmath76 , the classical ground state manifold possesses an emergent @xmath77 symmetry despite the presence of soc . however , both classical and quantum order - by - disorder ( obd ) lift this emergent degeneracy and choose particular spatial directions for the moments.@xcite indeed , finite @xmath1 also breaks the emergent symmetry@xcite and pins pseudospin moments in particular spatial directions , causing the phases on the inside of the phase diagram to be non - collinear in general . in the hk limit , the four - sublattice transformation@xcite exactly maps the parameter point @xmath78 to @xmath79 even in the quantum limit for both the hyperhoneycomb@xcite and @xmath2@xcite systems , thus relating the left and right edges of the phase diagram . in the presence of finite @xmath1 , however , the four - sublattice transformation is not longer useful ( it does not map a parameter point with finite @xmath1 to another point within the phase diagram ) ; the left and right _ interiors _ of the phase diagrams are no longer related . the centers of the phase diagrams are the @xmath1-only points . these points , much like in the 2d - honeycomb case , possess highly degenerate ground state manifolds.@xcite therefore , similar to the kitaev points , small perturbations can drive different magnetic orders . as a result , we see that a number of phases converge at these highly - degenerate points . the effects of a finite @xmath1 is most pronounced near the kitaev points where the highly degenerate classical ground state manifolds can be lifted by small perturbations . as a result , finite @xmath1 in these regions generates new magnetic orders that are not continuously connected with those found along the hk limit . of the eight new magnetic orders , four are incommensurate spiral phases ( sp@xmath80 , sp@xmath81 , sp@xmath82 , and sp@xmath83 ) , two are multiple-@xmath3 states ( mq@xmath84 and mq@xmath4 ) , one is an antiferromagnetic state ( af - sz@xmath85 ) , and one has finite net moment ( fm - sz@xmath85 ) . of the four spiral states , two possess @xmath3-vectors that vary in length depending on the ratio of the exchange parameters while all of them contain counter - propagating spiral patterns ( see sec . [ sec : orders ] for details ) . away from the kitaev points , finite but small @xmath1 modifies existing magnetic orders found in the hk limit . several of the existing phases namely the hk - sz , hk - fm , and hk - ss phases continuously deform with increasing @xmath1 to acquire additional spin textures . the details of these additional spin textures depend on the particular lattice in consideration , which we will outline in sec . [ sec : orders ] . in this section , we provide detailed descriptions of the magnetic phases found in our phase diagrams . we first discuss the new magnetic orders , then we will describe the phases that are continuously connected with those at the hk limit . we also present the static structure factors of the spiral phases , which are relevant to x - ray and neutron scattering experiments . a summary of the phases discussed in this section can be found in table [ tbl : orders ] . [ [ spiral - orders ] ] spiral orders : + + + + + + + + + + + + + + the spiral phases are those that can be represented by the ansatz in eq . [ eq : singleqansatz ] with incommensurate @xmath3-vectors . they are located where the kitaev exchange dominates over the heisenberg exchange and where the two exchanges compete and have opposite signs . these spiral regions are connected to both the kitaev and the pure-@xmath1 points and border the @xmath86 line . the above description of the location of spirals states is valid for both @xmath73 and @xmath74 . there are three spiral phases shared by both lattices : sp@xmath81 , sp@xmath80 , and sp@xmath82 . in addition , sp@xmath83 is found in the hyperhoneycomb lattice in a small region near the sp@xmath80 phase . all these spiral phases do not have a net moment and they are all non - coplanar . in the fm heisenberg - af kitaev region ( quadrant ii of fig . [ fig : pds ] ) , we find the sp@xmath80 phase . this phase has a @xmath3-vector of the form @xmath87 ( in the orthorhombic hkl notation ) , signifying that the spirals run along the @xmath88 direction ; therefore , we label this state as sp@xmath80 . in the hyperhoneycomb lattice , there is also the nearby sp@xmath83 phase , which has a @xmath3-vector of the form @xmath89 . in the af heisenberg - fm kitaev region ( quadrant iv ) , there are two spiral phases : the smaller region has a @xmath3-vector of the form @xmath87 while the larger region has @xmath89 . we label the former sp@xmath81 and the latter sp@xmath82 . we illustrate an example of the real - space spin configuration for these phases in figs . [ fig : realspace_a ] and [ fig : realspace_b ] using relevant projections that highlight the spin spirals . for the sp@xmath90 phases , the spins have been projected on to the @xmath25 plane while , for the sp@xmath91 phases , the projection is on to the @xmath26 plane . although we have chosen parameter points that give commensurate @xmath3-vectors , we note that generically the phases are incommensurate spiral phases . prior to expounding the features of these spiral states , we emphasize that the quoted @xmath3-vectors are those that would appear in the single-@xmath3 ansatz of eq . [ eq : singleqansatz ] , which are defined up to reciprocal lattice vectors of the conventional brillouin zone . we stress that these @xmath3-vectors are related , but not necessarily the same as , the peak positions of the static structure factor , which we denote by @xmath92 . the static structure factor is given by @xmath93 where @xmath30 is the spin at site @xmath31 of the ground state configuration and @xmath51 is its real - space position . in the second line , we summed over the unit cells but retained the sublattice indices such that the static structure factor is written in terms of the matrix spin - spin structure factor @xmath94 and relative sublattice positions @xmath95 . in this second form , we see explicitly that single-@xmath3 ansatz can have peaks at @xmath96 where @xmath55 is a reciprocal lattice vector of the conventional brillouin zone ( one of the form @xmath97 where all indices are integers ) . however , because of the form factor @xmath98 and the matrix structure of @xmath99 , not all peaks will have the same intensity and some may even be extinguished . in figs . [ fig : sf_a ] and [ fig : sf_b ] , we show the static structure factors along relevant cuts in reciprocal space which highlights the difference between @xmath3 and @xmath92 . although the shown sp@xmath90 phases have the same @xmath100 , sp@xmath80 possesses a peak at @xmath101=(0.66,0,0)$ ] while sp@xmath81 does not . instead , sp@xmath81 possesses a peak at @xmath102=(1.33,0,0)$ ] . likewise for the sp@xmath91 phases : sp@xmath83 and sp@xmath82 differ in that sp@xmath83 possesses a peak at @xmath103 $ ] while the closest peak along the @xmath89 direction for sp@xmath82 is located at @xmath104 $ ] . this property is generic for all @xmath3-vectors encountered in the spiral phases and hence explain our chosen notation : the spiral phase sp@xmath105 possesses the lowest - order structure factor peak at @xmath106 , where @xmath107 is short form for @xmath108 or @xmath109 depending on whether @xmath15 is @xmath50 or @xmath110 . this result is correlated with the orientation of the _ spiral - planes _ of these phases , which we will address after introducing several properties of these spiral - planes . although these are non - coplanar spirals , spins on any given sublattice all lie in the same plane , which we term _ spiral - plane _ of the sublattice / site ( this may not be evident in figs . [ fig : realspace_a ] and [ fig : realspace_b ] due to the projections ) . in other words , the non - coplanar nature of these spiral phases is due to different sublattices having different spiral - planes . specifically , sublattices that share honeycomb - planes also share spiral - planes , , the spiral - planes amongst sites in the honeycomb strips are slightly askew : bridging - sites share spiral - planes as do non - bridging - sites , but these two symmetry - inequivalent sites have spiral - planes that are slightly off - angle relative to each other ( in a manner where the @xmath111-axis still lies in the spiral - planes ) . ] whereas sublattices that do nt share honeycomb - planes have spiral - planes related by symmetry : in the sp@xmath90 phases , both the spiral - planes and spins preserve the @xmath112 rotation symmetry about the @xmath88 direction ( hereby denoted as @xmath113 ) , while in the sp@xmath82 phase , both the spiral planes and spins preserve the @xmath114 rotation symmetry . in sp@xmath83 , the spins break all @xmath115 symmetries but the spiral - planes are related by the @xmath114 rotations . in figs . [ fig : realspace_a ] and [ fig : realspace_b ] , sites that have the same spiral - planes have the same color . we have also illustrated the preserved symmetry operations . despite possessing the same @xmath3-vector for each sublattice as evident in eq . [ eq : singleqansatz ] , nns that share spiral - planes have spirals that propagate with different handedness , which can be readily verified in figs . [ fig : realspace_a ] and [ fig : realspace_b ] . in other words , the angles formed by nn spins that share spiral - planes are not constant as the spiral propagates we say that the two spirals _ counter - propagate_. we note that the corresponding spiral phases in the @xmath74 regime also possess counter - propagating spirals . this is because the time - reversal operation on odd sublattices leaves the handedness of all spirals invariant . the orientations of the spiral - planes are related to the honeycomb planes and distinguish the @xmath116 from the @xmath117 phases . they are also correlated with the positions of the peaks in the structure factors . in both the sp@xmath82 and sp@xmath81 phase , the spiral - planes at each site is aligned with the honeycomb - plane of that site . on the other hand , in the sp@xmath80 and sp@xmath83 phases , the spiral - planes at each site is aligned with the other honeycomb - plane of the lattice .- axis still lies in the spiral - planes ) . the angle between a spiral - plane and its related honeycomb - plane is a monotonic function of @xmath118 : in the sp@xmath80 phase , the angle is approximately @xmath119 when @xmath120 is near @xmath121 and increases as @xmath120 moves towards @xmath122 . in the sp@xmath82 phase , the angle increases from approximately @xmath119 when @xmath120 is closest to @xmath119 and increases as @xmath120 decreases towards @xmath123 . ] this implies that states that have structure factor peaks at @xmath106 have honeycomb - planes aligned / not aligned with spiral planes . in fig . [ fig : realspace_a ] and [ fig : realspace_b ] , we illustrate the alignment of the spiral - planes and the honeycomb - planes by the hue of the sites : shades of blue indicates that the site s spiral - plane and honeycomb - plane are in alignment , shades of red indicate otherwise . the lengths of the @xmath3-vectors are illustrated as a guide for the eye by color contours in fig . [ fig : pds ] : the @xmath3-vectors lengthen from the lightest to the darkest of colors within each phase . in the sp@xmath81 and sp@xmath83 phases , the @xmath3-vectors are approximately constant at @xmath124 and @xmath125 , respectively . on the other hand , the @xmath3-vector varies continuously within the sp@xmath80 and sp@xmath82 phases . in the sp@xmath80 phase of the hyperhoneycomb lattice , the @xmath3-vector varies between approximately @xmath126 and @xmath127 while in the corresponding phase of the @xmath2 lattice , the @xmath3-vector varies between approximately @xmath128 and @xmath129 . the majority of the parameter space within the sp@xmath80 of both lattices has @xmath100 , which are the states shown in fig . [ fig : realspace_a ] . in both lattices , the @xmath100 states are located centrally within the phase and border the @xmath86 line . the longer @xmath3-vectors are found in a narrow region bordering the fm@xmath130 phase while the shorter @xmath3-vectors occupy a small region towards the af kitaev point . in the sp@xmath82 of the hyperhoneycomb lattice , the @xmath3-vector ranges approximately from @xmath131 to @xmath132 while for the corresponding phase in the @xmath2 lattice , the @xmath3-vector range is smaller : approximately @xmath133 to @xmath134 . for this phase in both lattices , the majority of the parameter space possesses the @xmath3-vector @xmath125 and these are the states shown in fig . [ fig : realspace_b ] . the @xmath125 states are located centrally within the phase . the longer @xmath3-vectors are localized near the fm kitaev point , while the shorter @xmath3-vectors border the ss@xmath135 phase . [ [ mq_1 ] ] mq@xmath84 : + + + + + + + + + + + + + + + + + + + in quadrant iii of the phase diagram where both heisenberg and kitaev exchanges are ferromagnetic , there are two multiple-@xmath3 states that can not be expressed as a single-@xmath3 ansatz given by eq . [ eq : singleqansatz ] . the first of which is a non - coplanar state near the fm kitaev point and borders both the fm@xmath130 and sp@xmath82 phases . it has significant fourier components in @xmath136 for multiple values of @xmath15 and does not have a finite net moment . [ [ mq_2 ] ] mq@xmath4 : + + + + + + + + + + + + + + + + + + + the second of the two multiple-@xmath3 states lies closer to the pure-@xmath1 point than the first multiple-@xmath3 state . it also borders both the fm@xmath130 and the sp@xmath82 phases and is a non - coplanar state . this phase has significant fourier components in @xmath137 , @xmath138 , as well as @xmath139 , the last of which indicates that it possesses a finite net moment , which lies in the @xmath88 direction . a more detailed analysis of these multiple-@xmath3 states described here are left for future study . [ [ af - sz_af ] ] af - sz@xmath140 : + + + + + + + + + + + + + + + + + + + + + + + + + bordering the af@xmath141 phase is this coplanar phase with vanishing net moment , which is composed of antiferromagnetic chains along the skew - zigzag directions . in the case of the hyperhoneycomb lattice , this phase can be verified via lta as the exact ground state . this phase is connected to the af kitaev point and also partially borders the sp@xmath81 phase . in both lattices , the projections of the spins along the @xmath20 direction vanishes . in the hyperhoneycomb lattice , the af chains along the skew - zigzag directions are collinear and are nearly parallel to the honeycomb - plane of the skew - zigzag . the angle of deviation between the collinear spins and the honeycomb - plane increases as one moves aways from the af@xmath141 phase . this trend can be captured by the magnitude of the projection of the spins in the @xmath88 , which is depicted as colour contours in fig . [ fig : pds ] : the larger the projection along @xmath88 , the darker the color . in the @xmath2 lattice , the spins along the skew - zigzag directions are not perfectly collinear , but nevertheless the magnitude of the spin projections in the @xmath88 direction also follows the same trend . furthermore , the non - bridging-@xmath17-bonds are antiferromagnetically correlated . in both lattices , the phase preserves the @xmath142 rotation symmetry . [ [ fm - sz_af ] ] fm - sz@xmath140 : + + + + + + + + + + + + + + + + + + + + + + + + + this coplanar phase only exists in the @xmath2 lattice and is the version of the af - sz@xmath140 phase with a finite net moment . this phase is bordered by the sp@xmath80 , sp@xmath81 , and af - sz@xmath140 phases in the af - heisenberg regime and is connected to both the af kitaev and pure-@xmath1 points . like the af - sz@xmath140 phase of the @xmath2 lattice , spins have vanishing projections along the @xmath20 direction , they form near - collinear af chains along the zigzag directions , and the @xmath142 rotation symmetry is preserved . these af chains become progressively less collinear as one approaches the @xmath86 line . unlike the af - sz@xmath140 phase , however , the non - bridging-@xmath17-bonds are ferromagnetically correlated . as a result , despite being in the af - heisenberg regime , there is a net moment in the @xmath88 direction . the size of the net moment is small , especially compared to the fm@xmath130 phase , and decreases as we approach the sp@xmath81 phase or the af kitaev point . we illustrate this decrease in net moment by the lightening of the color contours . [ [ af_c ] ] af@xmath143 : + + + + + + + + + + + + + + + + + + + + + this collinear phase is the exact ground state in the af - heisenberg region when @xmath144 and @xmath145 : the lta succeeds in finding this exact ground state for both lattices . this phase is the hk - af state with moments aligned antiferromagnetically and locked in the @xmath20 direction due to the presence of @xmath1 . [ [ fm_a ] ] fm@xmath146 : + + + + + + + + + + + + + + + + + + + + + this coplanar state with finite net moment encompasses a large fraction of both phase diagrams . only in the hyperhoneycomb does the lta succeed in identifying this phase as the exact ground state . in both lattices , the projection of the spins along the @xmath88 direction is ferromagnetic while the projection along the @xmath20 direction vanishes . the projection along the @xmath147 direction behaves differently for the hyperhoneycomb and the @xmath2 models . in the hyperhoneycomb lattice , the @xmath147 component orders in the skew - zigzag order , while in the @xmath2 lattice , the @xmath147 component is ferromagnetic within each honeycomb strip . this state is connected to the hk - fm phase , and in the case of the hyperhoneycomb lattice , it is also connected to the @xmath148 segment of the hk - sz phase , where obd studies have shown that the hk - sz phase orders in the @xmath147 direction , much like the fm@xmath146 phase . the @xmath142 rotation symmetry is preserved in this phase . since the @xmath147 component of the phase in both lattices have a vanishing net moment , only the @xmath88 component contributes to the total moment . this total moment becomes saturated ( spins point entirely in the @xmath88 direction ) when approaching the hk - fm phase and decreases smoothly as we approach the sp@xmath80 boundary . in the hyperhoneycomb case , the total moment further decreases and vanishes smoothly as we approach the hk - sz phase . the magnitude of the moment along @xmath88 is depicted by the color contours in fig . [ fig : pds ] where the largest projection is colored darkest . [ [ ss_xy ] ] ss@xmath135 : + + + + + + + + + + + + + + + + + + + + + + + wedged within the hk - ss , af@xmath141 , and sp@xmath82 phases are two skew - stripy phases , the first of which to be discussed is the non - coplanar ss@xmath135 phase . this phase has the largest projection along the @xmath15(@xmath16 ) direction and this component orders in a skew - stripy fashion ( these two orientations are degenerate ) . in the hyperhoneycomb lattice , the other two cartesian components of the spins are small but finite and ensure that the spins along each zigzag chain are collinear . in the @xmath2 lattice , the @xmath16(@xmath15 ) component also orders in a skew - stripy fashion . this phase does not have a net moment and breaks all @xmath115 symmetries . [ [ ss_b ] ] ss@xmath149 : + + + + + + + + + + + + + + + + + + + this other skew - stripy phase is coplanar and lies farther away from the fm kitaev point relative to the ss@xmath135 phase . this phase borders the af@xmath141 phase and can be identified via the lta in the hyperhoneycomb lattice . in both lattices , the projection of the spins along the @xmath147 direction orders in a skew - stripy pattern while the @xmath20 projection vanishes . in the hyperhoneycomb lattice , the @xmath88 projection orders antiferromagnetically , while in the @xmath2 lattice , spins order ferromagnetically along non - bridging-@xmath17-bonds while spins along zigzag chains order antiferromagnetically . this state has a vanishing net moment and preserves the @xmath150 rotation symmetry . [ [ sz_xnegmedspacenegmedspace - y ] ] sz@xmath151 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + analogous to the ss@xmath135 phase , there exists a non - coplanar , skew - zigzag phase with a vanishing net moment near the af kitaev point bordering the sp@xmath80 phase that breaks all @xmath115 symmetries . this phase has the largest projection along the @xmath15(@xmath16 ) direction and this component orders in a skew - zigzag pattern ( these two configurations are degenerate ) . in the hyperhoneycomb lattice , the other two cartesian spin components ensure that the spins are collinear along zigzag chains . in the @xmath2 lattice , the @xmath16(@xmath15 ) component is also ordered in the skew - zigzag pattern . this phase is closely related to the sz@xmath151 phase : these two phases map onto each other under the @xmath72 classical transformation . [ [ sz_b ] ] sz@xmath149 : + + + + + + + + + + + + + + + + + + + in the @xmath2 model , an additional coplanar , skew - zigzag phase with vanishing net moment exists . this phase borders the fm@xmath130 , sz@xmath151 , sp@xmath80 , and hk - sz phases . it forms ferromagnetic skew - zigzag chains with antiferromagnetic non - bridging-@xmath17 bonds and preserves the @xmath142 rotation symmetry . the component along @xmath20 vanishes while the projection along @xmath147 is the greatest . this phase is closely related to the ss@xmath149 phase : these two phases map onto each other under the @xmath72 classical transformation . as mentioned previously , the classical transformation of applying time - reversal on the odd sublattices maps @xmath72 . using this transformation , the classical phase diagrams for @xmath68 in fig . [ fig : pds ] can be inverted to yield the phase diagrams for @xmath71 in fig . [ fig : pds_neg ] , where the radial coordinate now maps to @xmath153 $ ] . under this transformation , ferromagnetic components become antiferromagnetic , while stripy components become zig - zag ( these transformed phases are labeled in fig . [ fig : pds_neg ] ) . for the spiral and multiple-@xmath3 states , we use an overhead bar ( @xmath154 versus sp@xmath80 ) to emphasize the close relationship between the @xmath155 and the @xmath68 phases . for the spiral phases , we note that the @xmath3-vector range remains invariant after this transformation , while the @xmath115 rotation that the spiral state was invariant under must now be followed by a time - reversal operation @xmath156 . we draw special attention to the @xmath154 phases found in the @xmath71 region because of their close relationship to the experimental magnetic orders ( the real - space configuration of these spiral phases can be seen in fig . [ fig : realspace_a_mgamma ] and can be directly compared with those found in refs . ) . the experimental magnetic ordering of the hyperhoneycomb li@xmath7iro@xmath8@xcite possesses the same symmetries as that of @xmath154 : following the analysis used in ref . , we discern that the @xmath154 phase on the hyperhoneycomb lattice is described by the magnetic basis vector combination @xmath157 axes in refs . is along the orthorhombic directions . in this work , we have labeled these as the @xmath158 directions . hence , we have replaced @xmath159 with @xmath160 in the basis vector notation to avoid confusion . ] , which is identical to that of the experimental magnetic ordering@xcite , indicating that these two phases are indeed the same . in addition , the experimental ordering wavevector@xmath161lies within the range found in our model ( when expressed using the same definition as @xmath162 from ref . , the wavevectors found within our present model is @xmath163 where @xmath164 ) . for the @xmath2 li@xmath7iro@xmath8@xcite , @xmath162 also lies within the range found in our model . using the notation defined in ref . , the magnetic basis vector combination that describes the @xmath154 phase is @xmath165 $ ] , which only differs from the experimental combination of @xmath166 $ ] in the @xmath167-component . for completeness , we also show the real - space configurations of the @xmath168 and @xmath169 phases in figs . [ fig : realspace_a_mgamma ] and [ fig : realspace_b_mgamma ] respectively , as a comparison with figs . [ fig : realspace_a ] and [ fig : realspace_b ] . magnetic static structure factors of these spiral phases can be found in appendix [ app : magneticstructurefactor ] . [ cols="^,^,^,^,^,^,^,^,^ " , ] in the model presented , we have implicitly assumed an ideal crystal structure . in principle , distortions of the oxygen octahedra as well as ir positions can bring about additional anisotropic exchange interactions between neighboring pseudospins.@xcite in addition , large distortions can admix the @xmath0 and the neglected orbitals , which can generate anisotropic @xmath170-factors or even destroy the @xmath0 description entirely.@xcite nevertheless , these distortion effects are likely to be smaller in the 3d compounds compared to the 2d honeycomb iridates , since the octahedral environments in the former are more ideal than those of the latter.@xcite . as such , these distortion effects may serve only to affect quantitatively the phases found within our treatment while the qualitative features of our phase diagrams can be expected to remain robust . we have also assumed that the 3d materials lie deep in the mott insulating regime ( @xmath171 ) such that other hopping pathways , namely ones involving multiple oxygen and lithium ions , can be neglected , thus resulting in a model with only nn exchanges . the role of further neighbors have been considered in the 2d honeycomb iridates : further neighbor hopping in _ ab initio _ treatments@xcite , as well as further neighbor exchanges in the localized picture@xcite have been investigated . in addition , experimental work has suggested the possibility of significant further neighbor interactions in the case of the 2d li@xmath7iro@xmath8.@xcite meanwhile , there have also been studies on effects of further neighbors in the case of the hyperhoneycomb iridate in both the itinerant picture@xcite and in the mott limit@xcite . the effects of further neighbors have revealed interesting consequences in these cases , which suggests that additional studies on this subject and its role in a minimal model for the hyperhoneycomb and @xmath2 lattices may be warranted . on the other hand , if the ideal nn localized picture is adequate in capturing the magnetic order of the whole family of honeycomb iridates , then we may expect the 2d honeycomb , hyperhoneycomb , and @xmath2 iridates to be describable by nn @xmath172 models . the 2d - li@xmath7iro@xmath8 and the two 3d - li@xmath7iro@xmath8 are then expected to have comparable exchange parameters , since their chemical composition and local crystal structures are similar , while the 2d - na@xmath4iro@xmath5 is more distorted and may lie in a different region in parameter space . since the 2d - na@xmath7iro@xmath8 orders in a zigzag fashion@xcite and the 2d - li@xmath7iro@xmath8 is believed to order in a spiral phase@xcite , this line of reasoning suggests that the 3d - li@xmath7iro@xmath8 s may order in one of the spiral phases , especially since the spiral regions are relatively larger and the zigzag region smaller in the 3d phase diagrams . indeed , the two 3d - li@xmath7iro@xmath8 have been experimentally reported to order in a non - collinear fashion , with counter - propagating spirals@xcite . the ordering pattern in the case of the hyperhoneycomb li@xmath7iro@xmath8 case is precisely that of the hyperhoneycomb @xmath154 phase@xcite , while that of the @xmath2 li@xmath7iro@xmath8 has striking similarities with the @xmath2 @xmath154 phase@xcite . moreover , the similar thermodynamic properties of the two 3d - li@xmath7iro@xmath8 lattices may be due to the combination of the near - identical phase diagrams and the fact that both systems are near ideal ( distortion - free ) and largely driven by nn physics . in summary , motivated by the similar behaviors and local structure of the hyperhoneycomb and @xmath2 li@xmath7iro@xmath8 , we studied the nn pseudospin @xmath11 model , which contains the heisenberg , the kitaev , and the off - diagonal @xmath1 exchange interactions . we argue that with strong on - site interaction and atomic spin - orbit coupling , these models are minimal effective models for the @xmath0 pseudospins in these 3d honeycomb - based materials . the resulting classical phase diagrams that emerge in the two cases are very similar . on the other hand , when compared with the phase diagram of the 2d honeycomb lattice , the zigzag region diminished while spiral phases with counter - propagating spirals expanded in the 3d phase diagrams . we characterized all the magnetic orders in detail and provided static structure factor results of the spiral phases . we believe that our phase diagrams and detailed analysis will be relevant in identifying the phases of the 3d - li@xmath7iro@xmath8 compounds , and our model can serve as the minimal starting point in identifying the common thread which runs through this family of iridium - based materials . we thank r. schaffer and j. rau for discussions . we also thank r. coldea for his useful comments on our work and for making some of the results in refs . and available prior to publication . computations were performed on the gpc supercomputer at the scinet hpc consortium . scinet is funded by : the canada foundation for innovation under the auspices of compute canada ; the government of ontario ; ontario research fund - research excellence ; and the university of toronto . this research was supported by the nserc , cifar , and centre for quantum materials at the university of toronto . upon initial submission of this manuscript to the pre - print arxiv , independent experimental results on the magnetic structure of @xmath2 li@xmath7iro@xmath8 were released@xcite which reported the existence of a spiral order in the compound . in addition , independent experimental report of the existence of a spiral order in hyperhoneycomb li@xmath7iro@xmath8 was released@xcite after the submission of this work . although this work is an exploratory study of our pseudospin model , we have since made modifications to reference these experimental results throughout the current manuscript wherever relevant . octahedral crystal fields arising from the oxygen ions split the @xmath28 @xmath173 orbitals into the lower energy @xmath39 and higher energy @xmath174 multiplets . large atomic spin - orbit coupling further split the @xmath39 orbitals into the lowest energy @xmath175 quadruplet and higher energy @xmath0 doublets . the five electrons at each @xmath28 site fully - fill the @xmath175 orbitals and half - fill the @xmath0 orbitals , leaving the @xmath174 orbitals unoccupied . in the limit of large atomic spin - orbit coupling and strong on - site electron correlations , the @xmath175 and @xmath174 orbitals can be projected out and the relevant low - energy degrees of freedom are the localized @xmath0 pseudospins . in the presence of hopping between neighboring ir orbitals , the perturbed ground states can be captured by an effective hamiltonian derived by a strong coupling expansion . we assume the unperturbed , atomic hamiltonian to be comprised of atomic soc and on - site interactions . the atomic soc takes the form @xmath176 , where @xmath177 and @xmath178 are the spin and effective @xmath39 orbital angular momentum of electron @xmath50 . the on - site interaction consists of hund s coupling and intra / inter - orbital repulsion , which we write in the kanamori form @xmath179\ ] ] where @xmath180 , @xmath181 , and @xmath182 are the total number , spin , and effective orbital angular momentum at site @xmath31 ; @xmath40 and @xmath41 are the coulomb repulsion and hund s coupling respectively . the perturbation comes in the form of nn hopping amplitudes between @xmath39 orbitals : we introduce both direct orbital overlaps and effective hopping between neighboring orbitals mediated by the edge - shared oxygens . the hopping amplitudes can be written as @xmath183,\end{aligned}\ ] ] where @xmath33 is the label for bond @xmath184 , and @xmath185 are the other two components . the creation operator is given by @xmath186 and creates electrons on site @xmath31 in orbital @xmath187 , where the @xmath39 orbitals are labeled according to @xmath188 , @xmath189 , and @xmath190 . the hopping amplitudes are given by @xmath191 where @xmath192 , @xmath193 , @xmath194 , and @xmath195 are the slater - koster amplitudes parametrizing direct ir orbital overlaps and o - ir orbital overlaps respectively , while @xmath196 is the energy gap between oxygen and iridium . in the large soc limit , we perform the strong coupling expansion with the limit @xmath197 then project into the @xmath0 manifold . the resulting @xmath0 pseudospin hamiltonian then takes the form of eq . [ eq : ham ] with exchanges related to hopping amplitudes by eq . [ eq : exchanges ] . for more detail on the derivation of the expansion , we refer the reader to the appendix in ref . . from eq [ eq : exchanges ] , we note that the exchanges depend sensitively on the relative signs and magnitudes of the hoping amplitudes @xmath198 . hence , a detailed analysis of the underlying band structure would greatly aid the realistic estimations of these exchanges for the 3d li@xmath7iro@xmath8 systems . although the relative signs of @xmath1 among symmetry-_equivalent _ bonds can be determined by applying the appropriate symmetry transformation , the relative signs among symmetry-_inequivalent _ bonds are not constrained by crystal symmetries . however , these symmetry - inequivalent bonds have identical local environments ( _ e.g. _ nearest - neighbor oxygen and cation hopping pathways ) , and hence , the sign structure of @xmath1 among symmetry - inequivalent bonds are expected to follow transformations that preserve these local environments . for example , to relate the signs of @xmath1 on the @xmath15- and @xmath17-bonds emanating from a site , a @xmath115 rotation through the site and along the connecting @xmath16-bond would preserve the local oxygen and cation environments while mapping the @xmath15-bond to the @xmath17-bond . this microscopic justification of the relative signs of @xmath1 was used in assigning the sign structures in fig . [ fig : sign_lattice ] . we emphasize that this convention only fixes the _ relative _ signs @xmath1 in the lattice : @xmath1 can still be either positive or negative , leading to the phase diagrams in figs . [ fig : pds ] and [ fig : pds_neg ] respectively . we leave other possible choices of relative signs in @xmath1 among symmetry - inequivalent bonds for future study . the magnetic static structure factors along certain momentum cuts of the sp@xmath90 , sp@xmath91 , @xmath199 , and @xmath169 phases of both the hyperhoneycomb and @xmath2 lattices are presented in figs . [ fig : sf_a ] , [ fig : sf_b ] , [ fig : sf_a_mgamma ] , and [ fig : sf_b_mgamma ] respectively for reference . 71ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1143/jpsj.70.2880 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.96.087204 [ * * , ( ) ] link:\doibase 10.1143/jpsj.76.043706 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.064412 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.220403 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.127203 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.127204 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/24/34/345601 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.82.3045 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.83.1057 [ * * , ( ) ] link:\doibase 10.1146/annurev - conmatphys-062910 - 140432 [ * * , ( ) ] link:\doibase 10.1038/nphys1606 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.085111 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.085145 [ * * , ( ) ] link:\doibase 10.1146/annurev - conmatphys-020911 - 125138 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.017205 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.256403 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.027204 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.180407 [ * * , ( ) ] http://iopscience.iop.org/1367-2630/14/7/073015/ [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.180403 [ * * , ( ) ] http://iopscience.iop.org/0953-8984/24/38/382201 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.266406 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.106401 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.197201 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.224417 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.035107 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.195131 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.076402 [ * * , ( ) ] link:\doibase 10.1103/physrevb.87.220407 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.097204 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.220414 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.aop.2005.10.005 [ * * , ( ) ] link:\doibase 10.1103/physrevb.76.180404 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.247203 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.nuclphysb.2008.06.009 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.024426 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.180412 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.235102 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.112.077204 [ * * , ( ) ] http://iopscience.iop.org/1367-2630/16/1/013056 [ * * , ( ) ] http://arxiv.org/pdf/1402.1030v1.pdf [ ( ) ] http://arxiv.org/abs/1404.5818 [ ( ) ] link:\doibase 10.1103/physrevb.87.165117 [ * * , ( ) ] http://arxiv.org/abs/1403.3296 [ ( ) ] http://www.nature.com/ncomms/2014/140627/ncomms5203/full/ncomms5203.html [ * * ( ) ] http://arxiv.org/abs/1407.3954 [ ( ) ] http://arxiv.org/abs/1408.0246 [ ( ) ] link:\doibase 10.1103/physrevb.89.205132 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.045117 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.014424 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.115125 [ * * , ( ) ] http://arxiv.org/abs/1403.2724 [ ( ) ] http://arxiv.org/abs/1309.1171 [ ( ) ] http://arxiv.org/abs/1406.5415 [ ( ) ] link:\doibase 10.1103/physrevlett.101.076402 [ * * , ( ) ] http://arxiv.org/abs/1303.5922 [ ( ) ] link:\doibase 10.1103/physrev.94.1498 [ * * , ( ) ] link:\doibase 10.1103/physrevb.87.064508 [ * * , ( ) ] link:\doibase 10.1103/physrev.70.954 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0031-8914(74)90257-2 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.100406 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.024410 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.89.014414 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.245113 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.241102 [ * * , ( ) ] @noop |
hard - core dipolar bosons trapped in a parallel stack of 1d optical lattices ( tubes ) can develop several phases made of composites of particles from different tubes : superfluids , supercounterfluids and insulators as well as mixtures of those . bosonization analysis shows that these phases are threshold - less with respect to the dipolar interaction , with the key control knob " being filling factors in each tube , provided the inter - tube tunneling is suppressed . the effective _ ab - initio _ quantum monte carlo algorithm capturing these phases is introduced and some results are presented . | 1405.4328 | the unprecedented level of control in ultra - cold atom experiments has allowed for the realization of paradigmatic condensed matter models @xcite . in these systems the inter - particle interactions can be tuned by varying the scattering length through feshbach resonances and the atoms can be trapped in various geometries @xcite . such a flexibility makes ultra - cold atoms an almost ideal candidate for the study of strongly correlated many - body quantum systems as well as a playground for emerging new states of matter . these , in particular , include paired superfluids . recent experimental success in trapping ultra - cold bosonic atomic mixtures @xcite has rendered the study of pairing between components very timely . the impressively rapid experimental progress towards controlling polar molecules @xcite gives hope for accessing quantum many body systems with long range and anisotropic interaction in a very near future @xcite . a prominent example of bosonic systems currently available experimentally consists of an array of coupled one - dimensional tubes , with the interaction between tubes provided by dipolar forces . in the absence of the inter - tube tunneling , this system can be relevant to multi - component atomic mixtures . in general , when such a tunneling is finite , it represents coupled spin chains , which is one of the central topics in low - dimensional condensed matter physics @xcite . [ fig : setup ] of them are shown ) ; dashed arrows indicate the intra - plane @xmath1 and the inter - plane interactions @xmath2 . , scaledwidth=35.0% ] in the present work we discuss possible phases in a system of hardcore bosons confined to a stack of @xmath3 one - dimensional lattices tubes ( see fig . [ fig : setup ] ) . bosons in neighboring tubes interact via inter - tube interaction ( either nearest - neighbor or dipole - dipole ) , with the inter - tube tunneling suppressed ( this can be achieved experimentally with a deep optical lattice potential along the direction perpendicular to the tubes ) . our focus is on quantum many - body phases of self - assembled chains of molecules from different tubes @xcite . in previous theoretical studies , mostly variational methods have been used . in ref . @xcite the multilayered system has been mapped to a model amenable to classical monte carlo technique , and it has been shown that bosons in a stack of one - dimensional tubes can form superfluids of multi - atomic complexes chain superfluids ( csf ) @xcite , each chain consisting of one molecule from each tube , and there is , in general , a threshold for the csf formation . an interesting opportunity for the emergence of exotic parafermions , as a generalization of majorana fermions , in layered systems has been proposed in refs . @xcite and tested by monte carlo simulations @xcite . it was also suggested that in two parallel 1d lattices with no inter - tube tunneling an exotic superfluid , consisting of @xmath4 molecules from one tube and @xmath5 ones form the other , should also be possible to realize @xcite . in contrast to what was previously done , here we study the actual quantum hamiltonian of hard - core bosons by means of the _ ab initio _ path integral quantum monte carlo ( qmc ) simulations ( in continuous time ) using a multi - worm algorithm @xcite an extension of the worm algorithm @xcite and its two - worm modification @xcite . as we mentioned above , our algorithm is equally relevant to atomic mixtures and coupled spin chains . we will show that csf and other phases can be induced by infinitesimally small inter - layer interaction . our study is a first step toward _ ab - initio _ simulations of more involved cases including spin ladders and polar molecules with inter - tube tunneling . using this algorithm it should be possible to provide accurate recommendations for the experimental realizations of the complex dipolar phases . the system under consideration is described by the single - band tight - binding hamiltonian @xmath6 in grand canonical ensemble . here @xmath7 stands for the intra - tube tunneling amplitude , @xmath8 ( @xmath9 ) is the creation ( annihilation ) operator for a hard core boson at site @xmath10 , where @xmath11 labels the tubes and @xmath12 is the coordinate along a tube . here , @xmath13 denotes summation over nearest neighbors , @xmath14 , and @xmath15 is the chemical potential , which can be different in different tubes . the interaction @xmath16 can be arbitrary . in the case of the dipole - dipole interaction , with the polarization axis being perpendicular to the tubes and belonging to the plane of the tubes , it takes the form @xmath17 where @xmath18 sets the energy scale . in this geometry , the interaction along the @xmath19-axis is attractive . as we will discuss below , in 1d , _ arbitrary small _ @xmath20 can induce superfluidity of quasi - molecular complexes . this result follows from the bosonization analysis and has previously been noted for the case of pairing of hard core bosons in ref . @xcite . the repulsive part of the interaction along the @xmath21-axis favors solidification . a special role is played by the filling factor @xmath22 . as we will show later , in the case @xmath23 the insulating phase featuring 1d checkerboard order emerges in the limit @xmath24 even if no intra - layer repulsion is explicitly introduced . when dipoles are polarized perpendicularly to the tubes plane , the interaction becomes purely repulsive , @xmath25 and can result in super - counterfluid ( scf ) @xcite phases which are also thresholdless with respect to the interaction . a system of hard - core bosons , trapped in one - dimensional tubes with no inter - tube josephson coupling , forms @xmath3 independent superfluids characterized by @xmath3 quasi - condensate order parameters @xmath26 with phases @xmath27 . the hard - core nature of bosons in each tube plays a special role . as we will see below , an arbitrary small inter - tube interaction can induce multiplicity of various superfluid and insulating phases depending on the filling factors @xmath28 in the tubes . the counter intuitive threshold - less nature of the phases simply means that observing them is possible for arbitrary small @xmath20 on correspondingly large spatial scales . it is worth noting that , depending on a combination of the filling factors , various types of mixtures of such phases can exist as well . here we introduce a description in terms of the generalized superfluid stiffness @xmath29 and superfluid compressibility @xmath30 . this language of the generalized superfluid response turns out to be very helpful in defining ground states of the bosonic complexes as well as in characterizing ground states numerically . the response matrices are defined through contributions to the system action as a result of imposing infinitesimal thouless phase twists @xmath31 on the space - time boundaries of the tubes . such twists can be viewed in terms of the corresponding gauge potentials @xmath32 along space and @xmath33 along time , where @xmath34 stand for tubes length and inverse temperature in atomic units , respectively . it is important that , in the case of the periodic boundary conditions on the phases @xmath35 of the fields , such gauge potentials can not be absorbed into the phases . in general , the infinitesimal contribution of the twists to the action is given by : [ th ] e= _ zz. the quantities @xmath29 and @xmath30 can be expressed in terms of topological properties of the particle world - lines , windings @xmath36 , and can be measured numerically . global gauge invariance of the system implies that the total partition function @xmath37 can be represented as a statistical sum over all possible winding numbers of closed world lines of particles as [ z ] z= _ \{(z)}z[\ { } ] , where @xmath38 $ ] stands for a functional of windings in all tubes . the superfluid stiffnesses can be obtained as second derivatives of @xmath39 with respect to @xmath40 in the limit @xmath41 as [ r ] r_zz= [ w_x(z ) w_x(z)- w_x(z)w_x(z ) ] , [ c ] c_zz= [ w_(z ) w_(z)- w_(z)w_(z ) ] . as long as the tubes are identical , @xmath29 and @xmath30 depend on the difference @xmath42 , @xmath43 and @xmath44 . hence , the fourier transform along @xmath19-axis can be used , @xmath45 , where @xmath46 . then , eq . expressed in terms of the fourier transforms @xmath47 gives [ furr ] ( q_z)= [ _ x(q_z)_x(-q_z)- _ x(q_z)_x(-q_z ) ] and [ furc ] ( q_z)= [ _ ( q_z)_(-q_z)- _ ( q_z)_(-q_z ) ] these equations represent an extension of the ceperley and pollock expression @xcite for the superfluid stiffness and compressibility . in full analogy with the case @xmath48 , the ratio @xmath49 has the meaning of the speed of sound propagating along tubes with dispersion along the @xmath19-axis . extending the analogy , the product @xmath50 gives the luttinger parameter " ( rather , luttinger matrix ) as [ lut ] ( q_z ) = . thus , the action for arbitrary ( small ) phase fluctuations of the translationally invariant ( along @xmath19-axis ) system renormalized by the interactions becomes [ full ] s_r= _ 0^ddx _ q_z , where @xmath51 are the fourier components of the phases @xmath35 with respect to the tube index @xmath19 . the speed of sound is not significantly renormalized compared to the strong renormalization of superfluid stiffness @xmath52 and compressibility @xmath53 @xcite . this simply means that the space - time symmetry of the superfluid - insulator transitions is preserved in translationally invariant system . thus , for all practical purposes , the dispersion of the speed of sound @xmath54 vs @xmath55 can be ignored so that @xmath56 in units of @xmath57 . in this limit , the luttinger matrix and the matrix of stiffnesses are equivalent to each other . then , the generalized linear response can be fully described by the following translationally invariant action @xmath58 , or in the direct z - space as @xmath59 where @xmath60 stands for luttinger parameter matrix with the dimension @xmath61 . it is worth mentioning that this form features the non - viscous drag between superfluid flows in different tubes . it is responsible for the formation of the complex superfluid and supercounterfluid phases . we will be referring to the form ( [ rphi ] ) and , specifically , to the properties of the luttinger matrix @xmath60 while identifying the ground states of the bosonic complexes . if the matrix @xmath60 in ( [ rphi ] ) is non - degenerate ( in the case when all filling factors are different and not complimentary to unity ) , there exists the standard algebraic ( or 1d superfluid ) order in the correlators @xmath62 ( and @xmath63 for @xmath64 ) . a sketch of this phase is shown in fig . [ fig : nsf ] . as we will see below , should some filling factors become the same or complimentary to unity , the inter - tube interaction can easily destroy such atomic orders in favor of composite superfluids or supercounterfluids . if all @xmath3 tubes are characterized by the same incommensurate filling factor @xmath65 , the nature of the superfluid correlations changes dramatically as long as there is an arbitrary small attraction between the tubes . specifically , the matrix @xmath60 becomes degenerate so that @xmath66 decays exponentially . the algebraic decay will be observed only in the @xmath3-body density matrix @xmath67 where @xmath68 . in other words , the matrix elements of @xmath60 in eq.([rphi ] ) become all identical to each other , so that @xmath69 eigenvalues are equal to zero and only one remains finite corresponding to a finite superfluid stiffness of the sum of the phases @xmath70 . in terms of the fourier components of the matrix kernel , @xmath71 while @xmath72 . this defines the csf , a superfluid of quasi - molecular complexes , each complex consisting of @xmath3 bosons one from each tube . a sketch of this phase is shown in fig . [ fig : csf ] . if @xmath23 , it is possible to have a situation when only @xmath73 tubes have identical filling factors . then , the composite superfluid will be formed among these tubes while others carry the standard atomic superfluids . in general , a group of @xmath74 layers with the same filling factor adds degree of degeneracy m-1 to the matrix @xmath60 . in other words , the number of the remaining superfluid phases is equal to @xmath3 minus the total degree of degeneracy . this means that the matrix @xmath60 will have as many zero eigenvalues as there exist restored u(1 ) symmetries . as we will discuss below , such phases can be realized for arbitrary small inter - tube interaction @xmath20 . the concept of supercounterfluidity ( scf ) has been introduced for two - component systems in ref.@xcite . scf can exist in a lattice when the filling factors @xmath75 and @xmath76 for both components complement each other to an integer filling , @xmath77 . then , the repulsive interaction can induce binding of atoms of sort 1 " to holes of sort 2 " . using the language of broken symmetries , the u(1)@xmath78 u(1 ) symmetry becomes partially restored so that only one u(1 ) symmetry remains broken . in terms of fields , the field @xmath79 is condensed while @xmath80 becomes disordered . accordingly , the superflow can only exist in the counterflow manner when transfer of one atom of sort 1 " is compensated by motion of one atom of sort 2 " in the opposite direction . this property can naturally be extended to a general case of @xmath3 sorts of atoms when the superflow of , say , @xmath81 components is ( partially ) compensated by the counter - flow of the remaining components . the scf phase is sketched in fig . [ fig : scf ] . tubes . similarly to fig . [ fig : csf ] , the fuzzy cloud depicts a single algebraic order for all tubes . the dashed lines indicate binding between atoms and holes in the tubes with complementary fillings and between atoms in the tubes with the same fillings.,scaledwidth=30.0% ] in general , there could be @xmath82 tubes all with the same filling factors @xmath75 and @xmath83 tubes also with identical filling factors @xmath84 so that @xmath85 . thus , there are two groups of the composite superfluids , consisting of @xmath82 and @xmath83 complexes . accordingly , there are @xmath86 restored symmetries . moreover , the backscattering ( bs ) interaction between particles from the first and the second groups restores one additional symmetry . the corresponding composite operator which characterizes the algebraic order is @xmath87 , where @xmath88 label tubes from the first group and @xmath89 from the second . in such a phase a transfer of @xmath82 atoms from the first group is compensated by the counter - motion of @xmath83 atoms from the second group , so that there is a net transfer of @xmath90 atoms . accordingly , @xmath91 eigenvalues of the matrix @xmath60 in eq.([rphi ] ) are zero . in other words , the resulting state can be thought of as a bound state of two composite superfluids in the counter - flow regime a natural generalization of the two - component scf @xcite . in the special case when all , e.g , odd tubes have filling factor @xmath65 and all even ones have @xmath92 ( as exemplified in fig . [ fig : scf ] ) , the fourier transform can be used . in this case all fourier harmonics but @xmath93 are equal to zero , so that there is no net transfer of atoms . as we will discuss below , such a phase can also be realized for arbitrary small @xmath20 . the easiest way to form an insulator in 1d lattices is at filling factor @xmath22 . in a single tube ( @xmath48 ) at @xmath22 the checkerboard ( cb ) type solid can exist only if the two - body repulsion exceeds a certain threshold . the situation becomes dramatically different in the cases @xmath94 . as it will be discussed below , the bosonization analysis shows that , even in the absence of any repulsion , the inter - layer attraction induces the cb insulator in the limit @xmath24 as long as @xmath95 . this conclusion is consistent with our _ ab initio _ simulations . a sketch of this phase is shown in fig . [ fig : cb ] . .,scaledwidth=30.0% ] solids at other rational fillings @xmath96 are possible as well . to induce them , however , the interaction @xmath97 must exceed the corresponding thresholds determined by the denominators of the fractions . concluding this section we note that in an insulating state the renormalized luttinger matrix @xmath60 in the action is zero . here we will discuss the phases described in [ as ] , [ cs ] , [ scf ] , [ ci ] within the framework of the bosonization approach @xcite in order to reveal their threshold - less nature . the bosonic field operator @xmath98 is represented in terms of the superfluid phase @xmath99 and the density @xmath100 , which can be expressed in terms of haldane s angle " variable @xmath101 @xcite as @xmath102}\,,\ ] ] where @xmath103 is conjugate to the superfluid phase @xmath99 . the term @xmath104 gives the forward scattering ( fs ) interaction and the terms with @xmath105 account for the back scattering ( bs ) events . in the absence of inter - tube tunneling , the bosonized action corresponding to the hamiltonian is @xmath106,\ ] ] where @xmath107,\end{aligned}\ ] ] in units @xmath57 . here @xmath108 is the bare luttinger liquid parameter , that is , not yet renormalized by the interactions . for hard - core bosons and zero dipolar interaction , each tube is equivalent to a xxz spin @xmath109 chain with zero @xmath110-@xmath110 coupling . accordingly , @xmath111 ( see in ref . @xcite ) . in the following we use periodic boundary conditions along @xmath21- and @xmath19-coordinates . the second term in eq . is the fs part of the action , @xmath112 . in the long wave limit of the space - time fourier representation ( @xmath113 ) the fs part of the action becomes @xmath114 here , the summation is performed over the time - space harmonics , @xmath115 along each tube and @xmath116 in particular , for the dipole - dipole interaction given by eq . one finds @xmath117 for @xmath118 . for @xmath119 , the dipolar interaction is purely repulsive , with @xmath120 , and it must be cut off at some short distance . here we consider the same length scale along @xmath19 and @xmath21 , and choose the cutoff at @xmath121 such that @xmath122 . thus , in the long - wave limit , while the inter - layer interaction is attractive , the intra - layer one is repulsive . in the case of the purely repulsive dipolar interaction , eq . , the fs interaction given by eq . changes sign , that is , @xmath123 for @xmath64 ( with @xmath124 being unchanged ) . next , we introduce fourier harmonics along the @xmath19-coordinate , @xmath125 and @xmath126 , corresponding to @xmath127 and @xmath128 , respectively . we can now write the gaussian part @xmath129 of the action ( [ stotal ] ) as @xmath130|\tilde{\theta}(\vec{q},q_z)|^2 . \ ] ] eq . implies the renormalization of the speed of sound @xmath131 ( in units of the bare value ) as well as of the luttinger parameter @xmath132 thus , both quantities @xmath133 depend on the wave - vector @xmath55 counting the layers so that the action ( [ gauss ] ) takes the form ( [ full ] ) . as discussed above , in what follows we will ignore the renormalization of the speed of sound and will rather consider the form ( [ rphi ] ) . then , in terms of the dual variables @xmath101 the gradient part of the renormalized action becomes @xmath134 where @xmath135 is the inverse of the renormalized luttinger matrix @xmath60 introduced in eq . . finally , the third term in eq.([stotal ] ) accounts for the backscattering events @xcite which in the context of the system studied can be written as @xmath136\;,\ ] ] where the amplitudes @xmath137 are induced by the interaction and satisfy the renormalization flow ( to be derived in the standard one - loop approximation in appendix a ) . while the fs sets in the initial value of the luttinger matrix eq . , the bs is responsible for its further renormalization . in the limit @xmath24 the renormalization group ( rg ) flows can be found exactly . first we remind that for hard - core bosons @xmath138 and @xmath139 for @xmath64 at @xmath140 . therefore , as it will be clear below , in the limit of small interactions , only the lowest non - trivial values of @xmath141 ( @xmath142 ) can become relevant in the sum ( [ bs ] ) , provided @xmath143 . hence , the relevance of the backscattering for a particular pair @xmath144 of layers can be controlled by adjusting the bosonic populations in individual tubes . due to the spatially non - local nature of dipolar interactions , the composite superfluid phase , [ cs ] , can form between tubes with the same filling factors regardless of their geometrical positions . for example , in a system of @xmath145 tubes where @xmath146 ( here we consider @xmath147 ) , with all other values @xmath148 , the harmonics @xmath149 can become relevant , while all others remain irrelevant ( simply because of the oscillating phases @xmath150 , with @xmath151 , in the corresponding cos - harmonics in eq.([bs ] ) ) . the rg equations for the amplitudes of the corresponding harmonics ( @xmath64 ) between the tubes with identical filling factors are ( see appendix [ a ] ) @xmath152 v_{1;-1}(z , z'),\ ] ] where @xmath153 are the matrix elements of the matrix @xmath60 in eq.([rphi ] ) and the initial ( bare ) values of the amplitudes @xmath154 are determined by the dipolar interactions . in the limit of no interactions , that is @xmath155 , the rg flow starts from the critical point determined by the factor @xmath156 \to 0 $ ] ( since @xmath157 , with @xmath111 ) in eq.([rg11 ] ) . as explained in the appendix a , eq.([rg22 ] ) , this factor is @xmath158 and is positive for the case of the attractive inter - layer interaction ( [ vdip ] ) . thus , the relevant amplitudes diverge as @xmath159 with @xmath160 . while formally this implies that csf is induced by an arbitrary small interlayer attraction @xmath20 , a physical scale @xmath161 on which such a composite phase can be observed is actually exponentially divergent as @xmath162 , where `` ... '' means a coefficient @xmath163 ( see below ) . using the @xmath145 example from above , the formation of the csf between the tubes @xmath164 implies that the harmonics @xmath149 exhibit the runaway flow to @xmath165 in eqs.([rg11 ] ) , while all other combinations can be set essentially to zero . in other words , two u(1 ) symmetries are being restored so that the system is decsribed by the algebraic orders in @xmath166 and in the csf field @xmath167 . in the case of translational invariance along the @xmath19-axis , that is , when @xmath168 for all tubes , eq.([rg11 ] ) can be explicitly written in terms of the kernels of luttinger matrix and its inverse as @xmath169 v_{1;-1}(z),\ ] ] where we have taken into account that the amplitudes @xmath154 as well as the matrix elements @xmath153 are functions of the difference @xmath42 rather than of @xmath170 separately : @xmath171 , where @xmath172 with the corresponding inverse transformations . ignoring the renormalization of the luttinger matrix by the bs , the value of @xmath173 from eq . can be used in eq . in the limit @xmath24 . then , in the lowest order in @xmath20 we find @xmath174 where @xmath128 is given in eq . ) and @xmath108 is set to its value , @xmath111 , for non - interacting tubes . thus , for @xmath175 and @xmath176 , as it is in the case of the dipolar interaction between molecules polarized along the z - axis , eq.([vdip ] ) , the harmonics @xmath177 become relevant for arbitrary small @xmath20 . this implies that the superflow is only possible in the channel of the center of mass motion of all tubes because relative density fluctuations are gapped . it is also interesting to note that , in the case of the purely repulsive interaction ( [ eq : vdrep ] ) ( that is , when the molecules are polarized along y - axis ) , where @xmath178 for @xmath179 , the composite superfluid , csf , is also possible as long as @xmath180 . this binding caused by repulsion is a specific property of 1d geometry . the renormalization of the bs amplitudes , eq . , should be considered together with the renormalization of the matrix @xmath60 in eq.([rphi ] ) . as explained in the appendix [ a ] , these equations are @xmath181 ^ 2(k_{zz } + k_{z'z ' } - 2k_{zz'})\ ] ] for the off - diagonal terms , @xmath64 , and @xmath182 ^ 2(k_{zz } + k_{z'z ' } - 2k_{zz'})\ ] ] for the diagonal ones . here the constant @xmath183 depends on the type of the short - distance cutoff ( see in ref . @xcite ) . this constant can be absorbed into the definition of @xmath184 by simple rescaling of the amplitudes . it is worth noting that only the pairs @xmath185 such that @xmath186 are involved in eq . ) and eq . ) . in the case of the translational invariance , that is , @xmath168 , these equations become @xmath187 ^ 2\left(k(0 ) - k(z)\right),\ ] ] where @xmath188 and @xmath189 ^ 2\left(k(0 ) - k(z)\right ) , \ ] ] and they should be considered self - consistently together with eqs.([rg111 m ] , [ furm ] , [ fu2rm ] ) . an elementary inspection of eqs.([rg111 m ] , [ furm ] , [ fu2rm ] , [ rg2fm ] , [ rg0 m ] ) shows that @xmath190 is not affected by the rg because @xmath191 . this implies that the field @xmath192 always remains condensed . furthermore , as long as the initial flow of @xmath193 ( described by eq.([rg22 m ] ) ) drives the amplitudes away to @xmath194 , the fixed point for the luttinger matrix is given by @xmath195 , that is , by @xmath196 . if there is a pair of tubes @xmath170 ( @xmath64 ) with filling factors @xmath197 and @xmath198 , the bs harmonic @xmath199 can become relevant , while @xmath154 is irrelevant due to the mismatch of the filling factors . as a consequence , the gapless superflow is possible only in the counter - flow channel . in other words , it is the difference between the two phases which remains gapless . the rg equations for the counterflow can be written for each pair @xmath170 of tubes with the complementary filling factors by simply changing the sign in front of the @xmath153 term in the corresponding equations , eq . , derived above for the complex superfluids ( see details in the appendix [ a ] ) . specifically , we find @xmath200 v_{1;1}(z , z')\;.\ ] ] here , in full analogy with the composite superfluids , the @xmath201 channel can become gapped in the limit @xmath24 . for the case of more than two tubes in the counterflow regime , the dipolar interaction can induce an additional gap in the @xmath202 channel in tubes with identical filling factors . however , a simple count of the remaining gapless phases shows that the gap in the @xmath203 channel does not change their number . indeed , let s consider two sets of tubes , @xmath204 and @xmath205 , so that in the first one the filling factor in each tube is @xmath147 and in the second one it is @xmath92 . then , there will be gaps in the channels @xmath206 for each pair @xmath207 from the first set of @xmath82 tubes and in @xmath208 for each pair @xmath209 from the other set . as a result , there are two total phases from each group left gapless . then , the channels @xmath210 also become gapped due to the counter - flow bs . this leaves just one phase gapless . the described situation has a very simple interpretation : the gaps in tubes with equal filling factors imply formation of a _ pair _ of composite superfluids one per each group of tubes and these composite superfluids further bind in the counterflow regime , as discussed in the section [ scf ] . similarly to the composite superfluids , eq . , the luttinger matrix satisfies ( see appendix [ a ] ) @xmath211 ^ 2(k_{zz } + k_{z'z ' } + 2k_{zz'}),\ ] ] where @xmath64 , and [ rg2cdiag ] = c _ zz \ { [ v_1;1(z , z)]^2(k_zz + k_zz + 2k_zz ) + [ v_1;-1(z , z)]^2(k_zz + k_zz - 2k_zz)}. the first sum here is the contribution from the pairs of tubes with complementary filling factors , and the second one is due to the tubes with same filling factors . finally , we write the above equations for the case of translational symmetry along z - axis . this can be realized when , for example , tubes with even z - coordinates ( @xmath212 ) have filling factor @xmath65 and tubes with odd z ( @xmath213 ) have filling factor @xmath92 ( see fig . [ fig : scf ] ) . then , similarly to the composite superfluid case @xmath214 v_{1;1}(z),\ ] ] where the distance @xmath215 corresponds to pairs of tubes with the complementary filling factors . for the distances @xmath216 , that is , for layers with same filling factors , eq . has to be used . similarly the flow of the matrix of stiffnesses at odd distances @xmath19 is given by @xmath217 ^ 2(k(0 ) + k(z)),\ ] ] while even , non - zero distances @xmath19 are described by eq . , and the diagonal term has contribution from all the pairs of tubes @xmath218 ^ 2(k(0)+k(z ) ) + 2c\sum_{z=2,4,6 , ... } [ v_{1;-1}(z)]^2 ( k(0 ) - k(z)).\ ] ] it is instructive to ignore eq . and , and substitute the initial value into eq . in the limit @xmath24 . for @xmath188 this gives @xmath219 where we have only used the first order term in @xmath24 while expanding . thus , for purely repulsive interaction , eq.([eq : vdrep ] ) , the harmonic @xmath220 is relevant for arbitrary small @xmath20 in a direct analogy with the csf case . furthermore , it is interesting to note that inter - layer attraction @xmath221 , eq.([vdip ] ) , also induces the composite super - counter - fluid as long as the intra - layer repulsion is strong enough , that is , @xmath222 . the analysis of the above equations shows that for even number of layers , the fixed point corresponds to @xmath223 for @xmath224 and @xmath225 for @xmath213 . thus , while @xmath226 , the harmonic at @xmath227 remains condensed ( because @xmath228 ) . this , as discussed earlier , corresponds to the supercounterfluidity in the nearest neighbor layers , with the condensed field @xmath229 . in the absence of inter - tube interactions , hard core bosons can form a checkerboard ( cb ) insulator at filling factor @xmath22 only if the repulsive interaction is strong enough , so that the luttinger parameter @xmath108 is reduced from @xmath111 to @xmath231 ( see ref . this can also be seen from eq . written for @xmath119 , that is , for the intra - tube harmonic @xmath232 . in this case , eq . becomes @xmath233 . in the absence of inter - tube interaction the luttinger matrix becomes diagonal @xmath234 , so that one can write @xmath235 , implying that the critical value @xmath236 . the situation changes dramatically in the presence of inter - tube interaction . at filling @xmath230 harmonics @xmath237 with @xmath188 can become relevant for @xmath111 , as eqs.([rg22 m ] , [ rg44 ] ) indicate . this happens regardless of the sign of the inter - tube interaction even in the limit @xmath24 as long as @xmath238 . accordingly , all pairs of phases @xmath239 , with @xmath64 become gapped , which implies that all the individual phases @xmath101 are gapped . it is possible to make a much stronger statement : for @xmath23 , the @xmath22 insulating state occurs even in the absence of intra - tube interaction , i.e. @xmath240 , and for purely attractive inter - tube interaction @xmath241 . in order to see this , we analyze the rg eq . ( [ rg111 m ] ) which , as the initial flow ( [ rg22 m ] ) indicates , implies relevance of all @xmath193 for @xmath188 . accordingly , as eqs . ( [ rg2fm ] , [ rg0 m ] ) show , the matrix @xmath60 flows toward @xmath242 ( in the limit @xmath243 ) for all @xmath55 except @xmath244 . this means eq.([rg4 ] ) can be approximately rewritten as @xmath245 v_{1;1}(z ) \label{rg4lim}\ ] ] in the limit @xmath243 . keeping in mind that at small @xmath24 the initial value @xmath246 , this equation shows that , even if the renormalization of @xmath247 is ignored , @xmath248 flows to @xmath194 as @xmath249 as long as @xmath23 ( which means that the harmonic @xmath244 is also gapped and @xmath250 must actually flow to @xmath251 ) . as it will be seen below , this conclusion is also consistent with the simulations . the @xmath252 case is special because , in the one - loop approximation , the rg equations for @xmath203 and @xmath253 are independent from the equations for @xmath201 and @xmath247 . accordingly , the equation for @xmath247 predicts that it must flow to a stable fixed point @xmath254 as long as @xmath255 and @xmath256 . this issue will be discussed in greater detail elsewhere . below we will explicitly demonstrate numerically the thresholdless nature of the composite superfluid in the simplest case @xmath252 . the standard worm algorithm ( wa ) @xcite is based on the evaluation of one - particle correlators @xmath257 in imaginary time and the possibility to switch effectively to the functional space of the partition function of the closed world - lines of particles . if @xmath258 , particles ( or holes ) form bound complexes , the efficient simulations can be achieved only through evaluation of the @xmath74-particle correlators . in this case @xmath259 , such an algorithm has been developed in refs.@xcite . the situation becomes more complicated for @xmath260 , when no effective switching to the partition function space can , in general , be achieved . this problem has been resolved in ref . @xcite in the case of no inter - layer tunneling and in ref . @xcite in a more general setting . here the algorithm @xcite ( designed to work in a discrete space - time ) is extended to the quantum case , that is , to continuous time . while avoiding technical details , here we give a general overview of the quantities measured during the simulations . the most general correlator which can effectively describe a phase of @xmath74 bound complexes is the @xmath74-particle correlator a function of 6@xmath74 variables @xmath261 where @xmath262 stands for quantum - statistical averaging with the weight @xmath263 determined by the hamiltonian @xmath264 ( [ eq : h ] ) and @xmath265 , with @xmath266 being bosonic annihilation operator in the space - time point @xmath267 with @xmath268 . the imaginary time dependence is given by the interaction representation defined for an operator @xmath269 as @xmath270 , where @xmath271 is the operator in the schrdinger representation and @xmath272 is the part of the hamiltonian which is diagonal in the fock basis , that is , the interaction part of @xmath264 . evaluation of @xmath273 is based on the random walks of the @xmath274 open _ ends _ of the world - lines , _ worms _ @xcite , controlled by the famous metropolis prescription . the identification of the phases , then , stems from the statistics of the relative distances between the worms , as described in ref . for example , in the csf phase of complexes each composed of @xmath74 particles , all the correlators @xmath275 ( [ corr ] ) with @xmath276 exhibit exponential decay with respect to all the pairs of space - time distances @xmath277 , @xmath278 and @xmath279 where @xmath280 . this behavior is the key signature of insulators with no off - diagonal long range ( or algebraic ) order . a completely different behavior is demonstrated by the correlator @xmath273 . on the one hand , if all the _ ends _ from one set , e.g. , @xmath281 are kept inside a small region , the _ ends _ from the other set will automatically stay close together within some finite radius @xmath282 determining a typical extension of the constituents forming one complex , that is , @xmath283 . on the other hand , the dependence of @xmath273 on the relative space - time distance between the `` centers of mass '' @xmath284 ( defined as @xmath285/m$ ] and @xmath286/m$ ] ) features the off - diagonal long range ( or algebraic ) order . the transition from csf to the standard superfluid ( where @xmath257 is long ranged ) is marked by the divergence of @xmath282 . keeping in mind the specificity of the present system , we evaluated the correlator @xmath287 and kept only one pair of the variables @xmath288 , one from the first set and one from the other , in each tube ( there is no inter - tube tunneling so that each worm stays in its tube ) . in order to realize the `` confinement '' of the first set of variables @xmath289 , we have introduced an artificial configuration weight @xmath290 $ ] , where @xmath291 , @xmath292 are microscopic parameters chosen so that as to maximize the algorithm efficiency . accordingly , the expectation values @xmath293 are evaluated with respect to the weight @xmath294 . as demonstrated in refs.@xcite , the described approach turned out to be very effective in idenifying various phases as well as the universalities of the transitions . it can be easily adjusted to various systems . for example , if considering the bilayer system proposed in ref . @xcite the correlator @xmath295 should be used with @xmath296 pairs of the ends kept in one tube and @xmath297 pairs in the other . here we present results of _ ab - initio _ quantum monte carlo ( qmc ) simulations based on the path integral with @xmath264 from eq.([eq : h ] ) and focusing on demonstrating explicitly the absence of the threshold for the formation of the csf state . unless otherwise noted the simulations were performed in the case of nearest neighbor inter - layer attraction , and in the absence of intra - layer interactions . specifically , we considered the case @xmath252 and compared the result for the renormalized luttinger parameter determined numerically ( through the representations ( [ r],[c ] , [ lut ] ) ) with the prediction of rg . we have also performed simulations of the @xmath298 case within the approach described above and have demonstrated : 1.the formation of the csf phase ; 2 . the existence of the insulating cb state of the chains at the filling @xmath22 and provided data consistent with the absence of the threshold for its formation . absence of the threshold for the phases discussed above implies that , in order to realize them , there is no need to pursue strong dipole - dipole interactions . instead , the size of the system should be made large enough ( and temperature low enough ) so that the effects of small gaps are seen . here we will address the issue of no threshold in detail by _ ab - initio _ simulations of the bilayer system . the goal of this study is to demonstrate this property explicitly . we consider two identical layers located at @xmath299 with @xmath300 . then , the luttinger matrix consists of just two elements @xmath301 and @xmath302 . accordingly , the fourier representation along the z - axis has just two harmonics with @xmath303 , so that eq . ) yields @xmath304 and @xmath305 . as presented in eqs . , , [ pi ] ( ) = , [ 0 ] ( 0)= , in terms of space - time windings @xmath306 , @xmath307 , @xmath308 , @xmath309 . we have determined @xmath253 by qmc for various interactions and system sizes , where the rg scale @xmath310 was identified with the system size @xmath311 , provided the inverse temperature @xmath312 in the atomic units . practically , we have kept @xmath313 so that @xmath314 , in order to ensure space - time symmetry , that is , that the system is in its ground state . as a function of system size @xmath311 for different values of the inter - layer interaction @xmath315 , and in the absence of intra - layer repulsion . , scaledwidth=50.0% ] is shown vs @xmath316 with @xmath317 chosen in such a way as to achieve the collapse on a single curve . the size of the symbols is determined by the statistical error bars . the solid line is the rg solution for the separatrix with the critical value of the luttinger parameter being @xmath318 , eq . ( [ sepm ] ) . inset : plot of the rescaling parameter @xmath319 vs @xmath315 . solid line is the fit by @xmath320 , with @xmath321 . , scaledwidth=50.0% ] our purpose is comparing the numerical dependancies of @xmath322 vs @xmath311 for various interaction strengths @xmath20 with the rg flows . the raw data for @xmath323 is presented in fig . [ fig : k - n2 ] . as it turned out , within the statistical errors of the simulations , the curves of @xmath323 vs @xmath311 for various @xmath324 have been found to belong to one master curve the separatrix of the rg equations - ( discussed in the appendix b ) , which can be represented as [ sepm ] _ s -=2(),_s= -1 . where @xmath325 is a rescaling parameter which can be interpreted as the length @xmath282 the size of a bound dimer . this dependence has been found from rescaling @xmath326 $ ] for each value of the interaction . the result of this procedure is presented in fig . [ fig : master ] . as can be seen from the inset , @xmath327 as a function of the inter - tube interaction diverges as [ corm ] l_0 ~(),v_d 0 where @xmath328 is a constant ( @xmath329 ) . such a divergence proves that the critical value for the formation of the dimer superfluid is @xmath140 . thus , the accurate matching of the numerical data by the rg solution ( [ sepm ] ) over almost 50 orders of magnitude of the ( effective ) distances as well as the dependence ( [ corm ] ) indicate that paired superfluid is formed for infinitesimally small inter - layer interaction strength . such an approach matching numerical solution by the rg flow for finding critical point of berezinskii- kosterlitz - thouless transition @xcite has been pioneered in ref.@xcite . below we present qmc results obtained by the multi - worm algorithm for the case of @xmath298 tubes . as it has already been mentioned , achieving efficient numerical convergence by the approach @xcite in the cases @xmath23 is not possible . instead , the simulations should focus on evaluating the @xmath330body correlator @xmath287 , eq.([corr ] ) . then , the determination of the phases can be based on observing spatial dependencies of the corresponding correlators . we introduce two quantities @xmath331 and @xmath332 which can be viewed as spatial projections of the full correlator @xmath333 where the pair @xmath334 in @xmath335 belongs to the `` primed '' coordinates in the definition ( [ corr ] ) , and @xmath336 in @xmath337 are from the `` unprimed '' and the `` primed '' sets , respectively . specifically , @xmath338 and @xmath339 , with the artificial weight @xmath340 discussed at the beginning of the section [ section : qmc ] . given the definition , @xmath341 must exhibit exponential decay in the @xmath298 csf phase as well as in the insulating phases . the function @xmath337 , while demonstrating the exponential decay in the insulator , should show algebraic behavior in the csf phase . these features are clearly seen in fig . [ fig : d3 ] for @xmath298 identical tubes for two filling factors and in the presence of the full dipolar interaction at @xmath342 . while the main plot clearly shows the csf ( @xmath343 ) , the inset represents the cb insulator ( @xmath22 ) . tubes : qmc data ( points ) and their fits ( lines ) in the presence of dipolar interaction . the filling factors are shown . main pannel : while @xmath341 exhibits the exponential decay @xmath344 , @xmath345 shows algebraic order @xmath346 . inset : in the cb phase both functions are exponentially decaying , @xmath347.,scaledwidth=50.0% ] plotted for all sizes @xmath348 , in the absence of intra - layer repulsion and all inter - layer interaction strengths @xmath315 studied . solid line is the fit by the power - law dependence @xmath349.,scaledwidth=50.0% ] the cb phase is characterized by finite amplitude of the modulation of the density @xmath350 at the wavevector @xmath351 . the rg analysis conducted in sec . [ section : insulators ] indicates that such modulation can occur even in the absence of the intra - layer repulsion due to arbitrary small inter - layer attraction @xmath20 . in other words , the repulsive interaction causing the cb order is to be induced dynamically even if it is not present microscopiocally . this analysis , however , does not predict strength of such interaction . in our simulations without any intra - layer repulsion we were able to resolve the cb modulation only for @xmath352 . furthermore , the structure factor correlator @xmath353 showed a very weak dependence on the system size @xmath311 for the whole range of measurements @xmath354 . in these circumstances conducting the comparison with the rg flow like it was done for the case of the paired superfluid does not appear to be feasible . in other words , a weak dependence of the induced repulsion on @xmath20 does not allow approaching the critical region at small @xmath20 . thus , the observed cb order corresponds to the values of the renormalized luttinger parameter which are already so small that the structure factor @xmath355 becomes essentially independent of @xmath21 , with the factor @xmath356 being a non - universal coefficient ( cf . with the spin@xmath357-chain magnetization modulation in sec . 6 of ref.@xcite ) . however , despite such limitations , there is a feature which is consistent with the thresholdess nature of the cb . [ fig : str ] shows onsite cb contrast measured for all the system sizes and the inter - layer strength studied . the data can be fit by the power law dependence on @xmath20 , @xmath358 , which is consistent with no threshold in @xmath20 . superfluids of hard - core bosons in the multi - tube geometry turn out to be unstable toward forming composite superfluids , supercounterfluids and cb insulators for arbitrary small inter - tube interaction . this conclusion is supported by the bosonization and numerics based on the newly developed _ ab initio _ monte carlo alghorithm . thus , while realizing experimentally such phases , the smallness of the dipolar interaction can be to some extent compensated by enlarging system size . in the context of the emergence of parafermions @xcite in the multi - tube geometry with finite inter - tube tunneling , we find important conducting _ ab initio _ simulations of such systems in order to establish the requirements for their experimental realization . in particular , such simulations are needed to infer how the threshold for the transitions depends on the dipolar strength and the tunneling amplitude . another interesting system proposed in refs.@xcite also requires _ ab initio _ simulations for establishing practical ranges of the interaction and lattice parameters . as bosonization argument indicates , forming a superfluid consisting of complexes of @xmath359 hard core bosons from one tube with @xmath360 such bosons from the other requires exceeding some finite threshold in @xmath20 . indeed , the critical value of the luttinger parameter needed to make the bs harmonic @xmath361 ( [ bs ] ) relevant is @xmath362 , which is significantly smaller than @xmath111 for the hard core bosons even for the lowest non - trivial combination @xmath363 . increasing the intra - layer interaction reduces the @xmath108 value , so that , potentially , it may be possible to realize the phases @xcite . this , however , needs to be checked by the qmc . * acknowledgement * this work was supported by the nsf through a grant to itamp at the harvard - smithsonian center for astrophysics , grant phy1314469 , and by the grant from cuny hpcc under nsf grants cns-0855217 , cns-0958379 and aci-1126113 . here we outline derivation of the rg equations for general ( weak ) interactions . the procedure is a straightforward extension of the standard one ( see , e.g. , refs . @xcite ) . for small interactions , the only relevant harmonics in the backscattering terms ( [ bs ] ) can be those with the lowest integers @xmath141 , that is , @xmath364 , @xmath365 , with @xmath64 . the standard renormalization procedure consists of integrating out small oscillations of the haldane phases @xmath101 @xcite ( from the partition function @xmath366 ) within the spherical shell of @xmath367 between some cutoff @xmath368 and @xmath369 , and further rescaling @xmath370 and @xmath371 , with @xmath372 . in the lowest order ( one - loop approximation ) , this procedure implies independent renormalization of each harmonic . specifically , for the case of @xmath373 one finds @xmath374 v_{m ;- m}(z , z')\ ] ] where @xmath375 implies gaussian average with respect to the action ( [ rgauss ] ) , with the integration performed over the shell of the momenta defined above . in the @xmath376 dimensions , @xmath377 exhibits log - divergence , that is , @xmath378 and it is independent of @xmath379 . then , the rg flow is controlled by @xmath380 or , in a finite system of size @xmath311 , by @xmath381 so that @xmath382 . the mean @xmath383 can be represented in terms of the elements of the luttinger matrix @xmath60 from eq.([rphi ] ) which is the inverse of the matrix @xmath384 from the dual form ( [ rgauss ] ) . thus , eq . ( [ rg1 ] ) becomes @xmath385 v_{m ;- m}(z , z'),\ ] ] where @xmath153 are elements of the matrix @xmath60 . in the case of translational invariance along the @xmath19-axis this equation can be explicitly written in the form ( [ rg111 m ] ) . if one ignores the renormalization of the luttinger matrix , the value of @xmath173 from eq . can be used . for small @xmath20 one can expand eq . in powers of @xmath20 and rewrite eq.([rg11a ] ) as @xmath386 v_{m ;- m}(z),\ ] ] where @xmath128 is given in eq . ) . in the limit @xmath24 and for @xmath387 , the critical value of @xmath108 is @xmath318 . we also note that higher harmonics @xmath388 are irrelevant because the critical value for them in the limit @xmath389 is @xmath390 . the renormalization of the bs amplitudes , eq . ) , is considered together with the renormalization of the inverse of the matrix @xmath391 entering the quadratic form . in the one - loop approximation the main contribution is due to the same bs harmonic , @xmath392 . it generates the term @xmath393 ^ 2 $ ] in the second order with respect to the harmonics @xmath394 belonging to the rg shell , where the signs @xmath395 are correlated . thus , the contributions to the diagonal elements @xmath396 and to the off - diagonal ones @xmath391 where @xmath64 , should be considered independently . following the standard procedure ( see in refs.@xcite ) , the contribution to @xmath391 from the bs amplitude @xmath397 , with @xmath64 , can be represented as @xmath398 where the signs `` @xmath395 '' are correlated ; @xmath53 is a non - universal constant determining type of the short distance cut - off ( see in ref . @xcite ) . the contributions to @xmath399 come from all pairs . specifically , @xmath400 . \label{kzz}\ ] ] eqs . ( [ rg1k],[kzz ] ) , where @xmath375 implies averaging over the gaussian fluctuations within the momentum shell , lead to eqs.([rg2m],[rg2dm],[rg2c],[rg2cdiag ] ) , where it is taken into account that @xmath401 . at @xmath147 the only relevant harmonic is @xmath402 . thus the rg flow affects @xmath253 and @xmath203 only . the corresponding rg equations follow from eqs . , , as @xmath403 where we used the notations @xmath404 , @xmath405 . these equations are the standard kosterlitz - thoulless @xcite rg equations ( see in refs.@xcite ) . the channel @xmath410 is irrelevant as long as @xmath147 . at @xmath230 , or in the case of the complementary filling factors @xmath411 and @xmath412 , the channel @xmath413 must be considered as well . the corresponding rg equations follow from eqs . and eqs . , in the same form as eqs . where now @xmath414 , @xmath415 , with the initial value set as @xmath416 , with @xmath417 . thus , in the case @xmath252 , the channels @xmath201 and @xmath203 are decoupled from each other and are described by the same set of equations . a general solution of the system can be expressed in terms of two constants of integration , @xmath418 , @xmath419 , determined by the initial values of @xmath420 and @xmath421 , which in their turn are set by the microscopic model . if @xmath418 is real , the solution has a form @xmath422,\ , \xi = \frac{1}{g}-1 , \ , f_\eta=4\ln\left(\frac{l}{l_0}\right)\\ f_\eta&\equiv & \ln(\xi^2 ( l ) - \eta^2 ) + \frac{1}{|\eta| } \ln\left(\frac{\xi ( l ) - |\eta|}{\xi ( l ) + |\eta|}\right ) , \nonumber\end{aligned}\ ] ] where @xmath423 and @xmath424 . if @xmath425 , the solution becomes @xmath426,\,\xi = \frac{1}{g}-1 , \ , f_\eta=4\ln\left(\frac{l}{l_0}\right ) , \\ f_\eta & \equiv&\ln(\xi^2 ( l ) + |\eta|^2 ) - \frac{2}{|\eta| } \tan^{-1}\left(\frac{|\eta|}{\xi}\right ) , \nonumber\end{aligned}\ ] ] where @xmath424 . the constants @xmath427 are determined by the dipolar interaction , @xmath20 . if @xmath140 , that is , the hard - core bosons are non - interacting ( except for the hard - core constraint ) , the rg equations are trivially satisfied by @xmath428 , which implies that @xmath429 for @xmath140 . the critical solution ( @xmath430 ) belongs to the separatrix , @xmath431 , @xmath432 , @xmath433 : [ sep ] u_s=|_s| , _ s -=2 ( ) . algebraic order exists in the domain @xmath434 , where @xmath435 flows to the stable fixed point @xmath436 for real @xmath418 satisfying @xmath437 , and @xmath438 . all other initial values correspond to the runaway flows @xmath439 , that is , to the gapped state . as explicitly shown in eq . , small inter - tube attractive interaction @xmath20 lowers @xmath253 below @xmath111 , that is , the initial value of @xmath435 is @xmath440 . it is also clear that the initial bs interaction @xmath203 must also be @xmath441 in this limit . thus , @xmath442 , as follows from eqs . , . it is instructive to discuss the dependence @xmath443 vs @xmath20 . as mentioned already , @xmath444 for @xmath140 and it must become finite as @xmath445 . thus , @xmath443 has a meaning of the correlation length the size of a dimers forming paired superfluid . the type of the dependence can be established from , e.g. , eq .. starting from @xmath440 at @xmath446 , this equation becomes @xmath447 , where @xmath448 is some number of the order of unity . thus , [ cor ] l_0 ~ ( ) , where @xmath449 . as found in our simulations , eq . ( [ corm ] ) , this length , @xmath450 , determines the properties of the paired superfluid . the dependence ( [ cor ] ) should be , on one hand , contrasted with the temperature divergence @xmath451 of the correlation length in classical bkt transition on the approach to the critical temperature @xmath452 @xcite , and , on the other , it should be compared with the divergence of the two - body bound state size @xmath453 in 2d as the attractive potential @xmath454 . d. jaksch , c. bruder , j. i. cirac , c. w. gardiner , and p. zoller , . i. bloch , j. dalibard , and w. zwerger , rev . mod . 80 , 885 ( 2008 ) . p. windpassinger , k. sengstock , . j. catani , l. de sarlo , g. barontini , f. minardi , and m. inguscio , . b. gadway , d. pertot , r. reimann , and d. schneble , . d. m. weld , h. miyake , p. medley , d. e. pritchard , and w. ketterle , . j. m. sage , s. sainis , t. bergeman , and david demille , . ni , s. ospelkaus1 , m. h. g. de miranda , a. peer , b. neyenhuis , j. j. zirbel , s. kotochigova , p. s. julienne , d. s. jin , j. ye , . j. deiglmayr , a. grochola , m. repp , k. mrtlbauer , c. glck , j. lange , o. dulieu , r. wester , and m. weidemller , . k. aikawa , d. akamatsu , m. hayashi , k. oasa , j. kobayashi , p. naidon , t. kishimoto , m. ueda , and s. inouye , . s. ospelkaus1 , k. -k . ni , d. wang , m. h. g. de miranda , b. neyenhuis , g. qumner , p. s. julienne , j. l. bohn , d. s. jin , j. ye , . t. takekoshi , m. debatin , r. rameshan , f. ferlaino , r. grimm , h. c. ngerl , c. r. le sueur , j. m. hutson , p. s. julienne , s. kotochigova , and e. tiemann , . c. -h . wu , j. w. park , p. ahmadi , s. will , and martin w. zwierlein , . m. repp , r. pires , j. ulmanis , r. heck , e. d. kuhnle , m. weidemller , and e. tiemann , . b. yan , s. a. moses , b. gadway , j. p. covey , k. r. a. hazzard , a. m. rey , d. s. jin , j. ye , . t. giamarchi , _ quantum physics in one dimension _ , oxford university press , 2004 . wang , m. d. lukin , e. demler , . g. e. astrakharchik , g. de chiara , g. morigi , j. boronat , . m. klawunn , j. duhme , l. santos , . a. c. potter , e. berg , d .- w . wang , b. i. halperin , and e. demler , . baranov , a. micheli , s. ronen , p. zoller , . j. r. armstrong , n. t. zinner , d. v. fedorov , a. s. jensen , eur . j. d * 66 * , 85 ( 2012 ) . d. hufnagl , r. e. zillich , . b. capogrosso - sansone and a. kuklov , j. low temp . phys . * 165 * , 213 ( 2011 ) . a. tsvelik and a. b. kuklov , . p. lecheminant and h. nonne . e. burovski , g. orso , and t. jolicoeur , phys . * 103 * , 215301 ( 2009 ) . b. wunsch , n.t . zinner,2 , i.b . mekhov , s .- j . huang , d .- w . wang , e. demler , phys.rev.lett . * 107 * , 073201 ( 2011 ) . n. v. prokofev , b. v. svistunov , i. s. tupitsyn , phys . a * 238 * , 253 ( 1998 ) ; sov . a.kuklov , n. prokofev , and b. svistunov , phys . rev . lett . * 92 * , 030403 ( 2004 ) ; . g. syler , b. capogrosso - sansone , n. v. prokofev , b. v. svistunov , . l. mathey , . a. b. kuklov , b. v. svistunov , . e. l. pollock and d. m. ceperley , . |
understanding the nature of dense particle packings is a subject of intense research in the physical , mathematical and biological sciences . the preponderance of previous work has focused on spherical particles , and very little is known about dense polyhedral packings . we formulate the problem of generating dense packings of nonoverlapping , non - tiling polyhedra within an adaptive fundamental cell subject to periodic boundary conditions as an optimization problem , which we call the adaptive shrinking cell ( asc ) scheme . this novel optimization problem is solved here ( using a variety of multi - particle initial configurations ) to find the dense packings of each of the platonic solids in three - dimensional euclidean space , except for the cube , which is the only platonic solid that tiles space . we find the densest known packings of tetrahedra , icosahedra , dodecahedra , and octahedra with densities , , , and , respectively . it is noteworthy that the densest tetrahedral packing possesses _ no long - range order_. unlike the densest tetrahedral packing , which must not be a bravais lattice packing , the densest packings of the other non - tiling platonic solids that we obtain are their previously known optimal ( bravais ) lattice packings . we also derive a simple upper bound on the maximal density of packings of congruent nonspherical particles , and apply it to platonic solids , archimedean solids , superballs and ellipsoids . provided that what we term the asphericity " ( ratio of the circumradius to inradius ) is sufficiently small , the upper bounds are relatively tight and thus close to the corresponding densities of the optimal lattice packings of the centrally symmetric platonic and archimedean solids . our simulation results , rigorous upper bounds , and other theoretical arguments lead us to the conjecture that the densest packings of platonic and archimedean solids with central symmetry are given by their corresponding densest lattice packings . this can be regarded to be the analog of kepler s sphere conjecture for these solids . the truncated tetrahedron is the only non - centrally symmetric archimedean solid , the densest known packing of which is a non - lattice packing with density at least as high as . we discuss the validity of our conjecture to packings of superballs , prisms and anti - prisms as well as to high - dimensional analogs of the platonic solids . in addition , we conjecture that the optimal packing of any convex , congruent polyhedron without central symmetry generally is not a lattice packing . finally , we discuss the possible applications and generalizations of the asc scheme in the predicting the crystal structures of polyhedral nanoparticles and the study of random packings of hard polyhedra . | 0909.0940 | particle packing problems are ancient , dating back to the dawn of civilization . bernal has remarked that `` heaps '' ( particle packings ) were the first things that were ever measured in the form of basketfuls of grain for the purpose of trading or of collection of taxes @xcite . dense packings of hard particles have served as useful models to understand the structure of liquid , glassy and crystal states of matter @xcite , granular media @xcite , and heterogeneous materials @xcite . understanding the symmetries and other mathematical properties of the densest packings in arbitrary dimensions is a problem of long - standing interest in discrete geometry and number theory @xcite . a large collection of nonoverlapping solid objects ( particles ) in @xmath6-dimensional euclidean space @xmath7 is called a packing . the packing density @xmath8 is defined as the fraction of space @xmath7 covered by the particles . a problem that has been a source of fascination to mathematicians and scientists for centuries is the determination of the densest arrangement(s ) of particles that do not tile space and the associated maximal density @xmath9 @xcite . finding the maximal - density packing arrangements is directly relevant to understanding the structure and properties of crystalline equilibrium phases of particle systems as well as their ( zero - temperature ) ground - state structures in low dimensions in which the interactions are characterized by steep repulsions and short - ranged attractions . the preponderance of previous investigations have focused on dense packings of spheres in various dimensions @xcite . for congruent particles in three dimensions , the sphere is the only non - tiling particle for which the densest packing arrangements can be proved @xcite . it is only very recently that attention has turned to finding the maximal - density packing arrangements of nonspherical particles in @xmath0 , including ellipsoids @xcite , tetrahedra @xcite , and superballs @xcite . very little is known about the densest packings of polyhedral particles @xcite . + the platonic solids ( mentioned in plato s timaeus ) are convex polyhedra with faces composed of congruent convex regular polygons . there are exactly five such solids : the tetrahedron ( p1 ) , icosahedron ( p2 ) , dodecahedron ( p3 ) , octahedron ( p4 ) , and cube ( p5 ) ( see fig . [ platonic ] ) @xcite . one major concern in this paper is the determination of the densest packings of each of the platonic solids in three - dimensional euclidean space @xmath0 , except for the cube , which is the only platonic solid that tiles space . it is useful to highlight some basic geometrical properties of the platonic solids that we will employ in subsequent sections of the paper . the dihedral angle @xmath10 is the interior angle between any two face planes and is given by @xmath11 where @xmath12 is the number of sides of each face and @xmath13 is the number of faces meeting at each vertex . thus , @xmath10 is @xmath14 , @xmath15 , @xmath16 , @xmath17 , and @xmath18 , for the tetrahedron , icosahedron , dodecahedron , octahedron , and cube , respectively , where @xmath19 is the golden ratio . thus , since the dihedral angle for the cube is the only one that is a submultiple of @xmath20 , the cube is the only platonic solid that tiles space . we note in passing that in addition to the regular tessellation of space by cubes in the simple cubic lattice arrangement , there are an infinite number of other irregular tessellations of space by cubes @xcite . figure [ random - squares ] shows a portion of a realization of a two - dimensional analog of such an irregular tessellation . every polyhedron has a dual polyhedron with faces and vertices interchanged . the dual of each platonic solid is another platonic solid , and therefore they can be arranged into dual pairs : the tetrahedron is self - dual ( i.e. , its dual is another tetrahedron ) , the icosahedron and dodecahedron form a dual pair , and the octahedron and cube form a dual pair . + an archimedean solid is a highly symmetric , semi - regular convex polyhedron composed of two or more types of regular polygons meeting in identical vertices . there are thirteen archimedean solids : truncated tetrahedron ( a1 ) , truncated icosahedron ( a2 ) , snub cube ( a3 ) , snub dodecahedron ( a4 ) , rhombicosidodecahdron ( a5 ) , truncated icosidodecahdron ( a6 ) , truncated cuboctahedron ( a7 ) , icosidodecahedron ( a8 ) , rhombicuboctahedron ( a9 ) , truncated dodecahedron ( a10 ) , cuboctahedron ( a11 ) , truncated cube ( a12 ) and truncated octahedron ( a13 ) ( see fig . [ archimedean ] ) . note that the truncated octahedron is the only archimedean solid that tiles space . another important observation is that the tetrahedron ( p1 ) and the truncated tetrahedron ( a1 ) are the only platonic and archimedean solids , respectively , that are not _ centrally symmetric_. a particle is centrally symmetric if it has a center @xmath21 that bisects every chord through @xmath21 connecting any two boundary points of the particle , i.e. , the center is a point of inversion symmetry . we will see that the central symmetry of the majority of the platonic and archimedean solids ( p2 p5 , a2 a13 ) distinguish their dense packing arrangements from those of the non - centrally symmetric ones ( p1 and a1 ) in a fundamental way . some basic definitions concerning packings are given here . saturated _ packing is one in which there is no space available to add another particle to the packing . a _ lattice _ @xmath22 in @xmath7 is a subgroup consisting of the integer linear combinations of vectors that constitute a basis for @xmath7 @xcite . a _ lattice packing _ @xmath23 is one in which the centroids of the nonoverlapping particles are located at the points of @xmath22 , each oriented in the same direction . the set of lattice packings is a subset of all possible packings in @xmath7 . in a lattice packing , the space @xmath7 can be geometrically divided into identical regions @xmath24 called _ fundamental cells _ , each of which contains just the centroid of one particle . thus , the density of a lattice packing is given by @xmath25 where @xmath26 is the volume of a @xmath6-dimensional particle and @xmath27 is the volume of a fundamental cell . a more general notion than a lattice packing is a periodic packing . a _ periodic _ packing of congruent particles is obtained by placing a fixed nonoverlapping configuration of @xmath28 particles ( where @xmath29 ) with _ arbitrary orientations _ in each fundamental cell of a lattice @xmath22 . thus , the packing is still periodic under translations by @xmath22 , but the @xmath28 particles can occur anywhere in the chosen fundamental cell subject to the nonoverlap condition . the packing density of a periodic packing is given by @xmath30 where @xmath31 is the number density , i.e. , the number of particles per unit volume . the determination of the maximal - density arrangements of non - tiling polyhedral particles is a notoriously difficult problem , especially since such extremal structures will generally be non - bravais - lattice packings . computer simulations that seek the maximal - density packings can be an indispensable tool , especially if they can incorporate collective motions of the particles in order to obtain , in principle , the highest possible densities . however , the challenge presented by polyhedral particles in @xmath0 is the non - smooth ( i.e. , nonanalytic ) nature of the particle shape . in the case of smoothly - shaped particles , such as spheres , ellipsoids and superballs , one can construct analytic overlap potential functions " for the particles @xcite and hence one can employ efficient collision - driven molecular dynamics ( md ) growth packing algorithms that inherently involve collective particle motions @xcite . the fact that analytic overlap potential functions can not be constructed for polyhedral particles prevents us from using event - driven md growth methods to study such systems . in this paper , we devise a novel optimization scheme , called the adaptive shrinking cell ( asc ) , that can be applied to generate dense packings of polyhedra in @xmath0 . we employ it specifically to obtain the densest known packings of tetrahedra , icosahedra , dodecahedra , and octahedra with densities @xmath32 , @xmath2 , @xmath3 , and @xmath4 , respectively . the result for tetrahedra improves upon the density reported in our recent investigation @xcite . unlike the densest tetrahedral packing , which must not be a bravais lattice packing @xcite , the densest packings of the other non - tiling platonic solids that we obtain are their corresponding densest lattice packings @xcite . we also derive a simple upper bound on the maximal density of packings of congruent nonspherical particles , and apply it to platonic solids , archimedean solids , superballs and ellipsoids . we introduce the asphericity " parameter @xmath33 ( ratio of the circumradius to inradius ) to show that when @xmath33 is sufficiently small , the upper bounds are relatively tight and thus close to the corresponding densities of the optimal lattice packings of octahedra , dodecahedra and icosahedra as well as of the majority of the archimedean solids with central symmetry . our simulation results as well as theoretical arguments lead us to conjecture that the densest packings of platonic and archimedean solids with central symmetry are given by their corresponding densest lattice packings . this can be regarded to be the analog of kepler s sphere conjecture for these solids . the truncated tetrahedron is the only non - centrally symmetric archimedean solid , the densest known packing of which is a non - lattice packing with density as high as @xmath5 . our work also suggests that the optimal packings of superballs are their corresponding densest lattice packings . in a recent letter @xcite , we briefly reported the densest known packings of the non - tiling platonic solids obtained using the asc algorithm and proposed the aforementioned conjecture concerning the optimal packings of the platonic and archimedean solids . in this paper , we expand on theoretical and computational details and report additional new results . in particular , we provide comprehensive details about the asc scheme and the simulation results ( sec . ii ) , including a discussion about the various initial configurations we used for the asc algorithm ( sec . moreover , we have improved on the highest tetrahedral packing density reported in ref . @xcite ( i.e. , from @xmath34 to @xmath1 ) , by exploring a broad range of dynamical parameters for the algorithm and initial configurations . certain pair statistics of the densest known packing of tetrahedra ( e.g. , the contact number , the centroidal correlation function and the face - normal correlation function ) are given ( sec . it is noteworthy that the densest tetrahedral packing is a non - bravais structure with a complex periodic cell and possesses _ no long - range order_. the initial configurations for the icosahedral , dodecahedral and octahedral packings are described and the numerical challenges in producing dense packings of such polyhedral particles are discussed ( sec . in addition , a detailed derivation of the upper bound and tables containing the geometrical characteristics of the platonic and archimedean solids as well as their upper bound values are given ( sec . the upper bound is also applied to superballs and ellipsoids ( which was not done in ref . moreover , we provide the major elements of a possible proof of our conjecture ( sec . we also discuss how our conjecture could be generalized to other centrally symmetric polyhedral particles such as prisms and anti - prisms as well as high - dimensional analogs of the platonic solids . our work also naturally leads to another conjecture reported here for the first time , namely , the optimal packing of any convex , congruent polyhedron without central symmetry is generally not a ( bravais ) lattice packing ( sec . furthermore , we discuss the possible applications and generalizations of the asc scheme to predict the crystal structures of polyhedral nanoparticles and to the study of random packings of hard polyhedra . finally , we collect in appendices basic packing characteristics of various optimal lattice and non - lattice packings of polyhedra ( including lattice vectors ) that have been scattered throughout the literature and provide lattice vectors and other characteristics of the densest known packings of tetrahedra ( obtained here ) and truncated tetrahedra . we formulate the problem of generating dense packings of nonoverlapping , non - tiling polyhedra within a fundamental cell in @xmath0 subject to periodic boundary conditions as an optimization problem . in particular , the objective function is taken to be the negative of the packing density @xmath8 . starting from an initial unsaturated packing configuration of particles of fixed size in the fundamental cell , the positions and orientations of the polyhedra are design variables for the optimization . importantly , we also allow the boundary of the fundamental cell to deform macroscopically as well as compress or expand ( while keeping the particles fixed in size ) such that there is a net compression ( increase of the density of the packing ) in the final state @xcite . thus , the deformation and compression / expansion of the cell boundary , which we call the _ adaptive fundamental cell _ , are also design variables . we are not aware of any packing algorithm that employs both a _ sequential _ search of the configurational space of the particles and the space of lattices via an adaptive fundamental cell that shrinks on average to obtain dense packings . we will call this optimization scheme the adaptive shrinking cell ( asc ) . we will see that the asc optimization scheme allows for some desired collective motions of the particles to find the optimal lattice for the periodic cell . this is to be contrasted with previous treatments that use a fixed shape for the fundamental cell , which may or may not be the optimal shape . figure [ fig_cell ] illustrates a simple sequence of configuration changes for a four - particle system ( which is explained in more detail in sec . ii.a ) . by efficiently exploring the design - variable space ( dvs ) , which consists of the particle configurational space as well as the space of lattices via an adaptive fundamental cell , the asc scheme enables one to find a point in the dvs in the neighborhood of the starting point that has a higher packing density than the initial density . the process is continued until the deepest minimum of the objective function ( a maximum of packing density ) is obtained , which could be either a local or global optimum , depending on the particle shapes . @xmath35{deform_cell_a2.eps } & \includegraphics[width=5.0cm , keepaspectratio]{deform_cell_b2.eps } & \includegraphics[width=5.0cm , keepaspectratio]{deform_cell_c2.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } & \mbox{\bf ( c ) } \end{array}$ ] the asc optimization problem could be solved using various techniques , depending on the shapes of the particles . for example , for spheres , one can linearize the objective function and the constraints , and we have found that linear programming techniques can efficiently produce optimal solutions for such asc problems @xcite . however , for hard polyhedra the non - overlapping conditions given by the separation axis theorem ( discussed in detail below ) involve at least quartic inequalities , which makes it inefficient to solve even using nonlinear programming methods . here , for polyhedral particles , we solve the asc optimization problem using a stochastic procedure , i.e. , monte carlo ( mc ) method with a metropolis acceptance rule for trial moves to search the dvs efficiently . however , it is important to distinguish our procedure that incorporates deformation and compression / expansion of the fundamental cell ( i.e. , the space of lattices ) as design variables from previous mc hard - particle packing algorithms @xcite . in standard mc simulations , arbitrarily selected individual particle is given random displacement or rotation . this sequential movement method is not able to account for any collective motions of the particles , which is crucial to increasing the packing density , as pointed out in sec . i. in our procedure , the deformation / compression / expansion of the boundary at least in part allows for collective particle motions in a direction leading to higher packing density . moreover , it is the overall compression of the fundamental cell that causes the packing density to increase , not the growth of the particles as in most md and mc hard - particle packing algorithms @xcite . at first glance , one might surmise that an algorithm that employs particle growth with an adaptive non - shrinking fundamental cell is equivalent to our choice of fixing the particle size while allowing the cell to shrink on average . the former is computationally less efficient than the latter for polyhedral particles . specifically , growth of polyhedral particles requires manipulating the coordinates of the vertices of each particle , and thus involves at least @xmath36 numerical operations , where @xmath6 is the spatial dimension , @xmath28 is the total number of particles in the system and @xmath37 is the number of vertices per particle . it is much more computationally expensive to use growing / shrinking particles as trial moves for the optimization scheme , especially when the number of particles is large , compared to our adaptive fundamental cell approach , which only requires manipulating @xmath38 strain components . in the asc scheme , the macroscopic deformation and compression / expansion of the fundamental cell of the lattice is completely specified by a strain tensor . since we only consider small deformations , linear strain analysis can be applied here . starting from an initial configuration , a trial configuration can be generated by moving ( translating and rotating ) a randomly chosen particle or by a random macroscopic deformation and compression / expansion of the fundamental cell . if any two particles overlap , the trial configuration is rejected ; otherwise , if the fundamental cell shrinks in size ( which makes the density @xmath8 higher ) , the trial configuration is accepted . on the other hand , if the cell expands in size , the trial configuration is accepted with a specified probability @xmath39 , which decreases as @xmath8 increases and approaches zero when the jamming limit @xcite ( i.e. , locally maximal dense packing ) is reached . in particular , we find @xmath39 , with an initial value @xmath40 , decreasing as a power law with exponent equal to -1 works well for most systems that we studied . the particle motion is equally likely to be a translation or a rotation . the ratio of the number of particle motions to the number of boundary trial moves should be greater than unity ( especially towards the end of the simulation ) , since compressing a dense packing could result in many overlaps between the particles . depending on the initial density , the magnitudes of the particle motions and strain components ( e.g. , the dynamical parameters ) need to be chosen carefully to avoid the system getting stuck in some shallow local minimum . the parameters should also be adjusted accordingly as the packing density increases , especially towards the jamming point . the number of total monte - carlo moves per particle is of the order of @xmath41 . in the ensuing subsections , we describe in detail how we implement particle motions as well as the deformation and compression / expansion of the fundamental cell , and precise check for interparticle overlaps using the separation axis theorem @xcite . we also discuss the cell method and nearest - neighbor lists that are employed to speed up the simulations . the fact that a polyhedron is the convex hull of its vertices makes the set of vertices a useful geometrical representation of a such particle . it is convenient to choose the origin of the local coordinate system for the vertices to be the centroid of the polyhedron . other important geometrical properties of the polyhedron , such as its faces and edges , can be represented as certain subsets of the vertices . let the euclidean position of the centroid of the @xmath42th particle be @xmath43 ( @xmath44 ) . a translational motion of the particle centroid can be obtained by generating a randomly oriented displacement @xmath45 with small magnitude ( @xmath46 of the characteristic length of the particle ) , i.e. , @xmath47 a rotational motion can be generated by rotating the particle ( all of its vertices ) along a randomly chosen axis ( passing through its centroid ) by a random small angle @xmath10 . let the vector @xmath48 originating at the centroid denote the vertex @xmath49 . we have @xmath50 where @xmath51 and @xmath52 are the components of @xmath48 perpendicular and parallel to the rotation axis , respectively , and @xmath53 is the rotation matrix , i.e. , @xmath54.\ ] ] if this motion does not result in the overlap with another particle , the trial move is accepted ; otherwise it is rejected . for a lattice - based periodic packing , the fundamental cell is specified by the lattice vectors @xmath55 ( @xmath56 ) . recall that the euclidean coordinates of the particle centroids are @xmath43 ( @xmath44 ) . the relative coordinates of the centroids with respect to the lattice vectors are given by @xmath57 where @xmath58 $ ] . the adaptive fundamental cell allows for a small strain of the fundamental cell , including both volume and shape changes , which is represented by a symmetric strain tensor @xmath59 , i.e. , @xmath60 where @xmath61,\ ] ] and the new fundamental cell ( lattice vectors ) are given by @xmath62 substituting the above equation into eq . ( [ eq1 ] ) , we have @xmath63 thus , the strain of the fundamental cell corresponds to non - trivial _ collective _ motions of the particle centroids ( see fig . [ fig_cell ] ) . in general , the translational motions of the particles contain the contributions from a random independent part ( given by eq . ( [ eq1 ] ) ) and the collective motion imposed by the adaptive fundamental cell . it is this collective motion that enables the algorithm to explore the configuration space more efficiently and to produce highly dense packings . hard polyhedron particles , unlike spheres , ellipsoids and superballs , do not possess simple overlap potential functions . thus , the check of overlapping for such particles requires other techniques . two convex objects are separated in space if and only if there exist an axis , on which the line segments defined by projections of the two objects do not overlap ( see fig . [ fig_axis ] ) . this statement is usually referred to as the _ separation axis theorem _ ( sat ) @xcite . @xmath64{sep_axi.eps } \\ \end{array}$ ] for convex polyhedra , the above theorem has a simpler version . since a polyhedron is completely defined by its vertices , the line segment on the axis is defined by the most separated two projections of the vertices . moreover , the axis is either perpendicular to one of the faces or perpendicular to a pair of edges , one from each polyhedron . this reduces the number of axes that need to be checked from infinity to @xmath65 $ ] , where @xmath66 and @xmath24 is the number of edges and faces of the polyhedron , respectively . here we employ the sat to check for interparticle overlaps numerically up to the highest machine precision . for polyhedra whose circumscribed and inscribed spheres are well defined , two particles are guaranteed to overlap if the centroidal separation is smaller than the inscribed diameter and guaranteed not to overlap if the centroidal separation is larger than the circumscribed diameter . this pre - check dramatically speeds up the simulations starting from configurations with low densities . @xmath64{periodic.eps } \\ \end{array}$ ] the well - developed cell method @xcite for particle - system simulation is used here to speed up the process . however , the simulation box ( fundamental cell ) will not remain a cube during the simulation and several conventional techniques developed for cubic box need modifications . in particular , to obtain the minimum image distances , all surrounding boxes need to be checked explicitly ( see fig . [ fig_periodic ] ) . with the help of the relative - coordinate representation of the particle centroids , the separation vector of a pair of particles is given by @xmath67 where @xmath68 is a vector with the values of components to be 1 , 0 or -1 . note @xmath68 can also be considered as the index of the box that the image particle is in . the cells are taken to be the same shapes as the simulation box . partitions of the particle centroids into the cells are also convenient for relative coordinate representations , i.e. , the index of the cell ( represented by a vector ) @xmath69 for particle @xmath49 is given by @xmath70,\ ] ] where @xmath71 $ ] gives the smallest integer part of @xmath72 and @xmath73 is the number of cells . note that the boundary deformations and rotations of particles do not affect the cells in which the particles are situated ; only translations may cause transitions between the different cells . for dense packings , we make use of a near - neighbor list ( nnl ) to further improve the efficiency of the algorithm @xcite . in particular , when the packing density is high , each particle is trapped " in a `` cage '' formed by its near neighbors . these near - neighbor configurations are practically stationary , i.e. , the particles undergo very small jiggling " motions . thus , to check for overlapping between the particles , one only needs to consider a particles nearest neighbors . if a large magnitude translation is made , the nnl needs to be set up again . here we apply the asc scheme to obtain the densest known packings of the non - tiling platonic solids . due to its lack of central symmetry , the tetrahedron presents the greatest challenge for the numerical solution procedure of the asc scheme because of its tendency to get stuck in local ( density ) minima , which is a consequence of the associated rugged " energy or , more precisely , density " landscape ( e.g. , the packing density as function of the centroidal positions and orientations of all of the particles in the fundamental cell ) . hence , the choice of initial configurations becomes crucial in getting dense tetrahedral packings . by contrast , the central symmetry of the octahedron , dodecahedron and icosahedron results in density landscapes for the numerical procedure that are appreciably less rugged than that for tetrahedral packings . the determination of the densest packings of regular tetrahedra is part of the 18th problem of hilbert s famous set of problems . it is of interest to note that the densest ( bravais ) lattice packing of tetrahedra ( which requires all of the tetrahedra to have the same orientations ) has the relatively low density @xmath74 and each tetrahedron touches 14 others @xcite . recently , conway and torquato showed that the densest packings of regular tetrahedra must not be a bravais lattice packings , and found packings with density as large as @xmath75 @xcite . one such packing is based upon the filling of imaginary " icosahedra with the densest arrangement of 20 tetrahedra and then employing the densest lattice packing of icosahedra . a slightly higher density was achieved by a perturbation of the so - called welsh " packings with density @xmath76 @xcite . using tetrahedral " dice , chaikin et al . @xcite experimentally generated jammed disordered packings of such dice with @xmath77 . even though the dice are not perfect tetrahedra ( because the vertices and edges are slightly rounded ) , these experimental results suggested that the densest packings of tetrahedra could exceed the highest densities reported by conway and torquato . indeed , chen @xcite has recently discovered a periodic packing of tetrahedra with @xmath78 @xcite . we will call this the wagon - wheels " packing because the basic subunits consist of two orthogonally intersecting wagon " wheels . a wagon wheel " consists of five contacting tetrahedra packed around a common edge ( see fig . 1a of ref . ) . .some packing characteristics of certain known tetrahedral packings . here @xmath8 is the packing density , @xmath28 is the number of particles in the fundamental cell , and @xmath79 is the average contact number per particle . [ cols="^,^,^,^,^",options="header " , ] [ arch ] the upper bound ( [ bound ] ) in the case of superballs @xcite can be expressed analytically for all values of the deformation parameter @xmath12 . a three - dimensional _ superball _ is a centrally symmetric body in @xmath0 occupying the region @xmath80 where @xmath81 @xmath82 are cartesian coordinates and @xmath83 is the _ deformation parameter _ , which indicates to what extent the particle shape has deformed from that of a sphere ( @xmath84 ) . a superball can possess two types of shape anisotropy : cubic - like shapes for @xmath85 , with @xmath86 corresponding to the perfect cube , and octahedral - like shapes for @xmath87 , with @xmath88 corresponding to the perfect regular octahedron . in ref . , event - driven molecular dynamics growth algorithms as well as theoretical arguments led to the conjecture that the densest packings of superballs for all convex shapes ( @xmath89 ) are certain lattice packings , depending on the value of the deformation parameter @xmath12 . for convex superballs in the octahedral regime ( @xmath90 ) , the upper bound ( [ bound ] ) is explicitly given by @xmath91 similarly , for superballs in the cubic regime ( @xmath92 ) , the upper bound ( [ bound ] ) is given by @xmath93 where @xmath94 and @xmath95 is the euler gamma function . here and in the subsequent paragraphs concerning ellipsoids , we only state the nontrivial part of the upper bound ( [ bound ] ) . for @xmath12 near the sphere point ( @xmath84 ) , the densest lattice packings @xcite have densities that lie relatively close to the corresponding upper - bound values . for example , for @xmath96 and 0.97 , @xmath97 , and @xmath98 , respectively , which is to be compared to @xmath99 , and @xmath100 , respectively . for @xmath101 and 1.03 , @xmath102 , and @xmath103 , respectively , which is to be compared to @xmath104 , and @xmath105 , respectively . the fact that the constructed densest lattice packings of superballs have density @xmath106 relatively close to the upper bound value @xmath107 strengthens the arguments made in ref . that suggested that the optimal packings are in fact given by these arrangements . if the optimal packings around @xmath84 are indeed the lattice packings , it would be surprising that for other values of @xmath12 a continuous deformation of the superballs would result in a transition from lattice packings to denser non - lattice packings . in the case of ellipsoids in which the the ratio of the three semiaxes are given by @xmath108 ( @xmath109 ) , the upper bound ( [ bound ] ) becomes @xmath110 for prolate spheroids @xmath111 , and the bound becomes @xmath112 for oblate spheroids with @xmath113 , the bound is given by @xmath114 the upper bounds for ellipsoids generally do not work as well as those for the centrally symmetric platonic and archimedean solids , and superballs . this is due to the fact that the three principle directions ( axes ) are not equivalent for an ellipsoid , i.e. , it generally possesses different semiaxes . note that the three equivalent principle ( orthogonal ) axes ( directions ) of a centrally symmetric particle are those directions that are two - fold rotational symmetry axes such that the distances from the particle centroids to the particle surfaces are equal . in addition , the asphericity of an ellipsoid increases linearly as the largest aspect ratio @xmath115 increases without limit . by contrast , the three principle directions are equivalent for the centrally - symmetric platonic and archimedean solids as well as for superballs , for which the asphericity is always bounded and close to unity ( see tables [ plat ] and [ arch ] ) . thus , we see that an asphericity value close to unity is a necessary condition to have a centrally symmetric particle in which the three principle directions are equivalent . we have formulated the problem of generating dense packings of nonoverlapping polyhedra within an adaptive fundamental cell subject to periodic boundary conditions as an optimization problem , which we call the adaptive shrinking cell ( asc ) scheme . the procedure allows both a sequential search of the configurational space of the particles and the space of lattices via an adaptive fundamental cell that shrinks on average to obtain dense packings . we have applied the asc to generate the densest known packings of the platonic solids . for tetrahedra , we find a packing with density @xmath116 , which is a periodic ( non - bravais lattice ) packing with a complex basis . unlike the other platonic solids , finding dense packings of tetrahedra with our algorithm requires having good initial configurations . the densest packing was found using 314 particles in a rhombical fundamental cell that is similar to that of the hexagonal close packing . as we stressed in our earlier work @xcite and continue to confirm in this paper , it is possible that denser packings of tetrahedra will involve increasingly larger numbers of particles in the fundamental cell . in fact , the higher density found here is realized by a larger periodic packing ( 314 particles per cell ) than the one reported in ref . @xcite ( 72 particles per cell ) with density @xmath117 . it is apparent the obtained densest known tetrahedral packing is disordered in the sense that it possesses no long - range order , at least on the scale of the simulation box . this packing can be considered to be a disordered `` mixture '' of distorted wagon wheels and individual tetrahedra and it is distinct from the one reported in ref . @xcite in both the arrangement of the wagon wheels and the extent of the distortion of the wagon wheels . although we can not rule out the possibility of the existence of denser packings involving a more ordered arrangement of the wagon wheels and individual tetrahedra , it is reasonable to expect that wagon wheels will be key building blocks in denser packings and should at least be slightly distorted to fill the interparticle gaps in wagon - wheel clusters more efficiently , which necessarily introduces a certain degree of disorder to the packing . if denser packings of tetrahedra must involve larger number of particles without long - range order , than it raises the amazing prospect that the densest packings of tetrahedra might be truly disordered , due to the geometrical frustration associated with the lack of central symmetry of a tetrahedron and that tetrahedra can not tile space . this would be the first example of a maximally dense packing of congruent convex three - dimensional particles without long - range order . however , we can not offer definitive conclusions about this possibility at this stage . it is clear that in future work in the search for denser packings , increasingly larger number of tetrahedra must be considered , which can only be studied using greater computational resources . our simulation results and rigorous upper bounds strongly suggest that the optimal lattice packings of the centrally symmetric platonic solids ( octahedra , dodecahedra and icosahedra ) are indeed the densest packings of these particles , especially since these arise from a variety of initial `` dilute '' multi - particle configurations within the fundamental cell @xcite . it is noteworthy that a different simulation procedure ( an event - driven molecular dynamics growth algorithm with an adaptive fundamental cell ) has recently been used to demonstrate that the densest packings of octahedral - like superballs ( which contains the perfect octahedron ) are likely to be the optimal lattice packings @xcite . moreover , the fact that the optimal lattice packing densities of certain centrally symmetric nonspherical particles , such as the archimedean solids with central symmetry and convex superballs , are relatively close to their upper bounds as well as other theoretical arguments given below suggest that the densest packings of these particles may also be given by their optimal lattice packings . it is crucial to stress that the nonspherical particles in this family do not deviate appreciably from a sphere , i.e. , their corresponding asphericity @xmath33 is always bounded and relatively close to unity , and , moreover , their three principle axes ( directions ) are equivalent . this is in contrast to the ellipsoid , which , although centrally symmetric , generally possesses three principle axes ( directions ) that are inequivalent and its asphericity can increase without limit as the largest aspect ratio grows . these characteristics suggest that the densest ellipsoid arrangements are non - lattice packings , which indeed has been verified @xcite . our simulation results and the ensuing theoretical arguments lead to the following conjecture : conjecture 1 : _ the densest packings of the centrally symmetric platonic solids are given by their corresponding optimal lattice packings . _ we now sketch what could be the major elements of a proof of this conjecture . in the case of each of the platonic solids , face - to - face contacts are favored over vertex - to - face contacts to achieve higher packing densities , since the former allows the particle centroids to get closer to one another . such contacting neighbor configurations around each particle reduce the volume of the corresponding convex hull joining the centroids of the contacting particles , and the fraction of space covered by the particles within this convex hull should be increased ; see fig . [ octagon_pack ] for a two - dimensional illustration . @xmath118{octagon_a.eps } & \includegraphics[height=3.0cm , keepaspectratio]{octagon_b.eps } & \includegraphics[height=3.0cm , keepaspectratio]{octagon_c.eps } & \includegraphics[height=3.0cm , keepaspectratio]{octagon_d.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } & \mbox{\bf ( c ) } & \mbox{\bf ( d ) } \end{array}$ ] to achieve the densest packing , the fraction of space covered by the particles within the convex hull should ideally be maximized and so should the number of face - to - face contacts per particle . of course , this is only a local criterion that may not be consistent with the densest global packing . however , it will be seen that the equivalence of the three principle directions of a centrally symmetric platonic solid is crucial for this local optimization criterion to be consistent with the globally densest packing . it is noteworthy that orienting each of the particles in the packings of centrally symmetric platonic solids enable a larger number of face - to - face contacts and thus allows the maximal fraction of space covered by the particles within the convex hull . for example , a particle with @xmath24 faces , possesses @xmath119 families of axes that go through the centroid of the particle and intersect the centrally - symmetric face pairs such that the particles ( in the same orientation ) with their centroids arranged on these axes form face - to - face contacts . the requirement that the particles have the same orientation is globally consistent with a lattice packing . indeed , in the optimal lattice packings of the centrally symmetric platonic solids , each particle has the maximum number of face - to - face contacts that could be obtained without violating the impenetrability condition . it is highly unlikely that such particles possessing three equivalent principle directions and aligned in the same direction could form a more complicated non - lattice periodic packings with densities that are larger than the optimal lattice packings . such non - lattice packings would arise to take advantage of the rotational degrees of freedom . by requiring particle alignments , only translational degrees of freedom remain and hence optimization over these degrees of freedom would lead to the globally optimal packings , which should be lattice packings . constraining rotational degrees of freedom in this way in a non - lattice packing would at best lead to a local optimum in density . this conclusion is clearly supported by our simulations , which use multiple - particle configurations in the fundamental cell and only produce the optimal lattice packings . in the aforementioned arguments , a key step that is difficult to prove is the observation that alignment of each of the particles maximizes the number of face - to - face contacts and thus the fraction of space covered by the particles within the convex hull joining the centroids of the contacting neighbors around a central polyhedron . a rigorous verification of this step would lead to the consistency of the aforementioned local and global optimization criteria . it is noteworthy that in two dimensions , our local criterion to achieve the optimal packings of centrally symmetric regular polygons ( the two - dimensional counterparts of the centrally symmetric platonic solids ) amounts to the identification of optimal neighbor configurations with the maximal edge - to - edge contacts per particle such that the convex hull joining the centroids of the contacting neighbors is minimal . this statement is rigorously true in two dimensions because it is tantamount to a theorem due to fejes tth @xcite , which states that the densest packing of a centrally symmetric particle in two dimensions can be obtained by circumscribing the particle with the minimal centrally symmetric hexagon , which of course tiles @xmath120 . finding the minimal circumscribing centrally symmetric hexagon is equivalent to finding the minimal area convex hull joining the centroids of the six contacting particles ( the maximal number can be obtained in two dimensions for centrally symmetric particles ) . the fact that local optimality is consistent with the global optimality in two dimensions ( e.g. , a centrally symmetric hexagon can always tessellate the space ) does not hold in three dimensions . although we have presented a three - dimensional generalization of the fejes tth theorem , by replacing the minimal hexagon with the minimal convex - hull volume , as pointed out before , it is extremely difficult to provide a rigorous proof for it . although conjecture 1 applies to the centrally symmetric platonic solids , all of our arguments apply as well to each of the centrally symmetric archimedean solids . it can not be true for the truncated tetrahedron , which is the only non - centrally symmetric archimedean solid . specifically , it immediately follows from the results of ref . that a non - lattice packing of truncated tetrahedra can be constructed based on the `` primitive welsh '' tessellation ( i.e. , by removing the small regular tetrahedra ) that possesses the density @xmath121 , which is appreciably larger than the optimal - lattice - packing density of @xmath122 , and contains two particle centroids per fundamental cell . in fact , given that the truncated tetrahedra can not tile space , the density of the `` primitive welsh '' packing of truncated tetrahedra is so large that it may be the optimal packing of such particles . the lattice vectors of this periodic packing are given in appendix c. since the arguments used to justify conjecture 1 apply equally well to the 12 centrally symmetric archimedean solids , we are led to the following more general conjecture : conjecture 2 : _ the densest packings of the centrally symmetric platonic and archimedean solids are given by their corresponding optimal lattice packings . _ the aforementioned arguments can also be extended to the case of superballs , but here the local principle curvatures at the contacting points should be sufficiently small so as to maximize the fraction of space covered by the particles within the convex hull joining the centroids of the neighbors . the central symmetry and equivalence of the three principle directions of superballs means that dense packings of such objects are favored when the particles are aligned , which again leads to consistency between local and global optimality . the fact that the optimal lattice packing densities of superballs is relatively close to the upper - bound values , at least around the sphere point , as well as results from previous molecular dynamics simulations @xcite also strongly suggest that the densest packings of these particles are given by their corresponding optimal lattice packings . it is noteworthy that under the assumption that conjecture 2 is valid , one has upper bounds that are the complete analog of ( [ lemma ] ) , i.e. , @xmath123 , \label{lemma2}\ ] ] where @xmath124 is the volume of an appropriately chosen centrally symmetric platonic or archimedean solid and @xmath125 is the corresponding optimal density for such a solid , which according to conjecture 2 is the optimal lattice packing . this bound will generally be sharper than bound ( [ lemma ] ) because the reference optimal packing is less symmetric than the sphere . for example , the conjectured bound ( [ lemma2 ] ) will be sharp for slightly deformed platonic and archimedean solids or nonspherical particles derived by smoothing the vertices , edges and faces of polyhedra . @xmath126{cut_square_a.eps } & \includegraphics[height=4.35cm , keepaspectratio]{cut_square_b.eps } \\ \mbox{\bf ( a ) } & \mbox{\bf ( b ) } \end{array}$ ] our work also naturally leads to another conjecture : conjecture 3 : _ the optimal packing of any convex , congruent polyhedron without central symmetry generally is not a ( bravais ) lattice packing . _ in other words , the set of such polyhedra whose optimal packing is not a lattice is overwhelmingly larger than the set whose optimal packing is a lattice . we have seen that because the regular tetrahedron and truncated tetrahedron lack central symmetry , dense packings of such objects favor face - to - face contacts . such orientations immediately eliminate the possibility that lattice packings ( in which particles must have the same orientations ) are optimal . similarly , it is very plausible that dense packings of most convex , congruent polyhedra without central symmetry are facilitated by face - to - face contacts and hence the optimal packings can not be lattices . for example , consider a square with one missing corner , i.e. , an isosceles triangle with a right angle ( see fig . [ cut_square ] ) . at first glance , one might surmise that if the missing piece is sufficiently small , the original lattice packing should still be optimal or nearly optimal ( see fig . [ cut_square]a ) , since lattice packings are optimal for squares . however , no matter how small the missing piece may be , a periodic packing in which the fundamental cell contains two pentagons can be constructed that tile the plane ( see fig . [ cut_square]b ) . this is done by taking advantage of the asymmetry of the particle . thus , we see from this counterintuitive example that if the particle does not possess central symmetry , it is possible to exploit its rotational degrees of freedom to yield a periodic packing with a complex basis that are generally denser than the optimal lattice packing . on the other hand , there are special cases where the lattice will be optimal , such as for the rhombic dodecahedron that has one corner clipped @xcite . however , these special cases are overwhelmed in number by those whose optimal packings are not lattices . if conjecture 3 is valid , it also applies to nonspherical particles derived by smoothing the vertices , edges and faces of polyhedra provided that the particle curvature at face - to - face contacts is sufficiently small . it should not go unnoticed that the densest packings of all of the platonic and archimedean solids reported here as well as the densest known packings of superballs @xcite and ellipsoids @xcite have densities that exceed the optimal sphere packing density @xmath127 . these results are consistent with a conjecture of ulam who proposed , without any justification , that the optimal density for packing congruent spheres is smaller than that for any other convex body @xcite . the sphere is perfectly isotropic with an asphericity @xmath33 of unity , and therefore , as noted earlier , its rotational degrees of freedom are irrelevant in affecting its packing characteristics . on the other hand , each of the aforementioned convex non - spherical particles break the continuous rotational symmetry of the sphere and thus its broken symmetry can be exploited to yield the densest possible packings . however , broken rotational symmetry in and of itself may not be sufficient to satisfy ulam s conjecture if the convex particle has a little or no symmetry . it will also be interesting to determine whether our conjecture can be extended to other polyhedral packings . the infinite families of prisms and antiprisms @xcite provide such a class of packings . prism _ is a polyhedron having bases that are parallel , congruent polygons and sides that are parallelograms . an _ antiprism _ is a polyhedron having bases that are parallel , congruent polygons and sides that are alternating bands of triangles . prisms with an even number of sides and antiprisms are centrally symmetric . although prisms and antiprisms are naturally grouped with the archimedean solids ( i.e. , they are polyhedra in which the same regular polygons appear at each vertex ) , they are generally much less symmetric than either the platonic or archimedean solids . moreover , even the centrally symmetric prisms and antiprisms generally do not possess three equivalent directions . thus , it is less obvious whether bravais lattices would still provide the optimal packings for these solids , except for prisms that tile space ( e.g. , hexagonal prism or rhombical prisms ) . in future work , it would be desirable to test whether our conjecture extends to prisms and antiprisms that possess central symmetry and three equivalent directions using the asc scheme . it is worth noting that in four dimensions , the analogs of the tetrahedron , cube , octahedron , dodecahedron and icosahedron are the four - dimensional regular simplex , hypercube , cross polytope , 120-cell and 600-cell , respectively . all of these four - dimensional polytopes possess central symmetry , except for the simplex @xcite . while the hypercube and cross polytope tile @xmath128 , the optimal packings of simplices are still likely to be non - lattices . since our conjecture for the three - dimensional platonic solids should still apply in four dimensions , the densest packings of the 120-cell and 600-cell could be their corresponding optimal lattice packings . the cross polytopes for @xmath129 no longer tile space , and their optimal packings may still be their densest lattice packings provided that @xmath6 is sufficiently small . however , in sufficiently high dimensions , the densest lattice packings of the centrally symmetric polytopes are probably no longer optimal , since lattice packings in high dimensions are known to possess huge holes " into which additional particles can be inserted , yielding higher packing densities with possibly non - lattice arrangements . indeed , it has recently been argued that disordered sphere packings in very high dimensions could be denser than any ordered packing @xcite . recent progress in particle synthesis methods have enabled the production of a wide spectrum of nanoparticle shapes such as tetrahedra @xcite , cubes @xcite , icosahedra @xcite and prisms @xcite . in such applications , it would be of great interest to predict the corresponding crystal structures , which might possess unusual symmetries and properties . the idea of incorporating collective particle motions due to the adaptive fundamental cell should still make it efficient to search the desired crystal structures formed by those polyhedral nano - building blocks . finally , it is worth noting that the asc algorithm is also suitable to generate random packings of polyhedral particles . crucial dynamical parameters of the system , such as the strain rate can be properly controlled to produce packings with varying degrees of disorder , including the maximally random jammed ones @xcite . we will explore disordered packings in future work . we are grateful to henry cohn , john conway and alex jaoshvili for helpful comments on our manuscript . s. t. thanks the institute for advanced study for its hospitality during his stay there . this work was supported by the division of mathematical sciences at the national science foundation under award number dms-0804431 and by the mrsec program of the national science foundation under award number dmr-0820341 . the figures showing the polyhedra were generated using the antiprism package developed by adrian rossiter . here we collect fundamental packing characteristics ( i.e. , the lattice vectors , densities and contact numbers ) of the optimal lattice packings of the platonic and archimedean solids , most of which are obtained from refs . , and . the platonic solids are explicitly defined as the regions ( sets of points ) bounded by a set of linear equations of the coordinates . the archimedean solids are either defined as intersections of different platonic solids or delineated based on their symmetry . a point set @xmath130 representing the polyhedron multiplied by a number @xmath115 means an isotropic expansion of the polyhedron @xmath130 with linear ratio @xmath115 . @xmath131 represents the intersection region of two polyhedra @xmath132 and @xmath133 . the lattice vectors are given by column vectors ( i.e. , the basis vectors are the unit vectors along the coordinate axis ) . the naming code used below for the platonic and archimedean solids is the same one used in figs . 1 and 3 . a _ tetrahedron _ ( p1 ) has 4 vertices , 6 edges and 4 triangular faces . it is defined as the region @xmath134 the optimal lattice vectors are given by @xmath135 where the superscript `` t '' denotes transpose of a column vector . each tetrahedron of the packing contacts 14 others . the packing density is @xmath136 . an _ icosahedron _ ( p2 ) has 12 vertices , 30 edges and 20 triangular faces . it is defined as the region @xmath137 where @xmath138 is the golden ratio . the lattice vectors are given by @xmath139 where @xmath140 is the unique root of the polynomial @xmath141 it is found that @xmath142 and therefore the lattice vectors , up to nine significant figures , are given by @xmath143 each icosahedron of the packing contacts 12 others . the packing density is @xmath144 a _ dodecahedron _ ( p3 ) has 20 vertices , 30 edges and 12 pentagonal faces . it is defined as the region @xmath145 the optimal lattice vectors are given by @xmath146 each dodecahedron of the packing contacts 12 others . the packing density is @xmath147 . an _ octahedron _ ( p4 ) has 6 vertices , 12 edges and 8 triangular faces . it is defined as the region @xmath148 the optimal lattice vectors are given by @xmath149 each octahedron of the packing contacts 14 others . the packing density is @xmath150 . a _ cube _ ( p5 ) has 8 vertices , 12 edges and 6 square faces . it is defined as the region @xmath151 the optimal lattice vectors are given by @xmath152 each cube of the packing contacts 26 others ( which includes vertex - to - vertex contacts ) . the packing density is @xmath153 . a _ truncated tetrahedron _ ( a1 ) has 12 vertices , 18 edges and 8 face : 4 hexagons and 4 triangles . it is defined as the region @xmath154 the optimal lattice vectors are given by @xmath155 each truncated tetrahedron of the packing contacts 14 others . the packing density is @xmath156 . a _ truncated icosahedron _ ( a2 ) has 60 vertices , 90 edges and 32 faces : 20 hexagons and 12 pentagons . it is defined as the region @xmath157 the optimal lattice vectors are given by @xmath158 where @xmath159 ( @xmath56 ) are the lattice vectors of the optimal lattice packing of icosahedra . each truncated icosahedron of the packing contacts 12 others . the packing density is @xmath160 a _ snub cube _ ( a3 ) has 24 vertices , 60 edges and 38 faces : 6 squares and 32 triangles . let the snub cube orientated in a way such that its 6 square faces lie in the hyperplanes @xmath161 the optimal lattice vectors are given by @xmath162 where @xmath163 is the unique real solution of @xmath164 . each snub cube of the packing contacts 12 others . the packing density is @xmath165 a _ snub dodecahedron _ ( a4 ) has 60 vertices , 150 edges and 92 faces : 12 pentagons and 80 triangles . let the snub dodecahedron orientated in a way such that its 12 pentagonal faces lie in the hyperplanes of the faces of the dodecahedron @xmath166 . the optimal lattice vectors are given by @xmath167 each snub dodecahedron of the packing contacts 12 others . the packing density is @xmath168 a _ rhombicosidodecahdron _ ( a5 ) ( also known as small rhombicosidodecahedron ) has 60 vertices , 120 edges and 62 face : 12 pentagons , 30 squares and 20 triangles . it is defined as the region : @xmath169 \cap ( 4\phi+1)\cdot p2 \cap ( 3 + 3\phi)\cdot p3 \right\}.\ ] ] the optimal lattice vectors are given by @xmath170 each rhombicosidodecahedron of the packing contacts 12 others . the packing density is @xmath171 . a _ truncated icosidodecahdron _ ( a6 ) ( also known as great rhombicosidodecahedron ) has 120 vertices , 180 edges and 62 faces : 12 decagons , 20 hexagons and 30 squares . it is defined as the region @xmath172 \cap ( 6\phi+3)\cdot p2 \cap ( 5 + 5\phi)\cdot p3 \right\}.\ ] ] the optimal lattice vectors are given by @xmath173 each truncated icosidodecahedron of the packing contacts 12 others . the packing density is @xmath174 . a _ truncated cubeoctahedron _ ( a7 ) ( also known as great rhombicubeoctahedron ) has 48 vertices , 72 edges and 26 faces : 6 octagons , 8 hexagons and 12 squares . it is defined as the region @xmath175 the optimal lattice vectors are given by @xmath176 where @xmath177 . each truncated cubeoctahedron of the packing contacts 12 others . the packing density is @xmath178 . an _ icosidodecahedron _ ( a8 ) has 30 vertices , 60 edges and 32 faces : 12 pentagons and 20 triangles . it is defined as the region @xmath179 the optimal lattice vectors are given by @xmath146 each icosidodecahedron of the packing contacts 12 others . the packing density is @xmath180 . a _ rhombicuboctahedron _ ( a9 ) ( also known as small rhombicubeoctahedron ) has 24 vertices , 48 edges and 26 faces : 18 squares and 8 triangles . it is defined as the region @xmath181 the optimal lattice vectors are given by @xmath167 each truncated cubeoctahedron of the packing contacts 12 others . the packing density is @xmath182 . a _ truncated dodecahedron _ ( a10 ) has 60 vertices , 90 edges and 32 faces : 12 decagons and 20 triangles . it is defined as the region @xmath183\cdot p2 \right\}. \end{array}\ ] ] the optimal lattice vectors are given by @xmath167 each truncated dodecahedron of the packing contacts 12 others . the packing density is @xmath184 . a _ cuboctahedron _ ( a11 ) has 12 vertices , 24 edges and 14 faces : 6 squares and 8 triangles . it is defined as the region @xmath185 the optimal lattice vectors are given by @xmath186 each cuboctahedron of the packing contacts 14 others . the packing density is @xmath187 . a _ truncated cube _ ( a12 ) has 24 vertices , 36 edges and 14 faces : 6 octagons and 8 triangles . it is defined as the region @xmath188 the optimal lattice vectors are given by @xmath189 where @xmath190 . each truncated cube of the packing contacts 14 others . the packing density is @xmath191 . a _ truncated octahedron _ ( a13 ) has 24 vertices , 36 edges and 14 faces : 8 hexagons and 6 squares . it is defined as the region @xmath192 the optimal lattice vectors are given by @xmath193 each truncated octahedron of the packing contacts 14 others . the packing density is @xmath153 . here we report the lattice vectors of fundamental cell for the densest known tetrahedral packing with @xmath194 particles up to 12 significant figures , although our numerical precision is not limited by that . the side length of the tetrahedron is @xmath195 and its volume is @xmath196 . the lattice vectors are given by @xmath197 the volume of the fundamental cell is @xmath198 . thus , the packing density is readily computed as @xmath199 the coordinates of each of the 314 tetrahedra are given elsewhere @xcite . the `` primitive welsh '' tessellation of space consists of truncated large regular tetrahedra and small regular tetrahedra , as described in ref . . when the small tetrahedra are removed , the remaining truncated tetrahedra give the densest known periodic ( non - lattice ) packing of such objects . following ref . , the centroids of the truncated tetrahedra sit at the nodes * 0 * and * 1 * of the two adjacent cells of a body - centered cubic lattice shown in fig . 2(a ) of ref . . let length of the edges of the truncated tetrahedron be @xmath200 , then the lattice vectors are given by @xmath201 the basis vectors for the centroids of * 0 * -type and * 1 * -type truncated tetrahedra are respectively given by @xmath202 s. f. edwards , _ granular matter _ a. mehta , springer - verlag , new york , 1994 ) ; e. r. nowak , j. b. knight , e. ben - naim , h. m. jaeger , and s. r. nagel , phys . rev . e * 57 * , 1971 ( 1998 ) ; r. p. behringer , nature * 435 * , 1079 ( 2005 ) ; m. schr " oter , d. i. goldman , and h. l. swinney , phys . e * 71 * , 030301 ( 2005 ) . i. c. kim and s. torquato , j. appl . * 69 * , 2280 ( 1991 ) ; d. cule and s. torquato , phys . b * 58 * , r11829 ( 1998 ) ; m. c. rechtsman and s. torquato , j. appl . phys . * 103 * , 084901 ( 2008 ) ; t. zohdi , int . j. numer . eng . * 76 * , 1250 ( 2008 ) . g. parisi and f. zamponi , j. stat . mech . : theory exp . 2006 , p03017 . s. torquato and f. h. stillinger , phys . e * 73 * , 031106 ( 2006 ) ; s. torquato , o. u. uche and f. h. stillinger , phys . e * 74 * , 061308 ( 2006 ) . a. r. kansal , s. torquato , and f. h. stillinger , phys . e * 66 * , 041109 ( 2002 ) . this tiling - degeneracy example vividly illustrates a fundamental point made in ref . , namely , packing arrangements of nonoverlapping objects at some fixed density can exhibit a large variation in their degree of structural order . in the physical sciences and engineering , this is referred to as a _ bravais _ lattice . unless otherwise stated , the term lattice " will refer to a bravais lattice only , not the more general classification of a periodic lattice with a basis . the overlap potential function of a pair of strictly convex and smooth particles is an analytic function of the positions , orientations and shapes of the two particles , whose value indicates whether the two particles overlap or not , or whether they are tangent to one another ; see also ref . . in this paper , a deformation refers to a volume - preserving shape change of the fundamental cell ( i.e. , a shear ) , which is to be distinguished from a volume change of the cell , which can either be a compression or an expansion of the cell . w. s. jodrey and e. m. tory , phys . rev . a * 32 * , 2347 ( 1985 ) ; m. d. rintoul and s. torquato , phys . e * 58 * , 532 ( 1998 ) ; o. u. uche , f. h. stillinger , and s. torquato , physica a * 342 * , 428 ( 2004 ) ; a. donev , j. burton , f. h. stillinger and s. torquato , phys . rev . b * 73 * , 054109 ( 2006 ) . s. torquato and f. h. stillinger , j. phys . b , 105 , 11849 ( 2001 ) ; s. torquato , a. donev , and f. h. stillinger , int . j. solids struct . * 40 * , 7143 ( 2003 ) ; a. donev , s. torquato , f. h. stillinger and r. connelly , j. appl . phys . * 95 * , 989 ( 2004 ) . in ref . , the authors suggested ( not conjectured ) that the regular tetrahedron may not be able to pack as densely as the sphere , which would contradict a conjecture of ulam . " indeed , the results of ref . verifies that the tetrahedron does not violate ulam s conjecture . the fundamental cells of the initial configurations for the optimization algorithm involve periodic copies of the packings listed in table i. specifically , if the lattice vectors of the packings listed in table i are denoted by @xmath203 , @xmath204 and @xmath205 , then the lattice vectors associated with the initial configurations are given by @xmath206 , @xmath207 and @xmath208 , where @xmath209 are integers . thus , the number of tetrahedra in the fundamental cells of the initial configurations is an integral multiple , i.e. , @xmath210 , of that for the corresponding packings given in table i. for the optimal lattice packing , @xmath211 ; for the uniform packing , @xmath211 ; for the welsh packing , @xmath212 ; for the icosahedral packing , @xmath213 ; and for the wagon - wheels packing , @xmath214 . this bound was motivated by a question posed by robert connelly in may , 2005 at a banff international research station workshop entitled densest packings of spheres , " organized by k. bezdek , h. cohn and c. radin . specifically , he asked whether an upper bound on the maximal density of congruent ellipsoid packings in @xmath0 that is less than unity for any aspect ratios of the ellipsoid could be formulated . one of us in attendance ( s. t. ) provided the version of upper bound ( [ lemma ] ) for the special case of ellipsoids . of course , for ellipsoids , the bound ( [ lemma ] ) is less than unity provided that the eccentricity of the ellipsoid is sufficiently small . however , here we recognize that the same type of upper bound can yield a reasonably tight upper bound for other nonspherical shapes provided that the particle has a small asphericity @xmath33 . although we can not exclude the possibility that there exist other optimal packing particle arrangements that possess the same density as the optimal lattice packings , it is highly unlikely that those arrangements could be constructed in the same way way as the stacking variants of the optimal fcc lattice packings for spheres ( the so - called barlow packings ) . the polyhedral shape of the particles does not allow them to fit snugly into the `` holes '' formed in the particle layers above and below a particular layer , as in the case of spheres , and hence stackings of the layers can not result in packings with densities that are higher than that of the optimal lattice packing . if degenerate optimal packings exist , they can not be of the barlow " type . |
the local zero structure of a smooth map may _ qualitatively _ change , when the map is subjected to small perturbations . the changes may include births and/or deaths of zeros . the _ qualitative properties _ are defined as the invariances of an appropriate _ equivalence relation_. the occurrence of a _ qualitative change _ in the zero structures is called a _ bifurcation _ and the map is named a _ singularity_. the local bifurcation analysis of singularities has been extensively studied in _ singularity theory _ and many powerful algebraic tools have been developed for their study . however , there does not exist any available symbolic computer - library for this purpose . we suitably generalize some powerful tools from algebraic geometry for correct implementation of the results from singularity theory . we provide some required criteria along with rigorous proofs for efficient and cognitive computer - implementation . we have accordingly developed a maple end - user friendly library , named `` ` singularity ` '' , for an efficient and complete local bifurcation analysis of real zeros of scalar smooth maps . we have further constructed a built - in help for ` singularity ` and provided a comprehensive user - guide . recent progresses in maple have been used so that ` singularity ` generates the list of inequivalent persistent bifurcation diagrams for parametric singularities . the approach is independent of the _ singularity s codimension _ ( _ a measure of its complexity _ ) . the main features of ` singularity ` are briefly illustrated along with a few examples . our maple library is not only useful for research goals and engineering applications involving singularities but also for pedagogical purposes . + [ 0.1 in ] majid gazorcorresponding author . phone : ( 98 - 31 ) 33913634 ; fax : ( 98 - 31 ) 33912602 ; email : [email protected] . ] and mahsa kazemi _ keywords _ : singularity and bifurcation theory ; ideal membership problem ; standard and grbner bases . _ 2010 mathematics subject classification _ : 37g10 ; 13p10 ; 58k50 ; 58k60 . _ public access _ : ` singularity ` will be released for public access once this draft paper is peer - reviewed in a refereed journal . ' '' '' many real life problems may result in the analysis of local zeros ( around a zero solution , named a _ base point _ ) of a smooth map [ m0 ] f:^n^m^n , f(x,)=0 , x^n , ^m . we refer to by state variables , by state dimension and by parameters . note that locating singular base points of a smooth map is related to finding roots of nonlinear systems and is not the purpose of our work here ; see . hence , we may assume that the base point is the origin and . equation may demonstrate a surprising change on the solution set when the parameters vary . this occurs when the jacobian matrix of does not have a full rank . in this case we say that is _ singular_. equation may appear by direct mathematical modeling of a singular engineering problem or indirect , through reduction methods such as _ liapunov - schmidt reduction _ ; e.g. , see , ( * ? ? ? * pages 156162 ) and ( * ? ? ? * chapter vii ) . for example , equation appears in the study of equilibria and limit cycles of dynamical systems or steady - state solutions of partial differential equations . in fact , the theory described here is known as a `` natural framework '' for _ equilibrium bifurcation theory _ ; see . using liapunov - schmidt reduction , we can reduce the state dimension so that the jacobian matrix at the origin is the zero matrix . thus in this paper , we assume that [ m00 ] n=1 ( 0,0)=0.we will deal with the case of multi - state dimensional problems in a future project . when is a singular map and the parameters vary , the number of solutions for equation may change and any of such changes is called a _ bifurcation_. the equation is called a _ bifurcation problem _ and the set ( x,):f(x,)=0is called a _ bifurcation diagram_. singularities ( including those different from here ) and the theories to deal with them have appeared in several mathematical disciplines with the same or similar keywords . these theories have a long history dating back to the original works of hassler whitney , and then john mather , rene thom and v.i . arnold in late 1960s . depending on the context , space and equivalence relation , they are faced with essentially different tools , algebraic and geometric structures , applications and/or objectives ; e.g. , see . for an instance a _ geometric singularity _ refers to a variety whose tangent space is not regularly defined ; e.g. , a self intersecting curve or a cusp ; see . this leads to a theory of its own and it is named _ singularity theory_. arnold in uses _ - _ and _ right_-equivalences for his normal form classification of smooth functions . _ strategy equivalence _ is used for the analysis of _ strategy functions _ in . _ right - left _ equivalence is applied in to treat bifurcations of hamiltonian systems . the other instance is _ catastrophe theory _ and usually uses almost identical keywords as ours . catastrophe theory addresses basically different qualitative properties using _ right equivalence _ than the qualitative properties we pursue here using _ contact equivalence_. all of these mentioned subject areas follow different objectives and goals than what ours do . in this paper we only refer to the methodologies in dealing with the local bifurcation analysis of real zeros of smooth maps as singularity theory ; see for our main references of this subject . many powerful algebraic tools have been developed for local bifurcation analysis of zeros in equation . armbruster proposed a cognitive use of basis and encouraged a systematic implementations of the existing results in a computer . yet to the best of our knowledge , there does not exist any end - user symbolic library available ( neither free nor paid ) to everyone for the _ local bifurcation analysis of zeros of smooth maps_. this is a long overdue contribution despite its importance and wide applications . in the last two decades , there has been a considerable progress in development of computer algebra systems so that an efficient _ symbolic _ implementation of the results in singularity theory is now feasible ; see for numeric approaches . singularity theory uses two fundamental notions of _ intermediate order terms _ and _ high order terms _ for _ normal form computation_. recall that theory of normal forms here is to simplify singular maps and qualitative study of their zeros . intermediate order terms have a complicated structure and do not seem to hold a convenient algebraic structure . therefore , basis type of tools ( i.e. , its local versions ) alone are not sufficient to simplify removable intermediate order terms in a singular germ . furthermore as far as our information is concerned , there does not yet exist a complete algebraic characterization for _ high order terms _ in many cases of multi - state variables . this does not mean that high and intermediate order terms are computationally inaccessible but it does mean crude uses of basis type of tools fail to address _ normal form computations_. ( note that the basic and principal ideas in singularity theory start with normal form computation . ) indeed , the approach needs to be refined along with nonlinear normalization techniques . our approach efficiently simplifies removable intermediate order terms and is adoptable for simplification of high order terms even for the generalization of our results to multi - state variable cases . our main contributions here are some natural generalizations of known tools from algebraic geometry for correct symbolic implementation and providing certain criteria along with rigorous proofs for alternative and efficient symbolic implementation , an efficient symbolic implementation of the results of singularity theory ( one dimensional state variable ) in maple , developing an end - user friendly maple library named ` singularity ` , writing a user guide and constructing a comprehensive built - in maple help for ` singularity ` . a contribution here is to generalize some existing techniques and concepts from algebraic geometry to the context of ( locally ) smooth maps ( germs ) for a correct implementation . due to the infinite nature of taylor series of smooth maps , the computations are performed modulo a given degree . we provide a sufficient condition for a given degree whose truncation does not lead to error . the default work of ` singularity ` tests the condition and do the computations modulo an optimal degree . however , this approach adds a computational cost . further , smooth maps involve flat functions ( functions with zero taylor series ) and this may cause unnecessarily complicated formulas . thereby , it is fundamentally helpful to use a ring smaller than the ring of smooth maps when it is feasible . unlike the ring of formal power series , the associated computations in the ring of polynomials or fractional maps are exact and no truncation is required . we provide conditions along with rigorous proofs for the possible efficient implementations using ideals generated in either the rings of polynomials , fractional maps or formal power series . ` singularity ` is adopted accordingly . we accordingly provide guidelines for efficient symbolic implementation of the existing results from singularity theory ; see ( * ? ? ? * chapters 67 ) , ( * ? ? ? * sections 6.2 and 6.3 ) , ( * ? ? ? * chapters 14 ) and . this simply illustrates an interesting and realistic application of computational algebraic geometry to the equilibrium bifurcation analysis of real world problems . maple is one of the two major ( along with mathematica ) symbolic computer algebra systems . besides , maple is user - friendly and is widely used for mathematics and engineering research as well as educational purposes . thus , maple is a good choice for end - user libraries in general , and in particular for ` singularity ` due to various applications of singularity theory in different engineering disciplines . further , many useful and advanced techniques from computational algebraic geometry such as basis , elimination ideals for polynomial ring , regular chains and triangular decomposition method have already been developed and are now available in built - in packages of maple 18 . so far , the most efficient basis computation is due to maple algorithm implemented by jean - charles faugre ( see ) while regular chains and triangular decomposition method are only available in maple . ` singularity ` has benefited from an efficient use of these capacities of maple . besides , we have already developed some maple programs for ( parametric and orbital ) normal form computation of _ singular differential equations _ and their integration with ` singularity ` shall lead to a toolbox for local bifurcation control and analysis of singularities . ` singularity ` has been tested by all scalar examples ( nonsymmetric and without modal parameters ) and classifications given in and a few ( differences , error or not already reported data due to computational burden ) are verified in our favor . we use identical notations and terminologies from as far as it is feasible . ` singularity ` computes a variety of algebraic structures associated with singular scalar maps including _ tangent _ and _ restricted tangent spaces , _ _ high order term ideals , _ and the _ intrinsic ideals _ associated with ideals of both _ finite _ and _ infinite codimension . _ our maple library derives _ low and intermediate order terms , _ _ normal forms , universal unfolding , _ and _ transition sets . _ ` singularity ` efficiently simplifies _ intermediate _ and _ high order terms_. it , further , generates _ persistent bifurcation diagrams ( plot or animation ) , _ and estimates the _ transformations _ transforming _ contact - equivalent scalar maps to each other_. finally , ` singularity ` solves _ the recognition problem for normal forms and universal unfoldings_. an interesting capability of ` singularity ` is the classification of persistent bifurcation diagrams by generating an automatic list . this latter capability is in fact enabled by using a powerful built - in maple 18 package called regularchains . in order to develop ` singularity ` , we have discussed some modified concepts ( suitable in our context ) from computational algebraic geometry including _ division remainders , _ _ standard bases , _ _ elimination ideals , _ _ ideal membership problem _ and _ colon ideals _ for the local rings of fractional ( germ ) maps , formal power series and smooth maps . these are accordingly implemented in ` singularity ` . the rest of this paper is organized as follows . singularity theory and bifurcation analysis of equation - is discussed in section [ sec1int ] . we further explain how singularity theory is related to _ ideal membership problem _ in algebraic geometry . section [ sec2 ] describes how to treat the ideal membership problem . in this direction , computational algebraic tools such as standard and bases for ideals in three different rings , and the concept of _ finite codimension ideals are introduced . _ intrinsic ideals _ and their associated ideal representations are discussed in section [ sec3 ] . we further explain a procedure for computing the intrinsic part of an ideal . section [ sec4 ] gives our suggestions on how to implement some objects and results from singularity theory . these implementations include high order term ideals , tangent spaces , transition sets , persistent bifurcation diagram classifications , normal forms and the universal unfolding . the capabilities of the main features of ` singularity ` along with a few examples are sketched in section [ secfeatures ] . finally , our future and in - progress projects are outlined in section [ secfuture ] . | 1507.06168 | due to the local nature of the problem - , we recall the notion of _ smooth germs _ around a base point . two maps are considered as _ germ - equivalent _ when both maps are locally identical ; more precisely , when there exists a neighborhood of the base point so that both maps are equal on the neighborhood . a _ germ _ is a _ germ - equivalence _ class of a smooth map . we denote @xmath9 for the set of all scalar smooth germs whose base point is the origin . from now on we merely work with elements of @xmath9 rather than a scalar smooth map ; see @xcite . following ( * ? ? ? * chapter 7 ) and @xcite , we study the local zeros of maps when there is only one distinguished parameter denoted by @xmath10 , i.e. , [ eq1 ] g : , g(x,)=0,g(0,0)=g_x(0,0)=0 , m:=1 . the effect of additional parameters may be treated by study of their small perturbations . the main goal of this theory is to classify _ qualitative _ types of equation and its arbitrary small perturbations . in order to achieve this goal , we first define a _ qualitative property _ as a property that is invariant under an appropriate equivalence relation . here , we use _ contact - equivalence _ relation : * we say that the germs @xmath11 and @xmath12 are contact - equivalent when there exist a smooth germ @xmath13 and diffeomorphic germs @xmath14 and @xmath15 such that [ conteqvl ] g(x,)= s(x,)h(x(x , ) , ( ) ) , where @xmath16 @xmath17 and @xmath18 other alternative equivalences have also been reported in the literature based on their applications ; e.g. , see @xcite where _ right - left _ and _ strategy equivalences _ have been used . _ right equivalence _ is also used to study the isolated hypersurface of geometric singularities and to study the bifurcation analysis in _ catastrophe theory _ ; see @xcite . the contact - equivalence is here used , since it is the most natural equivalence relation preserving the zero structures . now we briefly describe the approach in singularity theory and how the computational algebraic tools may help to implement them . to study the local zeros of , we choose a representative ( say @xmath5 ) from contact - equivalent class of @xmath11 that is considered to be the _ simplest _ for the analysis and call it a _ normal from_. in order to compute the normal form of a singular germ , we need to compute certain ideals in the _ local ring of smooth germs_. this signifies the importance of the well - known ideal membership problem in algebraic geometry , that is , deciding what kinds of germs belong to an ideal generated by a given set . one may study the zero structures of the normal forms and then , conclude about the solution behavior of equation . for instance , let [ exmp1 ] g(x , ) : = ( x^2)+2(x)-3+().using the command @xmath19 in singularity , we obtain its normal form by @xmath20 the bifurcation diagrams of @xmath5 and @xmath11 are depicted in figure [ a ] and [ b ] . here , @xmath21 provides an approximation @xmath22 modulo degree 5 for @xmath23 which transforms @xmath11 into @xmath5 via equation . the locally invertible transformation @xmath24 sends the bifurcation diagram of @xmath5 into that of @xmath11 . real life problems can not be perfectly modeled by a system of equations and _ imperfections _ are always inevitable . furthermore , the singularity of a germ @xmath11 implies that the zeros of a small perturbation of @xmath25 say [ unfold ] h(x , , ) = 0 , h(x , , 0)=g ( x , ) , may behave substantially different than what the zero structure associated with @xmath26 does . the parameterized germ @xmath12 in is called an _ unfolding _ of @xmath27 hence , _ modeling imperfections _ and the possible existence of additional parameters in a model are the main obstacles of simply using normal forms of a singular germ for the qualitative understanding of a real life problem . thus , the approach needs to be refined through the notion of _ universal unfolding_. in fact , we are interested to find a parameterized family like [ eq4 ] g(x , , ) = 0 , ^p , such that for any small perturbation @xmath28 , the germ @xmath29 would be contact - equivalent to @xmath30 for some germ @xmath31 we call such germ @xmath32 a _ versal unfolding _ of @xmath27 a versal unfolding with the minimum possible number of parameters is called a _ universal unfolding _ for @xmath11 ; see ( * ? ? ? * definitions 6.4.2 and 6.4.3 , page 176 ) and ( * ? ? ? * definitions 1.1 and 1.3 , pages 120121 ) . the number of parameters in a universal unfolding is named the _ codimension _ of @xmath27 the universal unfolding accommodates any possible modeling imperfections , any arbitrary small perturbation and also the existence of any possible number of parameters ( in addition to the distinguished parameter @xmath33 in order to derive the universal unfolding of a given singularity , the computation of a vector space called _ tangent space _ or instead a basis for its complement is required . using singularity s command @xmath34 for @xmath11 in gives rise to [ univexmp ] g(x , , _ 1 , _ 2):= x^4++_1x+_2x^2 . _ bifurcation diagram classification _ of the universal unfolding gives an insight to the zero structure of a germ and any of its perturbations . this is studied by the notion of _ persistence _ in the bifurcation diagrams . in fact a bifurcation diagram is called _ persistent _ when all its small perturbations remain self contact - equivalent , and otherwise it is called _ nonpersistent_. finding nonpersistent systems and their associated subset of the parameter space ( the so - called _ transition set _ @xmath35 ) play a central role in this classification . more precisely , all parametric germs associated with parameters in a connected component of @xmath36 complement of @xmath37 are contact - equivalent . therefore by choosing a parameter from each connected component of @xmath36 a complete list of persistent bifurcation diagrams modulo contact - equivalent is obtained . the non - persistence may either originate from a local nature or be caused by the singular boundary conditions . local nonpersistent bifurcation diagrams are determined with families of germs associated with three semi - algebraic parameter spaces of codimension one ; i.e. , _ bifurcation _ @xmath38 , _ hysteresis _ @xmath39 and _ double limit point _ @xmath40 ; see ( * ? ? ? * page 140 ) for details . nonpersistent germs associated with boundary conditions add extra complications into the solution dynamics , when bifurcation diagrams cross the boundary ; see ( * ? ? ? * pages 154165 ) . given our description for any finite codimension singularity , the connected components of the complement set of the transition set , i.e. , @xmath41 provide the qualitative classification of persistent bifurcation diagrams . the command @xmath42)$ ] for @xmath32 in gives @xmath43 and @xmath44 the transition set @xmath35 is plotted in figure [ c ] . @xmath45 plots a complete list of all contact - inequivalent types of persistent bifurcation diagrams for @xmath32 ; see the second row in figure [ fig1 ] . given our description in the previous section , we mainly need to address the _ ideal membership problem _ in the ring of smooth germs . we call an _ ideal basis _ for the ( finite ) _ generators _ of a finitely generated ideal . the _ ideal membership problem _ refers to the question on whether or not an element belongs to a predefined ideal . although a _ finitely generated ideal _ is defined by a ideal basis , yet it is not an easy task to understand whether a given element belongs to the ideal or not . there are two different ways to facilitate the ideal membership problem . one is to find a convenient ideal presentation from which the ideal membership can be easily understood . this is feasible for certain ideals and will be addressed in section [ sec3 ] . the second is a computational approach based on a division algorithm . dividing a given element by the ideal basis ( generators ) , the element certainly belongs to the ideal when the remainder is zero . however for non - zero remainders with respect to an arbitrary ideal basis , the division may not help to conclude about the membership of the element in the ideal . b. buchberger in 1965 introduced an ideal basis ( called _ basis _ ) for ideals in the polynomial ring on which zero / nonzero remainders would indeed conclude the ideal membership ; see @xcite . armbruster and kredel @xcite suggested a cognitive use of basis for computing universal unfolding ; also see @xcite . then , gatermann and lauterbach @xcite extended the tools for equivariant bifurcation problems . wright and cowel in @xcite noticed that a local version of basis is indeed the appropriate tool to work with it in this theory . in fact , naive uses of basis yield wrong results for many singularities ( e.g. , see remark [ m15 ] ) . the main reason is that basis may work with polynomial germs , while the ring of smooth germs is a considerably larger ring than the polynomial germs . recall that a set may generate a larger ideal in a ring than what it generates in a subring . since wright and cowel s remark , the only result in this direction is due to gatermann and hosten @xcite which is a fundamental contribution . they used ( mixed ) _ standard basis _ for `` _ _ mixed modules _ _ '' over fractional maps and multi - dimensional state variables . they did not discuss computations with smooth maps and taylor series truncations . therefore , their suggested algorithms are limited to treat certain ( finite codimension ) fractional maps . their algorithms are useful for possible implementations in a computer algebra system to simplify fractional maps with finite codimension . however , they are not sufficient to compute _ normal forms_. the reason is that they attempted neither to simplify _ removable intermediate order terms _ nor to address the issue facing high order terms for multi - dimensional cases ( there does not yet exist a complete algebraic characterizations for high order terms ) ; for example see ( * ? ? ? * example 53 ) . the principal ideas of basis have been extended for localizations of the polynomial ring and formal power series . here , we are concerned with local rings and the terminology _ `` standard basis '' _ ( not basis ) is usually used in the context of local rings . following wright and cowel s remark , gatermann and hosten @xcite used the localizations of the polynomial ring ( a local ring ) to circumvent the problem . the localization of the polynomial - germ ring gives rise to the fractional germs , i.e. , germs with a fractional representative in their germ - equivalent class . although the set of all fractional germs is a local ring , yet it is still a much smaller ring than the ring of smooth germs . the other alternative is the local germ ring of all formal power series . this is a larger ring than the fractional germs and it is , perhaps , suitable due to borel lemma . ( borel lemma indicates the existence of a one - to - one correspondence modulo flat functions between smooth germs and formal power series through their taylor series expansion . ) one of the main obstacles in working with the local rings is termination of the algorithms . there are methods in the literature for computing the _ standard basis _ ( for fractional maps using mora normal form ) so that they solve the problem of algorithm terminations . yet for the case of formal power series , no computer program can compute and store infinite formal power series expansions and thus , their truncations up to certain degrees are mostly unavoidable . in other words , the computations are performed modulo a sufficiently high degree ( we will justify this in theorem [ thm2.18 ] ( part 4 ) ) . in this section we investigate and compare the computations in the local ( germ ) rings of polynomials , the fractional germs and formal power series with those in the ring of smooth germs . we discuss the circumstances on which they can be alternatively used . this helps to efficiently use the different algorithms and yet ensure about correctness of the results . let @xmath46 be a field of characteristic zero ; in particular we are interested in the field of real numbers . for our convenience , we simply identify a given germ with a convenient representative of that germ . for instance , we talk about the polynomial germ ring over the field @xmath46 and denote it by @xmath47 $ ] while we mean the ring of all smooth germs whose germ - equivalent class have a polynomial representative . the quotient ring of @xmath9 over the ideal of all flat germs is denoted by @xmath48 $ ] due to the fact that it is ring - isomorphic to the ring of formal power series . thus , we call @xmath48 $ ] the germ ring of _ formal power series_. further , we identify members of @xmath48 $ ] with the infinite taylor expansion of their representative . since @xmath49 $ ] and @xmath50 $ ] are noetherian rings , we can guarantee termination of most algorithms during the computations . denote @xmath51 when @xmath52 @xmath53 and @xmath54 any expression like @xmath55 is called a _ monomial germ _ while a _ term _ in @xmath9 means a monomial germ along with its coefficient . we define the lexicographic ordering @xmath56 on monomial germs @xmath55 as follows : x^_x^ _ i-_i i:= \{j| _ j-_j0}. a _ local order _ @xmath57 is a total ordering ( every two terms are comparable ) and furthermore , * for any @xmath58 , the condition @xmath59 implies @xmath60 . * @xmath61 for all @xmath62 . we further assume that @xmath63 for any @xmath64 due to dickson s lemma ( see ( * ? ? ? * page 251 ) and ( * ? ? ? * page 71 ) ) , every ( infinite ) set of monomials have a maximum with respect to any arbitrary local order ; here , the condition @xmath65 implies @xmath66 an important example of a local order is _ anti - graded lexicographic _ ordering @xmath67 defined by @xmath68 the localization of the polynomial germ ring is defined as : = \ { f , gk[x , ] , g(0,0 ) 0 } whose unique maximal ideal is generated by @xmath1 and @xmath69 it is common to denote @xmath70 with @xmath49_{\langle x , \lambda\rangle}.$ ] we call @xmath70 the _ ring of fractional germs_. throughout this paper , we denote k[[x , ] ] , , , unless it is explicitly stated . let @xmath71 and @xmath57 be a local order . the infinite jet and @xmath72-jet of @xmath5 are defined by its taylor series expansion around the origin and denoted by j^(f):= _ ( i , j ) a_ijx^i^j , j^k(f):= _ i+jk a_ijx^i^j a_ijk.the set of terms of @xmath5 are defined by ( f):= \ { a_ij x^i ^j| a_ij0},i.e . , all terms appearing in @xmath73 when @xmath74 the leading term of @xmath5 is lt(f)= ( f),i.e . , @xmath75 and @xmath76 for any @xmath77 the germ @xmath11 is flat iff @xmath78 we define @xmath79 when @xmath5 is a flat germ , i.e. , @xmath80 the coefficient and monomial of the leading term are respectively called the _ leading coefficient _ ( @xmath81 ) and _ leading monomial _ ( @xmath82 ) . for the case of @xmath83 $ ] and @xmath84 we may similarly define @xmath85 and @xmath86 now we present some known definitions , terminologies and theorems from @xcite these are suitably modified and generalized to fit in our purpose . [ m11 ] a _ remainder _ of a germ @xmath5 with respect to the set of germs @xmath87 and the local order @xmath57 is defined as a germ @xmath88 so that * @xmath89 , for some @xmath90 so that @xmath91 . * no term of @xmath92 is divisible by any of @xmath93 for @xmath94 . by replacing @xmath9 with @xmath50 $ ] in definition [ m11 ] , the remainders in the ring of @xmath50 $ ] is readily defined ; also see ( * ? ? ? * pages 251 - 252 ) . the same is true for the case of polynomial germ ring @xmath49 $ ] provided that the local order @xmath57 would be instead a monomial ordering like @xmath95 the remainders in the ring of @xmath70 is defined in ( * ? ? ? * page 170 ) using _ mora normal form algorithm_. usually , the terminology of _ normal form _ is used rather than _ remainder_. however , we choose _ remainder _ as the other may cause confusion with the _ normal forms _ of germs in singularity theory . the division here is related to malgrange preparation theorem and mather division theorem ; see ( * ? ? ? * corollaries a.6.2 and a.7.2 , theorem a.7.1 ) . when @xmath5 is flat , its remainder with respect to any set of germs and local order is flat . the remainder is not necessarily unique even modulo flat germs ; also see ( * ? ? ? * pages 251 - 252 ) . in fact , the remainder is unique modulo flat germs when @xmath32 is a _ standard basis _ ( or basis for the case of @xmath49 $ ] ) , defined as follows . for an ideal @xmath96 in @xmath97 we define the leading term ideal @xmath98 by @xmath99 [ sbasis ] let @xmath96 be an ideal in @xmath100 with a finite generating set @xmath101 . the set @xmath102 is called a _ standard basis _ of @xmath96 when @xmath103 the set @xmath104 is called a standard basis in @xmath100 when it is a standard basis for the ideal @xmath105 . [ new ] * the set @xmath106 $ ] is called basis with respect to @xmath56 when the condition holds . * any finite set of germs @xmath107 in @xmath9 is a standard basis in @xmath9 iff the set of formal power series @xmath108 is a standard basis in @xmath50.$ ] * let @xmath109 , k[[x , \lambda]]$ ] or @xmath70 and @xmath101 be a finite set . then , lt(i)_=lt(g_1 ) , , lt(g_m)_implies that @xmath110 is a generating set for @xmath96 . this has a simple argument as follows ; also see ( * ? ? ? * page 206 ) . for any @xmath111 r:=rem(f , \{g_i| i=1 , , m } , ) i. thus , @xmath112 and @xmath113 must be factored by @xmath114 for some @xmath115 the latter implies that @xmath116 otherwise this contradicts with @xmath117 being a remainder . computation of standard basis uses the notion of @xmath118-germs . let @xmath119 and @xmath120 be a local order . then , _ s - germ _ of @xmath5 and @xmath11 is defined by @xmath121 here @xmath122 stands for the least common multiple for a pair of monomials . ( also see hironaka theorem on ( * ? ? ? * page 252 ) ) [ m10 ] let @xmath123 or @xmath50 $ ] , @xmath124 @xmath125 and @xmath57 be a local order . then , * always @xmath5 has a remainder with respect to @xmath57 and @xmath126 * the set @xmath32 is a standard basis iff the remainder of @xmath5 with respect to @xmath32 and @xmath57 is unique modulo flat germs . * ( buchberger s criterion ) the set @xmath127 is a standard basis iff for all @xmath128 the expression @xmath129 is a flat germ . * the set @xmath32 is a standard basis iff @xmath92 is flat for all @xmath130 . when @xmath131 $ ] , the claim is _ hironaka theorem _ given on ( * ? ? ? * page 252 ) ; also see @xcite . therefore , we assume that @xmath123 . by hironaka theorem , the remainder always exists in the ring of @xmath50 $ ] , i.e. , there exists a rem(j^(f),j^(g_i)_i=1^m,)k[[x , ] ] and @xmath132 $ ] so that @xmath133 with @xmath134 . further for p(rem(j^(f),j^(g_i)_i=1^m,)),@xmath135 is not divisible by any of @xmath136 for @xmath137 . our proposed division algorithm for germs in @xmath9 follows @xmath138 where @xmath139 note that the last line in is merely a flat germ . equation implies that the division in @xmath9 merely differs by flat functions from its analog in @xmath50 $ ] . the proof of part ( a ) is complete by equation . let @xmath140 satisfying ( 1 ) and ( 2 ) in definition [ m11 ] . then , @xmath141 therefore , @xmath142 is unique if and only if @xmath143 is a standard basis in @xmath50 $ ] . this completes the proof by remark [ new ] . the proof of parts ( c ) and ( d ) are given by @xmath144 , equation and remark [ new](a ) . in order to compute the standard basis for an ideal in @xmath50 $ ] ( and also in @xmath9 ) , one needs to sequentially enlarge and update its set of generators by only adding the _ non - flat _ remainders ( with respect to the updated generators and a given local order ) of the @xmath118-germs of the generator pairs . this process is usually known as _ buchberger algorithm _ and it terminates when the ascending ideals generated by the leading terms of the updated generators stops any further enlargement . the buchberger algorithm is finitely terminated due to the fact that @xmath50 $ ] is a noetherian ring . * the division of a polynomial by a set of polynomials in our division algorithm may involve formal power series ; for example we have @xmath145 * the remainder of a polynomial divided by a polynomial may give rise to an _ infinite _ formal power series . for instance let @xmath146 then , h&:=&rem(x^2 , g , _ ) = x^5 + x^9 + 2x^13 + 5x^17 + 14x^21 + + & = & x^2-(x+x^2+x^5 + x^3 ^ 2 + 2x^6 + x^4 ^ 3 + ) ( x- x^2 ^ 2- x^4 ) . since the generator of an ideal with only one generator is always a standard basis , the remainder here is unique . * this example is to show that a finite set of polynomial ideal basis in @xmath9 may lead to a standard basis that includes non - polynomial germs . let @xmath147 . then , @xmath148 it is easy to verify that @xmath149 is a standard basis , where @xmath12 is defined in part ( b ) . the first and second buchberger criteria are applied for efficient computation of standard basis in ` singularity ` . however , we will not discuss them in this paper . let @xmath150 be a standard basis and @xmath151 for @xmath152 when * @xmath153 for all @xmath154 , except for when @xmath155 and @xmath156 , the set @xmath32 is called a _ reduced standard basis_. given a local order @xmath57 , ( * ? ? ? * theorem 2.1 , page 255 ) states that any ideal in @xmath50 $ ] has a unique reduced standard basis . [ m1 ] with respect to any local order @xmath57 , every finitely generated ideal @xmath157 has a reduced standard basis . the standard basis is unique modulo flat germs . put @xmath158 . by * theorem 2.1 , page 255 ) , there exists a unique reduced standard basis @xmath159 for @xmath160 where @xmath161 with @xmath162 . since @xmath163 for @xmath164 @xmath165}&=&\langle lt(g_{i } ) : i=1,\ldots , m \rangle_{k[[x , \lambda ] ] } = \langle lt(g_{i } ) : i=1,\ldots , m \rangle_{k[[x , \lambda]]}.\end{aligned}\ ] ] now we have \{lt(f)| fi}&=&\{lt(f)| f}lt(g_i):i=1 , , m_k[[x , ] ] + & & lt(g_i):i=1 , , m_. hence , @xmath166 and @xmath167 . assuming that @xmath168 the set @xmath169 is a reduced standard basis , since @xmath170 is flat for any @xmath171 due to part ( c ) in theorem [ m10 ] . the uniqueness follows ( * ? ? ? * theorem 2.1 ) and remark [ new ] ( a ) . the following theorem along with equation and buchberger algorithm provide computational guidelines on how to compute a reduced standard basis in @xmath172 [ m14 ] let @xmath173 . * there always exist germs @xmath174 for @xmath175 so that @xmath176 and @xmath153 for all @xmath154 , except for when @xmath156 and @xmath155 . * furthermore , assume that @xmath177 is a standard basis in @xmath172 then , the set @xmath178 ( for @xmath175 ) can be chosen so that it is a reduced standard basis . the proof here is similar to the idea on ( * ? ? ? * page 255 ) and gives us an algorithm for reduced standard basis computation . without loss of generality assume that @xmath179 for @xmath180 ; otherwise , @xmath181 for some @xmath182 satisfying conditions in definition [ m11 ] . thus , @xmath183 may be replaced by @xmath184 and this may cause @xmath185 for when @xmath186 . by theorem [ m10 ] , there exist @xmath187 for @xmath188 so that @xmath189 now for each @xmath190 we have @xmath191 define @xmath192 . this shows @xmath193 and the first claim follows . the second claim is valid due to @xmath194 [ m12 ] let @xmath5 be a non - zero flat germ , and @xmath195 where g:=g_1:=-(x ) , g_2:=x-(x ) , g_3:=x+^3+^2 ( 1+x)+f . since @xmath196 and @xmath197 , @xmath32 is not a standard basis with respect to @xmath67 . due to rem(s(g_1,g_2 ) , g , _ ) = 0 , rem(s(g_2 , g_3 ) , g , _ ) = -x^2 fand @xmath198 , we add @xmath199 to the basis , i.e. , @xmath200 . now @xmath201 is a standard basis , yet it is not a reduced standard basis . given theorem [ m14 ] and its proof , @xmath202 we define @xmath203 now @xmath204 is a reduced standard basis . finite codimensional ideals demonstrate an important role in ideal presentation in section [ sec3 ] . let @xmath205 , @xmath206 @xmath207}=\langle x , \lambda \rangle_{k[[x , \lambda]]},$ ] and @xmath208}=\langle x , \lambda \rangle_{k[x , \lambda]}$ ] . [ m4 ] an ideal @xmath96 in @xmath9 ( or in @xmath209 @xmath50 $ ] ) is said to have a _ finite codimension _ when @xmath210 ( @xmath211}}^k\subseteq i$ ] ) for @xmath212 equivalently , @xmath96 has a complement ( vector ) subspace in @xmath9 with finite dimension . now we compare the ideals in @xmath49 $ ] with those in @xmath209 @xmath50 $ ] and @xmath172 [ fcodremainder ] let @xmath96 be a finite codimensional ideal in @xmath172 * the ideal @xmath96 has a unique reduced standard _ polynomial _ germ basis . * assume that @xmath213 is a standard basis for @xmath96 and @xmath214 then , the remainder is always a _ polynomial germ _ , i.e. , @xmath215.$ ] in particular , we have @xmath216 when @xmath217 by theorem [ m1 ] , we assume that @xmath218 is a reduced standard basis for @xmath96 . assume that @xmath219 for some @xmath220 then , for any @xmath221 there exists @xmath222 so that @xmath223 on the other hand for any @xmath224 and @xmath225 @xmath226 and hence , @xmath227 the latter implies that @xmath228 for all @xmath229 finally , we recall that any finite codimension ideal includes the set of all flat germs . this allows us to replace @xmath230 with @xmath231 and @xmath232 is a reduced standard basis . the second claim similarly follows from @xmath233 and the fact that no term of @xmath234 is divisible by @xmath235 let @xmath236 be a reduced standard basis whose generated ideal is of finite codimension . then , @xmath237.$ ] however , @xmath238 may not always be a polynomial germ . consider the ordering @xmath239 and define i:=g_1:=2 ^ 3 - 3 ^ 2x+x^5 , g_2:=-3x^2 + 5x^5 , g_3:=-3 ^ 3 + 5x^4.the ideal @xmath96 is finite codimensional since @xmath240 here , @xmath241 is a standard basis and s = f_1:= x^2-x^4 , f_2:=-g_3 , f_3:=x^5-x^4is a reduced standard basis . now we have x^3 ^ 3=x^3 f_2 + x^5q f_1 + x^2q f_3,where @xmath242 , @xmath243 while @xmath244 and @xmath245 are not polynomial germs . the following theorem enables us to compute the standard basis for ideals in @xmath9 through computations in the fractional germs . [ thm2.17 ] consider @xmath246 @xmath67 and @xmath247 further , suppose that @xmath248 is a standard basis for @xmath249 then , @xmath250 is a standard basis for @xmath96 . see appendix [ 215 ] . the following theorem is one of our main contributions in this paper and provides important alternatives and criteria for our computations in different circumstances including ideals with infinite codimension . [ thm2.18 ] * suppose that @xmath251 then , the ideal @xmath252 has a finite codimension iff @xmath253 is a finite codimension ideal . assuming that @xmath253 has a finite codimension , for nonnegative integers @xmath254 and @xmath171 , @xmath255 iff @xmath256 * let @xmath257}}^{k}\subseteq\langle g_{1},\ldots , g_{n}\rangle_{k[x , \lambda]}\subseteq k[x , \lambda].$ ] then , @xmath257}}^{i}\langle \lambda^j\rangle_{k[x , \lambda]}\subseteq\langle g_{1},\ldots , g_{n}\rangle_{k[x , \lambda]}$ ] iff @xmath258 * for a finite codimension ideal @xmath253 in @xmath70 and an ideal @xmath96 in @xmath259 i= g_1 , , g_n _ i= g_1 , , g_n_.in particular , let @xmath260 be two ideals in @xmath259 @xmath96 have a finite codimension and @xmath261 then , @xmath262 * for an ideal @xmath96 in @xmath259 the following three conditions are equivalent . ^ki , ^k i , ^k i k[x , ] . * let @xmath263 and @xmath264 for @xmath265 then , either of the conditions ^kj^ng_1 , , j^ng_n _ ^kj^ng_1 , , j^ng_n_k[[x , ] ] , is equivalent to @xmath266 * consider the finite sequence of nonnegative integers @xmath267 so that the sequence @xmath268 is decreasing and @xmath269 is increasing . let @xmath270 be an ideal that is not necessarily of finite codimension and i:=^k_i^l_i_j^nf_1 , , j^nf_n_. then , either of the following conditions * * for each @xmath225 @xmath271 for some @xmath272 * * for each @xmath225 @xmath273 and @xmath274 for some @xmath272 + implies that ^k_i^l_i_f_1 , f_2 , , f_n_. part ( 1 ) . if _ part is trivial . thus , we assume that @xmath275 and prove that @xmath276 . using nakayama lemma ( * ? ? ? * lemma 5.3 , page 71 ) , it is enough to verify that @xmath277 . since @xmath278 , for some @xmath279 we have @xmath280 where @xmath281 and @xmath282 for the second claim , we do not need nakayama lemma . we assume that @xmath283 similar to above , for @xmath284 we have @xmath285 where @xmath286 and @xmath287 part ( 2 ) . proof is similar to the proof of the second claim in part ( 1 ) . part ( 3 ) . assume that @xmath288 trivially , g_1 , , g_n_i g_1 , , g_n_.let @xmath289 thus , @xmath290 for some @xmath291 so , @xmath292 for an @xmath293}}^{k+1}.$ ] since the left hand side belongs to @xmath294 the proof of the _ if _ part is complete . now assume that @xmath295 and @xmath296 by part ( 1 ) and @xmath297 @xmath298 for some @xmath212 since @xmath299 j^k(f)= a_i g_i , a_i . on the other hand @xmath300 this completes the proof of part ( 3 ) . part ( 4 ) is trivial . part ( 5 ) . now the assumption @xmath301 implies that for some @xmath302 and any @xmath303 @xmath304 where @xmath305 . this and nakayama lemma conclude that @xmath306 . the converse and rest of the proof use similar arguments . note that nakayama lemma given in ( * ? ? ? * lemma 5.3 ) is also true when @xmath9 is replaced with @xmath70 and @xmath50.$ ] part ( 6 ) is concluded by nakayama lemma and the fact that @xmath307 where @xmath308 the hypothesis of part ( 1 ) in theorem [ thm2.18 ] allows us to use fractional germs for computations while the condition in part ( 2 ) permits the use of basis . [ m15 ] part ( 1 ) from the previous theorem is not valid when @xmath70 is replaced with @xmath49 $ ] . for instance , consider the example given in ( * ? ? ? * table 1 , iii.1 for @xmath309 ) and ( * ? ? ? * page 77 ) , i:=g_1:=x^5+x^3+^2 , g_2:=5x^5 + 3x^3 , g_3:=5x^4 + 3x^2 ^ 2_.the ideal @xmath96 has a finite codimension since @xmath310 ; see ( * ? ? ? * page 77 ) . however , the ideal @xmath311}$ ] has an infinite codimension in @xmath49 $ ] . the reason for this is as follows . the reduced basis of @xmath96 with respect to @xmath312 is given by g:=3125 ^ 3 + 108 ^ 4 , 18 ^ 3 + 125 ^ 2x , 2x^3 + 5 ^ 2 , 2x^5 - 3 ^ 2.further for any natural number @xmath313 the remainder @xmath314 where @xmath315 therefore , @xmath316 does not belong to @xmath317 and @xmath317 is an infinite codimensional ideal . an elegant approach for answering the ideal membership problem is to provide a good presentation for ideals . in this section we define intrinsic ideals and use it for such representation . this plays a central role in developing ` singularity ` and is a prerequisite for most computations of objects presented in section [ sec4 ] . when only the trivial reparametrization @xmath318 is allowed in equation , the associated relation @xmath319 is called _ strongly equivalent relation _ ; see ( * ? ? ? * page 51 ) . then , an ideal @xmath96 is called _ intrinsic _ ( * ? ? ? * page 81 ) when @xmath96 includes all strongly equivalent classes of its members . by ( * proposition 7.1 ) , every finite codimension ideal @xmath96 in @xmath9 is intrinsic if and only if there exist nonnegative integers @xmath320 for @xmath321 so that [ intrinsic ] i=^s_i=0 _ ^m_i^n_i _ , @xmath322 and the sequence @xmath323 is strictly increasing while @xmath324 is strictly decreasing . the conditions on @xmath325 and @xmath323 make the representation unique . equation for intrinsic ideals gives a convenient answer for the ideal membership problem . the monomials @xmath326 for @xmath321 are called _ intrinsic generators _ of the intrinsic ideal @xmath327 for non - intrinsic ideals or more generally for a vector space @xmath328 we define their intrinsic part ; i.e. , we denote @xmath329 for the largest intrinsic ideal contained in @xmath96 . [ lemintdecom ] for an intrinsic ideal @xmath96 , there always exists a reduced monomial standard basis for @xmath96 that includes its intrinsic generators . let @xmath96 be a finite codimension vector subspace of @xmath9 or be a finitely generated ideal . then , there exist nonnegative integers @xmath330 and @xmath331 a reduced monomial standard basis @xmath332 for @xmath333 and @xmath334 so that [ intdecom ] i= itr(i)+ \{g_i}^m_i=1_,where @xmath335 or @xmath336 here , none of the terms in @xmath337 is divisible by @xmath338 i.e. , @xmath339 let the intrinsic ideal @xmath96 be given by and @xmath32 be the set of its intrinsic generators . then for @xmath340 we consecutively update @xmath32 by adding the monomials of type @xmath341 to @xmath32 where @xmath342 @xmath343 and @xmath341 is not divisible by the elements of @xmath126 this gives rise to a reduced standard monomial basis for @xmath327 for the second part , we divide the generators of @xmath96 by the reduced standard monomial basis of @xmath333 and define the nonzero remainders as @xmath344 a refinement of the decomposition for finite codimension ideals is given as follows . we denote @xmath345 for the set of all monomials not in @xmath328 while @xmath346 stands for the vector space generated by @xmath345 . now by ( * ? ? ? * corollary 7.4 ) , we have [ idealint ] i = itr(i)(i^_k ) . in this section we describe our suggestion on how to compute the intrinsic ideal representation and next , the representation for finite codimensional ideals . the first step here is to find a lower and upper bound for the maximum natural number @xmath72 so that @xmath347 for any ideal @xmath348 in a ring @xmath349 we define @xmath350 for @xmath351 obviously for j:=i : = , @xmath352 is a finite dimensional vector space and @xmath353 is a linear nilpotent map of nilpotent degree @xmath354 therefore , we have \{n_x , n_}k < n_x+n_. now we explain how to derive @xmath355 and @xmath356 computing a ( reduced ) standard basis for @xmath328 say @xmath357 we now introduce a vector space basis generator [ iperp ] b=\{w| w wlt(g_i)|g_is_}. for @xmath352 ; also see ( * ? ? ? * pages 177179 and proposition 4 ) and ( * ? ? ? * pages 128129 ) . [ e / ilem ] let @xmath96 be a finite codimension ideal in @xmath9 and @xmath118 be a standard basis for @xmath96 . then , @xmath358 is a @xmath46-vector space basis for @xmath359 since @xmath96 is a finite codimensional ideal , is so @xmath360 thus , @xmath361 is a finite set . assume that there exists @xmath362 for some @xmath363 where @xmath364 since @xmath118 is a standard basis , @xmath365 for some @xmath366 however , @xmath367 this contradicts with , unless @xmath368 for all @xmath115 this concludes that @xmath361 is @xmath46-linearly independent . for any @xmath369 we have @xmath370 ( rem(g , s , _ ) ) lt(i)_=,and @xmath371 is a polynomial germ . this concludes the proof . lemma [ e / ilem ] readily provides a matrix representation ( and thus , its nilpotent degree ) for @xmath372 the natural number @xmath373 satisfying @xmath374 is consecutively increased to obtain @xmath375 in fact , the remainders of @xmath376 divided by the standard basis @xmath118 conclude the result , thanks to theorem [ m10 ] ( part d ) . next we look for the maximum values of @xmath325 and @xmath323 such that @xmath377 our suggestion uses the concept of colon ideals . let @xmath378 and @xmath379 be two ideals in @xmath9 . we define the _ colon ideal _ of @xmath378 by @xmath379 ( see ( * ? ? ? * definition 5 , page 194 ) ) as @xmath380 by an inductive procedure and repeating the above for the colon ideal @xmath381 we obtain @xmath325 and @xmath323 as desired . ( note that @xmath382 and @xmath383 for any @xmath254 . ) therefore , the only remaining challenge is to compute the maximum value @xmath330 so that @xmath384 or equivalently , @xmath385 the following lemma and its follow up comments facilitate the computations of a finite generating set for colon ideals and next , their ( reduced ) standard basis readily determines @xmath386 the procedure mentioned above ( using equation and @xmath387 ) provides suitable lower and upper bound for @xmath386 [ colonlem ] suppose that @xmath96 is a finite codimensional ideal in @xmath9 , @xmath388 $ ] and @xmath389}$ ] . then , @xmath390 further , let @xmath391}=\langle h_{1 } , \ldots , h_{n}\rangle_{k[x , \lambda]}$ ] for a monomial germ @xmath392 $ ] . then , [ colon ] i : g _ = , , _ . let @xmath393 . since @xmath96 has a finite codimension , @xmath219 for some @xmath72 . hence , @xmath394=j.$ ] theorem [ thm2.18 ] ( part 3 ) completes the proof of the first part . for any @xmath395 , we have @xmath396 . thus , [ m9 ] ugig_. for any @xmath397,$ ] @xmath398 for some @xmath214 thereby , ig_k[x , ] = jg_k[x , ] = h_1 , , h_n_k[x , ] . by the first part , we have ig_=h_1 , , h_n_. this and equation imply that @xmath399 and @xmath400 . now let @xmath399 . hence , @xmath401 for any @xmath402 . on the other hand , @xmath403 shows that @xmath404 and eventually @xmath395 . given lemma [ colonlem ] , we would merely need to compute @xmath391}$ ] for the monomials @xmath405 note that the above lemma is a slightly modified version of ( * ? ? ? * theorem 11 , page 196 ) and ( * ? ? ? * theorem 5.5 , page 185 ) . we simply use @xmath391}$ ] rather than using the naive and alternative choice @xmath406 the reason for this is that the computation of @xmath391}$ ] using the lexicographic ordering is efficient and classic ( described below ) while such an approach for @xmath407 does not seem easy to work with standard basis with local orderings . finally , we recall the classical approach on the intersection computation of two ideals @xmath378 and @xmath379 in the polynomial germ ring @xmath49 $ ] ; see ( * ? ? ? * page 187 ) for more details . let @xmath408}$ ] and @xmath409}.$ ] we define j_3:= t f_i , ( 1-t)g_j| i= 1 , , n_1 j=1 , n_2 _ k[x , , t ] . thus , @xmath410= j_1\cap j_2.$ ] we may compute a basis for @xmath411 with respect to @xmath56 ( where @xmath412 ) g_3:= \{p_i(x , ) , q_j(x , , t)| i= 1 , , n j=1 , m},here @xmath413 represents those basis elements independent of @xmath414 while @xmath415 represents those explicitly depending on @xmath416 since @xmath417}= lt(j_3),$ ] we may conclude that g:= \{p_i(x , ) | i= 1 , , n } is a basis for @xmath418 the reason is as follows . for any @xmath419 , either @xmath420 or @xmath421 for some @xmath254 or @xmath171 must divide @xmath422 since @xmath414 divides the leading term of @xmath415 for any @xmath423 @xmath424}.$ ] given equation , we merely need to use the generators of @xmath328 a standard basis for @xmath333 and theorem [ m10 ] ( part d ) in order to obtain @xmath425 this completes the required approach for ideal representation of any finite codimension ideal in @xmath9 . applications of the above procedure is when @xmath96 is given by a finite set of generators @xmath426 thus , it is essentially useful ( when it is possible ) to instead use a truncated taylor expansion . this is justified when the hypothesis of theorem [ thm2.18 ] ( part 5 ) is satisfied . parts ( 1 , 3 ) in theorem [ thm2.18 ] indicate that for finite codimension ideals in @xmath259 we may simply replace @xmath335 with either @xmath70 or @xmath50.$ ] part ( 6 ) in theorem [ thm2.18 ] provides an important computational criterion for ideals with infinite codimension in @xmath9 . in this section we recall the algebraic tools and present our suggested approaches that are needed for computation of _ normal forms _ , _ universal unfolding , _ and _ persistent bifurcation diagram classification . _ we acknowledge frequent fruitful discussions with professor a. hashemi . the second author acknowledges b. m .- alizadeh s helps , ideas and lessons in the early stages of this project . they were essentially helpful in a fast learning of the concepts from algebraic geometry and their programming . given a singular germ @xmath369 , from ( * ? ? ? * pages 8889 ) we recall the intrinsic ideals @xmath427 and @xmath428 by ( g):= itr(xg , g , x^2g_x , g_x _ ) , and [ sg ] ( g):=_(m_i , n_i ) _ ^m_i^n_i _ , where @xmath429 and there would not exist nonnegative integers @xmath135 and @xmath430 such that @xmath431 @xmath432 @xmath433 the extra restrictions on @xmath325 and @xmath323 here make the presentation unique as of those in equation . * proposition 8.6 ) indicates that for any germ @xmath434 @xmath435 is contact - equivalent to @xmath27 terms in @xmath427 are called _ high order terms _ ; see ( * ? ? ? * page 89 ) . therefore , we may stay contact - equivalent to @xmath11 by removing all terms in @xmath436 from taylor expansion of @xmath27 therefore , @xmath427 has a finite codimension if and only if @xmath11 is _ finitely determined _ , i.e. , @xmath11 is contact - equivalent to a polynomial germ . * theorems 8.3 and 8.4 ) state that for any term @xmath437 in @xmath438 we must have @xmath439 @xmath440 while @xmath441 for any intrinsic generator @xmath442 in @xmath443 now we recall the _ intermediate order terms _ as terms belonging to a:=(g)^(^(g)\{x^m_i^n_i| x^m_i^n_i ( g ) } ) . now we are ready to provide an algorithm for computing _ normal form _ of a given germ @xmath27 using the procedure given in subsection [ compintrins ] , we may compute @xmath427 and remove all terms in @xmath436 from taylor expansion of @xmath11 to obtain a more simplified contact - equivalent germ , say @xmath5 . now it only remains to eliminate intermediate order terms @xmath444 from @xmath5 as many as possible . then , this gives rise to its _ normal form_. when @xmath444 is empty , @xmath5 will be called _ normal form _ of @xmath27 otherwise , we may use suitable polynomial change of variable @xmath24 and positive polynomial germ @xmath445 to eliminate some intermediate order terms from @xmath5 . for example , we may replace @xmath1 by @xmath446 in @xmath5 where @xmath447 and @xmath448 are arbitrary constant coefficients . this gives rise to a system of linear equations and a maximal solvable subsystem leads to the further elimination of negligible terms in @xmath5 and thus , the normal form computation of @xmath27 this approach needs to be systematically adopted along with standard basis computations for high order terms in multi - state variable cases . this is because complete algebraic characterization for high order terms is not yet known for many multi - state variable cases . this will be addressed in details in our upcoming result for multi - state variable cases . in this section we recall the algebraic formulation needed for computation of the universal unfolding of a singular germ @xmath27 the tangent space @xmath449 is defined by t(g):= g , xg_x , g_x _ k\{g_x , g_,g _ , , ^g _ } , for sufficiently large @xmath450 for a finite codimension @xmath25 there exists a natural number @xmath451 so that @xmath452 and @xmath453 for @xmath454 ; see ( * ? ? ? * page 127 ) . next , a universal unfolding of @xmath11 is defined by g(x , , ) : = g(x , ) + _ i=1^k_i p_i(x , ) , where @xmath455 @xmath456-s form a basis for a complement space of @xmath457 thus , we may choose @xmath458 the number @xmath72 is called _ codimension _ of @xmath449 or equivalently _ codimension _ of @xmath11 . the tangent space @xmath449 is computed via lemma [ lemintdecom ] and the procedure given in section [ compintrins ] . since equation is a direct summation here , a complement space basis can be obtained through equation and a linear algebra computation . given our description in section [ sec1int ] , persistent bifurcation diagram classification is complete by simplifying the defining equations of transition set @xmath35 and then , choosing one parameter from each connected component of @xmath459 ; see ( * ? ? ? * page 140 ) . the latter is enabled in singularity by using the maple package regularchains . let @xmath11 be a singular germ of finite codimension and @xmath460 be a universal unfolding of the germ @xmath27 we recall that the transition set @xmath461 where & : = & \{^kg = g_x = g_=0 ( x , , ) ( x , ) } , + [ trans ] & : = & \{^kg = g_x = g_xx=0 ( x , , ) ( x , ) } , + & : = & \{^kg = g_x=0 ( x_i , , ) i=1 , 2 x_1x_2}. possible reduction of variables in the defining equations ( in particular removing @xmath1 and @xmath10 from the equations ) in is desirable . in order to achieve this goal , we let @xmath462 $ ] and @xmath463 be the ideals generated by polynomial defining equations for either of @xmath38 and @xmath464 here @xmath465 stands for all germs of smooth functions of @xmath466 and @xmath64 for the case of @xmath40 , a new variable @xmath467 is introduced and the quadratic germ @xmath468 is also added to the generators of the ideal @xmath96 and @xmath348 . this is due to the fact that @xmath469 next , we compute the basis @xmath32 for @xmath96 in @xmath470 , \alpha\in \mathbb{r}^k,$ ] with respect to @xmath471 for @xmath472 thus by ( * ? ? ? * theorem 2 ) , @xmath473 $ ] is a basis for @xmath474.$ ] hence , @xmath473\subseteq j\cap { \mathscr{e}}_\alpha.$ ] additional restrictions on the transition sets are also obtained by the other elements of basis @xmath126 the reader are referred to our user - guide @xcite for a list of all functions , their capabilities , options and comprehensive information on how to work with ` singularity ` . in this section we merely describe the main features of ` singularity ` . the main functions ( not all ) are given in tables [ mainfuncsing ] and [ mainfuncgeom ] . to illustrate how these functions work , let @xmath475 then , @xmath476 returns @xmath477 @xmath478 @xmath479 @xmath480 @xmath481 @xmath482 and intrinsic generators of @xmath483 are @xmath484 and @xmath485 for an example of computations in infinite codimensional ideals , consider restricted tangent space of @xmath486 @xmath487 generates the restricted tangent space of @xmath11 as @xmath488 .`singularity ` s main functions in singularity theory . [ cols= " < , < , < , < " , ] let g:=\{g_1:=(^7+x)+(x^4)-x-1-^9 , g_2:=x^5-^2 , g_3:=(x^6)--1},@xmath489 $ ] and @xmath72 be a sufficiently large truncation degree . then , using the commands ` standardbasis`(g , ` vars ` , k , ` fractional ` ) , ` standardbasis`(g , ` vars ` , k , ` formal`),and @xmath490 we obtain the standard basis of @xmath32 in the rings of fractional maps , formal power series and smooth germs @xmath172 in this example , the standard basis in either of these rings is given by @xmath491 note that the options @xmath492 and @xmath493 in @xmath494 use a truncated taylor series expansion of germs ( for non - polynomial and non - fractional germs ) and are adopted according to theorem [ thm2.18 ] . our future and in progress projects are about multi - state dimensional singularities , singularities depending on symbolic coefficients and the study of moduli , equivariant singularities with compact symmetries , and an integration of ` singularity ` with the normal form analysis of singular ordinary differential equations . as our research progresses , we will update ` singularity ` and its user guide @xcite . in this paper , we have considered scalar singularities . however , multi - state dimensional singularities frequently occur in applications and an enhanced capability of ` singularity ` for their study is an important contribution . such enhancement enforces implementation of extended concepts and tools for module structures ; see @xcite . the study of local zero structures for parametric smooth maps with symbolic coefficients in the vicinity of their critical points is important for applications . here , we refer to parameters as small control parameters or perturbations while symbolic coefficients refer to those coefficients assumed to be constant . further , the complex dynamics associated with singular germs with positive modality requires certain techniques from algebraic geometry so that they can work with symbolic coefficients and generate their associated classifications . this requires some extensions of standard basis ; in our opinion , extensions similar to the notions of _ comprehensive basis _ and _ system _ in the context of local rings of germs ; see @xcite . this is in progress and will be addressed in details as a separate work . the possible enhancement of ` singularity ` for equivariant singularities with compact symmetries seems a challenging task but it is indeed a worthwhile contribution . we believe this should be closely related to a local germ ring version of sagbi basis ; e.g. , see @xcite . the contact equivalence relation is compatible with the equivalence relation arising from local invertible changes of variables and time rescaling in differential equations around an equilibrium ; see @xcite . thus , ` singularity ` can be integrated with the normal form computations of differential equations . the aim is to locate the local bifurcation of singular differential equations in terms of the coefficients and parameters of the original systems . this contributes to the local bifurcation control and to design efficient controllers for uncontrollable engineering problems ; see @xcite . arnold , s.m . gusein - zade , a.n . varchenko , _ singularities of differentiable maps _ , i : the classification of critical points caustics and wave fronts . arnold , _ local normal forms of functions , _ inventiones math . 35 ( 1976 ) 87109 . , _ bifurcation theory and computer algebra : an initial approach , _ lecture notes in computer science 204 , 1985 , pp . 126137 . , _ constructing universal unfolding using grbner basis _ , j. symbolic computation 2 ( 1986 ) pp . 383388 . , _ bases . a computational approach to commutative algebra _ , graduate texts in mathematics 149 , 574 pp . , springer verlag 1993 . , _ standard basis in power series rings : uniqueness and superfluous critical pairs _ , j. symbolic computation 15 ( 1993 ) pp . 251265 . , _ bifurcations in hamiltonian systems _ , computing singularities by bases , springer , berlin 2003 . mathematical institute , university of innsbruck , austria . phd thesis , 1965 . aequationes mathematicae 3 , ( 1970 ) 374383 . in : multidimensional systems progress , directions and open problems in multidimensional systems , n.k . bose ( ed . ) , chapter 6 ( 1985 ) 184232 . multidimensional systems and signal processing * 12 * ( 3/4 ) ( 2001 ) 223253 . lecture notes in computer science 4770 ( 2007 ) pp . 73101 . , _ computer algebraic tools for applications to catastrophe theory _ , europian conference on computer algebra . , lecture notes in computer science , springer 378 ( 1987 ) pp . , _ ideals , varieties , and algorithms _ , springer - verlag , new york 1991 . , _ using algebraic geometry _ , springer - verlag , new york 1998 . , _ a first course in computational algebraic geometry _ , cambridge university press , 2013 . a. dhooge , w. govaerts , y.a . kuznetsov , _ matcont : a matlab package for numerical bifurcation analysis of odes , _ acm transactions on mathematical software * 29 * ( 2003 ) 141164 . multiparameter bifurcation theory , contempary mathematics 56 ( 1986 ) pp . 97118 . , _ automatic classification of normal forms _ , j. nonlinear analysis 34 ( 1988 ) pp . 157190 . , _ computational algebra for bifurcation theory _ , j. symbolic computation 40 ( 2005 ) pp . 11801207 . , _ a user guide for singularity , _ arxiv:1601.00268 ( 2016 ) . , _ parametric normal forms for bogdanov takens singularity ; the generalized saddle - node case _ , discrete and continuous dynamical systems 35 ( 2015 ) pp . 205224 . , _ normal forms for hopf zero singularities with nonconservative nonlinear part _ , j. differential equations 254 ( 2013 ) pp . 15711581 . , _ volume - preserving normal forms of hopf zero singularity _ , nonlinearity 26 ( 2013 ) pp . 28092832 . , _ normal forms of hopf zero singularity _ , nonlinearity 28 ( 2015 ) pp . 311330 . , arxiv preprint arxiv:1412.5399 ( 2015 ) . , _ spectral sequences and parametric normal forms _ , j. differential equations 252 ( 2012 ) pp . 10031031 . , _ involutive bases algorithm incorporating @xmath495 criterion _ , j. symbolic computation 59 ( 2013 ) pp . 120 . , _ singularities and groups in bifurcation theory _ , volumes 1 - 2 , springer , new york 1985 and 1988 . , _ computation of singularities in large nonlinear systems _ , siam j. numer . 34 ( 1997 ) pp . 867880 . , _ _ n__umerical methods for bifurcations of dynamical equilibria , siam , 2000 . , _ a singular introduction to commutative algebra _ , springer - verlag , new york 2007 . , _ comprehensive involutive systems , _ lecture notes in computer science , volume 7442 , 2012 , pp . 98116 . bulletin of the iranian mathematical society 38 ( 2012 ) pp . 715724 . , _ resolution of singularities of an algebraic variety over a field of characteristic zero _ , annals math . 79 ( 1964 ) pp . 109326 . , _ a new algorithm for computing comprehensive groebner systems _ , proc . symp on symbolic and algebraic computation , munich , july 2010 . computers and mathematics ( cambridge , ma ) , springer , 1989 , pp . ( 1986 ) 11142 . nonlinearity 1 ( 1987 ) 215240 . , _ a new algorithm for discussing bases with parameters _ , j. symbolic computation 33 ( 2002 ) pp . 183208 . springer - verlag , new york , 2003 . j. murdock , _ hypernormal form theory : foundations and algorithms , _ j. differential equations * 205 * ( 2004 ) 424465 . j. murdock , d. malonza , _ an improved theory of asymtotic unfoldings , _ j. differential equations * 247 * ( 2009 ) 685709 . pitman , 1978 . , _ normal forms and unfoldings of singular strategy functions _ , dynamic games and applications 5 ( 2014 ) pp . let @xmath496 since @xmath497 part ( 1 ) in theorem [ thm2.18 ] implies that @xmath210 and @xmath498 hence , without loss of generality we may assume that @xmath499 thus , g= a_i lt(f_i)= j^ka_i lt(f_i)= j^k(a_i)lt(f_i)+ r , where @xmath500 and @xmath501 now we only need to prove that @xmath502 when @xmath503 since @xmath504 @xmath505 for some @xmath506 thus , j^k(f_i)= j^kj^k(b_ij ) q_j= j^k(b_ij ) q_j+s where @xmath507 therefore , @xmath508 on the other hand , lt(f_i ) = lt(j^k(f_i ) ) & & lt_q_1 , , q_m_= lt(q_1 ) , , lt(q_m ) _ + & & lt(q_1 ) , , lt(q_m)_. this completes the proof . |
we present black hole mass measurements from kinematic modeling of high - spatial resolution integral field spectroscopy of the inner regions of 9 nearby ( ultra-)luminous infrared galaxies in a variety of merger stages . these observations were taken with osiris and laser guide star adaptive optics on the keck i and keck ii telescopes , and reveal gas and stellar kinematics inside the spheres of influence of these supermassive black holes . we find that this sample of black holes are overmassive ( m ) compared to the expected values based on black hole scaling relations , and suggest that the major epoch of black hole growth occurs in early stages of a merger , as opposed to during a final episode of quasar - mode feedback . the black hole masses presented are the dynamical masses enclosed inpc , and could include gas which is gravitationally bound to the black hole but has not yet lost sufficient angular momentum to be accreted . if present , this gas could in principle eventually fuel agn feedback or be itself blown out from the system . | 1502.06617 | virtually every massive galaxy hosts a supermassive black hole in its core . these central black holes exhibit tight correlations with properties of their host galaxies bulges @xcite : bulge mass @xcite , bulge luminosity @xcite , and bulge stellar velocity dispersion @xcite . recently these correlations were re - explored using the wealth of new black hole mass measurements ( @xmath3 ) and galaxy parameters by @xcite and @xcite . however , the discovery that the @xmath4 relation may evolve with redshift ( e.g. * ? ? ? * ) has led to the suggestion that the most basic scaling relation may instead be with total stellar mass @xcite ; in this case , the evolution of the @xmath4 relation with redshift indicates the changing fractions of mass in galaxy bulges ( which contribute to @xmath5 ) versus disks ( which do not ) . see @xcite for a detailed review of our current understanding of black hole scaling relations . the mechanism through which black hole masses correlate with galaxy properties has been canonically associated with gas - rich galaxy mergers ( e.g. * ? ? ? gravitational torques funnel the gas into their centers , triggering two phenomena : an intense burst of star formation to feed the bulge , and accretion of gas on to the black holes in the centers of each galaxy . it has been postulated that black hole growth can regulate this process through agn feedback @xcite via massive winds that evacuate the gas from the galaxy on short timescales , cutting off star formation and future black hole growth . this sense of self - regulation has been confirmed observationally by @xcite , who find that the eddington ratio of a sample of agns depends on the supply of cold gas in the galaxy . if there is plenty of cold gas , the accretion rate does not depend on the quantity of gas available ; if the supply of cold gas is limited , the accretion rate depends on the rate at which stellar winds provide fuel for the agn . though the detailed mechanisms causing these correlations are still unconfirmed , star formation and black hole growth are fed by the same reservoir of inflowing gas ; their growth histories are intertwined . it is likely that these two processes compete for fuel in a predictable fashion . to understand this interplay , it is critical to look at systems in the midst of this increased fueling . one set of such galaxies are gas - rich mergers , which tend to have extreme bursts of star formation and a higher incidence of agn activity ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? during such a merger , does the black hole grow first , leaving the stars to slowly consume the remaining gas ? or is star formation quenched once the black hole reaches a bright quasar phase of extreme growth ? the position of a merger on black hole scaling relations would indicate the relative growth timescales , and confirm whether the putative quasar - mode feedback occurs at the end of a merger ( see figure [ scalingrelationtracks ] for the schematic example for the @xmath4 relation . ) . as star formation happens on few - kiloparsec spatial scales , accreting gas must lose less angular momentum to fuel a starburst than to feed a black hole . subsequent models therefore suggest that a black hole would grow substantially only after star formation has quenched itself and the galaxy bulge is in place @xcite , or at least that the peak black hole accretion time occurs later in a merger relative to peak star formation @xcite . this scenario would predict that gas - rich mergers would fall below black hole scaling relations . relation for isolated galaxies from @xcite ( black ) with three possible evolutionary tracks for merging galaxies overlaid . if the black hole grows first or more quickly than the galaxy bulge , mergers would lie above the relation ( as shown by the gold arrows ) . if the black hole growth lags the bulge growth and is responsible for curtailing evolution ( e.g. quenching through agn feedback ) , mergers would lie below the relation ( as shown by the purple arrows ) . if instead the black hole and the bulge grow in lockstep , the mergers would remain on the relation ( as shown by the blue arrow ) . ] merger - driven galaxy evolution is not a complete explanation for all black hole growth , however . though some agn studies find a correlation with major mergers @xcite , agns found with other selection techniques and at different redshift ranges do not show a higher rate of merging than field galaxies @xcite . instead of mergers bringing in gas , some systems probably are undergoing secular evolution , accreting their gas directly from the cold intergalactic medium . bar and spiral disk instabilities are capable of dissipating sufficient angular momentum from this gas to fuel black hole growth @xcite . the dividing line between these two processes is not yet well understood ; here we focus on understanding the black hole growth due to major mergers . nearby ( ultra-)luminous infrared galaxies ( ( u)lirgs ; * ? ? ? * ) are an excellent sample with which to study such gas - rich mergers . these galaxies have infrared luminosities upwards of @xmath6 ( @xmath7 for ulirgs ) , generally caused by a starburst and/or an agn heating up dust . their infrared luminosities correlate with merger rate , star formation rate , and agn fraction @xcite . in fact , in the local universe , such strong infrared activity is almost exclusively triggered by major mergers : @xcite showed that in a complete sample of iras ulirgs , 117 out of 118 galaxies are in the midst of strong tidal interactions . in a companion paper @xcite , we have studied the kinematics of both gas and stars in the inner kiloparsecs of a sample of ( u)lirgs , finding that nuclear disks on scales of a few tens to hundreds of parsecs are common . in two other papers @xcite , we demonstrated a technique that uses high spatial resolution integral field spectroscopy to measure black hole masses . with kinematic maps that resolve inside the sphere of influence of a black hole , the complex and unrelaxed large - scale dynamics are less important , and black hole masses can be measured to within a factor of a few . we note that this technique measures the unresolved central mass , which includes both the black hole and its accretion disk , and in some cases may also include a reservoir of gas feeding the accretion disk . in @xcite , the black hole mass in mrk273 n measured with this technique was consistent with the measurement made by oh maser kinematics @xcite . using a robust technique such as this to measure black hole masses in these gas - rich mergers is important because more traditional methods of black hole mass measurements rely on assumptions that are nt valid in the case of galaxy mergers . three - integral orbital superposition models ( as in * ? ? ? * ; * ? ? ? * ) are able to use large - field kinematics to separate different components to the mass profile of the galaxy ; however , this approach requires a dynamically relaxed system and is used therefore in isolated galaxies . another successful black hole mass measurement technique is reverberation mapping ( e.g. * ? ? ? * and references therein ) , which measures the time lag between flux variations of the continuum and the lines in the broad line region . since the cores of ( u)lirgs are so dusty , the broad line regions are too obscured to view . obtaining data at spatial resolutions sufficiently high to resolve inside the sphere of influence of a supermassive black hole at the typical redshifts of local ( u)lirgs ( @xmath8 ) requires adaptive optics systems , which are becoming available at an increasing number of ground - based observatories . though the hubble space telescope has excellent resolution in the visible bands , its relatively small mirror size limits the resolution at the longer wavelengths ( @xmath9 m ) necessary to look through the dust in these galactic nuclei ; even at the longest wavelength available to the wide field camera 3 ( @xmath10-band ) , these nuclei are still sometimes obscured . in order to achieve high spatial resolution in @xmath11-band , large ground - based telescopes have employed adaptive optics systems which measure turbulence in the earth s atmosphere and use a deformable mirror to correct for the resulting distortions . these distortion measurements require references , either a natural guide star ( ngs ao ) or a laser guide star plus a fainter natural `` tip - tilt '' star ( lgs ao ) . the addition of laser guide star adaptive optics has increased the area of the sky observable with this technique . throughout this paper we have adopted a cosmology of @xmath12kms@xmath13mpc@xmath13 , @xmath14 = 0.28 , and @xmath15 = 0.72 @xcite . in [ obs ] we present our data and reduction techniques . in [ kinematicfitting ] we briefly describe the kinematic fitting techniques demonstrated in @xcite and @xcite . in [ results ] we present the black hole masses measured from several tracers and in [ scalingrelations ] compare them to black hole scaling relations . [ conclusions ] contains our conclusions . we have selected 9 gas - rich merging galaxies in which to measure the black hole masses using gas and stellar kinematics . these galaxies represent the subset of merging galaxies presented in @xcite for which high quality kinematics data exist . the parent sample was drawn from the great observatories all - sky lirg survey ( goals ; * ? ? ? * ) , which targeted about two hundred of the brightest infrared galaxies in the sky ( log(l@xmath16/l@xmath1 ) @xmath17 11.0 ) . our targets were selected from that sample for the following criteria : available archival @xmath18- and @xmath19-band imaging from the hubble space telescope advanced camera for surveys ; visible double or extended cores or clumps in the central five arcseconds ( within our field of view ) ; large - scale morphology consistent with that of a major merger ; observable from keck observatory and with an appropriate guide star ( @xmath20 18th magnitude in @xmath21 , @xmath22 separation ) . our sample comprises ( u)lirgs involved in major gas - rich galaxy mergers , with a bias towards later stages of merging . these galaxies are nearby , with redshifts @xmath23 , which is required to achieve spatial resolutions of @xmath24 pc per pixel ( @xmath25 pc per resolution element ) . we obtained near - infrared integral field spectroscopy of the central kiloparsec of 9 merging galaxies with osiris , the oh - suppressing infrared imaging spectrograph @xcite on the w. m. keck i ( after august 2012 ) and ii ( before august 2012 ) 10-meter telescopes . all data presented here uses the 0.035 arcsec spaxel@xmath13 plate scale . this high spatial resolution is enabled by the keck observatory lgs ao system @xcite . our osiris observations were typically comprised of observing sets of object - sky - object ; each exposure was ten minutes ( 5 minutes for keck i observations ) . observations were taken either in the broad - band @xmath11 filter ( kbb : 1.965 - 2.381 @xmath26 m ) or in a narrow band targeting specific lines ( hn4 : 1.652 - 1.737 @xmath26 m ; kn5 : 2.292 - 2.408 @xmath26 m ) . when possible , the primary target lines were the co ( 2 - 0 ) and ( 3 - 1 ) bandheads at 2.293 @xmath26 m and 2.323 @xmath26 m in @xmath11 , which trace the kinematics of young stars . when available , we compare the stellar kinematics to those of various emission lines : [ fe ii ] 1.644 @xmath26 m , br@xmath27 2.16 @xmath26 m , h@xmath28 2.12 @xmath26 m . for some galaxies , stellar kinematics were not available because the co bandheads are redshifted out of @xmath11 , and we rely on the aforementioned emission lines for this analysis . total exposure times , observed filters , and parameters of the observed galaxies are listed in table [ tbl : observingparams ] . some of these data were presented in previous papers @xcite , these are indicated in table [ tbl : observingparams ] as well . we reduced our keck ii observations using the osiris data reduction pipeline version 2.3 ( using the updated osiris wavelength solution for data taken after october 2009 ) and our keck i observations using version 3 . this pipeline includes modules to subtract sky frames , adjust channel levels , remove crosstalk , identify glitches , clean cosmic rays , extract a spectrum for each spatial pixel , assemble the spectra into a data cube , correct for atmospheric dispersion , perform telluric corrections , and mosaic frames together . for some cases we utilized the scaled sky subtraction module based on the technique outlined in @xcite , which scales the thermal continuum and oh line groups separately to provide optimal sky subtraction ; we modified the module to include a smoother subtraction of the thermal continuum , as described in @xcite . we imaged the tip - tilt stars for each galaxy during the course of the observations in order to obtain an estimate of the point - spread function ( psf ) . a moffat function was fit to each tip - tilt star and then broadened according to the distance between the target and the tip - tilt star . this broadening accounts for the fact that the laser spot is not at infinity ; therefore it only probes most of the turbulence that affects the target . as a result , the further a target is from its tip - tilt star , the worse its psf will be . the isokinetic angle for the keck ii ao system , at which the strehl ratio is reduced to 37% of its peak value , is approximately 75 arcseconds . we also use this an estimate of the keck i ao system performance , which has not yet been characterized . lcccccccc cgcg436 - 030 & 0.0315 & 21.8 & 2012 sep 30 & kn5 & 30 & 5 + irasf01364 - 1042 & 0.0493 & 33.6 & 2012 oct 01 & kbb & 20 & 10 + iiizw035 & 0.0278 & 19.5 & 2011 dec 10 & kbb & 100 & 50 + mcg+08 - 11 - 002 & 0.0195 & 13.9 & 2012 jan 02 & kbb & 90 & 50 + ngc 2623 & 0.0196 & 13.9 & 2010 mar 04 , 2010 mar 05 & kn5 & 100 & 50 + ugc5101 & 0.0413 & 27.0 & 2010 mar 04 , 2010 mar 05 & kn5 & 80 & 50 + mrk273/ugc8696 & 0.038 & 26.4 & 2012 may 22 & hn4 & 60 & 30 + ngc 6240 n & 0.0244 & 17.2 & 2009 jun 17 & kn5 & 210 & 75 + ngc 6240 s & 0.0244 & 17.2 & 2007 apr 21 & kn5 & 20 & 10 + irasf17207 - 0014 & 0.0462 & 29.8 & 2011 may 23 , 2011 may 24 & hn4 & 40 & 20 + [ tbl : observingparams ] we measured the kinematics following the methods presented in detail in ( * ? ? ? * for stellar kinematics ) and ( * ? ? ? * for emission line kinematics ) , and reviewed again in @xcite . we briefly describe these techniques here . for each measurement , we first calculate the signal - to - noise ratio in each pixel and bin them using optimal voronoi tesselations @xcite in order to require a certain signal - to - noise ratio for reliable measurements . we used thresholds of 20 and 3 per resolution element for stellar kinematics and emission line kinematics , respectively . once binned , stellar kinematics were fit using the penalized pixel fitting routine @xcite and using @xmath11-band templates of late - type giants and supergiants from gnirs @xcite . emission lines were fit with gaussian profiles to determine the flux , velocity , and velocity dispersion in each bin . for emission lines where multiple lines exist in a specific band ( the 5 h@xmath28 transitions in the @xmath11-band , or br@xmath29 and br@xmath27 ) , those lines were fit simultaneously , requiring that the velocity and velocity dispersion be consistent between lines . our black hole mass measurement techniques have been demonstrated in two prior papers . in @xcite , we place limits on the mass of the black hole in the south nucleus of ngc 6240 using 2-d stellar kinematic maps and two methods . we calculate the lower limit to the mass by assuming the stars lie in a thin keplerian disk , and fit a density profile including a black hole and a smoothly - varying spheroidal mass component : @xmath30 . because of the high spatial resolution of our data , we probe radii small enough that the first term dominates over the second . the measured black hole mass here is a lower limit because this method ignores velocity dispersion ( assuming that all measured velocity dispersion is due to material along the line of sight , not intrinsic to the disk ) . any intrinsic velocity dispersion in the disk will be indicative of extra mass in the black hole . in the second scenario , we obtain an upper limit on the black hole mass by assuming the opposite : that all measured velocity dispersion is intrinsic to the disk . because the system is unrelaxed , there is unvirialized material along the line of sight increasing the velocity dispersion but not knowing about the central black hole mass . to find this limit , we use jeans axisymmetric mass models ( jam ; * ? ? ? we use both methods here for all galaxies with measured stellar kinematics , except for the north black hole in ngc 6240 . this nucleus contains patchy dust that prevents a consistent light profile from being produced by the jam models ; in order to use this measurement technique on that black hole , equivalently high resolution imaging at a longer wavelength ( @xmath31-band or beyond ) is required . in @xcite , we measure the mass of the black hole in the north nucleus of mrk273 in a similar manner . since no stellar kinematics were available , the jam models were not feasible . when only gas kinematics are available , we model the gas as a thin keplerian disk and proceed as in the first case above . the black hole mass measured for this galaxy agrees well with the measurement made using the kinematics of an oh maser @xcite , and therefore provides independent justification for this technique . we note that the thin keplerian disk approximation provides reasonable @xmath32 agreement to the maser measurement even though the disks analyzed in @xcite are thick rather than thin , with @xmath331 - 4 . we note that this thickness may cause our black hole mass measurements to underestimate the true black hole mass by up to @xmath34% . with this approach , we are measuring the unresolved central mass within @xmath35 parsecs . this is likely dominated by the black hole , and we therefore refer to it as such . however , this mass includes the accretion disk of the black hole and may include a reservoir of gas feeding the accretion disk . if the accretion disk is fueled smoothly by the nuclear disk , this mass will be accounted for by the radial profile ; we can not account for a pile - up of gas below our resolution limit . the possible implications for this will be discussed further in section [ scalingrelations ] . in each case , errors on the black hole mass measurement were obtained using a monte carlo approach . we created a model datacube and added random noise appropriate to the signal - to - noise of each spectrum and then refit the black hole mass for 100 iterations . the width of the resulting mass distribution is the error on the measurement . we have performed the above analysis for 9 additional black holes , listed in table [ tbl : observingparams ] , and include the results in table [ tbl : mbh ] and the fits in appendix [ bhmodelfigs ] . as in previous papers , these methods produce kinematic models in reasonable agreement with observations for most cases . the ability to measure black hole masses using different tracers , and additionally to use bracketing assumptions about velocity dispersion for stellar models , improves the robustness of these results . the numerical results from our black hole mass measurements are listed in table [ tbl : mbh ] , along with parameters relevant to black hole scaling relations , pulled from the literature : the stellar velocity dispersion of the bulge , the stellar mass of the bulge , and the total stellar mass of the galaxy . though consistent treatment of such parameters when drawing from different literature sources is difficult , care was taken to select measurements most appropriate to the scaling relation . the bulge luminosities and total stellar masses were taken from the goals papers @xcite and @xcite , respectively . @xcite measured the bulge luminosities using galfit @xcite to decompose @xmath10-band nicmos images into multiple srsic components . these galaxies are morphologically complex , which can produce an increased uncertainty in bulge fit parameters ; thus , the errors in bulge luminosities reported by @xcite and reproduced in table [ tbl : mbh ] may be larger than the equivalent measurements from hubble imaging of isolated galaxies . it is worth considering that we use the formal statistical errors produced by galfit ; the analysis of @xcite has shown that these underestimate the true errors , which can contain contributions from profile mismatch and nearby neighbors , both relevant to our sample . however , @xcite also find that galfit shows no systematic offset to fitting bulge luminosities of bright objects and that it is able to appropriately handle contamination by neighbors when fit simultaneously ( as was done in * ? ? ? still , for a more complete understanding of the errors associated with each bulge luminosity measurement , we refer the readers to the models and residuals for each system of @xcite , published online in figure set 7 . our total stellar masses were measured from seds using photometry masks @xcite designed to incorporate all of the light from the galaxy . we are thus not underestimating the total stellar mass , as can happen with fixed - aperture photometry , and we avoid potential biases related to varying host galaxy effective radius @xcite . the goals survey papers allow us to be confident that the bulge luminosities and total stellar masses were measured consistently across our sample . however , no such survey paper exists currently to measure the stellar velocity dispersions of the bulges . instead , we selected the stellar velocity dispersion measurement from the aperture that most closely aligned with the size of the bulge . unfortunately , no bulge information for mcg+08 - 11 - 002 is available in the literature . although it can not be placed on scaling relations at this time , we include our black hole mass measurements for completeness . lccccccc cgcg436 - 030 & br@xmath27 & @xmath36 & @xmath37 & @xmath38 & @xmath39 & 1,2,3 + irasf01364 - 1042 & br@xmath27 & @xmath40 & @xmath41 & @xmath42&@xmath43 & 1,2,3 + & h@xmath28 & @xmath44 & & + iiizw035 & stars - disk & @xmath45 & ... & @xmath46 & @xmath47&1,2 + [ 0.5ex ] & stars - jam & @xmath48 & & & + & h@xmath28 & @xmath49 & & & + [ 0.5ex ] & br@xmath27 & @xmath50 & & & + [ 0.5ex ] mcg+08 - 11 - 002 & stars - disk & @xmath51 & ... & ... & ... & ... + & stars - jam & @xmath52 & & & + & h@xmath28 & @xmath53 & & & + & br@xmath27 & @xmath54 & & & + [ 0.5ex ] ngc 2623 & stars - disk & @xmath55 & @xmath56 & @xmath57 & @xmath58 & 1,2,4 + [ 0.5ex ] & stars - jam & @xmath59 & & & + [ 0.5ex ] ugc5101 & stars - disk & @xmath60 & @xmath61 & @xmath62 & @xmath63 & 1,2,5 + & stars - jam & @xmath64 & & & + [ 0.5ex ] mrk273 n & [ fe ii ] & @xmath65 & @xmath66 & ... & @xmath67 & 1,6 + [ 0.5ex ] ngc 6240 n & stars - disk & @xmath68 & @xmath69 & @xmath70 & @xmath71&2,7,8 + [ 0.5ex ] ngc 6240 s & stars - disk & @xmath72 & @xmath73 & @xmath74 & @xmath71&2,7,8 + [ 0.5ex ] & stars - jam & @xmath75 & & & + irasf17207 - 0014 e & [ fe ii ] & @xmath76 & @xmath77&@xmath78 & @xmath79 & 2,6,7 + irasf17207 - 0014 w & [ fe ii ] & @xmath80 & @xmath77 & @xmath78 & @xmath79 & 2,6,7 + [ tbl : mbh ] using our measured black hole masses and host galaxy parameters from the literature , all listed in table [ tbl : mbh ] , we place the galaxies on several black hole scaling relations . in figure [ msigma ] , we show the @xmath4 relation from @xcite with our points overlaid . relation for isolated galaxies from @xcite ( black data and best - fit line : log(@xmath81 log@xmath82 km / s ) with our merging galaxies marked in large colored points . symbols indicate which method was used for that mass determination : gas disk modeling in circles , stellar disk modeling in upward triangles , and jam modeling in downward triangles . stellar disk models represent lower limits to the black hole mass , while jam models represent upper limits ; see text for details . n s point ( 6 ; brown ) is the mass measurement for only the north black hole , though there is a second in the system ( see * ? ? ? ] the black holes in our sample lie either within the scatter of the relation or above it . although a particular galaxy may fall within the scatter of the relation , we want instead to consider whether our sample as a whole is behaving as predicted by the @xmath4 relation . we calculate the distances from the @xmath4 relation for both our dataset and the reference population , and then perform a two - sided kolmogorov - smirnoff test on the two populations . we find a p - value of 0.003 , indicating that the chance of these two being drawn from the same population is less than 1% . to mitigate the possible effects of outliers in our small sample , we use the balanced bootstrap method to estimate the average offset from the relation . this reveals an average offset of @xmath83 in plot units ; we thus consider this offset a 2.6-sigma result . our sample of black holes are thus , on average , above the relation , suggesting with moderate significance that black holes may grow in mergers before the bulges are virialized . measuring the velocity dispersion of stars in the bulge can be difficult to define when two bulges are in the process of merging . this variation in measured velocity dispersion ( both due to the process of merging and due to geometric variations based on the line - of - sight ) has been quantified in the hydrodynamical galaxy merger simulations of @xcite . their analysis shows a maximum predicted mismeasurement in velocity dispersion of approximately 50% , which is insufficient to bring all our systems on to the relation . in figure [ mbulgelum ] , we show the @xmath84 relation . the background galaxies were plotted using @xmath85 from @xcite , with updated black hole masses from @xcite . we have updated the best - fit line with these numbers , producing the relation : @xmath86 relation . the black points have luminosities taken from @xcite and updated black hole masses from @xcite . the solid line is a log - log fit to the updated data : @xmath87 . galaxies presented here are indicated by large colored points , with parameters listed in table [ tbl : mbh ] . the bulge luminosities were not corrected for extinction ; the arrow in the top left shows the typical magnitude of such a correction ( 10% ) . symbols indicate which method was used for that mass determination : gas disk modeling in circles , stellar disk modeling in upward triangles , and jam modeling in downward triangles . stellar disk models represent lower limits to the black hole mass , while jam models represent upper limits ; see text for details . ] the black holes from our sample fall either within the scatter or above the @xmath84 relation . as with the @xmath4 relation , we perform a two - sided kolmogorov - smirnoff test to determine if our sample as a whole falls significantly above the relation ; we find a p - value of @xmath88 , indicating that these black hole masses are unlikely to be pulled from the sample distribution as the reference sample . we also calculate the mean offset using the balanced bootstrap method , @xmath89 in plot units , showing an offset with @xmath90-sigma significance . we note two possible calibration effects when considering our bulge luminosity measurements . first , although @xcite did not correct their bulge luminosities for internal extinction , we consider the effects such a correction would have on our sample because our galaxies are quite dusty . several of our galaxies have @xmath91 imaging published by @xcite , enabling a direct estimate of typical extinction over the central few kiloparsecs . these calculations estimate that the intrinsic luminosity of the bulges in our sample may be @xmath92% higher . this correction ( indicated by red arrow in the top left of figure [ mbulgelum ] ) is smaller than the size of our points and can be safely ignored . second , we investigate whether the varying bulge srsic indices of the fits in @xcite may be biasing our luminosity estimates . when fixing @xmath93 to 4 , the bulge luminosities of several of our galaxies change significantly , some higher and some lower ( s. haan , private communication ) . the net effect with this change is less scatter between our galaxies : a smaller but more statistically significant offset . we conclude that , while the srsic index fit may have a strong effect on an individual system s predicted black hole mass when using the @xmath84 relation , the offset in our data is not caused by the fitting technique of these bulge luminosities . we also note that the two galaxies which fall furthest above the relation , irasf01364 - 1042 and irasf17207 - 0014 , do not show atypically large residuals in the galfit bulge fits ( indeed , the residuals of the former appear smaller than average for this sample ) , which suggests that their offset is not due to errors in the bulge luminosities of @xcite . again , our sample of merging galaxies falls above this relation , suggesting that either stars have not been forming in the bulge as quickly as the black holes have been growing , or that a large mass of stars is still strewn about in tidal tails and unrelaxed features that will eventually become part of the bulge . as the @xmath10-band traces older populations of stars , it may be that some stars have recently formed in the bulge , but that this measurement misses them . although @xmath94 relations do exist for bluer colors , we are unfortunately unable to utilize them for this sample because the large amounts of dust obscure the galaxy cores , making bulge luminosity measurements in the optical highly uncertain . we have compared our bulge luminosities to those fit by @xcite in the @xmath19-band , and found that ( for typical mass - to - light ratios ) our @xmath10-band bulges are more massive . this simply confirms that dust is a bigger problem even in the @xmath19-band than missing stellar populations are in the @xmath10-band for this sample . in figure [ mtotalmass ] , we plot @xmath95 using points from @xcite and @xcite . these authors suggest that higher redshift galaxies sit on the local @xmath96 relation if you include total stellar mass instead of only bulge mass . in this way , they attribute evolution in the black hole scaling relation to the evolution of the fraction of stellar mass in the bulge . we include as a solid line the updated fit to the @xmath96 from @xcite , and plot our black hole masses in the same colored points as in figures [ msigma ] and [ mbulgelum ] . to determine if this sample of systems is in agreement with the @xmath95 relation , we perform a similar two - sided kolmogorov - smirnoff statistical test as above . we find that , compared to the local elliptical galaxies of @xcite , the p - value obtained is only 0.01 , less significant than above . when comparing to the higher - redshift samples of total stellar mass , we obtain a p - value of 0.0008 , and to the combined sample , a p - value of 0.004 . the balanced bootstrap method finds an offset from the mcconnell & ma relation of @xmath97 in plot units , similar to the @xmath94 relation . we thus conclude that our sample of galaxies on average lies above this scaling relation as well . we also note that the one point that falls below the relation , the brown circle mrk273 n ( labeled 6 ) , represents the mass in only one out of two black holes and is therefore artificially low . because this relation takes into account the entirety of stellar mass , the fact that many of our points fall above the relation suggests that star formation as a whole is not outpacing black hole growth during these mergers . relation . the grey points have luminosities and total stellar masses from @xcite and @xcite , who suggest that higher redshift galaxies sit on the local @xmath96 relation if you include total stellar mass instead of only bulge mass . in this way , they attribute evolution in the black hole scaling relation to the evolution of the bulge / disk fraction with redshift . we include also black points for local elliptical galaxies and the fit to for @xmath96 from ( * ? ? ? * log(@xmath98 log(@xmath99 ) . merging galaxies from this work are indicated by large colored points , with parameters listed in table [ tbl : mbh ] . symbols indicate which method was used for that mass determination : gas disk modeling in circles , stellar disk modeling in upward triangles , and jam modeling in downward triangles . stellar disk models represent lower limits to the black hole mass , while jam models represent upper limits ; see text for details . ngc 6240 s two nuclei were summed together and appear as one point ( 7 ; dark purple ) on this plot , paired with the entire system s total stellar mass . n s point ( 6 ; brown ) is the mass measurement for only one black hole , though there is a second in the system ( see * ? ? ? ] in the previous sections , we have compared our galaxies with the @xmath4 and @xmath96 relations from @xcite . we have also compared them to those from @xcite , and find that our offset remains . two galaxies appear to drive the offset in the @xmath4 relation : irasf01364 - 1042 and ngc 2623 . although we note that these are not outliers in a statistical sense , we must be certain that they are not unfairly biasing our results . because our sample for each scaling relation has only 8 points , throwing out these two galaxies has a significant effect in all cases through reducing the sample size . we note , however : when removing the two galaxies from the @xmath4 relation , the measured offset decreases quite a bit , while the uncertainty in the offset also increases ( producing only a 1.9-@xmath100 result ) . for the other two relations , the measured offset decreases only slightly , and the main effect is an increase in uncertainty ( so the significance still decreases from 3.3-@xmath100 to 2.3-@xmath100 ) . this demonstrates that the offsets are nt driven by these two galaxies in the @xmath94 and @xmath95 relations . due to the small sizes of our samples , the appropriate way to deal with potential outliers is to use the balanced bootstrap test , which we have done . the balanced bootstrap performs hundreds of resamplings in order to estimate the significance of a result , giving equal weight to all measurements . ( non-`balanced ' bootstrapping may resample some points more than others , which could cause a bias when one or more potential outliers exist . by balancing the bootstrapping , we force the equal treatment of all measurements ; this prevents a single point ( or two ) from biasing the results . ) by doing so , we are able to handle outliers ( or near - outliers ) without diminishing our sample size . the black holes in this sample of gas - rich major mergers consistently fall on or above the scaling relations for passive galaxies with stellar velocity dispersion , bulge luminosity and total stellar mass . although a number of our galaxies lie within the scatter of these scaling relations , our statistical tests show they are significantly more likely to fall on the upper side of the scatter than randomly distributed about the relations . though the offset from the @xmath4 relation is only a 2.6-sigma result , the agreement with the two other significant relations lends it confidence . taken together , these results show that increased black hole growth in these mergers has already begun and may be proceeding more rapidly than growth of the host galaxy . although we have a relatively small sample of black holes , it is striking that we already see a significant offset . as more black hole masses are measured in gas - rich mergers , the larger sample will provide a more accurate picture of how offset such systems are from black hole scaling relations . we caution the reader against considering the distance from a scaling relation as a direct proxy for merger stage . this is clearly evident when comparing where a system falls on one scaling relation versus another . for example , irasf17207 - 0014 falls closer to the @xmath4 relation than to the @xmath84 relation ; for ngc 6240s , the reverse is true . this discrepancy could be caused by at least two possible mechanisms : 1 ) the host galaxy parameters trace different aspects of galaxy growth , which may evolve differently . for example , the bulge velocity dispersion traces a convolution of mass growth with the virialization timescale whereas the bulge luminosity evolution may rely less stringently on the dynamical relaxation time . 2 ) as discussed above , the simulations of @xcite show that the measured stellar velocity dispersion of similar mergers oscillates during the course of the merger . analogous simulations would have to be performed for bulge luminosity and total stellar mass measurements to determine if they ought to oscillate as well . even if all three quantities were expected to oscillate during a merger , it seems contrived to require that they oscillate in lockstep . to determine precisely how galaxies evolve along a scaling relation ( and if they evolve the same way along each scaling relation ) , a larger sample of systems is required . by binning over merger stage , the range of distances from scaling relations can be seen , and key parameters affecting this can be identified . we leave this analysis for future studies . in this section , we discuss several plausibility arguments related to the above results . we consider the growth timescales and the available fuel for both black holes and the host galaxies . we posit that these systems followed black hole scaling relations before the merger began and that , since then , mass has been funneled to the center to enable black hole growth . for simplicity , we assume that each galaxy previously fell precisely on every scaling relation ; of course , as each relation contains intrinsic scatter , this is unlikely . however , we have no reason to believe that these systems were systematically offset from scaling relations before the mergers began , and thus this simplification should average out in the order - of - magnitude calculations of this and the following sections . with that assumption , we consider how long it would take for a black hole , growing at the eddington rate , to reach this far beyond its original mass . in table [ tbl : eddrate ] , we calculate this for our sample of galaxies . we see that the required times for these black holes to grow , while accreting at the eddington rate , are a few tens to hundreds of millions of years . as this is less than or comparable to a merger timescale , it is reasonable that these systems could have lain on scaling relations before the merger , and risen only after the merger began . as black hole growth is known to be episodic , we note that these systems could often have been accreting at sub - eddington rates for timescales less than those noted , and that the accretion times need not have been continuous . the merger of one of these objects , ngc 2623 , has been modeled by @xcite using h i kinematics from the very large array . their dynamical models indicate that this object is @xmath101 years past first pericenter passage , within the range of growth timescales we see for this object in table [ tbl : eddrate ] ( @xmath102 years ) . we also repeat that our black hole mass measurements produce dynamical masses of a central point source , which to our spatial resolution could be anything smaller than @xmath34pc . it is therefore plausible that some of the mass measured is in the accretion disk ( or , perhaps , in a nearby reservoir feeding the accretion disk ) . if that s the case , any mass that has not accreted directly onto the black hole would not cause the black hole to turn on as an agn . in such a scenario , the eddington rate may not be a relevant limiting factor in this growth phase . lcccccccc cgcg436 - 030 & 4.6e8 & 3.6e8 & 7.7e7 & 2.8e8 & 4.6e7 & 4.5e8 & 1.8e8 + irasf01364 - 1042 & 2.3e9 & 2.3e9 & 3.1e8 & 2.2e9 & 1.8e8 & 2.2e9 & 1.3e8 + iiizw035 & 2.6e8 & & & & & 2.1e8 & 8.5e7 + ngc 2623 & 3.0e8 & 2.9e8 & 2.3e8 & 1.9e8 & 5.3e7 & 2.3e8 & 7.6e7 + ugc5101 & 5.5e8 & & & & & 4.9e7 & 4.6e6 + mrk273 n & 1.0e9 & & & & & & + ngc 6240n & 8.8e8 & 7.8e8 & 1.1e8 & 7.1e8 & 8.2e7 & + & & & & & & 5.6e8 & 1.9e7 + ngc 6240s & 8.9e8 & 3.6e8 & 2.6e7 & 3.5e8 & 2.5e7 & + irasf17207 - 0014 & 2.6e9 & 2.1e9 & 8.8e7 & 2.5e9 & 1.9e8 & 2.1e9 & 8.8e7 + [ tbl : eddrate ] the gas - rich major mergers we ve studied here appear to have black holes significantly larger than would be expected based on their host galaxy properties . we assume that these systems , post - merger , ought to fall on scaling relations . as our systems are currently undergoing enhanced star formation , it is likely that the host galaxies will indeed continue to grow over the course of the merger , moving back towards the local scaling relations in figures [ msigma]-[mtotalmass ] . in table [ tbl : sfr ] , we show the relevant star formation rates from @xcite and calculate how long it would take at this star formation rate to bring the total stellar masses up to those predicted by the measured black hole masses . we see the typical timescale for these objects to reach predicted total stellar masses is 1 - 2 gyr at the present star formation rate . the dynamical models of @xcite find merger timescales for their sample of 4 systems to range from 140 myr - 1.4 gyr , depending on the initial conditions of the merger . apart from irasf10364 - 1042 , our galaxies are not far from this range . though star formation is a variable quantity , it is plausible for these galaxies to return to the @xmath95 relation by the time they have finished merging . as noted above , it is possible that not all the mass we measured has been accreted by the black hole yet . if this is the case , it allows for the possibility that some of this mass might be ejected via agn - driven winds and may indeed never be accreted ; we discuss this scenario in section [ outflows ] . if this were to happen , the stellar mass growth need not be so high to return these systems to the @xmath103 relation . lcccc cgcg436 - 030 & 1.5e11 & 85.87 & 1.7e9 & ... + irasf01364 - 1042 & 6.7e11 & 122.61 & 5.5e9 & ... + iiizw035 & 7.3e10 & ... & ... & ... + ngc 2623 & 7.9e10 & 69.19 & 1.1e9 & 1.66e9 + ugc5101 & 1.6e10 & 180.18 & 8.7e7 & 2.65e9 + mrk273 n & & & & + ngc 6240 n+s & 1.7e11 & 148.44 & 1.2e9 & 1.18e10 + irasf17207 - 0014 & 6.6e11 & 501.22 & 1.3e9 & 7.13e9 + [ tbl : sfr ] regardless of the rate of star formation discussed in section [ sfrsection ] , we might ask if there is enough molecular gas present in these galaxies to increase the @xmath104 sufficiently if all the gas were to be turned into stars eventually . we include the available mass in cold gas measured for each of these galaxies , when available , using h@xmath105 masses measured by @xcite . however , @xcite point out that the co - to - h@xmath28 conversion factor used is a primary source of uncertainty , and caution that their masses could be underestimated by up to a factor of 5 , and that agn contamination in ugc5101 could cause further underestimation . taking these uncertainties into account , the gas masses could be sufficient in ugc5101 and ngc 6240 , but perhaps not in ngc 2623 or irasf17207 - 0014 . we note that this estimate ignores atomic gas which may eventually cool and form molecular gas and then stars ; this quantity could be substantial , and that these systems could also continue to accrete gas from the circumgalactic medium . however , a number of ulirgs are also hosts to massive molecular outflows @xcite , which could instead further deplete the gas . thus , the low gas mass estimates may support the possibility that not all excess mass measured from the central regions will eventually be accreted to the central black hole . our technique produces a dynamical measurement of the mass produced by a point source in the centers of these galaxies . though the most straightforward interpretation of this is that the mass measured describes a black hole mass , it is possible that the mass distribution merely appears as a point source to our spatial resolutions . that is , anything significantly smaller than our pixel scales of @xmath34pc , and above the disk s radial density profile , would appear as a black hole to our measurements . as discussed in @xcite and @xcite , it is possible that some mass may be due to a nuclear star cluster . however , we rule out this interpretation for two reasons : 1 ) nuclear star clusters commonly have radii @xmath106pc @xcite , which would not appear as a point source to our dynamical modeling , and 2 ) the continuum luminosity present is not sufficient to account for a significant fraction of the central masses being due to stars . ( full spectral energy distribution fitting to the star clusters in and around the nuclear regions of these galaxies will be presented in a future paper . ) we can not , however , rule out the idea that gas present in the center could someday form a nuclear star cluster . galaxy merger simulations commonly assume bondi - hoyle accretion @xcite , a prescription in which effectively gas particles which enter within the sphere of influence of the black hole are automatically accreted so long as the eddington rate is not exceeded . however , gas inflow likely has angular momentum which must be discarded before accretion can occur , which can delay the accretion event by e.g. the viscous timescale @xcite . though we do nt have strong constraints on the properties of agn accretion disks , estimates predict that viscosity could slow down accretion by more than a hubble time @xcite . although the existence of agn confirms that accretion can occur on shorter timescales , the delay may not be negligible . this allows for the possibility that gas can build up in the core of a galaxy around the black hole without having yet been accreted . this accretion disk ( and , possibly , a reservoir feeding the accretion disk ) of gas could contain a substantial amount of mass en route to the black hole . this mass would all appear as a point source to our dynamical models . given the scaling arguments in the sections above , we also note the possibility that this putative accreting gas may never make it to the black hole . although the black hole dominates the gravitational pull on this gas , we find it plausible that when the black hole begins to accrete , its feedback may blow some of the rest of the gas away . thus , the extra mass that we see here could be the reservoir for massive outflows seen in a variety of local agn ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . at mass outflow rates of tens or hundreds of solar masses per year , the excess mass could easily be ejected in the time remaining in the merger . it is not currently known where the mass carried in these outflows is drawn from , or in particular , how much of it might come from within @xmath34 pc . simulations have also suggested that nuclear gas disks may leave behind remnant stellar disks on 1 - 10 parsec scales with masses 0.1 - 1 @xmath32 @xcite ; if this were to occur , the final black hole masses in our sample will be decreased by 10 - 50% . we note , however , that the black hole masses from which the scaling relations were calibrated did not have sufficient resolution to separate out remnant disks on 1 - 10 parsec scales either ; thus we only introduce a bias if our sample is more likely to host such disks than the typical galaxy . we note that such remnant disks were discovered in galaxy merger simulations but that they did not form until after the final coalescence of the two nuclei . thus it is not clear whether we would expect such disks ( broken from the larger nuclear disk on hundred parsec scales ) to be present in our sample . though the above possibilities may moderate the black hole masses measured by attributing some mass to other features , we still conclude that the black holes ( combined with their accretion disks ) in this sample of gas - rich mergers do not lie below black hole scaling relations , and in many cases are significantly more massive than their host galaxy properties would predict . this is directly opposite the predictions of quasar - mode feedback theories , in which black hole growth is delayed until the final stages of the galaxy merger when it cuts off star formation . as we note that ( u)lirgs are prototypically expected to evolve into quasars @xcite , we find this as further evidence of some delay in feedback timescale during a merger . such a delay ( due e.g. to a viscous accretion disk , as discussed above ) could reconcile the late - stage agn feedback paradigm with the early mass growth presented here . testing this scenario requires spatial resolutions higher by a factor of 10 - 20 , which may be feasible in the upcoming era of thirty - meter - class telescopes . these results differ from the conclusions drawn by @xcite , who show five merging galaxies falling below the @xmath4 and @xmath107 relations . they suggest that black hole growth might lag bulge formation in mergers , the opposite of the findings of this work . it is important to note the differing samples when considering these discrepant results . our sample of merging galaxies contains major gas - rich mergers ; the merging galaxies considered in @xcite are mainly minor mergers and/or gas - poor mergers . it is not unreasonable to imagine that such different types of mergers would have different effects on their central black holes . for example , a minor or gas - poor merger might use up all incoming new gas in star formation before it reaches the central black hole , causing a delay in black hole accretion until stellar feedback produces new gas to fuel it . a major gas - rich merger may have sufficient gas inflow to enable both star formation and black hole growth to begin relatively early . a sample of black holes in submillimeter galaxies appear to fall below the locally - defined @xmath96 relation by 1 - 2 orders of magnitude @xcite . these galaxies are thought to be the high - redshift ( z@xmath108 ) counterparts of local ulirgs and therefore might be reasonable to expect that they would behave in the same way as the sample presented here . however , the discrepancy may in fact be due to their usage of broad line velocity dispersions to infer @xmath32 ; indeed , when applying the same method in one local ulirg , @xcite find rough agreement with their submillimeter galaxies . it is also possible that differences between the samples such as eddington ratio , redshift , and merger stage might affect the black hole growth timescales . we have presented high spatial resolution near - infrared integral field spectroscopy of nuclear disks in nearby merging ( u)lirgs . these gas - rich mergers funnel gas to the galaxy centers where the gas forms disks ; stars then form _ in situ _ , creating a stellar disk as well . these disks , which were described in detail in @xcite , provide an alternative method to measuring black hole masses . three - integral orbital superposition models rely on an assumption of virialization and emission line black hole mass diagnostics require broad lines ; neither of these methods is feasible in such systems . however , with adaptive optics , we have produced two - dimensional kinematic maps at spatial resolutions close to or better than radius of the sphere of influence of the black hole . because the dynamical timescale this close to a black hole is short , the material quickly becomes virialized , and kinematic modeling is possible . we implemented two different techniques using these kinematic maps , originally presented in @xcite and @xcite , to measure 9 new black hole masses , producing a sample of black hole masses in 9 merging systems . by obtaining both gas and stellar kinematics for many of these systems , we are able to provide multiple independent measurements for a large fraction of our sample . gaseous disks are modeled as thin keplerian disks . stellar disks are more likely to have pressure support , but measuring this is not straightforward , so we place bracketing measurements using two assumptions . to provide a lower limit to the black hole mass , we assume the stars are also in a thin keplerian disk ; any pressure support would increase the black hole mass . we also use jeans axisymmetric mass models to take into account the velocity dispersion ; since this velocity dispersion may be inflated by intervening unvirialized material along the line of sight , this mass measurement is an upper limit . as the black hole masses measured here are in fact masses enclosed in @xmath34pc , some of the mass attributed to black holes here may be gas that has not yet accreted onto the black hole . although this mass is gravitationally bound to the black holes , we note that it is possible that future agn - driven outflows may carry a fraction of it away , thus diminishing the final black hole mass . though such outflows have now been seen in a number of ( u)lirgs , further studies will reveal how frequently they occur , how much mass they carry , and how much of the loaded mass comes from within @xmath34pc . we have placed these enclosed masses on three local black hole scaling relations , @xmath4 , @xmath84 , and @xmath95 , using host galaxy parameters from the literature . we find that though individual galaxies often fall within the scatter , as a population they are significantly offset towards higher black hole masses than isolated systems . several systems have central masses that sit considerably above scaling relations . under the assumption that these galaxies follow black hole scaling relations before and after their mergers , this suggests that black holes grow more quickly than their host galaxies during a major gas - rich merger . this does not line up with theoretical reasoning that star formation has easier access to inflowing gas ( e.g. * ? ? ? * ; * ? ? ? * ) or that a final burst of black hole growth is likely to shut off star formation , unless that feedback is delayed due to viscosity in the accretion disk @xcite or similar . we conclude that black hole fueling begins early in a gas - rich merger , and can initially outpace any simultaneous starbursting or bulge formation . these results have focused on a sample of gas - rich mergers in late stages of merging . to trace black hole growth through the entire merger sequence , a larger sample of black hole masses should be obtained for systems at a range of merger stages . such an observational sample could provide the basis for distinguishing between prescriptions for black hole accretion and feedback physics in galaxy merger simulations ( e.g. * ? ? ? a.m.m . acknowledges useful conversations with george privon , sebastian haan , tiantian yuan , stephanie juneau , kevin schawinski , david rosario , and simona gallerani . we enthusiastically thank the staff of the w. m. keck observatory and its ao team , for their dedication and hard work . data presented herein were obtained at the w. m. keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california , and the national aeronautics and space administration . the observatory and the keck ii laser guide star ao system were both made possible by the generous financial support of the w. m. keck foundation . the authors wish to extend special thanks to those of hawaiian ancestry on whose sacred mountain we are privileged to be guests . without their generous hospitality , the observations presented herein would not have been possible . this work was supported in part by the national science foundation science and technology center for adaptive optics , managed by the university of california at santa cruz under cooperative agreement ast 98 - 76783 . this material is based in part upon work supported by the national science foundation under award number ast-0908796 . _ facility : _ in this section , we show the best - fit dynamical mass models for each galaxy for each modeling technique in figures [ model1]-[model2 ] . as discussed in section [ results ] , each black hole was dynamically fit using a subset of the following methods : thin disk fitting to br@xmath27 , pa@xmath109 , or [ fe ii ] velocity map , thin disk fitting to h@xmath28 velocity map , thin disk fitting to stellar velocity map , jam modeling to stellar velocity and velocity dispersion maps . galaxies with multiple tracers show reasonable agreement in measured masses , which can be seen in table [ tbl : mbh ] and figures [ msigma][mtotalmass ] . residuals from each fit , divided by the measurement error , are included in these figures and demonstrate the goodness - of - fit of each model . the fitted disk parameters are also given in table [ tbl : disks ] . we note that many galaxies have complex kinematic structures that deviate from pure rotation in one or more of the tracers . these can be due to tidal streams or spiral arms along the line - of - sight , or to inflow or outflow when seen in gas tracers . where possible , we have masked these out to avoid biasing the disk models . masked spaxels appear white in the figures . a full analysis of the deviations from pure rotation and subsequent implications for inflow or outflow will be presented in a future paper . the positions of the black holes , plotted at ( 0,0 ) in each of the following figures , were determined primarily from the kinematic centers rather than the photometric centers . however , in the case of irasf17207 - 0014 , the kinematics are sufficiently complex that the two black holes were fixed to the photometric center of each disk measured with galfit in @xcite . velocities of the inner region of cgcg436 - 030 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . ] velocities of the inner region of irasf01364 - 1042 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . ] velocities of the inner region of irasf01364 - 1042 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . ] of iiizw035 . center : model @xmath110 of galaxy with black hole . right : residual map of model - data divided by errors in @xmath110 . all three panels are centered on the black hole and have been rotated so the major axis is horizontal . ] velocities of the inner region of iiizw035 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . ] velocities of the inner region of iiizw035 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . ] of mcg+08 - 11 - 002 . center : model @xmath110 of galaxy with black hole . right : residual map of model - data divided by errors in @xmath110 . all three panels are centered on the black hole and have been rotated so the major axis is horizontal . ] velocities of the inner region of mcg+08 - 11 - 002 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . ] velocities of the inner region of mcg+08 - 11 - 002 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . ] of ngc2623 . center : model @xmath110 of galaxy with black hole . right : residual map of model - data divided by errors in @xmath110 . all three panels are centered on the black hole and have been rotated so the major axis is horizontal . ] of ugc5101 . center : model @xmath110 of galaxy with black hole . right : residual map of model - data divided by errors in @xmath110 . all three panels are centered on the black hole and have been rotated so the major axis is horizontal . ] velocities of the inner region of iras f17207 - 0014 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . we have masked out regions in order to fit only the eastern black hole . ] velocities of the inner region of iras f17207 - 0014 . b ) map of model velocities of best fit black hole model . c ) residual map of best - fit model divided by errors in velocity . in panels a ) and b ) , velocity contours are marked in white . in each panel , a red x marks the position of the black hole . we have masked out regions in order to fit only the western black hole . ] one key feature of our thin disk black hole mass models is the inclusion of a radially - varying mass component to incorporate the mass profile of the nuclear disk . ( our jam models also incorporate galactic mass , but by fitting a mass - to - light ratio . ) we note that our fits use only a single power - law fit to the mass profile , so any pile - up of matter below our resolution limit will be considered part of the point mass . we include the mass profiles measured here for completeness , and for the purposes of calculating the implied radius of the sphere of influence of each black hole . recall that the mass profile is fit with the form @xmath111 . in table [ tbl : disks ] we report the normalization and power law fit for each of these parameters . for clarity , instead of reporting @xmath112 , we report the mass in the disk enclosed at 100 parsecs . a black hole s sphere of influence is defined as the region within which the black hole contributes 50% of the mass . we can therefore use our thin disk mass profiles to calculate the radius of the implied sphere of influence . this is interesting both for the purposes of understanding black hole dynamics and for confirming that we indeed have sufficient resolution to measure the black hole masses appropriately . as the fitted mass profiles of the disk and black hole masses are not independent , there is an issue of covariance to consider . that is , when a model fits a higher black hole mass , it will likely put less mass in the disk , and vice versa . when calculating errors on each of these parameters , we marginalize over the others : our black hole mass errors are calculated from the width of the black hole mass distribution in our monte carlo simulation , taking into account all possible fitted disk profiles . similarly , we report our errors on disk profile parameters using the widths of the distributions taking into account all fitted black hole masses . this is an appropriate way to handle covariance when drawing conclusions about the black hole masses , as is done in this paper . to account for the covariance issue in the measurement of @xmath113 , we calculate @xmath113 for every fit in the monte carlo simulation and draw our conclusions based on that . in one case ( irasf01364 - 1042 ) , the black hole mass is comparable or larger than the entire nuclear disk . the calculated radius of influence is thus artificially large , because presumably the mass profile of the galaxy changes beyond our field of view . the h@xmath28 fit to iiizw035 is also artificially large ; we attribute this to streaming motion in the h@xmath28 emission compromising the larger scale fit . this and other evidence of outflows will be presented in an upcoming paper . the affected radii are reported in parentheses in table [ tbl : disks ] . we see that , for our sample , we clearly resolve the spheres of influence of these black holes . we note that this method of calculating the size of the sphere of influence is uncommon , as a full mass profile is required to determine where @xmath114 . our capacity to resolve these scales exceeded our expectations because the black holes are more massive than predicted . lccccccc cgcg436 - 030 & br@xmath27 & @xmath36 & @xmath115 & @xmath116 & @xmath117 & @xmath118 + irasf01364 - 1042 & br@xmath27 & @xmath40 & @xmath119 & @xmath120 & ( @xmath121 ) & ( @xmath122 + & h@xmath28 & @xmath44 & @xmath123 & @xmath124 & ( @xmath125 ) & ( @xmath126 ) + iiizw035 & stars - disk & @xmath45&@xmath127 & 1.0 & @xmath128 & @xmath129 + & h@xmath28 & @xmath49 & @xmath130 & @xmath131 & ( @xmath132 ) & ( @xmath133 ) + [ 0.5ex ] & br@xmath27 & @xmath50 & @xmath134 & @xmath131 & @xmath135 & @xmath136 + [ 0.5ex ] mcg+08 - 11 - 002 & stars - disk & @xmath51 & @xmath137 & @xmath138 & @xmath139 & @xmath140 + & h@xmath28 & @xmath53 & @xmath141 & @xmath142 & @xmath143 & @xmath144 + & br@xmath27 & @xmath54 & @xmath145 & @xmath146 & @xmath147 & @xmath148 + [ 0.5ex ] ngc 2623 & stars - disk & @xmath55 & @xmath149 & @xmath138 & @xmath150 & @xmath151 + [ 0.5ex ] ugc5101 & stars - disk & @xmath60 & @xmath152 & @xmath153 & @xmath154 & @xmath155 + [ 0.5ex ] mrk273 n & [ fe ii ] & @xmath65 & @xmath156 & 2.0 & @xmath157 & @xmath158 + [ 0.5ex ] ngc 6240 n & stars - disk & @xmath68 & @xmath159 & @xmath160 & @xmath161 & @xmath162 + [ 0.5ex ] ngc 6240 s & stars - disk & @xmath72 & @xmath163 & @xmath164 & @xmath165 & @xmath166 + irasf17207 - 0014 e & [ fe ii ] & @xmath76 & @xmath167 & @xmath168 & @xmath169 & @xmath170 + irasf17207 - 0014 w & [ fe ii ] & @xmath80 & @xmath171 & @xmath153 & @xmath172 & @xmath173 + [ tbl : disks ] , j. , barczys , m. , krabbe , a. , et al . 2006 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 6269 , society of photo - optical instrumentation engineers ( spie ) conference series , p. l. , acton , d. s. , lai , o. , et al . 2000 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 4007 , society of photo - optical instrumentation engineers ( spie ) conference series , ed . l. wizinowich , 213 |
we introduce a new method to determine the relative contributions of different types of stars to the integrated light of nearby early - type galaxies . as is well known , the surface brightness of these galaxies shows pixel - to - pixel fluctuations due to poisson variations in the number of giant stars . differential spectroscopy of pixels as a function of fluctuation strength ( `` fluctuation spectroscopy '' ) effectively measures the spectral variation of stars as a function of their luminosity , information that is otherwise difficult to obtain for individual stars outside of the local group . we apply this technique to the elliptical galaxy , using hst / acs imaging in six narrow - band ramp filters tuned to spectral features in the range 0.8 m 1.0 m . pixels with% broad - band variations show differential color variations of 0.1% 1.0% in the narrow - band filters . these variations are primarily due to the systematic increase in tio absorption strength with increasing luminosity on the upper giant branch . the data are very well reproduced by the same conroy & van dokkum ( 2012 ) stellar population synthesis model that is the best fit to the integrated light , with residuals in the range 0.03% 0.09% . models with ages or metallicities that are significantly different from the integrated - light values do not yield good fits . we can also rule out several modifications to the underlying model , including the presence of a significant (% of the light ) population of late m giants . the current observations constitute a powerful test of the expected luminosities and temperatures of metal - rich giants in massive early - type galaxies . studies of pixels with much larger ( negative ) fluctuations will provide unique information on main sequence stars and the stellar initial mass function . | 1410.0007 | most of our knowledge of the stellar populations of galaxies comes from fitting stellar population synthesis models to their integrated light , as it is prohibitively difficult to characterize individual stars outside of the local volume . these models are commonly used to measure the total stellar masses of galaxies , as well as their ages , elemental abundances , star formation rates , dust content , stellar initial mass function , and other properties ( see the reviews by walcher et al . 2011 ; conroy 2013 , and references therein ) . many of the model ingredients carry significant uncertainties , which translate into uncertainties in derived stellar population parameters . the uncertainties in derived masses , star formation rates , and other parameters can be @xmath3dex or more ( see , e.g. , conroy , gunn , & white 2009 ; mitchell et al . 2013 ) . = 16.5 cm the predictions of stellar population synthesis models have historically been tested with star clusters , as stellar population parameters derived from the integrated light can be compared to those derived from the distribution of stars in the color - magnitude diagram ( cmd ) ( e.g. , worthey 1994 ; bruzual & charlot 2003 ; maraston 2005 ; schiavon 2007 ; conroy et al . 2009 ; thomas , johansson , & maraston 2011 ; nol et al . in fact , models are frequently tuned and calibrated to fit available data from open clusters in the milky way and the magellanic clouds , as well as the ancient , and generally metal - poor , globular clusters . thanks to programs such as the acs nearby galaxy survey treasury ( angst ; dalcanton et al . 2009 ) , some tests can now be done using individual stars in galaxies out to @xmath4mpc ( see , e.g. , girardi et al . 2010 ) . despite advances like these , the performance of the models remains essentially untested in important regions of parameter space . among the most important regimes is that of the metal - rich , old stellar populations which dominate the light of massive early - type galaxies . correctly modeling these stellar populations is critical for many aspects of astrophysics ; examples are the high mass end of the galaxy mass function ( croton et al . 2006 ) and the core mass function ( van dokkum et al . 2014 ) , the masses of central black holes ( van der marel et al . 1998 ) , the production of atomic elements in the universe ( worthey , faber , & gonzalez 1992 ; conroy , graves , & van dokkum 2014 ) , the stellar initial mass function ( van dokkum & conroy 2010 ; conroy & van dokkum 2012b ) , and the physical conditions in the densest star forming regions in the early universe ( krumholz 2011 ; nelson et al . 2014 ) . unfortunately , star clusters in the local group with similar ages and abundance patterns as giant ellipticals are rare ( e.g. , caldwell et al . 2011 ; van dokkum & conroy 2011 ) . we therefore largely rely on a combination of milky way stars and theoretical models to fit the spectra of these galaxies . of particular concern are the red giants that dominate the near - ir light . the spectra of these stars have strong and ubiquitous molecular bands of titanium - oxide ( tio ) and h@xmath5o , which often coincide with spectral features that are used to constrain abundances , ages , and the imf ( see , e.g. , lanon & wood 2000 ; rayner , cushing , & vacca 2009 ) . in this paper we introduce a new method to test the predictions of stellar population synthesis models for old , unresolved stellar populations . the method makes use of the fact that giants are sufficiently rare and luminous that poisson fluctuations in their number cause pixel - to - pixel fluctuations in the surface brightness distribution of nearby early - type galaxies . the existence of these surface brightness fluctuations is well known , and they have been used extensively to measure distances to galaxies : if a galaxy is closer , there are less giant stars in each pixel , and the pixel - to - pixel poisson fluctuations around the mean will therefore be larger ( e.g. , tonry & schneider 1988 ; tonry et al . 2001 ; mei et al . 2005 ; blakeslee et al . 2010 ) . pixels with high and low fluctuations allow us to do an experiment that gives unique information : we can study how the integrated spectrum changes when giants are added or subtracted to the galaxy . this is illustrated in fig . [ demo.fig ] , which shows the different stellar content of pixels with low and high fluctuations . the bright pixel at bottom left has many more luminous giants than the faint pixel at bottom right ; therefore , comparing spectra of pixels with different brightness enables us , in principle , to isolate spectral features of stars on the upper giant branch . furthermore , in the regime of very large fluctuations there will be pixels ( or more appropriately , lines of sight through the galaxy ) whose light has only a small contribution from luminous stars on the rgb and agb . spectroscopy of these pixels can provide information on the integrated light of stars that are still on the main sequence , which is difficult to access in any other way . this method builds on previous studies that compared the fluctuation amplitude in various broad - band filters to predictions from stellar population synthesis models ( e.g. , worthey 1994 ; vazdekis et al . 1997 ; blakeslee , vazdekis , & ajhar 2001 ; liu , graham , & charlot 2002 ; jensen et al . a central conceptual difference with previous work is that we do not use the absolute fluctuation signal in the analysis , which is difficult to measure ( as it requires an accurate distance , a thorough knowledge of the effects of the psf , and a careful treatment of contaminating objects ) . instead , the method relies on the relative change in the spectrum @xmath6 for a given relative change in the total flux @xmath7 . this is a key strength of the method : the distance , psf , and other effects that determine the absolute amplitude of the fluctuations are only relevant in that they set the dynamic range that is probed with the data . a second difference is that we identify the fluctuations of individual pixels rather than their statistical ensemble . here we take a first step toward fluctuation spectroscopy by using imaging through the narrow - band ramp filters in the advanced camera for surveys ( acs ) of hst . we study a regime where the observed fluctuations are dominated by systematic changes in the tio absorption strength with temperature on the upper giant branch . the target is , the most luminous galaxy in the virgo cluster ( with the possible exception of m87 depending on the chosen aperture ) . the main goal of the present paper is to introduce fluctuation spectroscopy as a method to dissect unresolved stellar populations . we also constrain certain aspects of the stellar synthesis model , and we show that more complete tests are possible with future data sets . the paper is organized as follows . in [ obs.sec ] the hst observations are described , with emphasis on the geometric constraints imposed by the ramp filters . in [ analysis.sec ] the variation in the ramp filters is measured as a function of the variation in the broad - band flux of a pixel . in [ model.sec ] we interpret the observed variation in the ramp filters , both in the context of a default isochrone and default population synthesis model and in the context of several variations of this model . the results are summarized in [ conclusions.sec ] . was observed for a total of 15 hst orbits in the period feb 22 march 1 , 2013 , in the context of program i d 12523 . we used the ramp filters in the acs camera to obtain images of the galaxy at six different wavelengths . the ramp filters transmit in a narrow optical bandpass , whose central wavelength can be tuned by rotating the filter . each filter consists of three segments , which are projected on different parts of the wfc detector array . the choice of segment ( and therefore the location of the image on the detector ) depends on the desired wavelength : within each segment only a limited range of wavelengths can be chosen . the six filters we used are listed in table [ obs.tab ] . their transmission curves , calculated with the synphot package , are shown in fig . [ filters.fig ] . the widths of the filters range from @xmath8 for the bluest filter to @xmath9 for the reddest , corresponding to a ( sparsely sampled ) spectral resolution ranging from @xmath10 to @xmath11 . the transmission curves are compared to a model spectrum of an old stellar population ( conroy & van dokkum 2012a ) , redshifted to 997and smoothed to a velocity dispersion of 300 . the central wavelengths were chosen to measure three redshifted spectral features that are important indicators of the imf : the nai @xmath12 doublet , the strongest calcium triplet line caii @xmath13 , and the feh @xmath14 wing - ford band ( e.g. , van dokkum & conroy 2011 ) . for each of these features we observed an `` on '' band and an `` off '' band , leading to six narrow - band filters . however , as we show later , the observed pixel - to - pixel variation in these bands in our data is actually _ not _ caused by variation in these absorption features , but by systematic changes in the tio absorption strength on the upper giant branch ( see [ model.sec ] ) . to avoid confusion we do not refer to these filters as , for example , `` nai on '' and `` nai off '' , even though that was how they were designed ; instead , we use names that simply refer to the central wavelength ( see table [ obs.tab ] ) . four of the wavelengths fall in the fr853n bandpass , which is part of the inner segment of its ramp filter unit . these images are therefore projected onto the iramp aperture , which is in the top left corner of the wfc array . the longest wavelengths fall in the fr1016n bandpass , which is part of the outer segment . the associated oramp aperture is in the bottom right corner of the array . .hst acs observations of ngc4472.[obs.tab ] [ cols="<,^,^,<,^,^",options="header " , ] @xmath15the aperture keyword specifies the pointing origin that is used by hst . depending on the wavelength , the inner ( `` iramp '' ) or outer segment ( `` oramp '' ) of the ramp filter unit has to be used . the galaxy is imaged on opposing corners of the wfc detector array for these two apertures . + @xmath16the name of the filter as it is used in the text . = 8.5 cm = 17.8 cm in addition to the narrow - band data we obtained deep broad - band images of . these broad - band images are critical as they have sufficiently high signal - to - noise ratio ( s / n ) to accurately measure the flux in each pixel , and therefore the contribution of giants to that pixel relative to its neighboring pixels . the narrow - band data do not have sufficiently high s / n for this measurement , but that is not needed : if it is known which pixels have high fluctuations and which pixels have low fluctuations , we can select the giant - rich and giant - poor pixels in the narrow - band data and average them to increase the s / n ratio . for this to work the broad - band data and narrow - band data have to be astrometrically - matched with an accuracy of @xmath17 pixel . in order to limit astrometric and resampling errors we observed twice in the broad - band f814w filter : once using the iramp pointing origin and once using the oramp pointing origin . this ensures that each narrow - band observation has a matching broad - band observation . pipeline - processed , drizzled data were obtained from the mast server . the charge transfer efficiency ( cte ) corrected _ drc files were used . cte effects are relatively small in our data as the background is relatively high ; nevertheless , the cte correction removes some faint streaks in the background and produces visibly cleaner images . the data were taken in nine visits : two visits for f995n and one visit for all other bands . we registered the f823n , f836n , f856n , f880n , and the iramp f814w images to a common reference frame . the shifts between these images are small ( @xmath18pixels ) , as they all use the same pointing origin . globular clusters and stars in the field were used to find the shifts ; no rotation or scale corrections were needed . we used the f814w image as a reference frame but first shifted it by @xmath19 , @xmath20 pixels . this non - integer shift ensures that the broad - band and narrow - band data are interpolated in the same way . we used a third - order polynomial for the interpolation to the reference grid . the f980n , f995n , and oramp f814w images were also aligned with each other , using the same procedure . for f995n the two visits were aligned and then added . the point spread function ( psf ) in the images needs to be exactly matched , as any differences will influence the measured signal in the ratio of two bands . we measured the width of the psf by fitting gaussians to point sources in the field in each band . it turns out that the width of the f814w psf is identical ( to within @xmath21 pixel ) to the f880n psf , but that the fwhm of the psfs in all other bands is narrower by @xmath22 pixels . it is not clear why the f880n psf behaves differently from the other ramp filters . we convolved the images to the f814w psf , and verified that the psfs were identical to @xmath21 pixel after the convolution . images of the eight observations ( the six ramp filters and the f814w images in each of the two apertures ) are shown in fig.[filter_im.fig ] . in order to determine the variation in a pixel we first need to know what the flux in that pixel would be in the absence of fluctuations . we used the deep f814w images to create models for the 2d light distribution of , as imaged at the iramp and oramp aperture locations . the models were made using the ellipse task in the iraf stsci package . all parameters were left free in the fit ; that is , the ellipticity , center , and boxiness / diskiness of the ellipses were all allowed to vary with radius . the fits were done in two steps . in the first step , only known bad and missing pixels were masked in the fit . this initial fit was subtracted from the galaxy , and strongly deviating pixels in the residual image ( @xmath23 or @xmath24 adus@xmath25 ) were added to the bad pixel mask . these deviating pixels are mostly at the locations of foreground stars and globular clusters , but also include some bad pixels that were not included in the data quality file . in the second step the fit was improved by adding these flagged pixels to the bad pixel mask . the surface brightness fluctuation in each pixel was determined in the following way . first , a conservative mask was created that includes all pixels that might be dominated or influenced by globular clusters or stars . in the inner regions of the galaxy these objects were identified by dividing the image by the smooth model , as the relative fluctuations in the center are typically @xmath26% ( see below ) . pixels in the central @xmath27 that deviate by more than 5% from the smooth model were identified as potential objects . at larger radii this is not a viable method as the surface brightness fluctuations approach these levels , and we identified contaminating objects by subtracting the model from the galaxy . away from the central regions , pixels deviating from the model by more than 0.5 counts / s were flagged as potential globular clusters . careful inspection of the residual images showed that the combined mask correctly identifies all `` visible '' globular clusters , stars , and background galaxies . the method identifies the central parts of such objects but not their faint wings , and the final mask was created by convolving the combined inner + outer mask with a @xmath28 pixel tophat . after creating the iramp and oramp masks the f814w images were divided by the smooth 2d models of the light distribution . pixels flagged in the object masks were set to zero . the resulting residual images contain the surface brightness fluctuations but also low level , large scale variation caused by subtle differences between the surface brightness distribution of the galaxy and the ellipse fit models . this large scale variation was removed using unsharp masking : the residual images were median filtered with a @xmath29 pixel mask , and divided by these heavily smoothed versions of themselves . the quantity `` surface brightness fluctuation '' , or `` sbf '' , in this paper refers to the fluxes in these residual images . the sbf of a pixel @xmath30 is therefore defined as sbf=@xmath31 , with @xmath32 the flux of a 2d model of ngc4472 that is constrained to fit the observed f814w surface brightness distribution of the galaxy exactly on @xmath33 scales . the final surface brightness fluctuation images in the iramp and oramp apertures are shown in fig . [ sbfdemo.fig ] . the images show clear variation , with some pixels being brighter than the mean and others fainter . the reality of the surface brightness fluctuations is demonstrated by their similarity in the top and bottom panels of fig . [ sbfdemo.fig ] . the iramp and oramp observations are entirely independent of each other : the two sets of exposures are independent and the galaxy is exposed on a different detector . nevertheless , the fluctuation pattern in the iramp and oramp data is identical , which means that the pattern is dominated by signal , rather than noise . the variation is highly correlated : clumps of pixels vary together . this in itself shows that the pattern is unlikely to be noise , as that is ( mostly ) uncorrelated in acs data . the correlations are caused by the hst psf , which smooths the intrinsically - uncorrelated variations . this is illustrated in fig . [ simpsf.fig ] , which shows an image with uncorrelated poisson noise along with the same image convolved with the acs f814w psf . the correlated noise pattern is very similar to that seen in fig . [ sbfdemo.fig ] . we find that the amplitude of the fluctuations is reduced by a factor of 4.2 ; that is , in actual acs images the surface brightness fluctuations are smaller by a factor of @xmath4 compared to an imaginary telescope which projects delta functions on a perfect detector with @xmath34 pixels . for reference , a gaussian with a fwhm of 2 pixels reduces the amplitude of random fluctuations by a factor of 3 . = 8.6 cm although the surface brightness fluctuations in the f814w images are dominated by signal , they are not noise - free . we quantify the per - pixel uncertainty empirically by shifting the iramp and oramp fluctuation images to a common reference frame and subtracting them . the comparison is done after binning the data by a factor of 2 in @xmath35 and a factor of 4 in @xmath36 , to limit the effects of sub - pixel errors in the shifts . figure [ traces.fig ] shows three randomly selected 1d traces extracted from the iramp and oramp f814w fluctuation images . the rms scatter in the difference is 0.003 , with no strong dependence on radius in the radial range of interest . correcting for the binning , and taking into account that the per - image uncertainty is smaller than the uncertainty in the difference of two images by a factor of @xmath37 , the per - pixel , per - image uncertainty is @xmath38 . = 8.6 cm we perform the analysis on relatively small sections of the iramp and oramp exposures , as the monochromatic patch is @xmath26arcmin when the ramp filter assembly is in place . the regions with uniform coverage are approximately @xmath39 and @xmath40 for the iramp and oramp respectively . the oramp monochromatic patch of the f814w surface brightness fluctuation image is shown in the left panel of fig . [ sbf_quant.fig ] . the `` mottled '' appearance is due to the surface brightness fluctuations . the amplitude of the fluctuations increases with distance from the galaxy center . the fluctuations are proportional to @xmath41 , with @xmath42 the average number of giant stars in a pixel , which means that the absolute amplitude of the fluctuations is proportional to the square root of the surface brightness . however , the _ relative _ fluctuations ( which we measure ) are proportional to @xmath43 , and therefore increase with decreasing surface brightness . the measured fluctuations in each of six annuli are shown in the right panel of fig . [ sbf_quant.fig ] . the rms increases with the distance to the center of the galaxy , from 0.006 in the inner annulus to 0.024 in the outer aperture . the orange curves in fig . [ sbf_quant.fig ] show the expected change with aperture under the assumption that the number of giants is proportional to the surface brightness of the galaxy . this simple expectation fits the observations well ( see tonry & schneider 1988 ) . the small asymmetry in the observed distributions may be partially caused by faint globular clusters and background galaxies . for the main analysis in the following sections we select all @xmath44 pixels with model surface brightness @xmath45magarcsec@xmath46 . in contrast to previous studies we are not concerned with an absolute calibration of the surface brightness fluctuation signal . the goal of this study is to measure the spectral variation as a function of the surface brightness fluctuation , that is , to determine by how much the spectrum of a pixel changes for a given change in its brightness . this relative nature of the method is an important aspect , as it makes the analysis independent of the distance of the galaxy , the effects of the psf , and other effects that determine the amplitude of the fluctuation signal . as discussed earlier , the only requirement is that the psf and astrometry of the narrow - band observations are matched to those of the broad - band observations . we measure the spectral variation by dividing adjacent narrow - band filters . four filter combinations are used : f823n / f836n , f836n / f856n , f856n / f880n , and f980n / f995n . for each of these combinations an `` index image '' was created by taking the ratio of the two adjacent narrow - band images . these index images were normalized by dividing by the median value of the index . in principle , the images now contain the relative index variation , with respect to a median index value that is identical to 1 , for each pixel ( in addition to noise ) . however , the median is not exactly 1 in the entire image , due to relative errors in flat fielding and a possible contribution from real large scale spatial variation in the index . as we are interested only in the relative fluctuations in the indices and not in their absolute values , we divided the index images by a `` correction flat '' that contains this large scale variation . this correction flat was created by filtering the index images with a median filter of @xmath29 pixels . the peak - to - peak variation in the correction flats is @xmath47% . the s / n ratio in the index images is much lower than in the f814w images , due to the much smaller width of the ramp filters . the typical per - pixel measurement uncertainty in the radial range of interest ranges from 0.025 0.06 , depending on the index . as we will show later , the indices need to be measured with an accuracy of @xmath48 to constrain model parameters two orders of magnitude better than the per - pixel uncertainty . fortunately , we can obtain the required s / n ratio by binning : as we have a precise measurement of the surface brightness fluctuation of each pixel from the f814w images , we can average all pixels in the index images that are known to have a particular fluctuation . this procedure is demonstrated for the f836n / f856n index in fig . [ sbf_raw.fig ] . the rms scatter in individual index measurements ( the @xmath49 width of the histograms ) is @xmath50 . nevertheless , in each surface brightness fluctuation bin the average index value can be determined with an uncertainty of only @xmath51 ( not including systematic errors ) , due to the large number of pixels that is averaged . we experimented with different averaging methods and chose the biweight ( beers , flynn , & gebhardt 1990 ) , as it is both robust and insensitive to outliers . using the median instead leads to nearly indistinguishable results . we also explored whether the results depend on the radius from the center of the galaxy , the flat fielding , sky subtraction , and errors in the shifts . the systematic error due to these effects is estimated at 0.0005 , that is , 0.05% . this uncertainty was added in quadrature to the poisson error of each index measurement . = 8.5 cm = 15.5 cm the relations between the averaged measured f823n / f836n , f836n / f856n , f856n / f880n , and f980n / f995n indices and the surface brightness fluctuation are shown in fig . [ fluct_meas.fig ] . there are clear relations , but the variation is small : over the full range of @xmath52 fluctuations the indices vary by maximally @xmath53 for f823n / f836n and f980n / f995n , @xmath54 for f836n / f856n , and @xmath55 for f856n / f880n . we note that the relation between the index strength and the surface brightness fluctuation is independent of the spatial resolution of the observations , as long as the quantity on the vertical axis is derived from data with the exact same spatial resolution as the quantity on the horizontal axis . the spatial resolution determines the dynamic range of the fluctuations that are probed at fixed surface brightness , not the form of the correlations between narrow - band indices and the surface brightness fluctuations .. ] we conclude from fig . [ fluct_meas.fig ] that pixels with high fluctuations have systematically different narrow - band indices than pixels with low fluctuations . in other words , sightlines through with a relatively high number of giant stars produce a different spectrum than sightlines with a relatively low number of giant stars . the fact that two indices show a positive correlation and one shows a negative correlation implies that we are seeing variation in spectral _ features _ , not an overall tilt in the spectrum . in the next section we interpret the correlations in fig . [ fluct_meas.fig ] in the context of expectations from the stellar spectra that go into the conroy & van dokkum ( 2012a ) stellar population synthesis models . the observed narrow - band index variations are modeled by mimicking the observations as closely as possible . we simulate the stellar population in a pixel by poisson sampling stars from a stellar library according to an isochrone , and then calculate the total luminosity and the integrated spectrum for each of these pixels . next , we determine the relations between the narrow - band indices and the surface brightness fluctuations in the same way as was done for the data , taking the redshift and velocity dispersion of the galaxy , the filter curves , and the uncertainties in the data into account . we use the same empirical stellar spectra that conroy & van dokkum ( 2012a ) use in their synthesis of old stellar populations . the number of templates depends on the model age ; in our default 13.5gyr model 106 templates are used . they are actually 70 distinct stellar spectra , with some of them doing `` double duty '' in different evolutionary phases that bring them to the same region of the hr diagram . they are carefully vetted solar metallicity stars from our own galaxy ; their near - ir spectra come from the irtf spectral library of cool stars ( rayner et al . details on their selection and the derivation of their bolometric luminosities and other parameters are given in conroy & van dokkum ( 2012a ) . the spectra are shown in fig . [ allstars.fig ] . in luminous stars the caii triplet lines are strong , whereas nai and feh ( wing & ford 1969 ) are weak . low luminosity dwarfs have weak caii and prominent nai and feh absorption . these characteristics are behind the widespread application of these three spectral features to constrain the stellar mass function from the integrated spectra of old stellar populations ( see , e.g. , conroy & van dokkum 2012b , and references therein ) . as discussed in [ obs.sec ] the wavelengths of the acs ramp filters were chosen to measure the strengths of these ( redshifted ) features.m 1.0@xmath0 m wavelength range . ] in the most luminous and the least luminous stars the spectra show broad bumps and troughs characteristic of molecular transitions . the most prominent of these are h@xmath5o and , particularly , tio , which has many bands in this wavelength range ( valenti , piskunov , & johns - krull 1998 ) . two bandheads are marked in fig . [ allstars.fig ] . the strength of the h@xmath5o and tio bands depends on temperature and surface gravity , and they are strongest in late m giants at the tip of the giant branch . the empirical templates were selected to have solar abundances , but as is well known elliptical galaxies have non - solar abundance ratios . in order to explore the effects of metallicity , and to create a default model that resembles as closely as possible , we have constructed two stellar libraries with alternative elemental abundances . each individual empirical stellar template was multiplied by the ratio of two theoretical spectra : one with the adjusted abundance , and one with solar abundance . in this way , the theoretical response to an abundance change is `` grafted '' onto the empirical spectra . this procedure is similar to that followed in conroy & van dokkum ( 2012a ) ; the difference is that in conroy & van dokkum ( 2012a ) we adjusted the integrated spectra and here we apply the corrections to individual stars . the first alternative library comprises `` standard '' @xmath56-enhanced spectra , with [ fe / h]=0 and @xmath57fe]=0.3.=[mg / fe]=[si / fe]=[ca / fe]=[ti / fe]=0.3 . ] the second , and default , library is tuned to match the individual elemental abundances of , as derived from integrated - light long - slit spectroscopy ( conroy & van dokkum 2012a and c. conroy & p.van dokkum , in preparation ) . has [ fe / h]=0.05 , [ mg / h]=0.3 , [ o / h]=0.3 , [ ti / h]=0.13 , [ na / h]=0.6 , [ ca / h]=0.05 , [ c / h]=0.1 , and [ n / h]=0.35 . in fig . [ zeffect.fig ] we show the differences between the spectra of two stars in the three libraries . for sun - like stars ( top panel ) the differences are subtle , but for luminous late type giants ( bottom panel ) they are substantial , with the spectrum that has the ngc4472 abundance pattern falling in between the ( original ) solar - metallicity spectrum and the @xmath58enhanced spectrum . = 8.5 cm each star in the library is assigned a weight , which is determined by a combination of a stellar evolution model ( as described in conroy & van dokkum 2012a ) and the imf . as we did for the metallicity , we base our default model on the best fit to the integrated - light spectrum of ( see conroy & van dokkum 2012a ) . the default weights are therefore calculated for a 13.5 gyr old stellar population with a salpeter ( 1955 ) imf . the weight is defined such that it is proportional to the average number of stars in a pixel . the relation between the weight and the bolometric luminosity is shown in fig . [ starweight.fig ] . the most luminous giants are rare ; only 1 in every 100,000 stars has @xmath59 . = 8.5 cm in practice , we simulate @xmath60 `` pixels '' with a per - pixel mass of @xmath61 . for a salpeter imf and an age of 13.5gyr , the number of stars in each pixel @xmath62 is related to the mass as @xmath63 . we assign @xmath62 to each pixel . each star in each pixel is labeled with a number from 1 to @xmath64 , with @xmath64 the number of templates in the library ( @xmath65 for the default library ) . this labeling is based on random numbers , with the probability of assigning a particular template proportional to its weight . the total bolometric luminosity of a pixel is calculated by summing the bolometric luminosities of its stars . we also sum the stellar luminosities integrated over the wavelength range 8150 8400 , as a proxy for the total f814w luminosity . the integrated spectrum of each pixel is constructed by summing the spectra of its stars . finally , the f814w luminosities were perturbed with a gaussian of width @xmath66 , to account for measurement uncertainties in the surface brightness fluctuation measurements ( see [ sbfmeas.sec ] ) . we first compare the distribution of f814w surface brightness fluctuations in our default model to the data . in this comparison , the number of stars per pixel is an adjustable parameter , although we note that our results are not very sensitive to the precise number of stars that is adopted . figure [ comp_sbf.fig ] compares the observed surface brightness fluctuations to those in our default model , for @xmath67 , @xmath68 , and @xmath69 . for this choice of @xmath62 there is a good match between the distributions , with an rms in both the data and the observations of 0.015 . the distributions have a slightly different shape because we did not take the surface brightness profile of into account in the modeling . = 8.5 cm = 16.5 cm the amplitude of the surface brightness fluctuations is particularly sensitive to the number of giants . as a result , for fixed @xmath62 , the model amplitude is a strong function of the imf , the age , and other parameters . in later sections we investigate the effects of changing the model parameters , and we always adjust the number of stars per pixel in order to match the observed rms of the f814w surface brightness fluctuations . in this context it should be noted that the parameter @xmath62 is not the true number of stars in a projected @xmath70 aperture , but the `` effective '' number of stars per pixel after smoothing by the acs psf ( roughly corresponding to uncorrelated areas of @xmath71 ) . we stress that , although we tune the models to reproduce the observed rms , this is actually not necessary . as explained in the introduction , our method is only sensitive to the _ relative _ change in the spectrum for a given _ relative _ change in the brightness of a pixel . specifically , a model with less stars per pixel , and therefore a larger absolute fluctuation signal , produces ( to first order ) the same relations between spectral variation and flux variation as our default model . the only reason why we adopt a number of stars in the model that produces the correct fluctuation rms is a subtle second - order effect : the smoothing of the modeled f814w luminosities with the observational error ( see 4.1 ) can only be done correctly if the absolute rms is matched reasonably well . although it is not important for our analysis , we can ask whether the observed rms of the fluctuations is similar to the expected rms based on the distance and surface brightness of and the acs psf . phrased differently , we can ask whether the stellar mass in our model is consistent with the observed surface brightness of . the median surface brightness of in the analysis region is @xmath72magarcsec@xmath46 . for a distance modulus of @xmath73 ( mei et al . 2007 ) this corresponds to @xmath74l@xmath75 per @xmath76 resolution element . the @xmath77 ratio of , as measured from our keck lris spectrum ( conroy & van dokkum 2012b , and in preparation ) , is @xmath78 . the expected stellar mass is therefore @xmath79 . the stellar mass in our model is a factor 1.5 higher , but the difference is barely significant : based on the surface brightness of we would have predicted an rms of @xmath80 , and we measure an rms of 0.015 . we now turn to the spectral variation as a function of the surface brightness fluctuation . the left panel in fig . [ fluctspec1.fig ] shows the average spectra of simulated pixels with fluctuations of @xmath81% and @xmath82% . the spectra are offset by @xmath83% , as expected , but are otherwise very similar : they are both dominated by the well - known spectral features of old stellar populations ( see , e.g. , the top panel of fig . [ filters.fig ] ) . the right panel shows the ratio of the two spectra , the `` fluctuation spectrum '' . the fluctuation spectrum is very different from the integrated - light spectra . there are no obvious features at the locations of the nai @xmath12 doublet , the ca ii triplet , or the feh @xmath14 wing - ford band . instead , the fluctuation spectrum is characterized by sharp bandheads at @xmath84 and @xmath85 , as well as a broad feature at @xmath86 . we identify these features as tio bandheads and h@xmath5o absorption , respectively . this is demonstrated by the thin red line , which is an approximation of a `` pure '' tio spectrum . this tio spectrum was generated by turning tio on and off in the modeling of the atmosphere of a single star with @xmath87k and @xmath88 and then taking the ratio of the resulting spectra . as can be seen in fig . [ allstars.fig ] the tio and h@xmath5o bands are strong in the most luminous giants , which have @xmath89l@xmath75 and @xmath90k 3600k . in the regime of small fluctuations , a pixel s surface brightness fluctuation is essentially determined by the number of these cool , luminous giants it contains . in our default model with @xmath68 stars per pixel , the average number of stars with @xmath89l@xmath75 per ( psf - convolved ) pixel ranges from 91 for pixels with sbf=0.95 to 150 for pixels with sbf=1.05 . we now compare the predicted fluctuation spectra to the observed relations between the narrow - band indices and the f814w surface brightness fluctuations . we averaged the model spectra in each of 11 surface brightness fluctuation bins , from 0.95 to 1.05 with steps of 0.01 . the spectra were redshifted to the distance of ( 997 ) and smoothed to a resolution of @xmath91 , which is the approximate velocity dispersion of . then , the spectra were divided by the spectrum of the central surface brightness fluctuation bin , with sbf=1 . this gives 11 differential fluctuation spectra , appropriate for fluctuations ranging from @xmath92 to @xmath93 . the central fluctuation spectrum is unity at all wavelengths , by construction . in fig . [ smooth.fig ] we show the ratio of the two extreme spectra , those with sbf=1.05 and sbf=0.95 . this spectrum can be compared to the unredshifted , unsmoothed version shown in fig . [ fluctspec1.fig ] . the six ramp filters are overplotted , after normalizing the filter curves by their peaks . the four blue filters sample wavelengths where there should be significant variation due to tio , with the two reddest filters effectively acting as a control . the 11 fluctuation spectra were integrated using these synthetic ramp filter curves . finally , the ratios between adjacent ramp filter fluxes were determined , creating synthetic indices that can be compared to the data . = 8.5 cm in fig . [ moddata1.fig ] the model predictions are compared to the observed fluctuations in the ramp filters . the data points are the same as those in fig . [ fluct_meas.fig ] ; the only difference is that the vertical axis spans the same range in all panels of fig . [ moddata1.fig ] . the lines are predictions from our default model , that is , a model that is based on the stellar library and stellar evolution tracks discussed in conroy & van dokkum ( 2012a ) , with an age of 13.5gyr , the abundance pattern as measured from its integrated - light spectrum , a salpeter ( 1955 ) imf , and a stellar mass of @xmath94per pixel . the models provide a remarkably good description of the data , even though they are determined in a completely independent way . both the sign of the correlations and their amplitude are a close match . the bottom panels of fig . [ moddata1.fig ] show the residuals , obtained by subtracting the model predictions from the data . the @xmath49 scatter in the residuals is @xmath95 for three of the indices . it is slightly higher at @xmath96 for the reddest index , and there is some evidence for a systematic trend in the residuals for this index . taking all four indices together , the reduced @xmath97 is 0.6 , which suggests that the systematic error ( 0.0005 ) is slightly overestimated . for a systematic error of 0.00025 the reduced @xmath97 is close to 1 . the excellent correspondence between the observed relations between the ramp filter indices and the f814w surface brightness fluctuations has several implications : it means that we understand the origin of the surface brightness fluctuations in this regime , and also that our stellar population synthesis of the upper rgb and the agb is correct to within the uncertainties of the method . in the previous section we have shown that our default model provides an excellent fit to the observed fluctuations . here we study the effects of changing various aspects of the model . we consider three distinct types of variation : changes to the underlying model ( specifically , the relative weights of stars in the stellar library ) ( [ library.sec ] ) ; varying the age , metallicity , and the imf ( [ age.sec ] ) ; and the regime of very large surface brightness fluctuations ( [ large.sec ] ) . we investigate four variations of the default underlying model . each model in this section has an age of 13.5gyr , the elemental abundance ratios , and a salpeter imf . the variations either change the spectra in the library of 106 templates or change the weights of these templates . changing the weights effectively means changing the time that stars spend in that particular phase of evolution . the first two of the variations are dramatic ( and effectively ruled out by many other observations of stellar populations ) , and are intended to illustrate the qualitative behavior of the fluctuation spectrum . the last two are more subtle ( and more reasonable ) . the first model variation we consider is to ignore the known differences between the spectra of dwarfs and giants and assume that the spectrum of a star depends only on its temperature . we replace the spectrum of every giant in the library by that of the dwarf star whose temperature is closest to that of the giant . as the dwarfs cover a broader range of temperatures than the giants ( see the inset of fig . [ starweight.fig ] ) , every giant in the library has a close temperature match on the main sequence . we scale each of the dwarf spectra in such a way that the average brightness in the @xmath98 m @xmath99 m wavelength range is the same as that of the giant it replaces . the resulting fluctuation spectrum , again obtained by dividing the averaged spectrum of + 5% f814w fluctuations by the averaged spectrum of @xmath82% fluctuations , is shown in the left panel of fig . [ fluctspec2.fig ] . the fluctuation spectrum is very different from that of the default model ( shown in orange ) . cool dwarfs generally have weaker molecular features than cool giants , and we therefore do not see the strong tio and h@xmath5o bands . instead , we see that high fluctuations have enhanced nai and feh absorption and suppressed caii triplet lines , as expected from the spectra of cool dwarfs . the fluctuation spectra for all 11 surface brightness fluctuation bins were redshifted and smoothed by the velocity dispersion of , just as was done for the default model . we calculated synthetic fluxes in the ramp filters and constructed model indices . in the top panels of fig . [ moddata2.fig ] we compared the predictions from this model to the data , and to the default model . the model can clearly be ruled out by the data : the spectra of giants are different from those of dwarfs of the same temperature . with this modification we highlight the fact that the most luminous giants , with @xmath89l@xmath75 , are responsible for the trends in fig . [ fluct_meas.fig ] . we switched the spectra of stars with @xmath89l@xmath75 with those of stars with @xmath100l@xmath101l@xmath75 . no changes were made to stars with @xmath102l@xmath75 . the rank order within each set of giants was conserved ; that is , the brightest star with @xmath89l@xmath75 was switched with the brightest star with @xmath103l@xmath75 . as before , we scaled each stellar spectrum so it has the same brightness in the @xmath98 m @xmath99 m range as the star it replaces . the fluctuation spectrum is shown in the second panel of fig . [ fluctspec2.fig ] . again , the broad molecular features have disappeared , as they now artificially occur in stars that are much more common and therefore do not contribute significantly to the surface brightness fluctuation signal . we now see that the caii triplet lines are stronger in pixels with a higher fluctuation . the reason for this behavior is that these lines are very strong in stars with @xmath100l@xmath101l@xmath75 ( see fig . [ allstars.fig ] ) . we will return to this in [ large.sec ] , where we discuss the behavior of the default model in the regime of large , negative fluctuations ( that is , sbf@xmath104 ) . in the second row of fig . [ moddata2.fig ] the observed ramp filter indices are compared to the predictions from this model . again , the model is a poor fit : the predicted trends are too weak , due to the absence of strong molecular bands in the fluctuation spectra . as noted earlier , stars with @xmath105l@xmath75 are rare but contribute significantly to the integrated light . their treatment is an important source of uncertainty in stellar population synthesis models , and our results provide strong confirmation of the quality of their implementation . next , we ask what the effect is of changing the contribution of stars with @xmath89l@xmath75 to the total luminosity . this contribution depends on the product of the time spent in their particular phase of evolution and their luminosity . the luminosities are generally better constrained than their timescales , and so we vary the weights of the stars ( i.e. , how many there are in a pixel ) rather than their individual luminosities . = 8.5 cm the third panel of fig . [ fluctspec2.fig ] shows the effect of increasing ( black ) and decreasing ( grey ) the number of luminous giants by a factor of 2 . there is an effect on the fluctuation spectrum , with the molecular features becoming stronger for an increasing number of luminous giants , but the effects are small . the reason for this modest change is that the spectral change is , in essence , a second - order effect . to first order , the variation in both the integrated luminosity of pixels and in the normalization of their spectra increases when the number of giants increases the variation in the spectra at fixed luminosity variation is a more subtle effect . this is also evident in the third row of fig . [ moddata2.fig ] , where the data are compared to the model predictions . the scatter in the residuals is @xmath106 , and similar to the scatter in the residuals from the default model . the reduced @xmath97 is 0.6 , whereas it is 10.7 and 9.4 for the models considered in 5.1.1 and 5.1.2 respectively . to constrain the absolute number of giants an accurate calibration of the absolute surface brightness fluctuation signal is required . finally , we consider the possibility that massive elliptical galaxies have a population of very cool giants that is not included in standard stellar population synthesis models at old ages . the strong tio bands in these stars overlap with several imf - sensitive features , which is why an arbitrary contribution of m giants is included as one of the nuisance parameters in the conroy & van dokkum ( 2012b ) model . the spectrum of such a cool giant is shown in fig . [ coolgiant.fig ] . this m7iii star , hd108849 , has a luminosity @xmath107l@xmath75 and an effective temperature @xmath108k . it has much stronger molecular bands than the average spectrum of @xmath105l@xmath75 stars in the default library ( which have spectral types m2 m6 ) . in the fourth panel of fig . [ fluctspec2.fig ] we show the fluctuation spectrum for a model where we add a population of these cool giants to the light . the weight of this new template is determined by setting the contribution of these stars to the total integrated light in the @xmath98 m @xmath99 m range to 5% , 10% , and 15% ( black , dark grey , and light grey in fig . [ fluctspec2.fig ] ) . as might have been expected the fluctuation spectrum shows stronger molecular features than the default model . as shown in the bottom row of fig . [ moddata2.fig ] , even the model with the smallest contribution from m7 giants is clearly a worse fit to the data than the default model . the scatter in the residuals from the fit is 0.0014 for the f856n / f880n index , compared to 0.0004 in the default model . the reduced @xmath97 is 1.5 for the 5% model ( with 43 degrees of freedom ) , 3.4 for the 10% model , and 6.0 for the 15% model . the 5% model can be rejected with 98% confidence , and the other models with @xmath109% confidence . from the four modifications that we considered we can infer that the data are sensitive to the spectra of the most luminous giants ( those with @xmath110l@xmath75 ) . this conclusion applies to the regime of small fluctuations ( with an rms of 0.01 0.02 ) , and the wavelength range 0.8@xmath0 m 1.0@xmath0 m . the data are remarkably _ in_sensitive to the _ number _ of these stars , as shown in the third row of fig.[moddata2.fig ] . as explained in [ pred.sec ] this is because varying the number of these stars `` stretches '' the relations between narrow - band indices and the surface brightness fluctuations , but does not change their slope . of the four modifications , the first two ( 5.1.1 and 5.1.2 ) mostly serve to illustrate the fact that the fluctuations are only sensitive to the most luminous giants . the latter modifications ( 5.1.3 and 5.1.4 ) are non - trivial and `` realistic '' , as the lifetimes and temperatures of luminous agb stars are not well - known from first principles . perhaps the most directly applicable result is that we can exclude a significant contribution from giants that are cooler than the lowest temperature ones in our default @xmath111gyr library . the formal upper limit on their contribution is 3% ( with 95% confidence ) , although we note that this ignores unknown systematic effects as well as possible degeneracies with other parameters . this constraint can be used in stellar population synthesis models to limit the allowed parameter space ( see , e.g. , supp . fig . 2 in van dokkum & conroy 2010 and the discussion in conroy & van dokkum 2012b ) . in this section we fix the underlying components of the model and consider variation in the stellar population parameters age , metallicity , and imf . in our default model these parameters are taken from the stellar population synthesis model that provides the best fit to the high quality keck integrated - light spectrum of ( conroy & van dokkum 2012a ; c. conroy & p. van dokkum , in preparation ) . this fit has an age of 13.5gyr , an imf that is close to salpeter , and the elemental abundance pattern that is described in [ ingredients.sec ] . as we showed in the previous section , the ramp filter data samples molecular bands in stars on the upper giant branch ; even though we use the same underlying model as conroy & van dokkum ( 2012a ) it is by no means guaranteed that the best - fitting model to the integrated light spectrum also provides the best fit to the surface brightness fluctuation signal . in addition to the default model we consider two younger ages ( 9gyr and 5gyr ) , two alternative abundance patters ( solar and the `` standard '' @xmath58enhanced model described in 4.1 ) , and two alternative imfs ( a very bottom - heavy imf with @xmath112 , and a bottom - light van dokkum ( 2008 ) imf with @xmath113 ) . the predicted fluctuation spectra are shown in fig . [ fluctspec3.fig ] , and the predicted ramp indices are compared to the observations in fig . [ moddata3.fig ] . we find that the indices are quite sensitive to age . the 9gyr model can formally be excluded at @xmath109% confidence . the reason for this sensitivity is that younger ages have a greater contribution of late m giants . in particular , the number of m6 giants is a factor of @xmath114 higher in the 9gyr model than in the 13.5gyr model . as these late type giants whose spectra have very strong molecular absorption bands dominate the surface brightness signal , the change in their contribution at old ages produces a readily detectable signature . the age constraint from the surface brightness fluctuations is fully consistent with the very old age derived from ngc4472 s integrated - light spectrum , which is mostly sensitive to the properties of stars near the main sequence turn - off . a caveat here is that the measurements are not truly independent , for two reasons : we are using the same model ingredients for both constraints , and late m giants would not contribute solely to the surface brightness signal but also to the integrated light . the indices are also sensitive to metal abundance , but not in a straightforward way . the standard @xmath58enhanced model predicts steeper relations than observed , and can be ruled out with high confidence.k ) . ] however , the predicted trends for solar metallicity are very similar to those of the default model ( and a good fit to the data ) , despite the fact that the default model has mg and o abundances that are much closer to the @xmath58enhanced model than to the solar model . as discussed in [ ingredients.sec ] , the only features that really matter are the tio bands , and the ti abundance of is relatively low .. if lines from stars of different temperature are on the same part of the curve of growth , abundance effects can cancel . ] = 16.5 cm the data provide no direct constraint on the imf . the amplitude of the broad - band fluctuations is sensitive to both the number of stars and the imf , but we adjust the number of stars so that all models match the observed rms of the broad - band fluctuations . stars per resolution element , the @xmath112 model has @xmath115 stars , and the bottom - light model has @xmath116 stars . ] in the regime of small fluctuations the trends in fig . [ moddata3.fig ] depend on the luminosity temperature relation on the upper giant branch , and are independent of the number of low mass stars . although there is no direct imf dependence , the results presented in this paper are relevant to measurements of the imf from integated light , as these rely on accurate modeling of the contribution of luminous giants to the spectra of early - type galaxies . in this paper we focused on a regime of relatively small surface brightness fluctuations , of @xmath1% . this range is dictated by the currently - available data : the poisson errors quickly increase for fluctuations @xmath117 and @xmath118 , as we simply do not have many pixels with variations that large . as a result of the small observed fluctuation range there are , on average , @xmath119 stars with @xmath89l@xmath75 even in the lowest fluctuation pixels . the fluctuation spectra are therefore completely dominated by the tio and h@xmath5o features that are characteristic of these luminous red stars . here we briefly consider what we can expect in a regime where there are far fewer stars per pixel , so that the surface brightness flucations are much larger . there is a wealth of information in the fluctuation spectra in this regime , as there are sightlines that are free of luminous giants . those pixels , with surface brightness fluctuations much smaller than 1 , provide a window of the lower giant branch and the main sequence that is unattainable in integrated light observations . we simulated fluctuation spectra in this regime , using our default model with @xmath120 per pixel rather than @xmath121 . in this model there are pixels that contain no stars with @xmath122l@xmath75 . the distribution of fluctuations in this model is shown in the top left panel of fig . [ largefluc.fig ] . the @xmath49 spread is @xmath123 but the distribution is asymmetric , with the high fluctuation tail containing multiple luminous giants . in the top right panel of fig . [ largefluc.fig ] we show fluctuation spectra in five regimes , going from surface brightness fluctuations much lower than 1 to fluctuations much greater than 1 . the fluctuation spectra are dramatically different in these regimes , as they probe different types of stars . low fluctuations ( 0.7/0.6 ) probe the variation of subgiants and low luminosity giants with temperature , and as these stars have no strong tio features the fluctuation spectra are dominated by nai and caii . intermediate fluctuations ( 1.1/1.0 ) are predicted to show very strong tio features in the fluctuation spectra , as the amplitude of their broad - band fluctuation depends on whether they have zero or one star with @xmath89l@xmath75 . the regime of strongly positive fluctuations ( 1.5/1.4 ) resembles the regime probed in our observations of , that is , the variation of the spectra of very luminous giants with temperature . the bottom panels of fig . [ largefluc.fig ] show the predicted relations between ramp filter indices and the f814w surface brightness fluctuation . the relations are not linear , and they depend on the imf . bottom - heavy imfs with @xmath112 show characteristic asymmetric trends , with little dependence on the surface brightness amplitude for sbf@xmath1241 and steep relations for sbf@xmath1251 . note that the horizontal and vertical ranges of these panels are much larger than in the previous sections . in this paper we introduced a new technique for studying old , `` semi - resolved '' stellar populations . we have shown that it is possible to identify individual pixels , or sightlines , in moderately deep hst imaging that have an increased or decreased number of giant stars due to poisson variations . by averaging carefully matched narrow - band filter observations in bins of surface brightness fluctuation , we have quantified how the spectrum of changes when giant stars are added or taken away . the resulting relations between narrow - band indices and broad - band surface brightness fluctuations are remarkably well reproduced with the basic ingredients of the conroy & van dokkum ( 2012a ) stellar population synthesis model . the residuals from the predicted relations are very small at 0.03% 0.09% . qualitatively , this means that the model passes an important test in a previously unexplored regime . it also means that we know what the most luminous stars in elliptical galaxies are : giants with temperatures @xmath126k and luminosities @xmath127l@xmath75 . their spectra are well understood , even to the extent that our default model with the abundance pattern is a significantly better match to the data than a `` standard '' @xmath58enhanced model with [ @xmath128fe]=0.3 . understanding stars with @xmath89l@xmath75 is important as they comprise @xmath129% of the bolometric luminosity of old stellar populations , and their strong tio spectral features coincide with the imf - sensitive features nai and feh . we stress , however , that our data do not provide information on lower luminosity stars , or on the absolute number of luminous giants ( see 5.1.3 ) . we have tested various perturbations of our default model . the most important result is that we can quantitatively constrain the contribution of late m giants to the integrated light . such stars are not expected to exist in substantal numbers in old stellar populations , but their presence can not easily be ruled out based on integrated light analysis . importantly , leaving the contribution of these stars as a free parameter in fitting stellar populations can lead to degeneracies with the derived stellar initial mass function , as several tio bands overlap with imf - sensitive features ( see 5.2 in conroy & van dokkum 2012b ) . here we show that , for this particular galaxy , they can not contribute more than 3% to the integrated light . taking the stellar templates and isochrones as a given , we have explored the dependence of the observed ramp filter data on stellar population parameters . in the regime of small fluctuations the data are insensitive to the imf , but they provide constraints on metallicity and age . in particular , the fluctuation spectra are quite sensitive to age , as the number of late m giants with strong tio features drops sharply at very old ages . it is generally difficult to distinguish stellar populations that are between 10 and 13 gyr old , and fluctuation spectroscopy is probably among the most sensitive diagnostics in this regime . the analysis in this paper can be extended in various directions . the two most obvious ones are to obtain actual spectra rather than narrow - band filter data , and to probe regimes of larger surface brightness fluctuations . as illustrated in fig . [ largefluc.fig ] we can expect qualitative changes in the spectra of fluctuations @xmath104 , as these sightlines offer a relatively unobstructed view of subgiants and the main sequence . the analysis could center on differential techniques such as those employed in this paper , or on direct modeling of the spectra of very negative fluctuations . in principle , it should be possible to empirically , and spectroscopically , reconstruct the entire upper rgb of elliptical galaxies something that is impossible to do in any other way . among other applications , this type of analysis should provide strong constraints on the imf ( bottom panel of fig . [ largefluc.fig ] ) . observations with integral - field spectrographs on existing telescopes , particularly when combined with adaptive optics , could provide this kind of information for galaxies within @xmath114mpc . with jwst and 30 m class telescopes on the ground these studies could be extended to the virgo cluster and beyond . progress can also be made by combining constraints from fluctuation spectroscopy with those from fits to integrated - light spectra . the methods provide complementary information , as traditional blue spectra are mostly sensitive to stars near the main sequence turn - off . we note here that none of the model variations discussed in this paper lead to dramatic ( that is , larger than a few percent ) variations in the integrated - light spectra , with the exception of the models with very young ages . finally , the stellar library can be augmented . we currently use only 70 distinct templates in our default library , and in some regimes incomplete coverage leads to large `` jumps '' from one template to the other . future improvements will be enabled by larger and more comprehensive stellar libraries . , t. c. , flynn , k. , & gebhardt , k. 1990 , , 100 , 32 , j. p. , cantiello , m. , mei , s. , ct , p. , barber degraaff , r. , ferrarese , l. , jordn , a. , peng , e. w. , et al . 2010 , , 724 , 657 , j. p. , vazdekis , a. , & ajhar , e. a. 2001 , , 320 , 193 , g. & charlot , s. 2003 , , 344 , 1000 , n. , schiavon , r. , morrison , h. , rose , j. a. , & harding , p. 2011 , , 141 , 61 , c. 2013 , , 51 , 393 , c. , graves , g. j. , & van dokkum , p. g. 2014 , , 780 , 33 , c. , gunn , j. e. , & white , m. 2009 , , 699 , 486 , c. & van dokkum , p. 2012a , , 747 , 69 , c. & van dokkum , p. g. 2012b , , 760 , 71 , d. j. , springel , v. , white , s. d. m. , de lucia , g. , frenk , c. s. , gao , l. , jenkins , a. , kauffmann , g. , et al . 2006 , , 365 , 11 , j. j. , williams , b. f. , seth , a. c. , dolphin , a. , holtzman , j. , rosema , k. , skillman , e. d. , cole , a. , et al . 2009 , , 183 , 67 , l. , williams , b. f. , gilbert , k. m. , rosenfield , p. , dalcanton , j. j. , marigo , p. , boyer , m. l. , dolphin , a. , et al . 2010 , , 724 , 1030 , j. b. , tonry , j. l. , barris , b. j. , thompson , r. i. , liu , m. c. , rieke , m. j. , ajhar , e. a. , & blakeslee , j. p. 2003 , , 583 , 712 , m. r. 2011 , , 743 , 110 , a. & wood , p. r. 2000 , , 146 , 217 , m. c. , graham , j. r. , & charlot , s. 2002 , , 564 , 216 , c. 2005 , , 362 , 799 , s. , blakeslee , j. p. , tonry , j. l. , jordn , a. , peng , e. w. , ct , p. , ferrarese , l. , west , m. j. , et al . 2005 , , 625 , 121 , s. , blakeslee , j. p. , ct , p. , west , m. j. , ferrarese , l. , jordn , a. , peng , e. w. , anthony , a. , & merritt , d. 2007 , , 655 , 144 , p. d. , lacey , c. g. , baugh , c. m. , & cole , s. 2013 , , 435 , 87 , j. 2012 , , 755 , l14 , e. , van dokkum , p. , franx , m. , brammer , g. , momcheva , i. , frster schreiber , n. , da cunha , e. , tacconi , l. , et al . 2014 , nature , 513 , 394 , n. e. d. , greggio , l. , renzini , a. , carollo , c. m. , & maraston , c. 2013 , , 772 , 58 , j. t. , cushing , m. c. , & vacca , w. d. 2009 , , 185 , 289 , e. e. 1955 , , 121 , 161 , r. p. 2007 , , 171 , 146 , d. , johansson , j. , & maraston , c. 2011 , , 412 , 2199 , j. & schneider , d. p. 1988 , , 96 , 807 , j. l. , dressler , a. , blakeslee , j. p. , ajhar , e. a. , fletcher , a. b. , luppino , g. a. , metzger , m. r. , & moore , c. b. 2001 , , 546 , 681 , j. a. , piskunov , n. , & johns - krull , c. m. 1998 , , 498 , 851 , r. p. , cretton , n. , de zeeuw , p. t. , & rix , h .- w . 1998 , , 493 , 613 , p. g. 2008 , , 674 , 29 , p. g. & conroy , c. 2010 , , 468 , 940 . 2011 , , 735 , l13 , p. g. , bezanson , r. , van der wel , a. , nelson , e. j. , momcheva , i. , skelton , r. e. , whitaker , k. e. , brammer , g. , et al . 2014 , , 791 , 45 , a. , peletier , r. f. , beckman , j. e. , & casuso , e. 1997 , , 111 , 203 , j. , groves , b. , budavri , t. , & dale , d. 2011 , , 331 , 1 , r. f. & ford , jr . , w. k. 1969 , , 81 , 527 , g. 1994 , , 95 , 107 , g. , faber , s. m. , & gonzalez , j. j. 1992 , , 398 , 69 |
the description of the tunneling of a macroscopic variable in the presence of a bath of localized spins is a subject of great fundamental and practical interest , and is relevant for many solid - state qubit designs . most of the attention is usually given to the dynamics of the `` central spin '' ( i.e. , the qubit ) , while little is known about the spin bath itself . here we present a detailed study of the dynamics of the nuclear spin bath in the mn-ac single - molecule magnet , probed by nmr experiments down to very low temperatures ( mk ) . the results are critically analyzed in the framework of the prokofev - stamp theory of nuclear - spin mediated quantum tunneling . we find that the longitudinal relaxation rate of themn nuclei in mn-ac becomes roughly-independent below k , and can be strongly suppressed with a longitudinal magnetic field . this is consistent with the nuclear relaxation being caused by quantum tunneling of the molecular spin , and we attribute the tunneling fluctuations to the minority of fast - relaxing molecules present in the sample . the transverse nuclear relaxation is also-independent for k , and can be explained qualitatively and quantitatively by the dipolar coupling between like nuclei in neighboring molecules . this intercluster nuclear spin diffusion mechanism is an essential ingredient for the global relaxation of the nuclear spin bath . we also show that the isotopic substitution ofh byh leads to a slower nuclear longitudinal relaxation , consistent with the decreased tunneling probability of the molecular spin . finally , we demonstrate that , even at the lowest temperatures - where only-independent quantum tunneling fluctuations are present - the nuclear spins remain in thermal equilibrium with the lattice phonons , and we investigate the timescale for their thermal equilibration . after a review of the theory of macroscopic spin tunneling in the presence of a spin bath , we argue that most of our experimental results are consistent with that theory , but the thermalization of the nuclear spins is not . this calls for an extension of the spin bath theory to include the effect of spin - phonon couplings in the nuclear - spin mediated tunneling process . | 0706.2760 | the understanding of quantum tunneling in mesoscopic systems has made huge progress in the past decades , to the point that nanofabricated devices are now being exploited as coherently tunneling two - level systems ( tlss ) for quantum information purposes.@xcite conceptually , a first breakthrough was the proper description of the coupling of an effective tls to an environment described by an oscillator bath.@xcite whether the system is an intrinsic tls ( e.g. a spin @xmath8 ) or the low - energy truncation of a more complicated entity ( e.g. the flux state of a squid ) , one can generally apply the oscillator bath theory when the environment is described by delocalized modes ( conduction electrons , phonons , photons , etc . ) and the couplings of the tls to each oscillator are weak . in many solid - states systems , however , it can be necessary to account for localized environmental excitations whose couplings to the tls are not weak . this type of environment is called `` spin bath '' @xcite and can not be mapped onto an oscillator bath . importantly , a spin bath environment can cause decoherence even at @xmath9 and is therefore of great relevance for quantum systems that are designed to show coherent dynamics , like qubits for quantum computation . the prototypical realization of a tunneling tls coupled to a spin bath is the giant spin of a single - molecule magnet ( smm).@xcite these molecular systems consist of a core of strongly interacting transition metal ions , surrounded by organic ligands . at sufficiently low temperatures the core of the molecule behaves effectively like a single large spin @xmath10 . when uniaxial magnetic anisotropy is present , the reversal of the spin direction requires - classically - a large energy , so that the spin direction can be frozen at very low @xmath3 . however , in the presence of a transverse magnetic field or a biaxial anisotropy , the spin direction can be reversed by tunneling through the anisotropy barrier.@xcite the electronic spins that form the smm are magnetically coupled to the nuclear spins that either belong to the magnetic ions themselves ( @xmath2mn , @xmath11fe , ) or to the surrounding ligand molecules ( @xmath6h , @xmath12c , ) . as a consequence of these couplings , the observation of macroscopic quantum tunneling of magnetization in smms @xcite can not be understood without invoking the dynamics of the nuclear spins themselves.@xcite the theoretical predictions for the role of nuclear spins in the magnetization tunneling of smms @xcite have been verified by a series of experiments on the fe@xmath13 compound.@xcite most remarkably , this material allows to change the isotopic composition of the sample , both by strengthening ( @xmath11fe @xmath14 @xmath15fe substitution ) and weakening ( @xmath6h @xmath14 @xmath7h substitution ) the hyperfine couplings , while leaving the electronic structure of the smms unaffected . as predicted , the rate of quantum relaxation of the magnetization was found to be directly related to the nuclear isotopic composition of the sample.@xcite more recently , the effect of isotopic substitution has been observed in the low-@xmath3 electronic specific heat of fe@xmath13 ( ref . ) and in the dephasing time of coherent electron spin precession in cr@xmath16ni.@xcite nuclear spin effects were also invoked in the interpretation of @xmath17sr data in isotropic molecules,@xcite and in an alternative description of the short - term magnetic relaxation in smms.@xcite all these works have analyzed the effect of the nuclei on the dynamics of the `` central spin '' , but a crucial aspect of the theory of the spin bath is that the tunneling of the central system has repercussion on the dynamics of the bath itself , so that the latter can not be simply regarded as an independent source of `` noise '' . until now , the experiments to probe the electron spin dynamics have not been able to test this delicate aspect of the theory . to understand the details of the nuclear spin fluctuations , one should then look _ directly _ at the nuclear spins by means of low - temperature nmr experiments , performed under different regimes for the quantum dynamics of the electron spin . these experiments have been carried out by several groups,@xcite but an accurate analysis of their implications for the more general theory of nuclear - spin mediated quantum tunneling is still lacking . in this work , we present a comprehensive set of experiments on the dynamics of @xmath2mn nuclear spins in the mn@xmath0-ac smm , and we use our results for a critical assessment of the theory of the spin bath . our data provide definitive proof that the nuclear spin dynamics is strongly correlated with that of the central spin , that is , it can not be treated as an independent source of noise . indeed , we find that the nuclear spin fluctuations change dramatically when the tunneling dynamics of the central spin is modified , e.g. by an external magnetic field . in addition , we shall demonstrate that the nuclear spins remain in thermal equilibrium with the phonon bath down to the lowest temperatures ( @xmath1 mk ) accessible to our experiment , where the thermal fluctuations of the electron spins are entirely frozen out . this implies that there is a mechanism for exchanging energy between nuclei , electrons and phonons _ through the nuclear - spin mediated quantum tunneling of the central spin_. this is the point where the current theoretical description of macroscopic quantum tunneling in the presence of a spin bath needs to be improved . as regards the `` macroscopicness '' of the quantum effects observed in smms , we adopt leggett s view that the most stringent criterion is the `` disconnectivity'',@xcite @xmath18 , which roughly speaking is the number of particles that behave differently in the two branches of a quantum superposition . for instance , while a cooper pair box@xcite is a relatively large , lithographically fabricated device , the quantum superposition of its charge states involves in fact only one cooper pair , i.e. two electrons , and its disconnectivity is only @xmath19 . the matter - wave interference in fullerene molecules,@xcite for instance , is a much more `` quantum macroscopic '' phenomenon , since it means that 60 @xmath20 ( 12 nucleons + 6 electrons ) = 1080 particles are superimposed between different paths through a diffraction grating . for the spin tunneling in mn@xmath0-ac smms discussed here , we have 44 electron spins simultaneously tunneling between opposite directions , which places this system logarithmically halfway between single particles and fullerenes on a macroscopicness scale . the paper is organized as follows . section [ experimental ] describes the physical properties of the sample used in the experiments , the design and performance of our measurement apparatus , and the methods of data analysis . section [ nsd ] presents the experimental results on the nuclear spin dynamics , starting with the nmr spectra , the longitudinal and transverse relaxation rates in zero field , and their dependence on a longitudinal external field . we also study the nuclear relaxation in different mn sites within the cluster , and the effect of isotopic substitution in the ligand molecules . in section [ spint ] we discuss the thermal equilibrium between nuclear spins and phonon bath , the experimental challenges in optimizing it , and the indirect observation of magnetic avalanches during field sweeps . in section [ theory ] we give an introductory review of the theory of the spin bath , and apply its predictions to the calculation of the nuclear relaxation rate as observed in our experiments . together with the information on the thermal equilibrium of the nuclear spins , this will allow us to draw clear - cut conclusions on the status of our current theoretical understanding of quantum tunneling of magnetization . we conclude with a summary and implications of the results in section [ conclusions ] . we chose to focus our study on the well - known [ mn@xmath0o@xmath0(o@xmath21cme)@xmath22(h@xmath21o)@xmath23 ( mn@xmath0-ac ) compound , which belongs to the family of smms with the highest anisotropy barrier . as we shall see below , the rationale for choosing a smm with high anisotropy barrier is that the electron spin fluctuations become slow on the nmr timescale already at temperatures of a few kelvin . the structure of the cluster @xcite ( fig . [ structure ] ) consists of a core of 4 mn@xmath24 ions with electron spin @xmath25 , which we shall denote as mn@xmath26 , and 8 mn@xmath27 ions ( @xmath28 ) on two inequivalent crystallographic sites , mn@xmath29 and mn@xmath30 [ fig . [ structure](a ) ] . within the molecular cluster , the electron spins are coupled by mutual superexchange interactions , the strongest being the antiferromagnetic interaction between mn@xmath26 and mn@xmath29 ( ref . ) . the molecules crystallize in a tetragonal structure with lattice parameters @xmath31 and @xmath32 . the ground state of the molecule has a total electron spin @xmath33 and , for the temperature range of interest in the present work ( @xmath34 k ) , we may describe the electron spin of the cluster by means of the effective spin hamiltonian : @xmath35 commonly adopted parameter values are @xmath36 k , @xmath37 mk and @xmath38 @xmath17k as obtained by neutron scattering data,@xcite and for the @xmath39 tensor the values @xmath40 and @xmath41 from high - frequency epr.@xcite the uniaxial anisotropy terms @xmath42 and @xmath43 can be attributed to the single - ion anisotropy of the mn@xmath27 ions,@xcite which is due to the crystal field effects resulting in the jahn - teller distortions of the coordination octahedra , where the elongation axes are approximately parallel to the @xmath44-axis of the crystal . considering only the diagonal terms , the energy levels scheme would be a series of doublets of degenerate states , @xmath45 , separated by a barrier with a total height @xmath46 k [ fig . [ structure](b ) ] . the transverse anisotropy terms , @xmath47 , lift the degeneracy of the @xmath48 states and allow quantum tunneling of the giant spin through the anisotropy barrier . we call @xmath49 the matrix element for the tunneling of the giant spin through the @xmath50-th doublet , and @xmath51 the corresponding tunneling splitting . the @xmath52 term arises from the fourfold @xmath53 point symmetry of the molecule , but there is now solid experimental evidence @xcite for the prediction @xcite that a disorder in the acetic acid of crystallization is present and gives rise to six different isomers of mn@xmath0 cluster , four of which have symmetry lower than tetragonal and therefore have nonzero rhombic term @xmath54 . epr experiments give an upper bound @xmath55 mk.@xcite for the purpose of nmr experiments , such isomerism may cause slight variations in the local hyperfine couplings , causing extra broadening in the @xmath2mn resonance lines . very recently , a new family of mn@xmath0 clusters has been synthesized , which does not suffer from the solvent disorder mentioned above , and yields indeed more sharply defined @xmath2mn nmr spectra.@xcite when adding spin - phonon interactions,@xcite the possible transitions between the energy levels of ( [ hamiltonianmn12 ] ) are sketched in fig . [ structure](b ) . we distinguish between _ intrawell _ spin - phonon excitations , where the spin state remains inside the same energy potential well , and the _ interwell _ transitions , which involve spin reversal by quantum tunneling through the barrier , allowed by the terms in ( [ hamiltonianmn12 ] ) that do not commute with @xmath56 . thermally - assisted tunneling involves both these types of transitions . the above discussion refers to the majority of the molecules in a real sample , but for our experiments the crucial feature of mn@xmath0-ac is the presence of fast - relaxing molecules ( frms),@xcite i.e. clusters characterized by a lower anisotropy barrier and a much faster relaxation rate , as observed for instance by _ ac_-susceptibility@xcite and magnetization measurements.@xcite it has been recognized that such frms originate from jahn - teller isomerism,@xcite i.e. the presence in the molecule of one or two mn@xmath27 sites where the elongated jahn - teller axis points in a direction roughly perpendicular instead of parallel to the crystalline @xmath44-axis . this results in the reduction of the anisotropy barrier to 35 or 15 k in the case of one or two flipped jahn - teller axes , respectively,@xcite and presumably in an increased strength of the non - diagonal terms in the spin hamiltonian as well . furthermore , the anisotropy axis @xmath57 of the whole molecule no longer coincides with the crystallographic @xmath44-axis , but deviates e.g. by @xmath58 in the molecules with 35 k barrier.@xcite the jahn - teller isomerism is very different from the above - mentioned effect of disorder in solvent molecules , and produces much more important effects for the present study . as will be argued below , the presence of the frms is essential for the interpretation of our results and , to some extent , may be regarded as a fortunate feature for this specific experiment . the sample used in the experiment consisted of about 60 mg of polycrystalline mn@xmath0-ac , with typical crystallite volume @xmath59 mm@xmath60 . the crystallites were used as - grown ( i.e. , not crushed ) , mixed with stycast 1266 epoxy , inserted in a @xmath61 6 mm capsule and allowed to set for 24 hours in the room temperature bore of a 9.4 t superconducting magnet . with this procedure , the magnetic easy axis of the molecules ( which coincides with the long axis of the needle - like crystallites ) ends up being aligned along the field within a few degrees . in addition , we shall report nmr spectra taken on a small single crystal ( mass @xmath62 mg ) . our experimental setup is based on a leiden cryogenics mnk126 - 400rof dilution refrigerator , fitted with a plastic mixing chamber that allows the sample to be thermalized directly by the @xmath60he flow . a scheme of the low - temperature part of the refrigerator is shown in fig . [ dilution+temperatures ] , together with the nmr circuitry . the mixing chamber consists of two concentric tubes , obtained by rolling a kapton foil coated with stycast 1266 epoxy . the tops of each tube are glued into concentric araldite pots : the inner pot receives the downwards flow of condensed @xmath60he and , a few millimeters below the inlet , the phase separation between the pure @xmath60he phase and the dilute @xmath60he/@xmath63he phase takes place . the circulation of @xmath60he is then forced downwards along the inner kapton tube , which has openings at the bottom side to allow the return of the @xmath60he stream through the thin space in between the tubes . both the bottom of the kapton tail and the outer pot are closed by conical araldite plugs smeared with apiezon n grease . a 2-turns copper coil is wound around the capsule containing the sample , mounted on top of the lower conical plug and inserted in the @xmath60he/@xmath63he mixture at the bottom of the mixing chamber tail , which coincides with the center of a 9 t superconducting magnet . the coil is then connected by a thin brass coaxial cable ( length @xmath64 m ) to two tunable cylindrical teflon capacitors , mounted at the still ( see fig . [ dilution+temperatures ] ) . at the frequency where the cable connecting capacitors and coil is precisely one wavelength , the circuit is equivalent to a standard lumped @xmath65-resonator . however , since the @xmath66-cable is a low - conductivity coax for low-@xmath3 applications , the quality factor of the resonator ( which includes the cable ) is drastically reduced . although this affects the sensitivity of the circuit , it also broadens the accessible frequency range without the need to retune the capacitors . cutting the cable for one wavelength at @xmath67 mhz , the circuit is usable between ( at least ) 220 and 320 mhz . as for the room - temperature nmr electronics , details can be found in ref . . the temperature inside the mixing chamber is monitored by two simultaneously calibrated speer carbon thermometers , one in the outer top araldite pot , and the other at the bottom of the kapton tail , next to the sample . at steady state and in the absence of nmr pulses , the temperature along the mixing chamber is uniform within @xmath68 mk . the effect of applying high - power ( @xmath69 w ) nmr pulses is shown in fig . [ dilution+temperatures](a ) and ( b ) . a sudden increase in the measured temperature is seen both at the bottom and the top thermometer , and can be attributed to the short electromagnetic pulse . the temperature at the lower thermometer , i.e. next to the sample and the nmr coil , quickly recovers its unperturbed value , whereas the upper thermometer begins to sense the `` heat wave '' carried by the @xmath60he stream with a delay of about 3 minutes . this has the important consequence that we can use the upper thermometer to distinguish the effect of sudden electromagnetic radiation bursts from the simple heating of the @xmath60he/@xmath63he mixture , as will be shown in [ superradiance ] below . the sample temperature is regulated by applying current to a manganin wire , anti - inductively wound around a copper joint just above the @xmath60he inlet in the mixing chamber . in this way we can heat the incoming @xmath60he stream and uniformly increase the mixing chamber temperature . for the @xmath60he circulation we employ an oil - free pumping system , consisting of a 500 m@xmath60/h roots booster pump , backed by two 10 m@xmath60/h dry scroll pumps . the system reaches a base temperature of 9 mk , and the practical operating temperature while applying @xmath70-pulses is as low as 15 - 20 mk . the @xmath2mn nuclear precession was detected by the spin - echo technique . a typical pulse sequence includes a first @xmath71-pulse with duration @xmath72 @xmath17s , a waiting interval of 45 @xmath17s , and a 24 @xmath17s @xmath73-pulse for refocusing . given the heating effects shown in fig . [ dilution+temperatures ] , a waiting time of 600 s between subsequent pulse trains easily allows to keep the operating temperature around @xmath74 mk . moreover , at such low temperature the signal intensity is so high that we could obtain an excellent signal - to - noise ratio without need of averaging , so that a typical measurement sequence took less than 12 hours . above 100 mk it proved convenient to take a few averages , but there the heating due to the _ rf_-pulses became negligible , and the waiting time could be reduced to @xmath69 s. and @xmath75 mk in the mn@xmath26 site . ( b ) the ( normalized ) integral of the echoes ( open dots ) is fitted to eq . ( [ recovery ] ) ( solid line ) to yield the lsr rate @xmath76.,width=302 ] the longitudinal spin relaxation ( lsr ) was studied by measuring the recovery of the longitudinal nuclear magnetization after an inversion pulse . we preferred this technique to the more widely used saturation recovery @xcite because it avoids the heating effects of the saturation pulse train , but we checked at intermediate temperatures that the two methods indeed lead to the same value of lsr rate . an example of echo signals obtained as a function of the waiting time after the inversion pulse is shown in fig . [ echo3d](a ) . by integrating the echo intensity we obtain the time - dependence of the nuclear magnetization , @xmath77 , as shown in fig . [ echo3d](b ) . for the ease of comparison between different curves , we renormalize the vertical scale such that @xmath78 and @xmath79 , even though usually @xmath80 , as could be deduced from fig . [ echo3d](a ) . this is just an artifact that occurs when the nmr line is much broader than the spectrum of the inversion pulse , and does not mean that the length of the @xmath73-pulse is incorrect . since the @xmath2mn nuclei have spin @xmath81 , we fitted the recovery of the nuclear magnetization with : @xcite @xmath82 \label{recovery}\end{aligned}\ ] ] where @xmath76 is the longitudinal spin relaxation rate . note that , in the simple case of a spin 1/2 , @xmath76 is related to the relaxation time @xmath83 by @xmath84 . the above multiexponential expression and its numerical coefficients are derived under the assumption that the @xmath85 multiplet is split by quadrupolar interactions , and it is possible to resolve the central transition within that multiplet . while earlier work indicated that all three manganese nmr lines are quadrupolar - split,@xcite more recent experiments on single crystal samples have questioned that conclusion,@xcite and thereby the applicability of eq . ( [ recovery ] ) to the present experiments . even if other sources of line broadening hinder the visibility of the quadrupolar contribution , the condition for the absence of quadrupolar splitting is an exactly cubic environment for the nuclear site , which is not satisfied here . for this reason , and for the ease of comparison with our@xcite and other groups earlier results,@xcite we choose to retain eq . ( [ recovery ] ) for the analysis of the inversion recovery data . the transverse spin relaxation ( tsr ) rate @xmath86 was obtained by measuring the decay of echo intensity upon increasing the waiting time @xmath87 between the @xmath88- and the @xmath73-pulses . the decay of transverse magnetization @xmath89 can be fitted by a single exponential @xmath90 except at the lowest temperatures ( @xmath91 k ) , where also a gaussian component @xmath92 needs to be included : @xmath93 as regards the experiments to determine the nuclear spin temperature , the measurements were performed by monitoring the echo intensity at regular intervals while changing the temperature @xmath94 of the @xmath60he/@xmath63he bath in which the sample is immersed . recalling that the nuclear magnetization is related to the nuclear spin temperature @xmath95 by the curie law : @xmath96 and assuming that @xmath97 at a certain temperature @xmath98 ( e.g. 0.8 k ) , we can define a calibration factor @xmath99 such that @xmath100 and use that definition to derive the time evolution of the nuclear spin temperature as @xmath101 while the bath temperature is changed . due to the strong magnetic hysteresis of mn@xmath0-ac , it is important to specify the magnetization state of the sample since , as will be shown below , this parameter can influence the observed nuclear spin dynamics . therefore we carried out experiments under both zero - field cooled ( zfc ) and field - cooled ( fc ) conditions , which correspond to zero and saturated magnetization along the easy axis , respectively . heating the sample up to @xmath102 k is sufficient to wash out any memory of the previous magnetic state . when the sample is already at @xmath103 k , the field - cooling procedure can be replaced by the application of a longitudinal field large enough to destroy the anisotropy barrier , e.g. @xmath104 t. importantly , the shift of the @xmath2mn nmr frequency with external field depends on the magnetization state of the sample:@xcite in a zfc sample each resonance line splits in two , one line moving to @xmath105 and the other to @xmath106 . conversely , in a fc sample only one line is observed , shifting to higher or lower frequency depending on the direction of @xmath107 relative to the magnetization direction . therefore , by measuring the intensity of the shifted lines in a moderate longitudinal field , typically @xmath108 t , we can check the magnetization of the sample _ as seen by the nuclei that contribute to the nmr signal_. mn nmr spectra of the ( a ) mn@xmath26 and ( b ) mn@xmath29 lines in mn@xmath0-ac , at @xmath75 mk . open circles : oriented powder . solid squares : single crystal . the mn sites corresponding to each line are shown in the central drawing of the molecular structure . all the spectra are measured in a field - cooled sample.,width=321 ] the basic feature of the @xmath2mn nmr spectra in mn@xmath0-ac is the presence of three well - separated lines , that can be ascribed to three crystallographically inequivalent mn sites in the molecule . the mn@xmath26 line , centered around @xmath109 mhz , originates from the nuclei that belong to the central core of mn@xmath24 ions , whereas the mn@xmath29 and mn@xmath30 lines , centered at @xmath110 and @xmath111 mhz , respectively , have been assigned to the nuclei in the outer crown of mn@xmath27 ions.@xcite in fig . [ spectra ] we show the mn@xmath26 and mn@xmath29 spectra at @xmath75 mk , both in the oriented powder and in the single crystal , in a fc sample . note that , whereas single - crystal spectra of mn@xmath0-ac have been recently published,@xcite the present spectra are the only ones measured at subkelvin temperatures so far . as argued already in ref . , the single - crystal spectra indicate that the width of the mn@xmath26 line may not originate from a small quadrupolar splitting . instead , at least two inequivalent mn@xmath24 sites may exist , supporting the growing amount of evidence about the lack of symmetry of the mn@xmath0-ac compound . we also note that the highest peak in the mn@xmath29 line at @xmath75 mk is found at a frequency @xmath112 mhz about 8 mhz higher than most of the previously reported spectra at @xmath113 k,@xcite with the exception of ref . , whereas the position of the mn@xmath26 line is consistent with all the previous reports . of the mn@xmath26 line , in zero external field and zfc sample . the inset shows some examples of recovery of the nuclear magnetization after a time @xmath114 from an inversion pulse , at the indicated temperatures . these curves have been fitted to eq . ( [ recovery ] ) to extract @xmath76.[wzf],width=321 ] the lsr rate as a function of temperature for the mn@xmath26 line , in zero field and zero - field cooled ( zfc ) sample , is shown in fig . the most prominent feature in these data is a sharp crossover at @xmath4 k between a roughly exponential @xmath3-dependence and an almost @xmath3-independent plateau . we have previously attributed the @xmath3-independent nuclear relaxation to the effect of tunneling fluctuations within the ground doublet of the cluster spins,@xcite and we shall dedicate most of the present paper to discuss our further results supporting this statement . here we shall also argue that , even in the high - temperature regime , thermally assisted quantum tunneling plays an essential role , and the experimental results can not be understood simply in terms of lsr driven by intrawell electronic transitions.@xcite it should be noted that the crossover from thermally activated to ground - state tunneling has also been observed by analyzing the @xmath3-dependence of the steps in the magnetization hysteresis loops.@xcite the important advantage of our nmr measurements is that the nuclear dynamics is sensitive to _ fluctuations _ of the cluster electron spins without even requiring a change in the macroscopic magnetization of the sample . clearly , no macroscopic probe ( except perhaps an extremely sensitive magnetic noise detector ) would be able to detect the presence of tunneling fluctuations in a zero - field cooled sample in zero external field , since the total magnetization is zero and remains so . below @xmath115 k the steps in the hysteresis loops of mn@xmath0-ac can be observed only at relatively high values of external field,@xcite which means that the spin hamiltonian under those conditions is radically different from the zero - field case . therefore , that both our data and the previous magnetization measurements show a crossover around @xmath116 k should be considered as a coincidence . the roughly @xmath3-independent plateau in the lsr rate below @xmath117 k is characterized by a value of @xmath118 s@xmath119 which is surprisingly high , which at first sight may appear like an argument against the interpretation in terms of tunneling fluctuations of the electron spin . experimentally it is indeed well known@xcite that the relaxation of the magnetization in mn@xmath0-ac in zero field may take years at low @xmath3 , which means that the tunneling events are in fact extremely rare . based on this , we are forced to assume that tunneling takes place only in a small minority of the clusters , and that some additional mechanism takes care of the relaxation of the nuclei in molecules that do not tunnel . this is a very realistic assumption , since all samples of mn@xmath0-ac are reported to contain a fraction of frms,@xcite as mentioned in sect . [ sample ] . moreover , since we are also able to monitor the sample magnetization , we verified that e.g. a fc sample maintains indeed its saturation magnetization for several weeks _ while nuclear relaxation experiments are being performed ( at zero field)_. this confirms that any relevant tunneling dynamics must originate from a small minority of molecules . on the other hand , it also means that the observed nmr signal comes mainly from nuclei belonging to frozen molecules , thus there must be some way for the fluctuations in frms to influence the nuclear dynamics in the majority of slow molecules as well . one possibility is to ascribe it to the fluctuating dipolar field produced by a tunneling frm at the nuclear sites of neighboring frozen molecules . in that case we may give an estimate of @xmath76 using an expression of the form : @xmath120 where @xmath121 is the perpendicular component of the fluctuating dipolar field produced by a tunneling molecule on its neighbors and @xmath122 is the tunneling rate . the highest value that @xmath121 may take is @xmath123 mt in the case of nearest neighbors , which leads to the condition @xmath124 s@xmath119 @xmath125 s@xmath119 . such a high rate is of course completely unrealistic . we must therefore consider the effect of a tunneling molecule on the nuclei that _ belong _ to the molecule itself , and look for some additional mechanism that links nuclei in frms with equivalent nuclei in frozen clusters . it is natural to seek the origin of such a mechanism in the intercluster nuclear spin diffusion , and in the next section we shall provide strong experimental evidences to support this interpretation . ( color online ) temperature - dependence of the tsr rate @xmath86 ( squares ) rates for a zfc sample in zero field and @xmath126 mhz . the solid line in the @xmath3-independent regime is a guide for the eye . inset : normalized decay of transverse nuclear magnetization , @xmath127 , for zfc ( full squares ) and fc ( open squares ) sample , at @xmath128 mk . the solid lines are fits to eq . ( [ t2lg ] ) , yielding the ratio @xmath129 . the sketches in the inset represent pictorially the fact that intercluster spin diffusion is possible in a fc sample since all the nuclei have the same larmor frequency , contrary to the case of a zfc sample.,width=321 ] the @xmath3-dependence of the tsr rate @xmath130 is shown in fig . [ t2fczfc ] . one may observe that below 0.8 k the tsr , just like the lsr , saturates to a nearly @xmath3-independent plateau . in particular , @xmath131 s@xmath119 , which is a factor @xmath132 larger than the low-@xmath3 limit of the lsr rate @xmath76 . the values plotted in fig . [ t2fczfc ] are all obtained by fitting the decay of the transverse magnetization with eq . ( [ t2 ] ) , i.e. with a single exponential . while this is very accurate at high @xmath3 , we found that for @xmath91 k a better fit is obtained by including a gaussian component , as in eq . ( [ t2lg ] ) . in any case , the single - exponential fit does capture the relevant value for @xmath86 at all temperatures . a point of great interest is the measurement of the tsr at @xmath75 mk in a fc and a zfc sample , as shown in the inset of fig . [ t2fczfc ] . the decay of the transverse magnetization is best fitted by eq . ( [ t2lg ] ) , whereby the gaussian component , @xmath92 , is separated from the lorentzian one , @xmath133 . from the gaussian component of the decay we can extract directly the effect of the nuclear dipole - dipole interaction , whereas the other mechanisms of dephasing ( e.g. random changes in the local field due to tunneling molecules ) contribute mainly to the lorentzian part . the fit yields @xmath134 s@xmath119 and @xmath135 s@xmath119 . these results can be understood by assuming that , at very low @xmath3 , the main source of tsr is the dipole - dipole coupling of like nuclei in neighboring molecules . then we can estimate @xmath86 from the van vleck formula for the second moment @xmath136 of the absorption line in dipolarly - coupled spins:@xcite @xmath137 yielding @xmath138 s@xmath119 if we take for @xmath139 the distance between centers of neighboring molecules . the estimated @xmath86 would obviously be much larger if one would consider the coupling between nuclei within the same cluster . as we argued when discussing the @xmath2mn spectra , it is possible that the cluster symmetry is low enough to prevent intracluster nuclear spin flip - flops . this may explain why eq . ( [ vanvleck ] ) yields the right order of magnitude when only coupling between nuclei in neighboring molecules is considered . an alternative argument is that , given the small number ( 4 at best ) of like @xmath2mn spins within one cluster , the dipolar coupling between them does not yield a genuine decay of the transverse magnetization for the entire sample . the macroscopic @xmath140 decay measured in the experiment reflects therefore the slower , but global , intercluster spin diffusion rate . a similar observation was recently made also in a different molecular compound , al@xmath141c@xmath142h@xmath143 ( ref . ) . we also note that , in the case of a zfc sample , the sum in eq . ( [ vanvleck ] ) should be restricted to only half of the neighboring molecules , since on average half of the spins have resonance frequency @xmath144 and the other half @xmath145 , and no flip - flops can occur between nuclei experiencing opposite hyperfine fields . this is equivalent to diluting the sample by a factor 2 , which reduces the expected @xmath86 in zfc sample by a factor @xmath146 . indeed , we find in the experiment @xmath147 which , together with the good quantitative agreement with the prediction of eq . ( [ vanvleck ] ) , constitutes solid evidence for the presence of intercluster nuclear spin diffusion . this is precisely the mechanism required to explain why the tunneling in a minority of frms can relax the whole nuclear spin system . the need for intercluster nuclear spin diffusion could already have been postulated by analyzing the lsr rate , and the magnetization dependence of the tsr rate gives an independent confirmation . for comparison , in a recent study of the @xmath15fe nmr in fe@xmath13 , baek _ _ et al.__@xcite attributed the observed tsr rate to the dipolar interaction between @xmath15fe and @xmath6h nuclei . they analyzed their data with the expression @xmath148 , where @xmath149 is the proton tsr time due to their mutual dipolar coupling and @xmath150 is the second moment of the @xmath15fe - @xmath6h coupling . however , the same model@xcite predicts the echo intensity to decay as @xmath151 . this function fails completely in fitting our echo decays , therefore we do not consider the @xmath2mn - @xmath6h dipolar coupling as an alternative explanation for the tsr we observe . finally we stress that , in our view , the fact that the lsr and the tsr are both roughly @xmath3-independent below 0.8 k does not find its origin in the same mechanism . rather , we attribute them to two different mechanisms , both @xmath3-independent : the quantum tunneling of the electron spin ( for the lsr ) and the nuclear spin diffusion ( for the tsr ) . having argued that the lsr in mn@xmath0-ac in driven by tunneling fluctuations of the frms , which are peculiar of the acetate compound , it s interesting to note that other varieties of mn@xmath0 molecules have meanwhile become available . in particular the mn@xmath0-__t__buac@xcite is a truly axially symmetric variety that does not contain any frms , and could provide an interesting counterexample for our results if studied by low-@xmath3 nmr . the mn@xmath0brac molecule is also thought to be free of frms,@xcite and some low-@xmath3 nmr experiments have been performed on it@xcite that show indeed very different results from what we report here . however , as we shall argue in [ spint ] , a definite conclusion on the meaning of nmr experiments at very low @xmath3 should only be drawn when the analysis of the nuclear spin thermalization is included . further insight in the interplay between the quantum tunneling fluctuations and the nuclear spin dynamics is provided by the study of the dependence of the lsr on a magnetic field @xmath107 applied along the anisotropy axis . it is clear from the hamiltonian ( [ hamiltonianmn12 ] ) that , in the absence of other perturbations , such a longitudinal field destroys the resonance condition for electron spin states on opposite sides of the barrier and therefore inhibits the quantum tunneling . in the presence of static dipolar fields , @xmath152 , by studying the tunneling rate as a function of @xmath107 one may in principle obtain information about the distribution of longitudinal @xmath152 , since at a given value of @xmath107 there will be a fraction of molecules for which @xmath153 and will therefore be allowed to tunnel just by the application of the external bias . ( color online ) longitudinal field dependence of the lsr rate @xmath76 in the zfc ( solid dots ) and fc ( open dots ) sample at @xmath75 mk . the measuring frequencies are @xmath154 mhz . the solid line is a lorentzian fit with hwhm @xmath155 mt . the dotted line through the fc data is a guide for the eye.,width=302 ] we show in fig . [ wvsb - lt ] the lsr rate @xmath156 at @xmath75 mk in the zfc sample , obtained while shifting the measurement frequency as @xmath157 with @xmath158 mhz , in order to stay on the center of the nmr line that corresponds to the molecules that are aligned exactly parallel with the applied field . since for a zfc sample the magnetization is zero , the field dependence should be the same when @xmath107 is applied in opposite directions , as is observed . the data can be fitted by a lorentzian with a half width at half maximum ( hwhm ) @xmath155 mt : this differs both in shape ( gaussian ) and in width ( @xmath159 mt ) from the calculated dipolar bias distribution in a zfc sample.@xcite an alternative experimental estimate , @xmath160 mt , can be found in magnetization relaxation experiments,@xcite but only around the first level crossing for frms ( @xmath161 t ) in the fc sample . for comparison , [ wvsb - lt ] also shows @xmath156 in the fc sample : the shape is now distinctly asymmetric , with faster relaxation when the external field is opposed to the sample magnetization . interestingly , @xmath156 in the fc sample falls off much more slowly on the tails for both positive and negative fields , while the value at zero field is less than half that for the zfc sample . we therefore observe that in zero field the recovery of longitudinal magnetization in the fc sample is faster than in the zfc , whereas the opposite is true for the decay of transverse magnetization ( inset fig . [ t2fczfc ] ) . if the lsr rate @xmath156 is to be interpreted as a signature of quantum tunneling , its hwhm is clearly larger than expected . part of the reason may be the fact that the width of the mn@xmath26 line is already intrinsically larger than both @xmath162 and the distribution of dipolar fields created by the molecules . indeed , the width of the mn@xmath26 line , @xmath163 mhz , translates into a local field distribution of width @xmath164 mt for @xmath2mn . the observed hwhm does depend , for instance , on the choice of @xmath165 . as soon as @xmath166 the presence of slightly misaligned crystallites in our sample may also contribute to the width of the resonance . in any case , all of the mechanisms mentioned above ( distribution of internal dipolar fields , width of the nmr line , distribution of crystallite orientations in the sample ) would yield a @xmath3-independent linewidth for @xmath156 . [ wvsb - t ] shows @xmath156 in zfc sample at three different temperatures , @xmath167 k , covering the pure quantum regime , the thermally - activated regime , and the crossover temperature . the nmr frequency in these datasets is @xmath168 . the data have been fitted by lorentzian lines yielding a hwhm @xmath169 increasing with temperature . we note immediately that the hwhm at @xmath75 mk is much smaller than the one obtained from the data in fig . [ wvsb - lt ] , the only difference between the two sets being @xmath165 and , subsequently , all other measurement frequencies at @xmath166 . indeed , we found that already in zero field the lsr rate does depend on @xmath170 , reaching the highest values at the center of the line and falling off ( up to a factor 5 ) on the sides . this dependence , however , becomes much weaker at high temperatures . it is therefore rather difficult to make strong statements about the meaning of the observed increase in @xmath162 with temperature . at any rate , however , the field dependencies observed here at low-@xmath3 are much stronger than those previously reported in the high-@xmath3 regime.@xcite goto _ et al . _ also reported @xmath156 for the `` lower branch '' of the mn@xmath26 line , viz . for the nuclei whose local hyperfine field is opposite to the external field ( ref . , fig . 6 , closed squares ) . that situation is equivalent to our fc data ( fig . [ wvsb - lt ] , open dots ) for @xmath171 . at large fields an overall increase of @xmath76 with @xmath107 is observed in ref . , but for @xmath172 t the lsr rate does decrease , in agreement with our results . ( color online ) longitudinal field dependence of the lsr rate in zfc sample at @xmath75 mk ( down triangles ) , @xmath173 mk ( diamonds ) and @xmath174 k ( up triangles ) . the measuring frequency in these datasets is @xmath175 mhz . the lines are lorentzian fits yielding hwhm @xmath169 mt , respectively.,width=302 ] we also noted , both in fig . [ wvsb - t ] and in the fc data in fig . [ wvsb - lt ] , that a small increase in @xmath156 occurs at @xmath176 t , which is approximately the field value at which the @xmath177 and @xmath178 electron spin states come into resonance . this feature is barely observable , but nevertheless well reproducible . as a counterexample , in another dataset ( not shown ) we investigated @xmath156 more carefully in the fc sample at @xmath75 mk for positive values of @xmath107 , and found no increase around @xmath179 t , as one would expect since the fully populated state , @xmath178 , is pushed far from all other energy levels . a similarly small peak in @xmath156 at the first levels crossing has been recently observed in fe@xmath13 as well.@xcite the role of the fluctuating hyperfine bias on the incoherent tunneling dynamics of smms , predicted by prokofev and stamp,@xcite has been clearly demonstrated by measuring the quantum relaxation of the magnetization in fe@xmath13 crystals in which the hyperfine couplings had been artificially modified by substituting @xmath11fe by @xmath15fe or @xmath6h by @xmath7h ( ref . ) . for instance , the time necessary to relax 1% of the saturation magnetization below 0.2 k was found to increase from 800 s to 4000 s by substituting protons by deuterium , whereas it decreased to 300 s in the @xmath15fe enriched sample . more recently , evangelisti _ _ et al.__@xcite showed that the @xmath15fe isotopic enrichment of fe@xmath13 causes the magnetic specific heat to approach its equilibrium value within accessible timescales ( @xmath69 s ) . since in mn@xmath0-ac the only possible isotope substitution is @xmath6h @xmath180h , we performed a short set of measurements on a deuterated sample . the sample consists of much smaller crystallites than the `` natural '' ones used in all other experiments reported here . although a field - alignment was attempted following the same procedure as described in [ measures ] , the orientation of the deuterated sample turned out to have remained almost completely random , probably due to the too small shape anisotropy of the crystallites . we therefore report only experiments in zero external field , where the orientation is in principle irrelevant . the results are shown in fig . [ t1t2d ] : the @xmath2mn lsr rate at @xmath75 mk in zero field and zfc sample is indeed reduced to @xmath181 s@xmath119 , i.e. 6.5 times lower than in the `` natural '' sample . this factor is the same as the reduction of the electron spin relaxation rate seen in deuterated fe@xmath13 ( ref . ) , and it coincides with the ratio of the gyromagnetic ratios of @xmath6h and @xmath7h . this finding unequivocally proves that the proton spins are very effective in provoking the tunneling events via the prokofev - stamp mechanism , and confirms that the lsr rate of the @xmath2mn nuclei is a direct probe of the electron spin tunneling rate . as regards the tsr , the result is quite intriguing : slow but rather ample oscillations are superimposed to the decay of transverse magnetization , and the overall decay rate appears slower than in the natural sample . this behavior is reminiscent of the change in tsr rate upon application of a small longitudinal magnetic field in the natural sample . the latter has a rather complicated physical origin and is still under investigation . some rather interesting results emerge from the analysis of extra measurements performed on the nmr line of the mn@xmath29 site , i.e. a mn@xmath27 ion . [ t1t2mn34 ] shows a comparison between the recovery of the longitudinal magnetization and the decay of the transverse magnetization in mn@xmath26 and mn@xmath29 sites , at @xmath75 mk in the fc sample and zero external field , at a frequency @xmath182 mhz . the tsr is very similar in both sites , although a closer inspection evidences that the gaussian nature of the decay is less pronounced in the mn@xmath29 sites , which leads to @xmath183 s@xmath119 instead of the @xmath184 s@xmath119 found in mn@xmath26 . more importantly , the lsr is three times slower in the mn@xmath29 site , as seen in fig . [ t1t2mn34](b ) . this is opposite to the high-@xmath3 regime , where the mn@xmath27 sites were found@xcite to have much faster relaxation . furthermore , the field dependence of the lsr rate appears sharper in the mn@xmath29 site , as shown in fig . [ wvsb34 ] . the asymmetry in @xmath156 for a fc sample is still present , but less evident than in the mn@xmath26 site due to the more pronounced decrease of @xmath76 already for small applied fields . the similarity between the tsr rates in the mn@xmath26 and the mn@xmath29 sites is indeed expected if @xmath140 is determined by intercluster nuclear spin diffusion . conversely , the difference in lsr is more difficult to understand if one assumes that the process that induces longitudinal spin relaxation is the tunneling of the molecular spin . however , one clear difference between mn@xmath26 and mn@xmath29 is the width of the nmr line , much larger in mn@xmath29 . since the integrated intensity of both lines is identical , the mn@xmath29 has an accordingly lower maximum intensity . we have verified for both sites that the lsr rate is the fastest when measuring at the highest intensity along each line . thus , the factor 3 slower lsr in mn@xmath29 could simply be another manifestation of the apparent dependence of the measured @xmath76 on the nmr intensity along each line . we point out , however , that the measured lsr rate is independent of the @xmath88 pulse length , which determines the spectral width of the pulse and thereby the fraction of spins being manipulated and observed . this means that the difference in w for the two sites can not be simply attributed to a difference in the number of spins excited during a pulse of given length but that other ( more complex ) factors must play a role . ( color online ) longitudinal field dependencies of the lsr rates in mn@xmath26 ( circles ) and mn@xmath29 ( diamonds ) sites , normalized at the zero - field value . the data are taken at @xmath75 mk in fc sample with central measuring frequencies @xmath185 mhz and @xmath186 mhz.,width=302 ] having demonstrated that the @xmath2mn longitudinal spin relaxation below 0.8 k is driven by @xmath3-independent quantum tunneling fluctuations , a natural question to ask is whether or not the nuclear spins are in thermal contact with the lattice at these low temperatures . let us recall that any direct coupling between phonons and nuclear spins is expected to be exceedingly weak , due to the very small density of phonons at the nuclear larmor frequency.@xcite relaxation through electric quadrupole effects , if present , would show a temperature dependence @xmath187 for direct process or @xmath188 for raman process ( @xmath189 is the debye temperature ) , which is not consistent with our observations . therefore the thermalization of the nuclei will have to take place via the electron spin - lattice channel . since in the quantum regime the only electron spin fluctuations are due to tunneling , the question whether the nuclear spins will still be in equilibrium with the lattice temperature is of the utmost importance . ( color online ) comparison between bath temperature @xmath94(solid lines ) and nuclear spin temperature @xmath95 ( circles ) , while cooling down the system ( main panel ) and while applying step - like heat loads ( inset ) . the waiting time between nmr pulses was 60 s in the main panel and 180 s in the inset . both datasets are at zero field in zfc sample.,width=321 ] we have addressed this problem by cooling down the refrigerator from 800 to 20 mk while monitoring simultaneously the temperature @xmath94 of the @xmath60he/@xmath63he bath in the mixing chamber ( just next to the sample ) and the nmr signal intensity of the mn@xmath26 line , in zero external field and on a zfc sample . the signal intensity was measured by spin echo with repetition time @xmath190 s. the nuclear spin temperature@xcite @xmath95 is obtained as described in [ measures ] , and plotted in fig . [ tnuclcooldown ] together with @xmath94 . we find that the nuclear spin temperature strictly follows the bath temperature , with small deviations starting only below @xmath191 mk . this result is crucial but rather paradoxical , and we shall discuss its implications in detail in [ thermaleq ] . experimentally , however , it certifies the effectiveness of our cryogenic design in achieving the best possible thermalization of the sample , since the nuclear spins are the last link in the chain going from the @xmath60he/@xmath63he bath via the phonons in the sample to the electron spins and finally to the nuclei . the lowest spin temperature that can be _ measured _ appears to depend on the pulse repetition time @xmath192 . to measure @xmath95 with the pulse nmr method we need a @xmath88 pulse to create a transverse nuclear magnetization , and after a time @xmath140 the spins are effectively at infinite @xmath3 so enough time must elapse before taking the next @xmath95 measurement . for the data in fig . [ tnuclcooldown ] , @xmath190 s was barely longer than the observed time for inversion recovery [ see fig . [ echo3d](b ) ] , and the lowest observed spin temperature is @xmath193 mk . this improved when using longer waiting times between pulses , e.g. @xmath194 mk with @xmath195 s , as shown in the inset of fig . [ tnuclcooldown ] . however , no matter how long the waiting time , we never observed a @xmath95 lower than @xmath196 mk . .[tabletnucl]experimental conditions and relaxation rates for the nuclear spin temperature experiments in fig . [ spintcompare ] [ cols="^,^,^,^,^,^,^,^ " , ] by looking at the theoretical curves alone we find that , given a fixed set of parameters for the frms spin hamiltonian ( [ hfrm ] ) , the nuclear - driven tunneling process always dominates over the phonon - driven one , both in the low-@xmath3 and in the high-@xmath3 regime . the situation may be reversed in the high-@xmath3 regime by assuming the sound velocity is lower than the literature value used here , but @xmath197 would never hold at low-@xmath3 under realistic circumstances . a comparison with the experimental data shows that both nuclear- and phonon- driven mechanisms yield a correct slope of @xmath198 in the thermally - assisted regime , @xmath199 k , whereas the phonon process can never reproduce the low-@xmath3 plateau . an almost perfect fit of the data is obtained by assuming that the frms have a total spin @xmath200 , while the @xmath33 case has a @xmath3-dependent region systematically starting at too high temperatures . in the nuclear - driven case , we used an optimal value of @xmath201 mk which seems very reasonable , since the main contribution to @xmath202 arises from the coupling to protons . conversely , the values used for @xmath203 appear very high , since to fit the nmr data we need to assume @xmath204 . as mentioned before , however , such a high value of @xmath203 should not be interpreted as the spectroscopic rhombic term in the spin hamiltonian , since it is used here as the parameter that tunes the tunneling splitting of the ground doublet , and therefore incorporates the effect of @xmath205-mixing@xcite if a manifold with different total spin value is energetically close to the @xmath205 ground state . an important remark is that our calculations , which only account for tunneling fluctuations as a source of nuclear lsr , can accurately reproduce the observed lsr rate _ also in the thermally activated regime _ , @xmath206 k. all the previous nmr experiments in that regime@xcite have been interpreted in terms of the `` intrawell '' electron spin fluctuations , that arise from thermal excitation of the electron spin state on the same side of the anisotropy barrier [ see fig . [ structure](b ) ] . as pointed out by goto _ et al._,@xcite that reasoning is inappropriate when applied to the lsr of nuclei belonging to mn@xmath24 ions , since the hyperfine coupling tensor is diagonal . in that case , a fluctuation in the @xmath207 projection of the electron spin does not result in a fluctuating field perpendicular to the nuclear quantization axis ( which is @xmath207 itself ) , and can not account for longitudinal nuclear spin relaxation . as we show in fig . [ fitw1 ] , including the effect of electron spin tunneling through @xmath208 doublets solves what appeared to be a paradox , since the tunneling fluctuations induce nuclear lsr also in the absence of non - diagonal hyperfine coupling terms . for the nuclei in mn@xmath27 ions , the intrawell electron spin transitions do provide an additional channel for nuclear relaxation , which may explain why the lsr rate at @xmath209 k in the mn@xmath27 sites is larger than that in the mn@xmath24 ones.@xcite in the preceding discussion , it may seem that we have not explicitly used the condition that the nuclear spins are in thermal equilibrium , which is what we observe in the experiment . this condition , however , is automatically implied in the application of eq . ( [ wmodel ] ) to the lsr rate . for eq . ( [ wmodel ] ) to actually represent the rate at which the nuclear spins exchange energy with a thermal bath and thereby return to the equilibrium magnetization after a perturbing nmr pulse , one needs to include the detailed balance condition : @xmath210 in other words , if the nuclear spin temperature has to reach equilibrium with the thermal ( phonon ) bath via the process described by the rates @xmath211 and @xmath212 , the latter _ must _ satisfy ( [ detbal ] ) . the detailed balance condition is often taken for granted , but in this case one needs to be more careful . @xmath211 and @xmath212 represent the rates of _ electron spin transitions _ that increase or lower the nuclear energy , respectively . the crucial point is that both are rates for _ tunneling _ transitions , which occur when the _ total energy _ of the `` electron plus nuclear spins '' system is the same before and after the electron spin flip . thus , the difference between @xmath211 and @xmath212 is simply that , e.g. , @xmath211 is the rate for a tunneling transition that increases the nuclear spin energy while reducing the electronic one . that is , after the flip most of the nuclei are oriented against their local hyperfine field , while the electron spin is favorably aligned with respect to the local field ( in particular the dipolar one , when @xmath213 ) . @xmath212 does the opposite and , interestingly , this means that the instantaneous local spin temperature ( local referring to the nuclei belonging to a specific molecule that has just flipped ) is negative . this situation is clearly very different from the standard nmr picture of nuclear relaxation by coupling to paramagnetic centers , where the latter make spin - phonon transitions between zeeman - split levels having different thermal populations . now we can summarize the meaning of our experimental results for the description of electron spin tunneling in the presence of a nuclear spin bath : \(i ) the prokofev - stamp theory of the spin bath , as developed so far and reviewed in [ spinbath ] , quantitatively and qualitatively reproduces the nuclear lsr rate in the whole temperature regime of our measurements , by assuming that the lsr is due to tunneling events in a minority of fast - tunneling molecules ; \(ii ) the additional observation that the nuclear spins are in thermal equilibrium with the phonon lattice at all temperatures implies that _ the rates @xmath211 and @xmath212 must be different . _ for this to happens , it is necessary to explicitly include the role of spin - phonon interactions in the nuclear - spin mediated tunneling process . importantly , the results of the calculations shown in fig . [ fitw1](b , d ) indicate that it is not sufficient to attribute the thermal relaxation to a phonon - assisted tunneling process as described in [ phononrate ] , working `` in parallel '' to the nuclear - spin mediated tunneling process . at the lowest temperatures , even the longest thermalization times observed in our experiments ( [ timespint ] ) are still much shorter than @xmath214 as calculated from phonon - assisted tunneling alone ( [ phononrate ] ) , thus reinforcing the need for a theory that includes nuclear - spin and phonon mediated tunneling _ at the same time _ ; \(iii ) our statement that the nuclear relaxation has to be mediated by inelastic electron spin tunneling processes , is further supported by specific heat experiments that show that a system of dipolarly - coupled tunneling molecules can relax to the long - range ferromagnetically ordered state , provided that the tunneling rate is fast enough for the experimental detection of the ordering anomaly.@xcite this means that _ there is a mechanism for the ensemble of electron spins to find its thermodynamic ground state , even at temperatures so low that the relaxation can only proceed by quantum tunneling . we argue here that such inelastic tunneling mechanism is the same mechanism that is responsible for the thermalization of the nuclear spins . _ indeed , by extending the specific heat measurements below the ordering temperature , one even observes the equilibrium specific heat contribution of the nuclear spins , meaning that both the electron spins and the nuclear spins can attain thermal equilibrium within the time - scale of the specific heat experiment ( 10 - 100 s ) at all temperatures reached . all the work presented here is dedicated to the small-@xmath215 , incoherent tunneling regime for the central spin . having shown that the description of the nuclear - spin mediated tunneling is incomplete without the inclusion of spin - phonon couplings , some concerns may be raised also on the current description of the spin bath effects on the electron spin in the large-@xmath215 , coherent tunneling regime.@xcite this work can not address that issue , and we think that the answer will have to come from low-@xmath3 nmr experiments in large transverse field , and _ quantitative _ analysis of pulsed - esr experiments . the purpose of the research presented here is to illustrate and analyze a prototypical example of quantum tunneling of a macroscopic variable ( the giant spin of single - molecule magnet ) in the presence of a spin bath environment . instead of looking at the macroscopic variable itself and deducing the effect of the environment on its dynamics , as is most often done , we have directly observed the behavior of the spin bath by means of low - t nmr experiments . we have provided compelling evidence that the longitudinal nuclear spin relaxation in the @xmath2mn nuclei of mn@xmath0-ac is driven by electron - spin quantum tunneling fluctuations . the nuclear lsr rate , @xmath76 , indeed contains all the features that are expected to be associated with tunneling of the molecular spin : i ) a @xmath3-independent plateau of the lsr rate for @xmath216 k ; ii ) a strong dependence of @xmath76 on a longitudinal magnetic field , that destroys the resonance condition for electron spin tunneling ; iii ) the slowing down of the nuclear lsr upon isotropic substitution of @xmath6h by @xmath7h in the ligands , by an amount identical to the slowing down of the quantum relaxation of the magnetization observed in similar systems . because of the short timescale of the observed lsr , we argued that the tunneling fluctuations must take place in a minority of fast - relaxing molecules , which are indeed known to be present in mn@xmath0-ac . for these fluctuations to relax the nuclear magnetization in the entire sample , an additional mechanism is required which equilibrates the nuclear spin polarization across neighboring molecules , i.e. intercluster nuclear spin diffusion . our data on the transverse nuclear spin relaxation show that the intercluster spin diffusion is indeed present and effective . all the above observations confirm and support the picture of nuclear - driven quantum tunneling of magnetization as originally formulated by prokofev and stamp . however , a crucial outcome of our experiments is the demonstration that the nuclear spins are in thermal equilibrium with the lattice phonons down to the lowest temperatures , where only quantum tunneling fluctuations of the electron spins are still present . this observation can not be explained within the present theory of the spin bath . the implications of our results are potentially very profound , particularly because of the growing interest toward a coherent manipulation of spins for quantum information processing . the spin - bath environment , describing localized two - level systems , has been repeatedly identified as the most important source of decoherence in solid - state qubits . this includes superconducting systems,@xcite quantum dots,@xcite nv centers in diamond@xcite and , of course , molecular magnets.@xcite we have investigated here the incoherent tunneling regime , but the theoretical formalism to describe the coupling between central spin and spin bath is identical in the case of coherent spin dynamics . therefore , the main finding of our work - that the role of phonons in the nuclear - spin mediated tunneling is currently lacking a proper description - suggests that also the contribution of the nuclear spin bath to the decoherence rate of realistic spin qubits may need to be revisited . we are indebted to o. n. bakharev , h. b. brom , d. bono , n. j. zelders and g. frossati for experimental help and extensive discussions . continuous and illuminating theoretical support from p. c. e. stamp and i. s. tupitsyn is gratefully acknowledged , and so are discussions with w. wernsdorfer , s. hill , n. v. prokofev , b. v. fine , m. evangelisti , a. j. leggett , y. imry , m. schechter and a. l. burin . we also thank k. awaga and k. takeda for useful correspondence about their results in ref . . the mn@xmath0-ac samples were supplied by r. sessoli and a. caneschi ( crystallites , natural and deuterated ) and a. millan ( single crystal ) . + this work is part of the research program of the `` stichting fom '' and is partially funded by the ec - rtn `` quemolna '' and ec - network of excellence `` magmanet '' ( no.515767 - 2 ) . |
for , let denote the infinite planar sector of opening , and be the laplacian in , , with the robin boundary condition , where stands for the outer normal derivative and . the essential spectrum of does not depend on the angle and equals , and the discrete spectrum is non - empty iff . in this case we show that the discrete spectrum is always finite and that each individual eigenvalue is a continous strictly increasing function of the angle . in particular , there is just one discrete eigenvalue for . as approaches , the number of discrete eigenvalues becomes arbitrary large and is minorated by with a suitable , and theth eigenvalue of behaves as and admits a full asymptotic expansion in powers of . the eigenfunctions are exponentially localized near the origin . | 1607.06848 | for @xmath0 , let @xmath1 denote the infinite sector of opening @xmath2 , @xmath21 in the present paper , we are interested in the spectral properties of the associated robin laplacian , to be denoted @xmath4 , which is defined as follows : for @xmath9 , the operator @xmath4 acts in @xmath5 as @xmath22 on the functions @xmath23 satisfying the boundary condition @xmath24 where @xmath25 stands for the outer unit normal . more precisely , @xmath26 is defined as the unique self - adjoint operator in @xmath5 corresponding to the quadratic form @xmath27 where @xmath28 is the one - dimensional hausdorff measure . during the last years , the spectral analysis of robin laplacians attracted a considerable attention . as shown in @xcite , the spectral properties of such operators are sensible to the regularity of the boundary , and the case of smooth domains has been studied in detail @xcite . on the other hand , only partial results are available for domains with a non - smooth boundary , cf . @xcite . the study of the above operator @xmath4 has several interesting aspects from the point of view of the existing results . first , it can be viewed as the simplest non - smooth domain in two dimensions and depending in an explicit way of the single geometric parameter @xmath10 . second , its spectral properties play an important role in the study of more general non - smooth domains in the strong coupling limit , see @xcite , as the sectors @xmath1 exhaust the whole family of possible tangent cones to the boundary in two dimensions . third , the domain @xmath1 and its boundary are non - compact , which may potentially lead to quite unusual spectral properties , such as the presence of an infinite discrete spectrum , and the respective studies in higher dimensions @xcite do not extend directly to the planar case : we will see below that the results are actually quite different . in fact , the only spectral result on @xmath4 available in the existing literature is as follows , see @xcite : @xmath29 and for @xmath30 the value indicated is an eigenvalue with an explicitly known eigenfunction @xmath31 . the aim of the present work is to provide a more detailed spectral analysis . our main results are as follows . first , as the essential spectrum of @xmath4 does not depend on the angle @xmath10 and equals @xmath11 , see theorem [ spec ] , it follows from that the discrete spectrum is non - empty if and only if @xmath30 , i.e. if and only if the sector is strictly smaller than the half - plane . in theorem [ thm31 ] we show that the discrete spectrum is always finite , which is a non - trivial result due to the non - compactness of the boundary . in subsection [ monotone ] we obtain more detailed results : in theorem [ increase ] we show that each individual eigenvalue is a strictly increasing continuous function of the angle @xmath10 and , moreover , that there is just one discrete eigenvalue for @xmath13 , see theorem [ thmpi6 ] . in section [ small1 ] we discuss the behavior of the discrete eigenvalues for small @xmath10 . we show that the @xmath17th eigenvalue @xmath18 behaves as @xmath32 see corollary [ corol43 ] , and , moreover , admits an asymptotic expansion up to any order with respect to the powers of @xmath20 , see theorem [ full ] . the number of discrete eigenvalues becomes arbitrary large and is minorated by @xmath15 , @xmath16 , as @xmath10 approaches @xmath14 , see corollary [ cor42 ] . in theorem [ agmon ] we show that the associated eigenfunctions are localized , in a suitable sense , near the vertex . our proofs are mostly variational and based on the min - max characterization of the essential spectrum and the eigenvalues . the proof of the finiteness of the discrete spectrum uses an idea proposed in @xcite for a different operator involving a similar geometry , while the continuity and the monotonicity of the eigenvalues are established using a suitable change of variables . the asymptotics for small @xmath10 is based on the well - known born - oppenheimer strategy @xcite and , similar to various problems involving small parameters @xcite , on a reduction to a one - dimensional effective operator , which in our case acts in @xmath33 as @xmath34 with a suitable boundary condition at the origin , while an additional work is required due to the singularity of the potential . we remark that the use of an improved hardy inequality @xcite allows one to perform the reduction in a rather direct way , without any preliminary localization argument for the associated eigenfunctions . the properties of the eigenfunctions are studied with the help of the standard agmon - type approach using a suitable decomposition of the domain . we remark that in the present paper we are not discussing the ( non-)existence of eigenvalues embedded into the continuous spectrum . some partial information can be easily obtained , for example , the recent rellich - type result @xcite implies the absence of positive embedded eigenvalues for @xmath35 , but a specific separate study is needed in order to cover all possible cases . this will be discussed elsewhere . for @xmath9 denote by @xmath36 the self - adjoint operator acting in @xmath33 by @xmath37 one easily checks that @xmath38 and that the unique eigenvalue is @xmath39 with @xmath40 the associated eigenfunction . remark that the quadratic form for @xmath36 is @xmath41 hence @xmath42 for @xmath43 and @xmath44 , consider the operator @xmath45 in @xmath46 acting by @xmath47 remark that the associated quadratic form @xmath48 is @xmath49 the operator @xmath45 has a compact resolvent and its spectrum is purely discrete and consists of the simple eigenvalues @xmath50 , @xmath51 , numbered in the increasing order . remark that due to the scaling we have @xmath52 the maps @xmath53 are real - analytic , in particular @xmath54 coincides with the @xmath55th neumann eigenvalue of @xmath56 , hence @xmath57 by ( * ? ? ? * proposition a.3 ) the following assertions hold true : * for any @xmath9 there holds @xmath58 , and the associated eigenfunction is @xmath59 * the inequality @xmath60 holds if and only if @xmath61 . * for large @xmath62 there holds @xmath63 let us introduce some quantites to be used later in the text . for @xmath64 , denote @xmath65 then we have @xmath66 and a simple direct computation gives @xmath67 the first eigenfunction @xmath68 of @xmath69 , chosen positive and normalized , is then given by @xmath70 for any fixed @xmath62 and @xmath71 the functions @xmath72 and @xmath73 are smooth , and , by direct computation , @xmath74 moreover , as @xmath75 is analytic and simple , we can compute its first derivative . consider the implicit equation satisfied by @xmath76 and @xmath77 : @xmath78 we take the derivative of with respect to @xmath79 and , after multiplying by @xmath76 , we integrate it over @xmath80 . after two integrations by part we obtain the following equality @xmath81_{-l}^l + \big[\big(\partial_{\gamma}\phi_{l,\gamma}(t)\big)\big(\partial_t\phi_{l,\gamma}(t)\big)\big]_{-l}^l \\ = \partial_\gamma e_1(l,\gamma ) \int_{-l}^l\lvert \phi_{l,\gamma}(t)\rvert^2 dt + e_1(l,\gamma ) \int_{-l}^l \partial_\gamma\phi_{l,\gamma}(t ) \phi_{l,\gamma}(t ) dt.\end{gathered}\ ] ] we now take the derivative of with respect to @xmath79 and we get @xmath82 after replacing these expressions in we finally have @xmath83 in particular , @xmath84 furthermore , due to the preceding consideration , the following representation is valid : [ prop2.5 ] there exists @xmath85 such that @xmath86 for all @xmath87 . define @xmath88 by @xmath89 . due to and to the equality @xmath90 , the function @xmath91 is bounded at infinity . for @xmath79 near @xmath14 one has , due to and and due to the analyticity , @xmath92 , which shows that @xmath91 is bounded near @xmath14 . as @xmath91 is continuous , the result follows . recall first the min - max principle giving a variational characterization of eigenvalues . let @xmath93 be a self - adjoint operator acting in a hilbert space @xmath94 of infinite dimension . we assume that @xmath93 is sembounded from below , @xmath95 , @xmath96 , and denote @xmath97 by @xmath98 we denote its @xmath55th eigenvalue when ordered in the non - decreasing order and counted with multiplicities . the domain and the form domain of @xmath93 will be denoted by @xmath99 and @xmath100 , respectively , and by @xmath101 we denote the associated sesquilinear form . recall that @xmath100 is a hilbert space when considered with the scalar product @xmath102 . the following result is a standard tool of the spectral theory , see e.g. ( * ? ? ? * section xiii.1 ) . let @xmath103 and @xmath104 be a dense subset of the hilbert space @xmath100 . let @xmath105 be the @xmath17th _ rayleigh quotient _ of @xmath93 , which is defined by @xmath106 then one and only one of the following assertions is true : 1 . @xmath107 and @xmath108 . 2 . @xmath109 and @xmath110 for all @xmath111 . in order to use the min - max principle for the robin laplacians we need the following density result , which is quite standard ( see appendix [ appa ] for the proof ) . [ density ] the set @xmath112 is dense in @xmath113 . we remark that most of the subsequent constructions are heavily based on estimates for the rayleigh quotients of various operators . the following proposition summarizes the known results on the spectra of the robin laplacians @xmath4 . to keep the presentation self - contained , we give a complete proof in appendix [ appa2 ] . [ spec ] for any @xmath114 and any @xmath9 the quadratic form @xmath115 given by is closed and semibounded from below , hence , the associated operator @xmath4 is self - adjoint in @xmath5 . furthermore , * @xmath116 for any @xmath114 . * if @xmath117 , then @xmath118 and @xmath119 is an associated eigenfunction . * for @xmath120 , the discrete spectrum of @xmath4 is empty . therefore , the discrete spectrum of @xmath4 is non - empty if and only if @xmath30 , _ which will be assumed in the rest of the paper_. our principal aim is to obtain a more detailed information on the number of discrete eigenvalues and on their behavior with respect to the angle @xmath10 . remark that the domain @xmath1 is invariant by dilations , hence , the operator @xmath4 is unitarily equivalent to @xmath121 . therefore , in what follows we restrict our attention to the operator @xmath122 our first objective is to show that the discrete spectrum of @xmath26 is finite . remark that the result is not trivial : as shown in @xcite , robin laplacians on dilation - invariant domains in @xmath123 , @xmath124 , may have an infinite discrete spectrum . later , in section [ monotone ] , we prove that the eigenvalues of @xmath26 are monotone continuous functions of @xmath10 . we start with a decomposition of @xmath26 due to the symmetry of the sector . consider the upper half @xmath125 of @xmath1 , @xmath126 , and the unitary map @xmath127 by direct computation , for @xmath128 one has @xmath129 with @xmath130 and @xmath131 is given by the same expression but acts on the smaller domain @xmath132 and @xmath133 and @xmath134 , where @xmath135 is the projection onto the @xmath55the component . hence , if @xmath136 and @xmath137 are the self - adjoint operators acting in @xmath138 and associated with @xmath139 and @xmath131 , respectively , then , by construction , @xmath140 , and it is sufficient to study separately the spectra of @xmath137 and @xmath136 . let us show first that @xmath141 to see this , consider the half - plane @xmath142 and remark that if one takes @xmath143 and denotes by @xmath144 its extension by zero to @xmath145 , then @xmath146 where @xmath147 if @xmath148 is the self - adjoint operator associated with @xmath149 and acting in @xmath150 , then @xmath151 . on the other side , by applying a rotation one sees that @xmath148 is unitarily equivalent to @xmath152 acting in @xmath153 , where @xmath154 is defined in subsection [ halfline ] and @xmath62 is the free laplacian in @xmath155 . in particular , @xmath156 , which proves . therefore , we have @xmath157 furthermore , in view of theorem [ spec ] we have @xmath158 in particular , @xmath159 and the eigenvalue multiplicities are preserved . it also follows that all eigenfunctions of @xmath26 associated with the discrete eigenvalues are even with respects to @xmath160 . [ thm31 ] the discrete spectrum of @xmath26 is finite for any @xmath161 . in view of , it is sufficient to show that the operator @xmath162 has only a finite number of eigenvalues in @xmath163 . during the proof , if @xmath164 is a self - adjoint operator associated to a semi - bounded from below quadratic form @xmath165 and @xmath166 , we denote by @xmath167 or @xmath168 the number of the eigenvalues ( counting the multiplicities ) of @xmath164 in @xmath169 for @xmath170 , and set @xmath171 otherwise . the proof scheme is inspired by ( * ? ? ? * theorem 2.1 ) . the idea is to perform a dimensional reduction in order to compare the operator with a one - dimensional one and to conclude using a bargmann - type estimate . we first introduce a decomposition of @xmath172 . let @xmath173 and @xmath174 be smooth real - valued functions defined on @xmath175 such that @xmath176 for @xmath177 , to be determined later , consider the functions @xmath178 and @xmath179 defined on @xmath125 by @xmath180 and the following subdomains of @xmath172 , @xmath181 see fig [ fig : covering ] . we get easily @xmath182 for all @xmath183 , so that @xmath184 where @xmath185 covering of @xmath172 by the domains @xmath186 ( surrounded by the dash line ) and @xmath187 ( surrounded by the solid line ) . the support of @xmath188 is hatched.,scaledwidth=60.0% ] we define the following quadratic forms : @xmath189 due to @xmath190 the min - max principle and give for for any @xmath191 : @xmath192 where @xmath193 and @xmath194 are the self - adjoint operators acting respectively in @xmath195 and @xmath196 and produced by the forms @xmath197 and @xmath198 . then , @xmath199 let us first estimate @xmath200 . we consider the following two domains : @xmath201 and the quadratic forms @xmath202 by the min - max principle we have @xmath203 on one hand , the operator @xmath204 associated to @xmath205 has a compact resolvent , which implies @xmath206 on the other hand , the operators @xmath207 associated to @xmath208 can be represented as @xmath209 where @xmath210 is the neumann laplacian in @xmath211 and @xmath212 is the self - adjoint operator in @xmath213 associated with the quadratic form @xmath214 one has @xmath215 and @xmath216 . setting @xmath217 and taking @xmath218 implies @xmath219 and then @xmath220 and we conclude by , and that @xmath221 , for all @xmath222 . now let us estimate @xmath223 for @xmath218 . let us introduce the quadratic form @xmath224 where @xmath145 is the half - plane given by , then @xmath225 by the min - max principle . if we make an anti - clockwise rotation of angle @xmath226 of @xmath145 , then we obtain the half - plane @xmath227 , and the operator @xmath228 associated with @xmath229 is then unitarily equivalent to the operators @xmath230 associated with the quadratic form @xmath231 where @xmath232 one has , for @xmath233 , @xmath234 where the operator associated to @xmath235 is @xmath236 with @xmath154 defined in subsection [ halfline ] and @xmath62 the free laplacian in @xmath155 . let us consider the orthogonal projections @xmath237 and @xmath238 in @xmath239 , @xmath240 remark that @xmath241 , where @xmath242 is the spectral projector of @xmath154 on @xmath243 . for @xmath244 there holds @xmath245 , and @xmath246 as @xmath247 by the spectral theorem . writing @xmath248 we have @xmath249 by the spectral theorem applied to @xmath154 , for a.e . @xmath250 one has , @xmath251 and , finally , @xmath252 for any @xmath253 one can estimate @xmath254 then , using the equality @xmath255 , we get @xmath256 where @xmath257 we can choose @xmath258 and @xmath259 such that @xmath260 , then for @xmath261 one arrives at @xmath262 we introduce the quadratic form @xmath263 then , by and the min - max principle we have @xmath264 in order to show that the number of negative eigenvalues of @xmath265 is finite , we want to use a bargmann - type estimate , see e.g. ( * ? ? ? * eq.(8 ) ) : @xmath266 we can write , using the fact that @xmath267 $ ] , @xmath268 using the boundedness of @xmath269 we finally get the following upper bound : @xmath270 hence , @xmath271 and implies that @xmath272 for @xmath261 . by we conclude that @xmath273 . let us discuss first the monotonicity of the rayleigh quotients of @xmath26 with respect to @xmath10 . [ prop - mon1 ] for any @xmath191 the function @xmath274 is non - decreasing and continuous . in view of the constructions of subsection [ parity ] it is sufficient to show the result for the rayleigh quotients @xmath275 of @xmath162 . we denote by @xmath276 the quadratic form obtained after the anti - clockwise rotation of angle @xmath226 of @xmath172 . then @xmath276 is unitarily equivalent to @xmath277 and we have @xmath278 for all @xmath279 , where @xmath280 . after the scaling @xmath281 and writing @xmath282 we have @xmath283 and @xmath284 . then , we can define @xmath285 for @xmath286 and @xmath287 the associated operator in @xmath288 . by construction , we have @xmath289 . the dependence of @xmath287 on @xmath10 only appears through the coefficient @xmath290 , which gives the result due to the min - max principle . let us now obtain a stronger result for the eigenvalues . [ increase ] assume that for some @xmath291 and @xmath292 the operator @xmath293 has at least @xmath17 discrete eigenvalues , then @xmath26 has at least @xmath17 discrete eigenvalues for all @xmath294 , and the function @xmath295 is strictly increasing . if @xmath293 has at least @xmath17 discrete eigenvalues , then by the min - max principle one has @xmath296 , which by proposition [ prop - mon1 ] implies @xmath297 for all @xmath294 , hence @xmath298 is the @xmath17th discrete eigenvalue of @xmath26 by the min - max principle . the weak monotonicity and continuity also follow from proposition [ prop - mon1 ] . let us show the strict monotonicity of the eigenvalues . let @xmath299 , @xmath300 such that @xmath301 and @xmath302 . we continue using the notation of the proof of proposition [ prop - mon1 ] , then @xmath303 , and we need to show the strict inequality @xmath304 . for all @xmath305 one has , with @xmath306 , @xmath307 let @xmath308 be an orthonormal basis in @xmath309 . on one hand , for all @xmath310 we have @xmath311 on the other hand , using the min - max principle and we have @xmath312 assume that @xmath313 , then the second term on the right - hand side is zero . as the unit ball of @xmath314 is compact , there must exist @xmath315 with @xmath316 such that @xmath317 i.e. @xmath318 . then @xmath319 depends on the @xmath320 variable only , but as @xmath321 we necessarily have @xmath322 , which contradicts the normalization @xmath316 . another important corollary is as follows : [ corol123 ] assume that for some @xmath323 the operator @xmath324 has a unique discrete eigenvalue , then @xmath26 has a unique discrete eigenvalue for all @xmath325 . a natural candidate for @xmath323 is @xmath326 . in fact , the respective operator @xmath327 admits a separation of variables and is unitarily equivalent to @xmath328 with @xmath154 given in subsection [ halfline ] . hence , @xmath327 has a unique discrete eigenvalue @xmath329 , which shows [ oneeigenv ] for @xmath330 , the operator @xmath26 admits a unique discrete eigenvalue . in fact , we can obtain a better estimate : [ thmpi6 ] the operator @xmath26 has a unique discrete eigenvalue for @xmath331 . in view of corollary [ corol123 ] it is sufficient to consider @xmath332 . we continue using the domain @xmath333 from the proof of proposition [ prop - mon1 ] . let @xmath334 be the laplacian in @xmath335 with the robin boundary condition @xmath336 at @xmath337 , the neumann boundary condition et the line @xmath338 and with the neumann boundary condition at the both sides of the lines @xmath339 and @xmath340 , then by the min - max principle for any @xmath43 and @xmath341 one has the inequality @xmath342 let us argue by contradiction . assume that @xmath343 , then it follows from that @xmath344 remark that @xmath345 , where @xmath346 is the laplacian in the triangle @xmath347 with the robin boundary condition at @xmath337 and with the neumann boundary condition on the other two sides , the operator @xmath348 in the laplacian in the half - strip @xmath349 with the robin boundary condition at @xmath337 and with the neumann boundary condition at the remaining part of the boundary , and @xmath350 is the neumann lalplacian in @xmath351 , see fig.[prop36](a ) . [ cols="^,^ " , ] the spectrum of @xmath348 can be easily estimated using the separation of variables , and then @xmath352 for @xmath353 . furthermore , @xmath354 , and eq . implies @xmath355 remark that each eigenfunction of @xmath346 can be extended , using the reflections with respect to the neumann sides , to an eigenfunction of the laplacian @xmath356 with the robin boundary condition @xmath336 on the equilateral triangle @xmath357 composed from six copies on @xmath358 , see figure [ prop36](b ) , where @xmath359 is an equilateral triangle of side length @xmath360 . therefore , @xmath361 , where @xmath362 is the restriction of @xmath356 to the functions which are invariant under the reflections with respect to the medians and with respect to the rotations by @xmath363 . the eigenvalues of @xmath356 in the limit @xmath353 were analyzed in ( * ? ? ? * section 7 ) . in particular , the first eigenfunction of @xmath356 has the above - mentioned symmetries , hence , @xmath364 . on the other hand , the second eigenvalue of @xmath356 is double - degenerate , and no associated eigenfunction has the required symmetries : there is just one eigenfunction , noted @xmath365 in ( * ? ? ? * section 7 ) , which is even with respect to one of the medians , but it does not possess the other symmetries . therefore , @xmath366 for large @xmath62 . on the other hand , it is shown in ( * ? ? ? * subsection 7.4 ) that @xmath367 for @xmath368 , which contradicts . this contradiction shows that the inequality @xmath369 is not possible , and @xmath370 has a unique discrete eigenvalue . the present section is devoted to the study of the discrete spectrum of @xmath26 as @xmath10 tends to @xmath14 . we are going to show first the following result giving an estimate for the rayleigh quotients of @xmath26 : [ thprinc ] there exists @xmath371 and @xmath372 such that for all @xmath373 and for all @xmath191 there holds @xmath374 before passing to the proof let us discuss the most important consequences . recall that @xmath375 is the number of discrete eigenvalues of @xmath26 , and it is finite in virtue of theorem [ thm31 ] . [ cor42 ] there exists @xmath16 such that @xmath376 as @xmath10 is small . in particular , @xmath375 tends to @xmath377 as @xmath10 tends to @xmath14 . for all @xmath373 and any @xmath191 one has @xmath378 by the min - max principle , the number @xmath298 is the @xmath17th eigenvalue iff it is strictly less than @xmath379 . notice that the right - hand side is smaller than @xmath379 for all @xmath380 , @xmath381 and then @xmath375 is not smaller than the integer part of @xmath382 . another obvious corollary is [ corol43 ] for any @xmath191 there holds @xmath383 in theorem [ full ] below we show a stronger result that @xmath384 admits a full asymptotic expansions in powers of @xmath20 . the main idea of the proof is to compare the operator @xmath26 with some one - dimensional operator . namely , for @xmath385 consider the following operator acting in @xmath33 : @xmath386 and @xmath387 be the associated quadratic form , @xmath388 denote by @xmath389 the friedrichs extension of @xmath390 . it is known ( see appendix [ appb ] ) that its essential spectrum equals @xmath391 and that the discrete spectrum consists of the simple negative eigenvalues @xmath392 while the respective eigenfunctions @xmath393 are @xmath394 with @xmath395 being the laguerre polynomials . denote @xmath396 and define a unitary operator @xmath397 , @xmath398 recall that the set @xmath399 was defined in lemma [ density ] . we have then @xmath400 we define the new quadratic form @xmath401 , @xmath402 defined initially on @xmath403 . as the set @xmath399 is dense in the form domain of @xmath26 , the operator @xmath287 corresponding to the closure of @xmath404 writes as @xmath405 , and @xmath289 for all @xmath191 . recall that the numbers @xmath406 are defined in subsection [ interval ] . the next proposition is a direct application of the min - max principle . for all @xmath407 we have @xmath408 let @xmath407 , then @xmath409 we recognize in the bracket the quadratic form associated to the robin laplacian @xmath410 defined in subsection [ interval ] . as @xmath411 , we have @xmath412 . then we can apply the min - max principle to obtain @xmath413 the conclusion is due to the equality . the operator @xmath410 defined in subsection [ interval ] will play a special role . recall that @xmath414 , and the associated normalized eigenfunction is @xmath415 in this section , we will omit the lower indices and denote @xmath416 define two orthogonal projections @xmath237 and @xmath238 in @xmath417 by @xmath418 during the proof , the functions @xmath419 and @xmath420 will always be related by . [ prop4.5 ] for all @xmath407 we have @xmath421 , and @xmath422 where @xmath423 . let @xmath407 , then @xmath424 . as @xmath425 is smooth , one has @xmath426 . moreover , @xmath427 the evaluation of @xmath404 on @xmath428 gives @xmath429 by , @xmath430 for all @xmath431 , @xmath432 , then @xmath433 applying fubini s theorem we obtain @xmath434 and the expression in the curly brackets equals @xmath435 due to the choice of @xmath425 . then , @xmath436 to finish the proof we use . in order to obtain an upper bound for the form @xmath404 we need to study the quantity @xmath437 . [ majork ] there exists @xmath438 such that for all @xmath439 and @xmath440 we have @xmath441 . we first notice that @xmath437 is continuous on @xmath175 . in addition , @xmath442 then , @xmath443 we also give an upper bound for the second term @xmath444 : @xmath445 finally , we arrive at @xmath446 with @xmath447 in order to conclude it is sufficient to show that @xmath448 is bounded on @xmath175 . as the function @xmath448 is continuous , we only have to prove that it admits finite limits in @xmath14 and @xmath377 . let @xmath449 . the function @xmath450 is @xmath451 with @xmath452 , @xmath453 for @xmath454 , and @xmath455 furthermore , we have by @xmath456 then we can write @xmath457 , where @xmath458 after a direction computation using the behavior of @xmath459 as @xmath460 and @xmath454 we get @xmath461 then @xmath448 admits finite limits too , which concludes the proof . [ cor4.9 ] for all @xmath462 and for all @xmath463 , @xmath464 the next proposition gives an upper bound for the rayleigh quotients . [ upperbound ] there exist @xmath465 and @xmath466 such that @xmath467 for all @xmath191 and all @xmath373 . by proposition [ prop2.5 ] and corollary [ cor4.9 ] , for all @xmath462 there holds @xmath468 let @xmath469 , then , @xmath470 let @xmath471 and @xmath472 such that @xmath473 . denote @xmath474 , then @xmath475 and @xmath476 . by we obtain @xmath477 and by min - max principle , @xmath478 recall that @xmath479 is dense in @xmath480 as @xmath481 is the friedrichs extension of @xmath390 ( see subsection [ op1d ] ) , hence , @xmath482 which concludes the proof . here we will still use the orthogonal projections @xmath237 and @xmath238 defined in . we recall that for @xmath462 we have @xmath426 , @xmath483 , and @xmath484 . [ sbb1 ] for all @xmath407 and for all @xmath440 , @xmath485 with the constant @xmath164 from lemma [ majork ] . let @xmath462 , then @xmath486 the first term is known thanks to proposition [ prop4.5 ] . then we have @xmath487 applying the spectral theorem to the operator @xmath410 we obtain @xmath488 using the equality @xmath489 we finally get @xmath490 to estimate the last term in we write @xmath491 the functions @xmath428 and @xmath492 are orthogonal with respect to the scalar product of @xmath493 , @xmath494 and with respect to the form @xmath495 , i.e. @xmath496 then becomes @xmath497 in addition , @xmath498 , and the derivative in @xmath499 gives @xmath500 the substitution into gives : @xmath501 to get a lower bound , we notice that @xmath502 and using lemma [ majork ] we obtain @xmath503 furthermore for any @xmath504 we have @xmath505 and using again lemma [ majork ] and @xmath506 we get @xmath507 the substitution of and into gives us the lower bound for the cross - term . combining it with finally gives us the following lower bound for @xmath404 : @xmath508 we notice that @xmath437 is positive , which concludes the proof . the next proposition gives us a lower bound of @xmath404 in terms of the one - dimensional operator @xmath389 defined in subsection [ op1d ] . [ sbb2 ] there exists @xmath509 such that for all @xmath510 and for all @xmath407 , @xmath511dr + k \lvert v \rvert ^2_{l^2(v_\alpha)}.\ ] ] in order to prove proposition [ sbb2 ] , we will need some preliminary constructions . let us introduce a new quadratic form , @xmath512 where @xmath513 [ a1sbb ] there exists @xmath514 such that , for all @xmath515 and for all @xmath516 , @xmath517 it is sufficient to check that , for all @xmath518 , the potential @xmath519 is uniformly semi - bounded from below on @xmath175 . notice that @xmath520 with @xmath521 clearly , @xmath522 is continuous on @xmath175 . in addition , by we have @xmath523 as @xmath454 . furthermore , @xmath524 by . hence , @xmath525 as @xmath526 , and we can conclude that @xmath522 admits a finite lower bound on @xmath175 . then , by , there exists @xmath514 such that @xmath527 for all @xmath518 . introduce another quadratic form , @xmath528 dr , \quad u\in c^\infty_c(\mathbb r_+).\ ] ] [ a2sbb ] there exists @xmath529 such that for all @xmath515 and for all @xmath530 , @xmath531dr + c_2 \lvert u\rvert^2_{l^2(\mathbb r_+)}.\ ] ] using the function @xmath91 from proposition [ prop2.5 ] we can write , for all @xmath530 , @xmath532dr -\|\phi\|_\infty \lvert u\rvert ^2_{l^2(\mathbb r_+)}.\ ] ] we want to study the right hand side of this inequality . to do that we separate the integral in two parts : a first integral on @xmath533 and a second one on @xmath534 . on one hand , we notice that @xmath535 and , by an improved hardy type inequality from ( * ? ? ? * lemma a.1 ) , as @xmath536 , @xmath537 then we have @xmath538dr,\ ] ] and we can write @xmath539dr \\ \geq ( 1-\alpha^2)\int_0 ^ 1 \lvert u'(r)\rvert^2 dr+ \alpha^2 \int_0 ^ 1 \left(\frac{1}{4r^2}+\frac{1}{4r}\right)\lvert u(r)\rvert^2 dr + \int_0 ^ 1 \left(-\frac{1}{4r^2}-\frac{1}{r\alpha } \right)\lvert u(r)\rvert^2 dr.\end{gathered}\ ] ] finally , @xmath540 dr \\ \geq ( 1-\alpha^2)\int_0 ^ 1 \bigg [ \lvert u'(r)\rvert^2 -\left(\frac{1}{4r^2}+\frac{1}{\alpha(1-\alpha^2)r}\right ) \lvert u(r)\rvert^2 \bigg]dr.\end{gathered}\ ] ] on the other side , the integrals on @xmath534 can be estimated by @xmath541 dr \\ \geq ( 1-\alpha^2 ) \int_1^{+\infty } \bigg[\lvert u'(r)\rvert ^2 -\left(\frac{1}{4r^2(1-\alpha^2 ) } + \frac{1}{r\alpha(1-\alpha^2 ) } \right ) \lvert u(r)\rvert^2 \bigg ] dr .\end{gathered}\ ] ] using @xmath542 we obtain @xmath543dr \\ \geq ( 1-\alpha^2 ) \int_1^{+\infty } \bigg[\lvert u'(r)\rvert^2 -\left(\frac{1}{4r^2}+\frac{1}{r\alpha ( 1-\alpha^2)}\right ) \lvert u(r)\rvert^2 \bigg]dr -\frac{\alpha^2}{4 } \lvert u \rvert ^2_{l^2(\mathbb r_+)}.\end{gathered}\ ] ] putting and together we obtain the result whith @xmath544 . let @xmath462 , then @xmath545 and @xmath546 , and , by proposition [ sbb1 ] , @xmath547 we can conclude thanks to lemmas [ a1sbb ] and [ a2sbb ] and by writing @xmath548 . [ lowerbound ] there exists @xmath549 and @xmath550 such that , for all @xmath373 and for all @xmath191 , @xmath551 the proof is again based on the use of the min - max principle , and we need a preliminary assertion . [ diag ] we define , for all @xmath552 , the quadratic form @xmath553 dr,\ ] ] and let @xmath554 be the self - adjoint operator associated with its closure in @xmath555 . then @xmath556 for all @xmath191 . let @xmath471 . by the min - max principle , @xmath557 moreover , thanks again to te min - max principle , we also have : @xmath558 as @xmath559 for all @xmath191 , one has @xmath560 , and @xmath561 namely @xmath562 . we are now able to give the proof of the lower bound of the eigenvalues . let @xmath191 . by proposition [ sbb2 ] and the min - max principle we have @xmath563 consider the map @xmath564 , where @xmath420 and @xmath492 are defined in . then , @xmath565 thanks to lemma [ diag ] . as @xmath566 we can estimate @xmath567 now we are able to finish the proof of theorem [ thprinc ] . thanks to proposition [ upperbound ] and [ lowerbound ] , there exist @xmath568 , @xmath569 and @xmath570 such that , for all @xmath373 and for all @xmath191 one has @xmath571 . taking @xmath572 we arrive at the result . theorem [ thprinc ] and corollary [ corol43 ] give a first order asymptotics for the eigenvalues . in particular , it follows that each discrete eigenvalue is simple as the angle is small . this can be used to apply the standard perturbation theory to obtain a full asymptotic expansion . [ full ] for any @xmath471 there exist @xmath573 , @xmath574 , such that for any @xmath575 one has the asymptotics @xmath576 and @xmath577 . let us consider the operator @xmath287 acting on @xmath578 and defined in section [ polarcoord ] : @xmath579 using the scaling , @xmath580 , @xmath581 , one shows that @xmath287 is unitarily equivalent to @xmath582 , where @xmath583 acts in @xmath584 by @xmath585 and is associated with the quadratic form @xmath586 for the eigenvalues one has @xmath587 , @xmath191 , and we prefer to work with @xmath583 in what follows . remark that , in view of theorem [ spec ] , for any @xmath9 one has @xmath588 in particular , there exist @xmath589 and @xmath590 such that for small @xmath10 there holds @xmath591 introduce the following differential expressions @xmath592 then @xmath593 . we look for a formal approximate eigenpair @xmath594 of @xmath583 of the form @xmath595 satisfying in the sense of formal series the following eigenvalue problem @xmath596 by collecting the terms according to the powers of @xmath10 , one arrives at an infinite system of partial differential equations . to determine @xmath597 we collect the terms containing @xmath598 , then we have to solve @xmath599 as a consequence , @xmath600 only depends on @xmath71 . collecting the terms corresponding to @xmath601 we get @xmath602 let us denote by @xmath603 the operator acting as @xmath604 on @xmath584 whose domain is @xmath605 we already know that @xmath606 and we notice that the orthogonal projections on @xmath607 and the differential expression @xmath608 commute . thus we can integrate on @xmath56 and obtain @xmath609 using the boundary condition , the equality becomes @xmath610 here , we recognize a one - dimensional differential operator @xmath611 defined in section [ op1d ] . hence , we are lead to choose @xmath612 where @xmath191 is fixed for the rest of the proof , and @xmath613 is the associated normalized eigenfunction , see appendix [ appb ] . we then can get an expression for @xmath614 rewritting as @xmath615 integrating it two times in @xmath616 and using the boundary condition we obtain @xmath617 where @xmath618 has to be determined in the next step . notice that the function @xmath619 belongs to @xmath33 as @xmath613 decays exponentially , see appendix [ appb ] . we now can give the proof of the existence of the further terms . let @xmath620 and suppose that @xmath621 and @xmath622 are known and satisfy @xmath623 and the functions @xmath624 and @xmath625 decay exponentially , @xmath626 . we want to determine @xmath627 . we first use the equation obtained by collecting the terms in @xmath628 , i.e. @xmath629 using and the hypotheses we have @xmath630 we integrate it two times in @xmath616 and we use the boundary condition to cancel the term corresponding to @xmath616 , then there exists @xmath631 such that @xmath632 we set @xmath633 notice that each @xmath634 , @xmath635 , is then in @xmath33 and decay exponentially due to the hypothesis on @xmath636 , @xmath637 , and @xmath638 . then can be written in the form @xmath639 we have now to determine @xmath640 and @xmath641 . let us consider the equation obtained after collecting the terms in @xmath642 : @xmath643 the integration of on @xmath56 with respect to @xmath616 and the boundary conditions give , @xmath644 this equation admits a solution @xmath645 iff the right hand side belongs to @xmath646 . thus , @xmath640 is uniquely determined by @xmath647 as @xmath641 satisfies the inhomogeneous equation , a standard application of the variation of constants shows that it is exponentially decaying , which concludes the construction of the formal asymptotics . now we are going to show that the above formal expression for @xmath648 provides an asymptotics for the eigenvalues of @xmath583 . now let us fix @xmath649 and consider the finite sums @xmath650 by the preceding constructions one has @xmath651 remark that @xmath652 does not belong to the domain of @xmath583 as it does not satisfy the boundary condition , but belongs to the form domain of @xmath583 , and for any @xmath653 one has , using the integration by parts , @xmath654 recall that @xmath655 . furthermore , without loss of generality we may assume that the constant @xmath656 in is such that @xmath657 for small @xmath10 and consider the associated shifted sesquilinear form @xmath658 using the preceding estimates and the cauchy - schwarz inequality one deduces from and that for any @xmath659 there holds , with some @xmath660 , @xmath661 as @xmath10 is sufficiently small . remark that for small @xmath10 the values @xmath662 and @xmath663 can be estimated as @xmath664 . on the other hand , @xmath665 for small @xmath10 , and it follows from that , with some @xmath666 , @xmath667 a simple application of the spectral theorem , see proposition [ propd1 ] in appendix [ appd ] , shows that @xmath668 by theorem [ thprinc ] , the only point of the spectrum of @xmath583 which can satisfy the above estimate is the @xmath17th eigenvalue @xmath669 . as @xmath670 is arbitrary , the result follows . let @xmath161 be fixed . the following proposition ( agmon - type estimate ) shows that the eigenfunctions of @xmath26 corresponding to the discrete eigenvalues are localized near the vertex of the sector . [ agmon ] let @xmath671 be a discrete eigenvalue of @xmath26 and @xmath672 be an associated eigenfunction , then for any @xmath673 one has @xmath674 let @xmath673 and @xmath43 . define @xmath675 . we are going to show first that there exists @xmath676 such that @xmath677 the proof follows essentially the same steps as ( * ? ? ? * proposition 2.8 ) showing a similar result for the lowest eigenvalue of corner domains . let @xmath173 and @xmath174 be smooth functions ob @xmath175 satisfying @xmath678 for @xmath679 , consider the functions @xmath680 defined on @xmath1 , @xmath681 . we get easily , for @xmath682 and for @xmath683 , @xmath684 in particular , there exists @xmath685 such that @xmath686 for @xmath687 we have @xmath688 , and leads to @xmath689 we are going to provide a lower bound for the first two terms in the bracket . as @xmath690 , we have immediatly by the min - max principle @xmath691 to estimate @xmath692 we introduce the domain @xmath693 , and let @xmath694 be the sector obtained by translation of vector @xmath695 of @xmath1 . define @xmath696 see figure [ dessinea2 ] , then @xmath697 . partition of @xmath698.,scaledwidth=40.0% ] consider the quadratic forms @xmath699 and denote by @xmath700 and @xmath701 the associated self - adjoint operators in @xmath702 and @xmath703 respectively . as @xmath704 and @xmath705 are unitarily equivalent and @xmath706 is non - negative , we have @xmath707 we define @xmath708 remark that , if one takes @xmath709 and denotes by @xmath144 its extension by zero to @xmath710 then , @xmath711 , where @xmath712 then @xmath713 , where @xmath714 is the self - adjoint operator associated to @xmath715 acting in @xmath716 . by applying an anti - clockwise rotation of angle @xmath226 one sees that @xmath714 is unitarily equivalent to @xmath717 , where @xmath718 is the ( robin - neumann ) laplacian acting in @xmath719 and defined on @xmath720 and @xmath62 is the free laplacian in @xmath155 . then , @xmath721 . by ( * lemma a.1 ) we have @xmath722 and there exist @xmath723 and @xmath724 such that for all @xmath725 we have @xmath726 finally , for all @xmath682 we have @xmath727 combining with and we get , for all @xmath682 and @xmath725 , @xmath728 denote @xmath729 . the functions @xmath730 and @xmath731 belong to @xmath732 and @xmath733 an integration by parts , see ( * ? ? ? * lemma 2.7 ) for details , gives @xmath734 as @xmath735 , we obtain by , for all @xmath725 , @xmath736 which can be transformed in virtue of into @xmath737 with @xmath738 as @xmath739 , we have @xmath740 . in addition , @xmath741 . therefore , one can find @xmath742 and @xmath743 such @xmath744 and @xmath745 that for all @xmath746 . furthermore , there exists @xmath747 such that @xmath748 for @xmath746 . for the same @xmath749 we can show that there exits @xmath750 such that @xmath751 for all @xmath746 . the inequality implies then @xmath752 and , finally , @xmath753 with @xmath754 . notice that , @xmath755 which gives . now let us pass from to . we have @xmath756 using the inequality @xmath757 with @xmath758 we get @xmath759 and then @xmath760 by combining with we arrive at @xmath761 as the right - hand side does not depend on @xmath62 , one can pass to the limit as @xmath353 using the monotone convergence , which gives the result . by standard arguments , @xmath762 is dense in @xmath113 . in order to prove lemma [ density ] it is then sufficient to show that any function @xmath763 can be approximated by the functions from @xmath399 in the norm of @xmath113 . let @xmath763 . pick a smooth function @xmath764 with @xmath765 and set , for small @xmath259 , @xmath766 finally , set @xmath767 . notice that @xmath768 is radial and then there exists @xmath769 sucht that @xmath770 . in addition , @xmath771 satisfies @xmath772 and @xmath773 . let us show that @xmath774 as @xmath775 . denote @xmath776 and @xmath777 , then @xmath778 and the right - hand side tends to @xmath14 as @xmath779 is small due to the dominated convergence . furthermore , @xmath780 and the first term tends to 0 by the dominated convergence . on the other hand , @xmath781 and @xmath782 the functions @xmath23 and @xmath783 are bounded , and we can get the following upper bound : @xmath784 { } 0,\ ] ] which concludes the proof . let @xmath9 , @xmath785 and @xmath786 . using we have @xmath787 writing @xmath788 we conclude that for all @xmath789 and for all @xmath790 we have @xmath791 which means that the trace can be extended to a bounded linear map from @xmath113 to @xmath792 . furthermore , the boundary term in @xmath115 is then infinitesimally small with respect to the gradient term , hence , @xmath115 is closed on @xmath113 due to the klmn theorem , see ( * ? ? ? * theorem x.17 ) . let us show that @xmath793 for @xmath30 . by we have the inequality @xmath794 . on the other hand , by the explicit computation , @xmath795 for @xmath796 . it follows that the lower bound is optimal and that the bottom of the spectrum is an eigenvalue , and @xmath600 is an associated eigenfunction . let us show that @xmath797 for @xmath798 . decompose @xmath1 into the following three pieces : @xmath799 then @xmath800 and we define , for @xmath801 , @xmath802 by the min - max principle and the inclusion @xmath803 we have @xmath804 where @xmath104 is the selft adjoint operator acting on @xmath5 associated to the quadratic form defined for @xmath805 by @xmath806 . we have then @xmath807 , where @xmath808 are the self - adjoint operator associated with @xmath809 and acting in @xmath810 . notice that @xmath811 is positive and @xmath812 and @xmath813 are unitarily equivalent and have the same spectrum . furthermore , @xmath812 is unitarily equivalent to @xmath814 , where @xmath210 is the neumann laplacian in @xmath33 and @xmath36 is defined in subsection [ halfline ] , which gives @xmath815 . therefore , @xmath816 , and gives the result . let us show the lower bound @xmath817 let @xmath818 with @xmath385 . we denote by @xmath819 the sector obtained after a translation of @xmath1 along the vector @xmath820 . let @xmath821 be the orhogonal projection of @xmath164 on the half - line @xmath822 and @xmath823 , and , in particular , @xmath824 for @xmath825 . in the same way , we define @xmath826 the orthogonal projection of @xmath164 on @xmath827 . consider the following four domains : @xmath828 see fig [ dessinspecess ] . clearly , @xmath829 . partition of @xmath1.,scaledwidth=40.0% ] define for @xmath830 the quadratic forms @xmath831 by the min - max principle , the inclusion @xmath832 implies the inequality @xmath833 where @xmath104 is the self - adjoint operator acting in @xmath5 and associated to the quadratic form defined for @xmath834 by @xmath835 . ( remark that we use the convention @xmath836 . ) we have then @xmath837 , where each @xmath808 is the self - adjoint operator acting in @xmath810 and associated to @xmath809 , and implies @xmath838 the operator @xmath811 has a compact resolvent , then its essential spectrum is empty and @xmath839 , and the operator @xmath840 is non - negative , hence , @xmath841 . the operator @xmath812 et @xmath813 are unitarily equivalent , so we have to study @xmath812 only . by applying an anticlockwise rotation by angle @xmath842 we see that @xmath812 is unitarily equivalent to the self - adjoint operator @xmath843 acting in @xmath844 and associated with the quadratic form @xmath845 clearly , @xmath846 , where @xmath847 is the robin - neumann laplacian acting in @xmath848 on the domain @xmath849 and @xmath210 is the neumann laplacian in @xmath33 . as @xmath850 , we have @xmath851 , and and an easy computation , see e.g. ( * ? ? ? * lemme a.1 ) , shows that @xmath852 as @xmath824 . hence , for any @xmath790 there exists @xmath853 such as @xmath854 as is valid for any @xmath43 , we have @xmath855 for any @xmath856 , which implies . due to the results of subsections [ b3 ] and [ b4 ] we have @xmath857 for any @xmath10 . in order to conclude the proof , it is sufficient to show the inclusion @xmath858 it is more convenient to work with the rotated sector @xmath859 and the associated robin laplacian @xmath860 in @xmath861 corresponding to the quadratic form @xmath862 clearly , @xmath860 is unitarily equivalent to @xmath4 . the proof of consists in finding , for each @xmath863 , a family of functions @xmath864 with @xmath865 then the result follows by the spectral theorem for self - adjoint operators . let @xmath866 satisfying @xmath867 for @xmath868 and @xmath869 for @xmath870 . for @xmath649 , define a function @xmath871 by @xmath872 where @xmath873 for @xmath874 and @xmath385 is arbitrary otherwise . the function @xmath875 is smooth , compactly supported and satisfies the robin boundary condition @xmath876 at @xmath877 , hence , it belongs to @xmath878 , and @xmath879 . easy estimates show that for large @xmath670 we have @xmath880 which gives . we consider the operator acting on @xmath882 defined by : @xmath883 and the associated quadratic form @xmath387 , @xmath884 some parts of the analysis of the operator @xmath4 are based on the spectral properties of the friedrichs extension @xmath389 of @xmath390 . various parts of the description and of the spectral analysis of @xmath389 are spread through the literature , and the present section we give a compact presentation of the necessary results . let @xmath885 be the adjoint of @xmath390 . recall the deficiency subspaces of @xmath390 are defined by @xmath886 and their dimensions @xmath887 are called the deficiency indices of @xmath390 . the following proposition gives us the existence of self - adjoint extensions of @xmath390 . the operator @xmath885 is given be the same differential expression as @xmath390 and acts on the domain @xmath888 the deficiency indices of @xmath390 are equal to @xmath360 , and @xmath390 admits self - adjoint extensions . denote by @xmath889 the set on the right - hand side of . let @xmath890 , then for all @xmath891 one has @xmath892 and @xmath893 in particular , @xmath894 . let @xmath895 satisfy @xmath896 , then , for all @xmath259 , we have @xmath897 and @xmath898 . due to @xmath899 we also have @xmath900 for all @xmath790 . this shows the inclusion @xmath901 . let us prove the reverse inclusion . let @xmath902 . after an integration by parts we have , for all @xmath790 and for all @xmath891 , @xmath903 notice that the boundary terms are well defined as @xmath904 are continuous . moreover , @xmath905 as @xmath906 . hence , in the limit @xmath775 we obtain @xmath907 for all @xmath891 . then @xmath908 and , finally , @xmath909 . recall that a symmetric operator admits self - adjoint extensions if and only if his deficiency indices are equal . in our case , @xmath910 commutes with the complex conjugation , which gives immediately @xmath911 , as @xmath912 and @xmath913 are mutually complex conjugate . it remains to determine the deficiency indices . the functions @xmath914 are the solutions to the differential equation @xmath915 represent @xmath916 , then the function @xmath917 is a solution to the whitakker s equation , @xmath918 whose linearly independent solutions are expressed in terms on the confluent hypergeometric functions , and the space of @xmath919 solutions is one - dimensional , see ( * ? ? ? * eq . 13.1.31 ) . to describe the self - adjoint extensions of the operator @xmath390 we use the boundary triple approach , see e.g. @xcite . the integration by parts gives @xmath920 the self - adjoint extensions of @xmath390 are restrictions of @xmath910 to maximal subspaces @xmath921 satisfying @xmath922 with the property that @xmath923 for all @xmath924 . notice that @xmath925 is finite but @xmath926 or @xmath927 could be infinite . as shown in e.g. @xcite , based on the asymptotic behavior of the confluent hypergeometric functions , there exists linear maps @xmath928 such that for any @xmath929 one has , as @xmath460 , @xmath930 or , equivalently , @xmath931 and a direct computation gives the equality @xmath932 . furthermore , for any @xmath933 there is @xmath934 such that @xmath935 and @xmath936 . in the language of @xcite , the triple @xmath937 is a boundary triple for @xmath390 , and any self - adjoint extension of @xmath390 is a restriction of @xmath885 to the functions @xmath938 satisfying the boundary condition @xmath939 , where @xmath940 is a real - valued parameter . by ( * ? * theorem 3.1 ) , the friedrichs extension @xmath389 corresponds to the boundary condition @xmath941 , i.e. to @xmath942 . the spectral properties for this case are completely analyzed in ( * ? ? ? * subsection 8.3.3 ) : the essential spectrum is @xmath391 , and the discrete spectrum consists of the simple eigenvalues @xmath943 with the associated eigenfunctions @xmath393 given by @xmath944 where @xmath395 are the laguerre polynomials ( * ? ? ? * section 22.1 ) . at reader s convenience we provide a complete proof of the following simple assertion . [ propd1 ] let @xmath93 be a self - adjoint operator in a hilbert space @xmath94 such that @xmath945 , and let @xmath71 be the associated sesqulinear form . assume that there exist a non - zero @xmath946 and numbers @xmath947 and @xmath948 such that @xmath949 then @xmath950 . due to the spectral theorem for self - adjoint operators we have @xmath951 , and for @xmath952 there holds @xmath953 . hence , we can rewrite the inequality as @xmath954 and , @xmath955 as the vectors @xmath956 cover the whole of @xmath957 as @xmath419 runs through @xmath958 , taking the supremum over @xmath419 in gives @xmath959 . as @xmath960 , by the spectral theorem for self - adjoint operators we have @xmath961 , which means that there exists @xmath962 such that @xmath963 . hence , @xmath964 . in particular , @xmath965 and @xmath966 , which concludes the proof . m. abramowitz , i. a. stegun : _ handbook of mathematical functions with formulas , graphs , and mathematical tables . _ 10th printing ( volume 55 of applied mathematics series , us national bureau of standards , 1972 ) . d. m. gitman , i. v. tyutin , b. l. voronov : _ self - adjoint extensions in quantum mechanics . general theory and applications to schrdinger and dirac equations with singular potentials . _ birkhuser , basel , 2012 . b. simon : _ on the number of bound states of two body schrdinger operators : a review . _ in e. h. lieb , b. simon , a. s. wightman ( eds . ) : _ studies in mathematical physics . essays in honor of valentine bargmann . _ princeton university press , princeton , nj , 1976 , pp . |
we describe a procedure to identify stars from nearby moving groups and associations out of catalogs of stars with large proper motions . we show that from the mean motion vector of a known or suspected moving group , one can identify additional members of the group based on proper motion data and photometry in the optical and infrared , with minimal contamination from background field stars . we demonstrate this technique by conducting a search for low - mass members of the pictoris moving group in the tycho-2 catalog . all known members of the moving group are easily recovered , and a list of 51 possible candidates is generated . moving group membership is evaluated for 33 candidates based on x - ray flux from , h line emission , and radial velocity measurement from high - resolution infrared spectra obtained at . we confirm three of the candidates to be new members of the group : tyc 1186 - 706 - 1 , tyc 7443 - 1102 - 1 , and tyc 2211 - 1309 - 1 which are late - k and early - m dwarfs 45 - 60pc from the sun . we also identify a common proper motion companion to the known pictoris moving group member tyc 7443 - 1102 - 1 , at a 26.3 separation ; the new companion is associated with the x - ray source 1rxs j195602.8 - 320720 . we argue that the present technique could be applied to other large proper motion catalogs to identify most of the elusive , low - mass members of known nearby moving groups and associations . | 0812.1159 | young stars are critical in understanding not just the star - formation process itself but also the formation of planetary systems , through observation of their circumstellar environment @xcite . this is most efficiently achieved from the nearest possible vantage point , which makes nearby young stars very sought - after targets @xcite . nearby young stars have become prime targets to achieve direct exoplanet imaging @xcite because massive planets are expected to be much more easily detected in their youth at a time when they still shine bright from their own gravitational collapse . young stars are known to exist in the vicinity of the sun as members of `` moving groups '' , each consisting in loose associations of stars moving in the same approximate direction within the solar neighborhood . proposed nearby moving groups and associations include the castor moving group @xcite , the @xmath0 pictoris moving group @xcite , the tucana / horologium association @xcite , the ab doradus moving group @xcite , the carina - near moving group @xcite , and notably the very young tw hydrae association @xcite whose probable members include several brown dwarfs @xcite . all of these nearby moving groups and associations are suspected to be dispersed fragments from a larger neighboring ob association @xcite . stars from nearby moving groups and associations are identified and confirmed through a combination of kinematic data and evidence of youth . most of the original members were identified based on their full ( three dimentional ) kinematics provided by accurate proper motion and parallax from the hipparcos catalog @xcite , complemented with radial velocity observations . while it is relatively straightforward to identify moving group members among samples of stars for which both proper motion _ and _ parallax data exist , it is more complicated to search for members among field stars with no parallax data in hand . proper motion measurements are now widely available for vast numbers of stars from all - sky proper motion catalogs such as tycho-2 @xcite , usno - b1.0 @xcite , or superblink @xcite , but parallax and radial velocity measurements remain relatively rare , particularly for stars with visual magnitude @xmath5 mag . as a result , one must rely on secondary diagnostics , such as unambiguous signs of youth , to search for moving group members . young low - mass stars are expected to have faster rates of rotation and thus higher coronal activity through a stronger dynamo effect @xcite . high coronal activity is associated with strong emission lines of atomic hydrogen and large x - ray emission . spectroscopic follow - up of x - ray sources from the _ rosat _ catalogs has yielded positive identification of many nearby young stars , including new members of the tw hydrae association @xcite and the @xmath0 pictoris moving group @xcite . however , the _ rosat _ catalogs are flux limited , and though they are useful in identifying nearby young stars , they can not serve as the basis for a _ complete census _ of moving group members . it is telling that most confirmed members of young moving groups consist in stars of spectral subtype f , g , or k , and include few low - mass stars of spectral type m. assuming those groups to follow the standard initial mass function , one would expect them to have significantly larger numbers of associated low - mass stars . such low - mass members arguably are particularly promising targets for exoplanet surveys @xcite because the detectability of planets of a given mass is improved when the parent star has low intrinsic luminosity . in this paper , we develop a general technique to identify new members of known moving groups among stars listed in large proper motion catalogs . we demonstrate the technique by performing a search for new members of the @xmath0 pictoris moving group ( bpmg ) , using a subsample of stars from the tycho-2 catalog @xcite with proper motions @xmath6 mas yr@xmath7 . the general method is described in section 2 . a list of candidates members of the @xmath0 pictoris moving group is identified in section 3 . follow - up spectroscopic observations are described in section 4 , and new @xmath0 pictoris members are identified . conclusions follow in section 5 . let @xmath8 be the mean motion of a local moving group , in the local standard of rest , with @xmath9 a unit vector pointing in the direction of the galactic center , @xmath10 in the direction of the sun s galactic orbital motion , and @xmath11 pointing toward the north galactic pole . at any point on the sky described by the galactic coordinates @xmath12 , one can calculate the local projected motion in the plane of the sky @xmath13 of the moving group in the direction of galactic longitude @xmath14 and galactic latitude @xmath15 from @xmath16 @xmath17 in galactic coordinates , the angle @xmath18 that this vector @xmath13 subtends with a star s local proper motion vector @xmath19 is @xmath20 wich is easily derived from the definition of the scalar product of the two vectors . stars which are actual members of the moving groups will have their proper motion vector closely aligned with the projected motion of the group , and will thus have @xmath21 . one can thus sift through a proper motion catalog and search for potential members of the moving groups by considering only the stars whose proper motion vector is within a few degrees of the expected orientation of the projected motion of the moving group . this restriction can be expressed simply as : @xmath22 the particular choice of @xmath23 ( or @xmath24 ) will depend on the accuracy of the catalog s proper motions , and also on the dispersion in velocities between stars of the moving group . for example , @xmath25 would restrict the sample only to stars whose proper motion is within @xmath26 of the expected projected motion of the moving group . a value of @xmath27 closer to @xmath28 will considerably restrict the search , but may overlook some actual members . a smaller value of @xmath27 will be more likely to include all members of the group , but will also produce a larger sample with more contaminants . proper motion aligned in the direction of the projected motion of the moving group is however not a sufficient condition for a star to be a member of the group , because a given proper motion vector may correspond to a range of transverse velocities , depending on the star s distance . if the distance @xmath29 to the star is known , then an actual member of the moving group will also verify : @xmath30 with @xmath31 and @xmath32 given by eqs.1 - 2 above , and where the velocities are in km s@xmath7 and the distance @xmath29 in parsecs . however , distances are generally not known for stars in proper motion catalogs , the one notable exception being the _ hipparcos _ catalog . for a star with no recorded distance , it is however possible to constrain the distance based on color - magnitude relationships . if one assumes that a star is a member of the moving group , then it is possible to get an accurate distance estimate for that star based on the magnitude of the proper motion alone ; this hypothetical distance can then be checked for consistency using color - magnitude relationships . assuming that one given proper motion star is also a member of the moving group , then its hypothetical , _ kinematically derived _ distance @xmath33 will be @xmath34 if @xmath33 is found to be inconsistent with some other distance estimate , such as a photometric distance , a spectroscopic distance or better yet a parallax distance , then the star can be ruled out as a possible member of the moving group . unless parallax measurements are already available for the candidate star , it is useful to compare @xmath33 with the photometric distance . consider the visual magnitude @xmath35 of a proper motion star and its infrared @xmath36 magnitude , if the object is a main - sequence star , then it can be expected to follow a color - magnitude relationship @xmath37 whose form depends on the age of the star . if a star is a member of the moving group under consideration , and assuming interstellar extinction to be negligible ( generally valid for nearby stars ) , then one expects @xmath38 where @xmath39 is the mean dispersion about the color - magnitude relationship for members of that particular moving group . an additional test is to verify that the radial velocity of the prospective moving group member is also consistent with group membership . a moving group member can be predicted to have a radial velocity @xmath40 such that @xmath41 this prediction can be tested with radial velocity measurements . ultimately , one would want to confirm group membership by accurately measuring the @xmath42 components of velocity of the source , which requires accurate distance measurement through geometric parallax as well as accurate radial velocities from high - resolution spectroscopy @xcite . additionally , moderate- to high - resolution spectroscopy should be used to check for the presence of lithium in the atmosphere ( @xmath436708 in emission ) which is a strong diagnostic for young age as li is progressively depleted over time in low - mass stars @xcite . however such a detailed analysis of candidate members is expensive . high - resolution spectroscopy of moderately faint stars requires long observing times and/or large telescope apertures , and parallax measurements are intensive and require years of careful monitoring . it would be desirable to pre - select stars with the highest predicted likelihood of being moving group members . such stars are expected to display secondary evidence of youth , which includes evidence for atmospheric activity such as x - ray flux or h@xmath2 line emission . while activity in itself does not necessarily imply that a star is young , a high proportion of the most active stars are found to be young objects @xcite . the above equations and practical considerations provide a road map for identifying elusive members of a known moving group out of a statistically complete catalog of stars with large proper motion . 1 . isolate a subsample of high proper motion stars whose proper motion vector points in the expected direction for the moving group ( eq . 2 . identify stars from that subsample for which photometric data shows a match between the hypothetical kinematic distance and the photometric distance ( eq.7 ) . 3 . reduce the sample further by selecting targets showing evidence for activity , such as strong x - ray or h@xmath2 emission . 4 . compare the predicted radial velocity ( eq.8 ) to radial velocity measurements to identify probable kinematic members of the group . confirm kinematic membership through parallax measurements 6 . confirm young age through detection of @xmath436707 . in section 3 below , we provide a demonstration of steps 1 - 4 , applied to a subsample of bright high proper motion stars . we demonstrate the technique described above by performing a search for members of the @xmath0 pictoris moving group in a subsample of high proper motion stars from the tycho-2 catalog of @xcite . we have assembled a list of 86,626 stars with proper motions @xmath6 mas yr@xmath7 ; their distribution on the sky is shown in figure 1 ( _ top panel _ ) . this represents an all - sky sample of relatively bright stars . all objects are listed with tycho-2 visual magnitudes @xmath44 mag , with @xmath45 of the stars in the @xmath46 mag range . we have used the vizier service and additional software of our own to cross - correlate the entire list with the two micron all - sky survey ( 2mass ) all - sky catalog of point sources @xcite and obtain infrared @xmath47 , @xmath48 , and @xmath49 magnitudes for all the stars ( a 10@xmath4 search radius was used , and proper motions taken into account ; most counterparts were recovered within 2@xmath4 of their extrapolated position ) . the first selection cut selects for stars which have a proper motion vector whose orientation is consistent with bpmg membership . we adopt as the fiducial space motion of the bpmg the mean motion of its known members , estimated by @xcite to be @xmath50 km s@xmath7 relative to the sun . we apply eq . 3 , and select tycho-2 stars whose proper motion vector makes an angle @xmath51 with the local projected bpmg motion vector ( @xmath27=0.978 ) . this reduces the sample to 15,989 stars ( 18.4% of the initial sample ) whose proper motion vector is aligned with the local projected motion of the bpmg . their distribution on the sky is shown in figure 1 ( _ bottom panel _ ) . the distribution shows a dipolar moment , whose axis is aligned along the apex of the bpmg motion . we have compared our list of proper motion selected objects to the 51 known members of the bpmg listed in @xcite and @xcite . we find that we recover 30 of the known members . of the 22 stars that are not recovered , we find nine stars that are too faint to be listed in the tycho-2 catalog ( 8 of them being faint companions of recovered group members ) . ten other stars have proper motions smaller than the proper motion limit of our initial sample ( @xmath6 mas yr@xmath7 ) and thus could also not possibly have been recovered . the two remaining stars are bright high proper motion stars which are listed in the _ hipparcos _ catalog but , for some reason , are missing from the tycho-2 . all stars being accounted for , we then find our proper motion selection algorithm to have a recovery rate of 100% . this suggests that any unrecognized groups member that is present in the initial sample will be selected in the proper motion cut . the 30 recovered members are listed in table 1 , and their location on the sky is indicated in figure 1 with solid triangles . a majority of the members are located in the southern sky , with a concentration around @xmath52 ) . however this moving group extends over a broad swath of the sky , and even has a few members north of the celestial equator . it even is possible that the sun could be situated within the confines of the group . hence , any of the @xmath53 other stars could potentially be a @xmath0 pictoris member , regardless of its position on the sky . the largest reduction in the number of possible bpmg candidates is achieved through kinematic distance selection . based on data from the _ hipparcos _ catalog , we calculate that the median transverse velocity of stars with proper motion @xmath6 mas yr@xmath7 is @xmath54 km s@xmath7 . the mean motion of the bpmg is only @xmath55 km s@xmath7 , which means that field stars are moving faster on average , and thus @xmath56 will tend to underestimate their distances , making field stars appear underluminous . figure 2 shows the color - magnitude diagram for all 15,989 tycho-2 stars identified as possible members , with @xmath33 used to calculate a pseudo absolute magnitude . the figure reveals that known group members occupy a distinct locus in this color - magnitude diagram , with most members following a relatively well defined color - magnitude relationship . this is not unexpected , as the calculated @xmath33 should be very close to the actual distance @xmath29 to those stars . we find that one can model the mean locus of the bpmg members with @xmath57 with all members but one falling within @xmath58 mag of this line . as expected , the vast majority of stars in the overall sample do not cluster along the line defined in eq.9 , but appear significantly underluminous in the color - magnitude diagram . nonmembers may fall off the well - defined locus of the group members for two main reasons : ( 1 ) their calculated @xmath33 either underestimate or overestimate their true distance , and/or ( 2 ) the stars are significantly older than bpmg members , and do not follow the same color - magnitude relationship as the young bpmg stars . from figure 2 , we see that most stars fall well below the standard color - magnitude relationship of bpmg members , most probably because @xmath33 significantly underestimate their true distance . all these background stars can be efficiently eliminated from the list of prospective candidates . we find that of the initial 15,962 prospective members , only 835 fall within @xmath58 mag of the mean color - magnitude relationship of bpmg members . this reduces the list of possible candidates by a factor of 20 . the much reduced subsample , however , remains contaminated by a significant number of background giants which are easily identified in the color - magnitude diagram . these appear to be red clump k giants , with optical - infrared colors 2@xmath593 . in figure 2 , they show up as a nearly vertical band of stars extending all the way up the diagram . these stars are a serious source of contaminants , however they occupy a fairly narrow range of colors . since we are mostly interested in identifying the elusive low - mass members of the bpmg , we can further restrict our search to stars with very red colors , thus eliminating most of those k giants . we introduce @xmath60 as additional color cut ( dashed line in figure 2 . this further reduces the sample to a short list of 55 probable _ low - mass _ bpmg members . such a sample is of a size very much manageable for follow - up spectroscopic programs . a search of the literature reveals that four of the 55 candidates are actually known members of the tw hya association ( twa ) . the reason for their presence among bpmg candidates is due to the fact that the twa mean motion vector @xmath61 km s@xmath7 @xcite is very close to the mean motion vector of the bpmg @xmath62 km s@xmath7 @xcite . in many directions on the sky , the two associations run nearly parallel to each other , which makes confusion possible . both groups also contain young stars , and occupy similar loci in the color - magnitude diagram . the four twa interlopers are listed in table 1 . of the 51 remaining candidates , we proceeded to investigate further 33 objects which were within range for our follow - up spectroscopic observations ( see below ) . this short list of 33 candidates is provided in table 2 . follow - up observations are now being planned for the remaining 18 targets , which will be presented in a separate paper . table 2 compiles data on the 33 candidates followed up in this paper . the table lists the tycho-2 catalog ( tyc ) identifier , hip numbers for stars listed in the _ hipparcos _ catalog , coordinates in the icrs system , proper motions , tycho-2 visual magnitude v@xmath63 and 2mass infrared k@xmath64 magnitude . the table also lists the predicted distances assuming bpmg membership ( @xmath33 , based on the proper motion ) . though parallaxes are not generally available for stars in the tycho-2 catalog , most of the brighter ( v@xmath65 ) entries are also _ hipparcos _ stars , and accurate parallaxes exist for them . in tables 1 and 2 , we identify all the _ hipparcos _ objects by their catalog number , and provide distances ( d@xmath66 ) based on their _ hipparcos _ parallaxes as obtained from the recent new reduction of @xcite . distances derived from these parallaxes are also tabulated ; they are generally accurate to @xmath671 - 2 parsecs , though a few have large uncertainties , including tyc 689 - 130 - 3 ( = [ rhg95]853 , = lp 476 - 207 ) whose estimated distance is [email protected] pc . additionally , the parallax to bd+28@xmath68382b , the resolved companion to hd 14082 , is clearly overestimated in the @xcite catalog , as it is quoted to be [email protected] mas , compared with [email protected] mas for hd 14082 . figure 3 plots d@xmath66 against the kinematic distance d@xmath69 , calculated from the proper motion assuming all stars to be moving with the mean motion of the @xmath0-pictoris moving group ( equation 6 ) . we provide uncertainties on d@xmath70 which are based on the uncertainties in the tycho-2 proper motions . the relative errors on the proper motions are systematically smaller than those on the parallaxes , which yield smaller errors on d@xmath70 than on d@xmath66 . one should keep in mind , however , that the d@xmath71 assume the stars to be moving exactly at the mean motion of the group , which may not be strictly accurate because group members have some dispersion about the mean velocity . we find that all confirmed bpmg members have d@xmath66 within 20% of d@xmath70 . the small scatter is due to parallax and proper motion accuracies , but also to the intrinsic velocity dispersion of individual group members which is unaccounted in d@xmath70 . of the 33 bpmg candidates , we find seven _ hipparcos _ stars whose parallax distances ( d@xmath66 ) are significantly different from the predicted kinematic distances ( d@xmath70 ) ; this eliminates them as possible group members . six have d@xmath70 overestimating their true distance ; these must be background stars with transverse velocities larger than the bpmg . the other star has d@xmath70 underestimating its true distance , in this case the star must be a foreground object moving slower than the bpmg . with those eliminated from the sample , our follow - up observations have focused on the remaining 26 candidates . we searched for counterparts to the bpmg candidates in the _ rosat _ x - ray catalogs , both the _ rosat _ all - sky bright source catalog of @xcite , and the _ rosat _ all - sky survey faint source catalog @xcite . we also searched for counterparts to the known bpmg members . according to @xcite , the positional accuracy of the _ rosat _ catalog for point sources is 13@xmath4 , 90% of the time , with some outliers possibly having errors as large as 40@xmath4 . to make sure we did not overlook possible x - ray counterparts , we have examined all _ rosat _ sources within 50@xmath4 of our bpmg candidates . we found possible x - ray counterparts for 22 of the known bpmg members , and also to all four of the tw hya association interlopers . we further found counterparts to six of the bpmg candidates . of the 32 possible x - ray counterparts , 25 have _ rosat _ positions placing them within 15@xmath4 of the tycho-2 star , and are thus convincing matches . the other seven sources have large offsets , and were investigated further . a counterpart to tyc 6412 - 1068 - 1 ( = hd 203 ) , the source 1rx j000648.9 - 230608 , is found with an offset of 24@xmath4 . an examination of scans from the digitized sky surveys ( dss ) and from 2mass did not reveal any optical of infrared source within 15@xmath4 of the quoted _ rosat _ position . furthermore , the _ rosat _ catalog reports a positional error of 19@xmath4 for that x - ray source . given the larger positional error and the absence of any alternative , we conclude that 1rx j000648.9 - 230608 is an actual counterpart of tyc 6412 - 1068 - 1 ( = hd 203 ) . likewise the _ rosat _ sources 1rxs j184523.4 - 645201 and 1rxs j184657.3 - 621037 were found to be near the stars tyc 9077 - 2487 - 1 ( = hd 172555 ) and tyc 9073 - 762 - 1 , with offsets of only 26@xmath4 and 33@xmath4 , respectively . the two sources are quoted to have _ rosat _ positional errors of 21@xmath4 and 24@xmath4 , and we could not find any other possible match to an optical or infrared source . we thus identify these sources as the actual x - ray counterparts to tyc 9077 - 2487 - 1 and tyc 9073 - 762 - 1 . the star tyc 7760 - 283 - 1 has a possible x - ray counterpart ( 1rxs j121527.9 - 394843 ) whose _ rosat _ position is 33@xmath4 from the tycho-2 position . the quoted _ rosat _ positional uncertainty is 14@xmath4 , which would suggest the x - ray source is not a match . however , examination of the dss and 2mass reveals no other possible optical counterpart . because the hardness ratios are consistent with the other bpmg and tw hya association members , we assume the x - ray source to be a match , although more accurate x - ray astrometry would have to confirm this . among the new bpmg candidates , we found the star tyc 643 - 73 - 1 to have an associated x - ray source ( 1rxs j024419.4 + 105707 ) with a _ position off by 45@xmath4 . again the positional uncertainty quoted in the _ rosat _ catalog is relatively large ( 43@xmath4 ) and we thus conclude that the two are the same object ; no other possible counterpart can be seen in the dss and 2mass images . the x - ray count rates of all identified counterparts are listed in tables 1 and 2 . we also give the two _ rosat _ hardness ratios @xmath72 and @xmath73 . finally , the star tyc 7443 - 1102 - 1 is found to be 26@xmath4 from a _ rosat _ source ( 1rxs j195602.8 - 320720 ) which has a quoted positional accuracy of only 9@xmath4 . the source is however not a counterpart to tyc 7443 - 1102 - 1 . a close examination of the dss images reveals that the x - ray source is a near - perfect match to a nearby , fainter star . interestingly , a search of the usno - b1.0 reveals that this star has a proper motion [ @xmath74=[30,-78 ] mas yr@xmath7 which is nearly coincident with the proper motion of tyc 7443 - 1102 - 1 , [ @xmath74=[30,-66 ] . the two stars appear to form a common proper motion pair . based on the usno - b astrometry , the companion is at an angular separation @[email protected] from tyc 7443 - 1102 - 1 , with a position angle @xmath76=316.5 degrees . assuming tyc 7443 - 1102 - 1 to be a member of the bpmg , the kinematic distance of [email protected] parsecs suggests a wide binary with projected orbital separation of 1,200@xmath6742 au . to summarize , we find that the majority ( 21 of 30 ) of the known bpmg members have x - ray counterparts in the _ rosat _ catalogs , which indicates chromospheric activity consistent with their young ages . likewise , all four twa interlopers have bright associated _ rosat _ counterparts . on the other hand , we find x - ray counterparts for only seven of the 33 candidates under investigation ( table 2 ) . this suggests that a significant fraction of the candidates are not young / active and are thus not genuine bpmg members but field stars with projected motions aligned with the projected bpmg velocity vector . interestingly , the counterparts of the bpmg candidates have hardness ratios consistent with those of moderately young stars in nearby moving groups . the counterparts are distributed on a locus which coincides roughly with the locus of the known bpmg members ( figure 4 ) . all of them in turn have hardness ratios intermediate between those of extremely young t tauri stars , and those of the older field k and m dwarfs @xcite . this suggests that bpmg candidates with x - ray counterparts are likely to be nearby stars of relatively young ages , whether or not they are actual bpmg members . because some of the known bpmg members also do not have counterparts in _ rosat _ , we find that the absence of a _ rosat _ counterpart is not sufficient to justify rejection . we thus complement the _ rosat _ data with a spectroscopic search for atomic line emission . we have carried out a follow - up program of low - resolution spectroscopy to measure the strength of the h@xmath2 emission line in bpmg candidates . targets in the northern sky were observed from mdm observatory on kitt peak , with the 2.4 m hiltner telescope . spectra were obtained with the mkiii spectrograph , using the thick frontside - illuminated 2k@xmath772k loral ccd camera ( `` wilbur '' ) , which has a pixel size of 15@xmath78 m . we used the 300 lines mm@xmath7 grating blazed at 8000 in first order to produce spectra with a resolution of 6.0 . with the telescope at f/7.5 focal , the spatial scale was [email protected] pixel@xmath7 . the 1.5 wide , longslit spectra were reduced using iraf , following standard flatfield correction , sky background subtraction , and spectrum extraction . the wavelength scale was calibrated using an arc spectrum of near . flux calibration was determined from same - night observations of a set of spectrophotometric standards ( feige 66 , feige 110 , wolf 1346 ) from @xcite . effective spectral coverage is 5800-8300 . in the southern sky , spectra were obtained using the 1.5 m telescope operated by the smarts consortium at the cerro tololo inter - american observatory ( ctio ) . the telescope was equipped with the r - c cassegrain spectrograph . we used grating # 47 ( 800 lines mm@xmath7 blazed at 8000 ) operated in the first order which , with the 110@xmath78 m slit , provides a resolution of 3.1 . all the spectra were obtained by service observers ( along with calibration arcs and spectrophotometric standards ) and kindly processed by f. walter using the pipeline procedures he wrote and described in @xcite . spectra were obtained for 31 of the stars under investigation ; a subsample is displayed in figure 5 . spectral types were also recovered from the literature for tyc 3762 - 1492 - 1 ( = eg cam ) and tyc 1942 - 2581 - 1 ( = ho cnc ) , though both stars were rejected a priori as possible members because of their _ hipparcos _ parallax clearly at odds with the predicted moving group distance . spectral types were determined for all the objects after comparison with spectra from classification standards . three stars were classified as g dwarfs , all others were found to be either late - type k or early - type m dwarfs . for the m dwarfs , subtypes were calculated based on the strength of the cah and tio molecular bands around @xmath79 , as measured by the tio5 , cah2 , and cah3 molecular band indices , and following the calibration of @xcite . subtypes for the m dwarfs are all accurate to @xmath80 . subtype assignment for the g and k dwarfs is based on a comparison with spectra from classification standards . classification of g and k stars using red spectra is however difficult , and some subtypes are accurate to 1 - 2 subclasses ; uncertain subtypes are noted with a colon ( `` : '' ) . spectral classification can be found in the literature for some of these stars ; the subtypes generally agree with ours to @xmath58 subtype . many of the former spectral types are based of photographic images from objective prism surveys @xcite some of which are off by more than 1 subtype ; in these cases we deem our ccd spectra to be more reliable . assigned spectral types appear in table 2 . h@xmath2 in emission was positively identified in seven stars . the equivalent widths of the h@xmath2 lines were calculated in iraf , using the onedspec package . the results are shown in table 2 ; stars for which we found no evidence of h@xmath2 in emission are noted as `` n.d . '' ( for `` nondetection '' ) . we rejected as possible bpmg member any star showing no significant h@xmath2 in emission _ and _ having no counterpart in _ rosat_. this reduced the list of possible candidates to only six stars , which were targeted for follow - up radial velocity measurement . radial velocity observations were obtained using the cryogenic echelle spectrograph ( cshell ) , the high - resolution infrared spectrometer at the nasa 3 m infrared telescope facility ( irtf ) . the detector is a 256@xmath77256 insb array . we set the grating angle to provide 1.5548@xmath78 m at the center of the array . at this wavelength , cshell s free spectral range is @xmath81730 km s@xmath7 ( 38 ) . we used the 0.@xmath45 slit for a resolution r@xmath8230,000 . the instrumental setup was set by the requirements of the part of our observing program directed toward pre - main - sequence spectroscopic binaries @xcite but they are equally well suited for the radial velocity measurements reported here . the spectra were taken as a series of beam - switched , 300 s integrations and were extracted using procedures described in @xcite . the radial velocities were measured by cross - correlation using the same templates as described in @xcite and @xcite . we observed six known members of the bpmg in table 1 , and six candidate members from the list compiled in table 2 . radial velocities were measured and compared to the radial velocities @xmath83 predicted assuming bpmg membership ; results are compiled in table 3 . all six previously reported bpmg members have predicted radial velocities @xmath83 within 4.5 km s@xmath7 of our observed values , corroborating their bpmg membership . three of the stars have radial velocity measurements quoted in the literature . tyc 85 - 1075 - 1 ( = v1005 ori ) is cited by @xcite to have a radial velocity of + 39 km s@xmath7 ; however this measurement was based on low - dispersion spectrograms and is much less reliable than our own estimate of + [email protected] km s@xmath7 . the star tyc 7460 - 137 - 1 (= at mic ) , was measured by @xcite to have a radial velocity v@[email protected] km s@xmath7 , this time using higher dispersion spectrograms . the reported value is largely consistent with our own ( [email protected] km s@xmath7 ) . the faint companion to at mic , the star nltt 49691 , is quoted in the literature with having a radial velocity v@[email protected] km s@xmath7 @xcite which is also consistent with our own measurement ( [email protected] km s@xmath7 ) . the close pair is part of a hierarchical triple with tyc 7457 - 641 - 1 ( = au mic ) . several radial velocity measurements exist for au mic ; the most recent documented one is from @xcite , quoting v@[email protected] km s@xmath7 which is only slightly off our own measured value . for candidate members , we find that three stars ( tyc 777 - 141 - 1 , tyc 6806 - 631 - 1 , and tyc 6849 - 1795 - 1 ) have predicted radial velocities which are inconsistent with the measurements , with differences larger than 15 km s@xmath7 ; this indicates that the three stars are interlopers , whose proper motions are only aligned with the projected motion of the bpmg by chance . the other three candidates ( tyc 1186 - 706 - 1 , tyc 7443 - 1102 - 1 , and tyc 2211 - 1309 - 1 ) have predicted radial velocities in excellent agreement with the observed values , all within 3.5 km s@xmath7 , comparable with the small differences observed for the previously known members . all three stars are hereby identified as highly probable members of the bpmg . all three stars also show h@xmath2 in emission and are x - ray emitters , which suggest they are relatively young and active . we predict that parallax measurements should confirm the distance estimates listed above , as an ultimate test of moving group membership . for all the stars observed at irtf , we also estimate the projected rotational velocities @xmath85 based of the width of the atomic lines : cross - correlation is optimized using a suite of slow - rotator templates , which are `` spun up '' using an algorithm based on @xcite . the templates are sampled every 2 km s@xmath7 ( which defines the uncertainty on @xmath85 ) . results are noted in table 3 . of the three new members , only one ( tyc 2211 - 1309 - 1 ) is a moderately fast rotator ( @xmath86=30@xmath672 km s@xmath7 ) . the other two have @xmath87 7@xmath672 km s@xmath7 , which suggest either that they are not fast rotators or that the rotation axis makes a small angle with the line of sight . a slow rotation in itself would not , however , rule out the star as a bpmg member . as one can see from table 3 , our measurements indicate that several of the known bpmg members also have relatively small values of @xmath85 , with tyc 5832 - 666 - 1 ( = bd-13@xmath686424 ) having @xmath85=7@xmath672 km s@xmath7 . the star tyc 1186 - 706 - 1 , which we identify as a k7.5v dwarf , has a predicted distance of 59.7pc . the star was first identified as a red dwarf in the objective prism survey of @xcite , based on the strength of its sodium d line ; it is star stkm 1 - 34 in the stephenson catalog . the star was classified as k5 , from a visual inspection of the photographic plate spectrogram . this subtype is broadly consistent with our own classification , although we believe our ccd classification should be considered more reliable . the star was also identified by @xcite to be the optical counterpart to the x - ray source 1rxs j002334.9 + 201418 from the _ rosat _ bright source catalogue ( rass - bsc ) of @xcite , but the star could not be classified because it was too bright and was saturated on the schmidt plates used in this objective prism survey . more recently , the star was observed as part of the superwasp wide field photometric survey , and discovered to be periodically variable @xcite . the star is cataloged under the name 1swasp j002334.66 + 201428.6 , and is found to be variable at the 2%-3% level with a period of 7.9 days . such variation is consistent with rotational modulation . the star tyc 7443 - 1102 - 1 is an m0.0v dwarf , at a predicted distance of 57.7pc . the predicted distance is similar to that of tyc 1186 - 706 - 1 , but the two stars are in very different directions on the sky ( see figure 1 ) and must be physically unrelated . tyc 7443 - 1102 - 1 does not have any previous mention in the literature . however , the common proper motion companion identified while searching for x - ray counterparts ( see section 4.2 above ) is , of course , the known x - ray source j195602.8 - 320720 @xcite . we associate this x - ray source with the proper motion star usno - b1.0 0578 - 1079977 . its position is listed in table 3 . the companion has a visual magnitude [email protected] mag ( estimated from photographic plate measurements ) , compared with v@xmath89=11.95 mag for tyc 7443 - 1102 - 1 . the companion should also be considered a new member of the @xmath0 pic moving group . the fourth new member , tyc 2211 - 1309 - 1 , is an m0.0ve red dwarf at a predicted distance of 45.6 pc . this object was first prosaically identified as a `` noncluster x - ray source ( star ? ) '' by @xcite , in a search for galaxy clusters among _ rosat _ sources . the object was identified by @xcite as a variable x - ray source with flaring behavior . this would be consistent with chromospheric activity on that star . this star was also identified in the superwasp wide field photometric survey of @xcite and found to be variable in the optical regime . the star is cataloged under the name 1swasp j220041.59 + 271513.5 , and was observed to be variable at the 2% level , with a period of 0.52 days . this behavior suggests rotational modulation in a fast rotator , and is consistent with the high value of @xmath85 we infer for the rotational velocity . an early m - type star with a rotation period of 0.5 days , having a radius of about 0.5 r@xmath90 @xcite , would be expected to have a rotation velocity at the equator of 50 km s@xmath7 , consistent with our inferred value of @xmath85 = 30 km s@xmath7 . we have presented an algorithm to identify probable members of any known nearby moving group from large catalogs of proper motion stars . the technique identifies prospective members based on the orientation of their proper motion vector , which aligns with the projected motion of the group in the plane of the sky , and uses the magnitude of the proper motion to estimate a distance of the assumption of group membership . that predicted distance is used to verify consistency with the known color - magnitude relationship of the group . we have tested the method by conducting a search for possible members of the @xmath0 pictoris moving group in a subsample of high proper motion stars @xmath6 mas@xmath7 from the tycho-2 catalog . a subset of 31 candidates was investigated for signs of youth , including low - resolution spectroscopy to look for h@xmath2 lines in emission , and a search for x - ray counterparts in the _ rosat _ catalogs . twelve stars showing possible evidence for youth were targeted for high - resolution infrared spectroscopy , in order to corroborate moving group membership from their radial velocities . three of the stars are found to be convincing new members of the bpmg , these are tyc 1186 - 706 - 1 , tyc 7443 - 1102 - 1 , and tyc 2211 - 1309 - 1 . the star tyc 7443 - 1102 - 1 is also found to have a common proper motion companion , with an angular separation of 26@xmath4 , which is most probably a wide companion with a projected separation @xmath881,200 au . we believe that the method can be successfully expanded to the identification of members from any other moving group and association . the use of a deeper proper motion catalog , such as the recent ppm - extended ( ppmx ) catalog @xcite , which extends to v=15 mag , or the superblink database @xcite that is statistically complete to v=19 mag , should allow for the identification of most low - mass members of these moving groups , at least down to the hydrogen burning limit . perryman , m. a. c. & esa 1997 , the hipparcos and tycho catalogues . astrometric and photometric star catalogues derived from the esa hipparcos space astrometry mission , publisher : noordwijk , netherlands : esa publications division rrlrrrrrrrrrrr 6412 1068 1 & 560 & hd 203 & 1.708454 & -23.107429 & 97.1 & -47.4 & 6.18 & 5.24 & [email protected] & [email protected] & 44@xmath6715 & [email protected] & [email protected] + 1777 1480 1 & 10679&bd+28@xmath68382b & 34.352851 & 28.741941 & 86.7 & -76.7 & 7.76 & 6.26 & [email protected] & [email protected] & 364@xmath6773 & [email protected] & [email protected] + 1777 1479 1 & 10680 & hd 14082 & 34.355136 & 28.745208 & 84.3 & -77.6 & 7.04 & 5.78 & [email protected] & [email protected] & & & + 45 990 1 & 11360 & hd 15115 & 36.567489 & 6.292664 & 100.7 & 5.5 & 6.79 & 5.82 & [email protected] & [email protected] & 77@xmath6720 & [email protected] & [email protected] + 2323 566 1 & 11437 & ag tri & 36.871692 & 30.973667 & 84.0 & -71.8 & 10.08 & 7.08 & [email protected] & [email protected] & 566@xmath6746 & [email protected] & [email protected] + 4739 1551 1 & 21547 & c eri & 69.400451 & -2.473389 & 43.6 & -64.1 & 5.21 & 4.53 & [email protected] & [email protected] & & & + 85 1075 1 & 23200 & v1005 ori & 74.895042 & 1.783751 & 38.1 & -94.4 & 10.10 & 6.26 & [email protected] & [email protected] & 651@xmath6742 & [email protected] & [email protected] + 8513 572 1 & 23309&cd-57@xmath681054 & 75.196235 & -57.257244 & 36.2 & 72.6 & 10.11 & 6.24 & [email protected] & [email protected] & 330@xmath6752 & [email protected] & [email protected] + 689 130 3 & 23418 & [ rhg95]853 & 75.494987 & 9.983299 & 17.2 & -82.0 & 11.74 & 6.37 & [email protected] & [email protected] & 661@xmath6759 & [email protected] & [email protected] + 8099 1392 1 & 27321 & @xmath0 pic & 86.821182 & -51.066703 & 4.1 & 83.3 & 3.85 & 3.52 & [email protected] & [email protected] & & & + 9172 690 1 & 29964 & ao men & 94.617607 & -72.045021 & -8.5 & 75.7 & 9.99 & 6.81 & [email protected] & [email protected] & 1030@xmath6730 & [email protected] & [email protected] + 8704 1271 1 & 76629 & v343 nor & 234.740005 & -57.707344 & -46.2 & -97.9 & 8.16 & 5.85 & [email protected] & [email protected] & 1420@xmath6772 & [email protected] & [email protected] + 6805 1909 1 & 79881 & d sco & 244.574661 & -28.613777 & -32.4 & -100.2 & 4.78 & 4.73 & [email protected] & [email protected] & & & + 9064 3514 1 & 84586 & v824 ara & 259.356384 & -66.950714 & -21.6 & -136.4 & 6.87 & 4.70 & [email protected] & [email protected] & 476@xmath6727 & [email protected] & [email protected] + 8369 1619 1 & 88399 & hd 164249 & 270.764190 & -51.648807 & 2.8 & -87.2 & 7.01 & 5.91 & [email protected] & [email protected] & 150@xmath6742 & [email protected] & [email protected] + 7911 5035 1 & 88726 & hd 165189 & 271.707855 & -43.424964 & 13.8 & -105.3 & 5.63 & 4.38 & [email protected] & [email protected] & & & + 9077 2487 1 & 92024 & hd 172555 & 281.361907 & -64.870918 & 32.9 & -148.2 & 4.77 & 4.29 & [email protected] & [email protected] & 97@xmath6731 & [email protected] & [email protected] + 9073 762 1 & & & 281.718963 & -62.176612 & 18.1 & -76.6 & 12.22 & 7.85 & [email protected] & & 137@xmath6738 & [email protected] & [email protected] + 7408 54 1 & & & 282.685302 & -31.796323 & 10.6 & -77.8 & 11.30 & 7.46 & [email protected] & & 308@xmath6737 & [email protected] & [email protected] + 8381 2435 1 & 92680 & pz tel & 283.274414 & -50.180332 & 15.8 & -84.1 & 8.41 & 6.36 & [email protected] & [email protected] & 1000@xmath6788 & [email protected] & [email protected] + 8765 2571 1 & 95261 & hd 181296 & 290.713256 & -54.423740 & 25.0 & -83.1 & 5.02 & 5.00 & [email protected] & [email protected] & & & + 8765 638 1 & 95270 & hd 181327 & 290.745513 & -54.537860 & 24.1 & -82.9 & 7.04 & 5.91 & [email protected] & [email protected] & & & + 6909 1892 1 & 99273 & hd 191089 & 302.271636 & -26.223878 & -52.4 & -58.8 & 7.19 & 6.07 & [email protected] & [email protected] & 73@xmath6718 & [email protected] & [email protected] + 7460 137 1 & 102141 & at mic & 310.462432 & -32.434368 & 261.3 & -344.8 & 11.24 & 4.94 & [email protected] & [email protected]&3910@xmath67120 & [email protected] & [email protected] + 7457 641 1 & 102409 & au mic & 311.288940 & -31.340036 & 278.8 & -360.0 & 8.77 & 4.52 & [email protected] & [email protected]&5950@xmath67121 & [email protected] & [email protected] + 6348 98 1 & 103311 & hd 199143 & 313.948486 & -17.114032 & 62.2 & -65.4 & 7.32 & 5.81 & [email protected] & [email protected]&1400@xmath67107 & [email protected] & [email protected] + 6349 200 1 & & az cap & 314.011260 & -17.181446 & 59.3 & -63.0 & 10.63 & 7.03 & [email protected] & & 236@xmath6726 & [email protected] & [email protected] + 9340 437 1 & & cpd-72@xmath682713 & 340.703155 & -71.705772 & 94.1 & -54.4 & 10.54 & 6.89 & [email protected] & & 727@xmath6778 & [email protected] & [email protected] + 5832 666 1 & & bd-13@xmath686424 & 353.128265 & -12.264112 & 138.1 & -83.2 & 10.71 & 6.56 & [email protected] & & 620@xmath6748 & [email protected] & [email protected] + 7190 2111 1 & & & 160.625610 & -33.671215 & -122.2 & -29.3 & 10.92 & 6.89 & [email protected] & & 324@xmath6731 & [email protected] & [email protected] + 7208 347 1 & 53911 & tw hya & 165.466476 & -34.704719 & -73.4 & -17.5 & 11.27 & 7.29 & [email protected] & [email protected] & 571@xmath6743 & [email protected] & [email protected] + 7201 27 1 & & & 167.307785 & -30.027683 & -90.0 & -87.7 & 11.16 & 6.71 & [email protected] & & 341@xmath6735 & [email protected] & [email protected] + 7760 283 1 & & & 183.878219 & -39.811759 & -75.9 & -26.3 & 11.43 & 7.30 & [email protected] & & 375@xmath6765 & [email protected] & [email protected] + rrlrrrrrrrrrrrrrc 1186 706 1 & & & 5.894279 & 20.241407 & 63.0 & -38.1 & 10.85 & 7.33 & [email protected] & & 176@xmath6727 & [email protected] & [email protected] & k7.5v & [email protected] & table 3 + 5853 933 1 & & bpm 47010 & 18.914964 & -21.514326 & 81.4 & -22.5 & 13.03 & 8.52 & [email protected] & & & & & k7.0v & n.d . & no + 643 73 1 & 12787 & & 41.088851 & 10.961535 & 73.5 & -57.3 & 11.40 & 7.11 & [email protected] & [email protected] & 301@xmath6794 & [email protected] & [email protected] & m0.5ve & [email protected] & no + 3333 1029 1 & & & 64.939537 & 47.758914 & 21.2 & -79.0 & 11.32 & 7.69 & [email protected] & & & & & m0.0v & n.d . & no + 2384 1106 1 & & hd 279890 & 66.975112 & 35.936912 & 29.8 & -64.5 & 10.82 & 7.20 & [email protected] & & & & & k0v : & n.d . & no + 3762 1492 1 & 28368 & eg cam & 89.907287 & 58.593612 & 11.2 & -252.9 & 10.27 & 6.21 & [email protected] & [email protected] & & & & m0.5v & & no + 8895 225 1 & 31878&cd-61@xmath681439 & 99.958557 & -61.478378 & -25.7 & 71.3 & 9.69 & 6.50 & [email protected] & [email protected] & 143@xmath6714 & [email protected] & [email protected] & k5v : & n.d . & no + 2443 845 1 & & stkm 1 - 603 & 100.331420 & 33.907360 & -25.6 & -68.1 & 11.33 & 7.96 & [email protected] & & & & & m0.0v & n.d . & no + 1903 1306 1 & 34222 & gj 265a & 106.426132 & 27.471057 & -49.2 & -97.2 & 10.26 & 6.78 & [email protected] & [email protected] & & & & k7.0v & n.d . & no + 1917 2452 1 & 35191 & gj 9227 & 109.082481 & 27.142967 & -39.0 & -191.6 & 10.75 & 6.18 & [email protected] & [email protected] & & & & m3.0v & n.d . & no + 4618 116 1 & & & 109.992683 & 84.077461 & -38.5 & -86.0 & 11.93 & 7.51 & [email protected] & & & & & m2.5v & n.d . & no + 777 141 1 & & & 113.734748 & 14.765309 & -80.1 & -106.4 & 10.74 & 6.39 & [email protected] & & & & & m3.0ve & [email protected] & table 3 + 9389 53 1 & & & 117.695777 & -79.867774 & -48.0 & 57.7 & 11.18 & 7.81 & [email protected] & & & & & k3v : & n.d . & no + 1942 2581 1 & 42253 & ho cnc & 129.232696 & 23.246904 & -108.7 & -106.7 & 9.51 & 6.41 & [email protected] & [email protected] & & & & k5v & & no + 1952 1263 1 & & stkm 1 - 764 & 139.756805 & 23.597393 & -49.6 & -55.0 & 11.92 & 8.36 & [email protected] & & & & & k7.5v & n.d . & no + 240 2164 1 & 48447 & & 148.163299 & 3.130231 & -61.6 & -33.7 & 10.56 & 7.07 & [email protected] & [email protected] & & & & k7.5v & n.d . & no + 8606 821 1 & & & 149.721710 & -57.824100 & -165.9 & 36.5 & 11.88 & 6.44 & [email protected] & & & & & m4.5v & n.d . & no + 7732 1096 1 & & & 164.767608 & -41.300849 & -118.8 & -75.4 & 11.84 & 7.12 & [email protected] & & & & & m1.5v & n.d . & no + 7322 67 1 & & ltt 6184 & 232.204833 & -34.332885 & -104.2 & -118.5 & 9.04 & 5.77 & [email protected] & & & & & g9v : & n.d . & no + 8304 1011 1 & & & 233.453536 & -49.300254 & -138.3 & -213.6 & 11.16 & 6.47 & [email protected] & & & & & k3v : & n.d . & no + 6778 560 1 & & bd-22@xmath684030 & 237.258209 & -23.042593 & -51.5 & -74.5 & 10.01 & 6.89 & [email protected] & & & & & g5v : & n.d . & no + 5046 481 1 & & bd-05@xmath684273 & 244.614517 & -6.074768 & -32.5 & -78.8 & 10.05 & 6.50 & [email protected] & & & & & k0v : & n.d . & no + 1513 989 1 & & hd 147926 & 245.998031 & 20.394264 & -63.6 & -47.1 & 8.13 & 4.96 & [email protected] & & & & & k2v : & n.d . & no + 6806 631 1 & & hd 147611 & 246.008605 & -29.178794 & -42.3 & -139.5 & 10.15 & 6.77 & [email protected] & & 505@xmath6739 & [email protected] & [email protected] & k5v : & [email protected] & table 3 + 7892 2635 1 & & & 264.482879 & -42.324363 & -11.6 & -107.2 & 11.60 & 7.45 & [email protected] & & & & & m1.5v & n.d . & no + 6832 549 1 & & & 266.221771 & -25.644212 & -14.7 & -82.5 & 11.03 & 7.49 & [email protected] & & & & & k2v : & n.d . & no + 6849 1795 1 & & & 268.725524 & -26.827995 & 18.1 & -103.5 & 10.28 & 6.91 & [email protected] & & 563@xmath6745 & [email protected] & [email protected] & k5v : & [email protected] & table 3 + 6262 1013 1 & & & 268.909790 & -22.320178 & -9.6 & -70.4 & 11.71 & 7.75 & [email protected] & & & & & k0v : & n.d . & no + 7899 6490 1 & & hd 324810 & 270.559997 & -37.682273 & 18.9 & -185.5 & 11.95 & 6.97 & [email protected] & & & & & & n.d . & no + 7443 1102 1 & & & 299.018127 & -32.126995 & 31.2 & -65.0 & 11.80 & 7.84 & [email protected] & & & & & m0.0v & [email protected] & table 3 + 2211 1309 1 & & & 330.173126 & 27.253810 & 76.3 & -14.7 & 11.37 & 7.72 & [email protected] & & 219@xmath6722 & [email protected] & [email protected] & m0.0ve & [email protected] & table 3 + 4275 1683 1 & & & 330.877685 & 66.683044 & 72.6 & 14.5 & 11.19 & 7.85 & [email protected] & & & & & k7.5v & n.d . & no + 4285 2546 1 & & & 357.159545 & 62.026763 & 77.9 & -8.6 & 11.34 & 7.58 & [email protected] & & & & & k5.0v & n.d . & no + rlrrrrrcr 85 - 1075 - 1 & v1005 ori & 74.895042 & 1.783751 & @xmath91 & @xmath92 & 12@xmath672 & m0.5ve & yes + 7408 - 54 - 1 & & 282.685302 & -31.796323 & @xmath93 & @xmath94 & 50@xmath672 & k7ve & yes + 7460 - 137 - 1 & at mic & 310.462432 & -32.434368 & @xmath95 & @xmath96 & 15@xmath672 & m4.0ve & yes + & nltt 49691 & 310.463112 & -32.436175 & @xmath97 & @xmath98 & 12@xmath672 & m4.0v & yes + 7457 - 641 - 1 & au mic & 311.288940 & -31.340036 & @xmath99 & @xmath100 & 12@xmath672 & m1.0ve & yes + 6349 - 200 - 1 & az cap & 314.011260 & -17.181446 & @xmath101 & @xmath102 & 20@xmath672 & k5ve & yes + 5832 - 666 - 1 & bd-13@xmath686424 & 353.128265 & -12.264112 & @xmath103 & @xmath104 & 7@xmath672 & m0.5ve & yes + 1186 - 706 - 1 & & 5.894279 & 20.241407 & @xmath105 & @xmath106 & 7@xmath672 & k7.5v & yes + 777 - 141 - 1 & & 113.734748 & 14.765309 & @xmath107 & @xmath108 & 6@xmath672 & m3.0ve & no + 6806 - 631 - 1&hd 147611 & 246.008605 & -29.178794 & @xmath109 & @xmath110 & 4@xmath672 & k4v : & no + 6849 - 1795 - 1 & & 268.725524 & -26.827995 & @xmath111 & @xmath112 & 25@xmath672 & k4v : & no + 7443 - 1102 - 1 & & 299.018127 & -32.126995 & @xmath113 & @xmath114 & 6@xmath672 & m0.0v & yes + & 1rxs j195602.8 - 320720 & 299.012262 & -32.121850 & & & & & yes + 2211 - 1309 - 1 & & 330.173126 & 27.253810 & @xmath115 & @xmath116 & 30@xmath672 & m0.0ve & yes + |
we study competition between s - wave order and d - wave order through two holographic superconductor models . we find that once the coexisting phase appears , it is always thermodynamically favored , and that the coexistence phase is narrow and one condensate tends to kill the other . the phase diagram is constructed for each model in terms of temperature and the ratio of charges of two orders . we further compare the behaviors of some thermodynamic quantities , and discuss the different aspects and identical ones between two models . | 1405.0382 | one of the most studied subjects in the ads / cft correspondence @xcite is holographic superconductors , which may shed light upon real word strongly coupled superconductivity . some holographic superconductor models with different symmetry of condensation have been constructed , including s - wave @xcite , p - wave @xcite and d - wave @xcite . such holographic setups indeed reveal some basic properties of the real superconductors . nevertheless , most of studies in the literature focus on the case with only a single order parameter . on the other hand , there are various orders in real high temperature superconductors @xcite . thus it is desirable to generalize the single order parameter case to multi order parameter case . indeed , the holographic framework provides us a convenient way to uncover the interaction among those orders by simply adopting dual dynamical fields in the bulk with appropriate couplings . following this strategy , several attempts on the competition of multi order parameters in the holographic superconductor models have already been made . in refs . @xcite , the authors considered the case of two competing scalar order parameters coupled to one u(1 ) gauge field in the bulk . they found the signature of a coexisting phase where both scalar order parameters appear at the same time . another holographic superconductor model with a scalar triplet charged under a su(2 ) gauge field in the bulk was built in ref . they showed that the s+p coexisting phase turns out to be thermodynamically favored once it appears . other related studies can be found in refs . @xcite . in this paper , we will study the competition mechanism between s - wave order and d - wave order . there are two acceptable holographic models describing the d - wave condensation in the literature , in which the d - wave order parameter is dual to a charged massive spin two field propagating in an asymptotically ads background . the authors of ref . @xcite first constructed a minimal gravitational model by introducing a symmetric , traceless rank - two tensor field minimally coupled to a u(1 ) gauge field in the background of the ads black hole . the d - wave condensate appears below a critical temperature via a second order phase transition , resulting in a superconducting phase with no hard gap for its optical conductivity . let us call it ckmwy d - wave model in terms of the initials of the five authors . the other holographic d - wave model was proposed soon after the first one with the same matter fields but much more complex interactions @xcite . the phase diagram , optical conductivity , as well as fermion spectral function were investigated in detail . with a fixed gravity background , this model has advantages such as being ghost - free and having the right propagating degrees of freedom . this model will be named as bhry d - wave model in what follows . to realize s - wave order , we will take advantage of the well known abelian - higgs model @xcite in terms of a complex scalar field charged under a u(1 ) gauge field in the bulk . in order to realize the condensation of s - wave order and d - wave order in one holographic model , we can simply combine the abelian - higgs model with a d - wave model . thus , we could have two holographic models with s - wave order and d - wave order . actually , we will study the competition between s - wave order and d - wave order for both cases in the probe limit where one neglects the back reaction of matter fields to the background geometry . the phase structures are given and the behaviors of the thermodynamic quantities for the s+d coexisting phase are also studied . the coexisting phase does appear in both models and is thermodynamically favored . apart from the above common features , the behavior of the ratio of superconducting charge density over the total charge density versus temperature in two models is different . we also analyze the optical conductivity of the coexisting phase and find some new features . , there also exists coexisting phase under the model parameters @xmath0 , @xmath1 and @xmath2 . both results are consistent with each other in that case . ] the paper is organized as follows . first we study the competition mechanism in the s - wave + bhry d - wave model in section [ sect : superfluid ] , by investigating including the phase transition , thermodynamics and optical conductivity . we discuss the competition between two orders for the s - wave + ckmwy d - wave model in section [ sect : superconductor ] . we will also give a comparison between the two models . conclusions and discussions are given in section [ sect : conclusions ] . to study the competition between s - wave and d - wave orders , let us first start with the holographic model by combining the abelian - higgs s - wave model @xcite and bhry d - wave model @xcite . the holographic model with a scalar field @xmath3 , a symmetric tensor field @xmath4 and a u(1 ) gauge field @xmath5 is described by the following action : @xmath6-i q_2 f_{\mu\nu } \varphi^{*\mu\lambda } \varphi^\nu_\lambda\\ -m_2 ^ 2\big(|\varphi_{\mu\nu}|^2-|\varphi|^2\big)+2r_{\mu\nu\rho\lambda } \varphi^{*\mu\rho}\varphi^{\nu\lambda } -r_{\mu\nu}\varphi^{*\mu\lambda}\varphi^\nu_\lambda-\frac{1}{4 } r |\varphi|^2 , \end{split}\ ] ] where @xmath7 and @xmath8 , @xmath9 , @xmath10 and @xmath11 is the riemann tensor of the background metric . @xmath3 is the scalar order and @xmath12 is the tensor order . the parameters @xmath13 and @xmath14 are the charges of the scalar and the tensor fields , respectively . one can perform a rescaling to set the charge @xmath13 of the scalar to be unity . then the phase structure of this theory is determined by the ratio @xmath15 by fixing the mass square of the scalar field @xmath16 and the mass square of the tensor field @xmath17 . we shall set @xmath18 without loss of generality in the following discussion . the corresponding equations of motion are as follows , @xmath19 \\ & & \qquad + 2 r_{\mu\rho\nu\lambda } \varphi^{\rho\lambda } - g_{\mu\nu } \frac { r}{4 } \varphi - i\frac{q_2}{2 } \big ( f_{\mu\rho } \varphi^\rho_\nu + f_{\nu\rho } \varphi^\rho_\mu \big ) , \\ & & \nabla_\mu f^{\mu\nu } = j^\nu,\end{aligned}\ ] ] where @xmath20 note that here there is no direct interaction between @xmath3 and @xmath4 , but they interact with each other via the u(1 ) gauge field and the strength is controlled by the ratio of charge @xmath21 . working in the probe limit , we choose the background metric to be the 3 + 1 dimensional ads - schwarzschild black hole with planar horizon , which reads @xmath22 where @xmath23 and the ads radius has been set to be unity . the horizon is located at @xmath24 and the hawking temperature for this black hole is @xmath25 , which is also the temperature of the dual field theory . we consider an ansatz where @xmath4 and @xmath5 depend only on the radial coordinate @xmath26 and the spatial components of @xmath4 are turned on only . according to ref . @xcite , it is consistent to turn on a single component of @xmath4 and to set other components of the gauge field except for @xmath27 to be zero . then our ansatz is @xmath28 with @xmath29 , @xmath30 and @xmath31 all real functions . with the above ansatz , the equations of motion for @xmath32 , @xmath3 and @xmath33 are given by @xmath34 here the prime denotes the derivative with respect to @xmath26 . from the above explicit equations of motion , we can easily get the s - wave or d - wave superconductivity by turning off the tensor degree of freedom @xmath33 or the scalar field @xmath3 , respectively . therefore , with this model at hand , we can study the competition mechanism between the s - wave order and d - wave order . it is easy to see that equations has a symmetry @xmath35 under this symmetry transformation , the role of s - wave and d - wave would interchange each other . without loss of generality , here we focus on the case @xmath36 . in order to find the solutions for all the three functions @xmath37 , one must specify suitable boundary conditions both at the ads boundary and at the horizon . we demand that the matter fields near the boundary @xmath38 should behave as @xmath39 where @xmath40 and @xmath41 . for spin two operators . therefore , the mass of @xmath4 has a lower bound , i.e. , @xmath42 . ] note that the fall - off of @xmath3 and @xmath33 is chosen so that the dual charged operators have no deformation but can acquire expectation value spontaneously . according to the holographic dictionary , up to a normalization , the coefficients @xmath43 , @xmath44 , @xmath45 and @xmath46 are interpreted as chemical potential , charge density , the expectation values of scalar operator @xmath47 and the spin two operator @xmath48 , respectively . at the horizon , in addition to @xmath49 , one must require @xmath50 in order that @xmath51 is finite at the horizon . regularity of the solution at the horizon @xmath52 requires that all the functions have finite value and admit a series expansion in terms of @xmath53 as @xmath54 by plugging the expansion into , one can find that there are four independent parameters at the horizon @xmath55 . note that the equations of motion have a useful scaling symmetry @xmath56 where @xmath57 is a real positive constant . taking advantage of the above scaling symmetry , we can set @xmath58 for performing numerics . then we have three independent parameters @xmath59 , where two of them will be chosen as shooting parameters to match the asymptotic expansion . after solving the set of equations , we can obtain the condensates @xmath60 and @xmath61 , chemical potential @xmath43 and charge density @xmath44 by reading off the corresponding coefficients in , respectively . the normal phase in the dual field theory is characterized by the vanishing vacuum expectation values of both condensates , which corresponds to vanishing scalar field @xmath3 and spin two tensor field @xmath33 in the bulk . the gravity background describing the normal phase can be solved exactly , which reads @xmath62 before solving the set of coupled equations numerically , we make a briefly qualitative analysis on the possible phases for such a model . following ref . @xcite , we rephrase the equations for the s - wave and d - wave as a potential problem . it is convenient to work in the z - coordinate where @xmath63 . in this coordinate , the infinite boundary is now at @xmath64 , while the horizon is at @xmath65 . with the transformation @xmath66 and @xmath67 , the evolution equations for s - wave and d - wave in equations can be rewritten as follows @xmath68 where @xmath69 and @xmath70 . after introducing a new variable @xmath71 , the above equations can be further expressed as @xmath72 where @xmath73 with @xmath74 as @xmath75 and @xmath76 as @xmath77 . now in terms of the new variable @xmath71 , the equations of motion for s - wave and d - wave are rephrased as a potential problem on a semi infinite line , i.e. , @xmath78 . we will analyze this potential problem in detail case by case . our discussion is base upon the lemma proven in ref . @xcite : for two potentials @xmath79 and @xmath80 over the same domain with @xmath81 , the lowest eigenvalue of @xmath79 would be strictly greater than the lowest eigenvalue of @xmath80 . the lemma implies that if the lowest eigenvalue mode for @xmath80 is a zero mode and then @xmath79 can not have a bound state or a zero mode . note that we focus on the case @xmath36 and have set @xmath18 . in this case , no matter which gauge field configuration we choose , we always have @xmath83 , which implies that a zero mode of s - wave should form before a zero mode of d - wave . although the condensate of s - wave changes the gauge field profile , according to the lemma , with the modified gauge potential @xmath83 and a s - wave has a node - less condensate , no zero mode or bound state of d - wave exists . that is to say , in the phase with s - wave condensed yet , the d - wave can not condense . therefore the phase structure of the system is the same as that of s - wave holographic superconductor with a single scalar . this case is much more complicated . one may expect that the d - wave field with large charge @xmath14 will always dominate . however , the potential @xmath85 diverges like @xmath86 near the boundary @xmath87 when we lower the temperature matters . lowering the temperature is equivalent to increasing the chemical potential . here we choose to vary the temperature and keep the chemical potential fixed through the whole paper . ] . therefore , lowering the temperature possibly makes the mass dependent potential more important and hence the s - wave tends to dominate . we will confirm this with the numerical calculation . our main purpose is to observe the phase diagram of the model in terms of temperature and the charge of the tensor field @xmath14 . as the ratio of charges between the tensor field and the scalar field . ] we set the mass square @xmath0 and @xmath88 in this paper . we expect that the model would admit three different superconducting phases . the first superconducting phase corresponds to the pure s - wave with @xmath89 and @xmath90 . the second one is the pure d - wave with @xmath91 and @xmath92 . the third superconducting phase admits the coexisting of the s - wave and d - wave orders . here we take @xmath93 as a typical example . the condensations for pure s - wave and pure d - wave superconducting phases are depicted in figure [ sd ] . as we lower the temperature , the normal phase becomes unstable to developing scalar / tensor hair at a certain critical temperature @xmath94 . for the given charge , one can see that the critical temperature of pure s - wave is lower than the one for the d - wave case . thus when we lower the temperature , the d - wave order phase should first appear . once the d - wave order appears , if one goes on lowering the temperature , an interesting question arises : whether the other condensate happens or not ? the left plot shows the condensate as a function of temperature for the pure s - wave phase and the right plot is for the condensate of the pure d - wave phase . when one lowers the temperature , the s - wave order or d - wave order emerges at a critical temperature.,title="fig : " ] the left plot shows the condensate as a function of temperature for the pure s - wave phase and the right plot is for the condensate of the pure d - wave phase . when one lowers the temperature , the s - wave order or d - wave order emerges at a critical temperature.,title="fig : " ] condensate of the operators in the s+d coexisting phase . the blue curve is for the condensate of the d - wave operator , while the red curve is for the s - wave operator . we see that the d - wave order first condenses , then both orders coexist , finally the d - wave order disappears , leaving only the s - wave order.,title="fig : " ] our numerical results confirm that the model does admit the coexistence region of two orders with different symmetry , which is drawn in figure [ co ] . we can see that as one lowers the temperature , the d - wave order first condenses at @xmath94 where the superconducting phase transition happens . when we continue lowering the temperature to a certain value , say @xmath95 , the s - wave order begins to condense , while the condensate of d - wave order decreases , resulting in the state with both orders ; if one further lowers the temperature , the d - wave condensate quickly goes to zero at a temperature at @xmath96 . when temperature is lower than @xmath96 , there exists only the s - wave order . the coexisting phase with both s - wave order and d - wave order can only appear in a narrow range @xmath97 . based on the above discussion , we have totally three different superconducting phases in our model . in order to determine which phase is thermodynamically favored , we should compare the free energy of the system for each phase . here we will work in grand canonical ensemble , where the chemical potential is fixed . in the gauge / gravity duality the grand potential @xmath98 of the boundary thermal state is identified with temperature times the on - shell bulk action with euclidean signature . because we work in the probe limit , we only need to consider the contribution from the matter fields to the free energy . the gibbs free energy can be expressed as @xmath99 where @xmath100 . the left plot shows the difference of gibbs free energy between the superconducting phase and the normal phase . the blue curve is for the d - wave phase , the green line is for the s - wave phase , while the red curve is for the s+d coexisting phase . the right plot is an enlarged version of the left one to show the s+d phase more clearly.,title="fig : " ] the left plot shows the difference of gibbs free energy between the superconducting phase and the normal phase . the blue curve is for the d - wave phase , the green line is for the s - wave phase , while the red curve is for the s+d coexisting phase . the right plot is an enlarged version of the left one to show the s+d phase more clearly.,title="fig : " ] we plot the difference of the gibbs free energy between the superconducting phase and the normal phase in figure [ free ] . the green curve is shown for the pure s - wave phase , and the blue curve represents the pure d - wave phase . the free energy for the d - wave phase is lower when @xmath101 , while the free energy for the s - wave is lower when @xmath102 . when @xmath103 , the s+d coexisting phase has the lowest free energy , indicating that once the s+d phase exists , it is thermodynamically favored . as we know , there is only a small window admitting the two orders to coexist . outside the region , it reduces to phases with only a single order . this means that the system is dominated by the d - wave order when @xmath101 , while dominated by the s - wave order when @xmath102 . when @xmath103 , the @xmath104 coexisting phase dominates . the total charge density as a function of temperature . the red curve is for the normal phase , while the blue one corresponds to the superconducting phase . there are three special temperatures at which the derivative of charge density with the temperature are discontinuous.,title="fig : " ] as we have seen , for suitable @xmath14 , the coexisting phase can appear . once the coexisting solution exists , it is thermodynamically favored , compared to the pure s - wave and pure d - wave phases . next we give a further investigation on the @xmath104 coexisting phase . we study the behavior of the charge density and the ratio of the superconducting charge over the total charge density @xmath105 with respect to the temperature . our numerical results are summarized in figure [ rho ] and [ ratio_1 ] . from figure [ rho ] , it can be seen clearly that there exist three particular points at which the derivative of the charge density with respect to temperature is discontinuous , indicating a second order phase transition . the one with the highest temperature is the critical point for the superconducting phase transition , while the remaining two points are inside the superconducting phase , indicating the appearance and disappearance of coexisting phase . the ratio of the superconducting charge density over the total charge density @xmath105 versus temperature . the red curve describes the ratio @xmath105 when the system transfers from the d - wave phase to the s - wave phase through the s+d coexisting phase . the green dashed curve is for the ratio @xmath105 of the pure s - wave phase and the blue dashed curve is the ratio for the pure d - wave phase.,title="fig : " ] we can also see the signal of phase transition from the behavior of the ratio @xmath105 versus temperature in figure [ ratio_1 ] . the superconducting charge density @xmath106 can be obtained following ref . more precisely , the normal charge density is carried by the black hole and can be read from the electric field at the horizon @xmath107 . the total charge density is just @xmath44 in ( [ boundary condition_1 ] ) , determined by the gauge field at the ads boundary . thus we can obtain the superconducting charge density @xmath108 . the ratio @xmath105 has a small kink in the region for the coexisting phase . when we lower the temperature , the ratio @xmath105 increases in the s+d coexisting phase . in order to ensure the system is indeed in a superconducting state , and to see whether there are any new phenomena occurring in such coexisting phase , we would like to calculate the optical conductivity @xmath109 . to compute the frequency dependent conductivity in the @xmath110-direction , we consider a set of self - consistent time dependent fluctuations of the fields @xmath111 , @xmath112 , @xmath113 , @xmath114 and @xmath115 . the coupled linearized algebra - differential equations for the @xmath116 component of these perturbations are [ e : eomlinear ] @xmath117 \nonumber \\ & \quad - \frac{iq_2\psi_2}2 \ , \big ( { \varphi_{zy , z}^ * } - \varphi_{zy , z } \big ) + \frac{iq_2}{2f } \ , ( \psi_{2,z } f - \psi_2 f_{,z } ) \big ( \varphi_{zy}^ * - \varphi_{zy } \big ) -\frac{2 q_1 ^ 2 a_x \psi_1 ^ 2}{z^2 f}\ , \\ \label{e : eomty } 0 & = \varphi_{ty , zz } + \frac2z \ , \varphi_{ty , z } - \frac{2f + m_2 ^ 2}{z^2 f } \ , \varphi_{ty } + \frac{q_2\omega + 2q_2 ^ 2 \phi}{4z^2 f } \ , \psi_2 a_x + \frac i2\ , \big [ 2(\omega + q_2 \phi ) \varphi_{zy , z } + q_2 \phi_{,z } \varphi_{zy } \big ] \ , \\ \label{e : eomrzy } 0 & = \big [ ( \omega + q_2\phi)^2 z^2 - m_2 ^ 2 f \big ] \ , \varphi_{zy } + \frac i4\ , q_2 f \psi_2 a_{x , z } + \frac i2 q_2 f \psi_{2,z } a_x \nonumber \\ & \quad - i(\omega + q_2\phi ) z^2 \varphi_{ty , z } - \frac i2 \ , \big [ 4(\omega + q_2 \phi)z + q_2 \phi_{,z } z^2 \big ] \ , \varphi_{ty } \ , \end{aligned}\ ] ] where we have made a coordinate transformation @xmath63 and all quantities in above equations of motion are expressed in terms of coordinate @xmath118 . the equations for @xmath119 and @xmath120 are obtained by complex conjugation and an additional transformation @xmath121 to @xmath122 . the functions @xmath114 and @xmath120 can be eliminated from the first two equations using ( [ e : eomrzy ] ) , leaving three coupled differential equations for @xmath111 , @xmath112 and @xmath119 . the boundary conditions we impose on are as follows . since the conductivity is related to the retarded green s function for the charge current , we should impose the ingoing boundary condition for each fluctuation near the black hole horizon @xmath123 , i.e. , @xmath111 , @xmath112 and @xmath119 have the behavior as @xmath124 near the boundary @xmath64 , the asymptotical behavior for the perturbation variables @xmath111 , @xmath112 and @xmath113 is @xmath125 where @xmath126 . here @xmath127 and @xmath128 are the sources of the perturbation fields . after looking for solutions where the source term in the series expansion of @xmath112 and @xmath113 vanishes , one can obtain the conductivity as @xmath129 the real part ( left ) and imaginary part ( right ) of the conductivity as a function of frequency at temperature @xmath130 . the red curve is for the s+d coexisting phase , the green line is for the pure s - wave phase and the blue curve for the pure d - wave phase . , title="fig : " ] the real part ( left ) and imaginary part ( right ) of the conductivity as a function of frequency at temperature @xmath130 . the red curve is for the s+d coexisting phase , the green line is for the pure s - wave phase and the blue curve for the pure d - wave phase . , title="fig : " ] the numerical results for the conductivity are shown in figure [ aa ] . the green , blue and red curves are for the pure s - wave , pure d - wave and the s+d coexisting phases , respectively . for sufficiently large frequency , re(@xmath131 ) has a very simple behavior . much more interesting phenomena happen in the low frequency region . unlike the s - wave case which only has a bump at @xmath132 in figure [ aa ] , for pure d - wave condensate , apart from a much more obvious bump at @xmath133 , re(@xmath131 ) has an additional spike at a lower frequency . in pure d - wave phase , one can observe two spikes in the conductivity in the small frequency region ( see figure 2 in ref . however , for the case with @xmath134 in this paper , we can only see one spike . our results suggest that the behavior of conductivity for the bhry d - wave model @xcite might depend on the mass of the tensor field . ] this spike may indicate the existence of a bound state @xcite . one can see clearly that the peak becomes much more sharp in the s+d coexisting state , thus the bound state is enhanced due to the additional condensate of s - wave order . in addition , the real part of the conductivity has a direct delta function at @xmath135 since the imaginary part of conductivity shown in figure [ aa ] has a pole at the origin . the calculations of free energy and the conductivity uncover that the coexisting phase is indeed a thermodynamically favored superconducting phase once it appears . however , we do not rule out the possibility that only one order parameter exists for other choices of @xmath14 . thus , it is helpful to construct the phase diagram in terms of temperature @xmath136 and the charge @xmath14 , which tells us in which region the coexisting phase appears . let us first make a qualitative discussion from the side of the free energy . since we fix @xmath137 and @xmath138 and set @xmath18 , the critical temperature from the normal phase to the s - wave superconducting phase is fixed , while the critical temperature for the d - wave superconducting transition is proportional to @xmath14 . for sufficiently large @xmath14 , the critical temperature of the d - wave condensate is higher than the s - wave case , so as one lowers the temperature , the d - wave order will condense first . the s - wave order can only condense behind the d - wave order . however , we have checked that in this case the free energy of the s - wave case is always larger than the one for the d - wave phase . thus , one can expect that the s - wave order would not dominate the system for very large @xmath14 . on the other hand , for small enough @xmath14 , the s - wave order condenses before the d - wave order . as one lowers @xmath14 , the critical temperature of d - wave order decreases and the free energy of the d - wave order becomes higher and higher and will finally be always larger than the one for the s - wave condensate . therefore , there can be only s - wave order condensation for sufficiently small @xmath14 . for intermediate range of @xmath14 , as we show in figure [ free ] , the free energy for s - wave case and the one for d - wave case has an intersection at some temperature . if no new phase appears , there should be a first order phase transition from one order to the other order . nevertheless , the competition between two orders results in the state with both orders coexisting near the crossing point ( see figure [ co ] ) . to summarize , in the @xmath139 phase diagram , the phase boundary between the pure s - wave phase and normal phase should be a line parallel to the @xmath14 axis , and the line separating the pure d - wave phase from normal phase is a straight line passing through the original point @xmath140 . the s+d coexisting phase can only appear in some region of @xmath14 , above which there is only d - wave order , while below which there is only s - wave order . the precise boundary among different phases can only be determined by numerical calculation . the @xmath14-@xmath136 phase diagram . the four phases are colored differently and we label the most thermodynamically favored phase in each region.,title="fig : " ] the complete phase diagram with @xmath0 and @xmath88 is constructed in figure [ phase_1 ] . the phase diagram is divided into four parts and the corresponding phase we named in each region is the most thermodynamically favored phase . indeed , the phase boundary between the normal phase and pure d - wave phase ( s - wave phase ) is a straight line . the red curve describes the phase transition between the s - wave phase and the s+d coexisting phase . this phase transition occurs when the single s - wave phase becomes unstable to developing a d - wave hair . therefore , we can derive this red curve using the d - wave as a perturbation on the s - wave superconducting background @xcite . with the same method , we draw the blue curve which corresponds to the phase transition between @xmath104 coexisting phase and d - wave phase . the full phase diagram is divided into four phases by these curves as boundaries . from figure [ phase_1 ] , we see that the coexisting phase exists only in a narrow region in the phase diagram . we denote the critical temperature for a single s - wave or d - wave starting to condense as @xmath141 and @xmath142 . if we set the charges of the s - wave and d - wave fields to unity , then @xmath143 and @xmath144 . we see that * in the regime @xmath145 and @xmath146 , the @xmath104 coexisting phase does not exist at all . this is different from the case in ref . @xcite . in ref . @xcite , for the coupling between the scalar field and the tensor field @xmath147 , the @xmath104 coexisting phase indeed exists , but the free energy of such coexisting phase is larger than those of single order . thus , from the point of view of thermodynamics , the @xmath104 coexisting phase ceases to exist for @xmath147 in ref . , the s - wave dominates the system and there is no condensation of the d - wave order . * as @xmath14 increases beyond @xmath148 , the @xmath104 phase appears , which emerges from the d - wave phase . more precisely , as we continue to lower the temperature to @xmath95 , the s - wave order begins to condense , while the condensate of the d - wave order decreases , resulting in the phase with both orders ; if one further lowers the temperature to @xmath149 , the d - wave condensate quickly goes to zero ; when the temperature is lower than @xmath149 , there exists only the condensate of s - wave order . there are three second order phase transition as we lower the temperature of the system . the first phase transition happens at @xmath94 when the d - wave order condenses . the second phase transition is at @xmath95 when s - wave order starts condensing . the third one occurs at @xmath96 when the condensation of the d - wave order becomes vanishing . * if we continue increasing @xmath14 to the case @xmath146 , the s - wave order never condenses and the resulting phase diagram is the same as that of model with only d - wave order . finally , we try to give a qualitative explanation on the mechanism through which the condensation of one order affects the dynamics of the other order . note that here the back reaction is not taken into account . thus the two fields interact only through their effect on the gauge field once one or both has ( have ) condensed . through looking at the gauge field we may give some insight into the competing mechanics between two orders . the blue curve is the effective mass square of s - wave without the condensation of d - wave . the red curve is the effective mass square of the s - wave under the condensation of d - wave . it can be seen clearly that the effective mass of s - wave increases after the condensation of d - wave.,title="fig : " ] * first , after the d - wave order condenses , if one keeps lowering the temperature and reaches the critical temperature at which the pure s - wave would condense , this condensation does not happen . this is due to the fact that the condensation of the d - wave increases the effective mass of the s - wave , thus prevents the instability of the s - wave to happen , which can be seen from figure [ eff_mass ] . this reflects the competition between s - wave and d - wave . * however , if we go on decreasing the temperature , the condensation of s - wave does happen . this is due to the fact that the effective mass of the s - wave is lowered and ultimately even if the condensation of the d - wave depleted the gauge potential , the background with only d - wave order becomes unstable . * at last , the condensate of the s - wave order kills the first one . this may be thanks to the effective mass of the s - wave being lower . it should be noted that this phenomenon is model dependent . this narrow coexistence region of two superconducting orders and the fact that one condensate can eventually kill the other also happen for two s - wave orders in ref . @xcite and @xmath150 case in ref . however , it should be noted that this is not the case for the @xmath151 phase studied in ref . @xcite and the double s - wave scenario in ref . @xcite , since in both cases the coexisting phase survives even down to a low temperature . furthermore , the competition diagram here is similar to the competition between the conventional s - wave and the triplet balian - werthamer or the b - phase pairings in the doped three dimensional narrow gap semiconductors , such as @xmath152 and @xmath153 in the condensed matter system @xcite . although in ref . @xcite the competition is apparently between a s - wave order and a p - wave order , d - wave and p - wave are similar in some circumstances , for example , their excitations of the normal component can be probed using low frequency photons . with the same method , we generalize our above analysis to the case with different masses . for convenient , we keep the mass square of the d - wave @xmath88 unchanged . we increase the mass square of @xmath16 up to @xmath154 . we give the parameter space for the @xmath104 coexisting phase with d - wave condensed first in figure [ diff_mass ] . figure [ diff_mass ] can be derived as follows . here we want to find the critical ratio @xmath14 such that @xmath136 is a critical temperature at which the d - wave order @xmath33 begins to vanish . at such a temperature , @xmath33 is very tiny and can be treated as a perturbation on the background where only @xmath3 condenses , i.e. , @xmath155 where the profile of @xmath32 comes from the hairy ads black hole with only @xmath3 condensed . we demand @xmath33 to be regular at the horizon and to fall off as in near the ads boundary . then this equation can be considered as an eigenvalue problem with positive eigenvalue @xmath156 . the numerical result for the lowest eigenvalue versus temperature is presented in figure [ diff_mass ] . every point in each curve gives the value of @xmath14 and the corresponding temperature below which the d - wave order tends to vanish . from upper to down , different curves correspond to @xmath157 , and @xmath158 , respectively . we clearly see that as the mass square @xmath16 of s - wave increases , the maximal critical temperature brings down . we also find that the value of @xmath14 for a @xmath104 coexisting phase lowers when @xmath16 increases . especially when @xmath154 , the value of @xmath14 is always one , which corresponds to the orange line in figure [ diff_mass ] . this is due to the symmetry of equations mentioned before . with the symmetry , we have @xmath159 . gives the parameter space for the @xmath104 coexisting phase with the d - wave condensed first . by the symmetry , we can easily obtain the opposite solution , which is also a coexisting phase but with the s - wave condensed first . ] the parameter space for @xmath104 coexisting phase with d - wave condensed first . here @xmath88 is fixed and different curves from upper to down correspond to @xmath160 , respectively . for each @xmath16 , we plot the ratio @xmath14 as a function of the critical temperature below which the d - wave order becomes vanishing.,title="fig : " ] with the same strategy , in this section we study the competition between s - wave order and d - wave order in the model combining the abelian - higgs s - wave model @xcite with the ckmwy d - wave model @xcite . the full action including a u(1 ) gauge field @xmath5 , a complex scalar field @xmath3 and a symmetric , traceless tensor field @xmath161 takes the following form @xmath162 with @xmath163 here @xmath164 and @xmath8 . in the probe limit , matter fields can be treated as perturbations in the 3 + 1 dimensional ads black hole background . let us consider the following ansatz @xmath165 with all other field components being turned off and @xmath30 , @xmath31 and @xmath29 being real functions . then the explicit equations of motion are @xmath166 we use the shooting method to solve the coupled equations of motion . most of our calculations are the same as those in section [ sect : superfluid ] , for the sake of brevity , we will omit the details about numerical analysis . in this section , we set @xmath18 , @xmath0 and @xmath167 . the general fall - off of the matter fields near the boundary @xmath38 behaves as @xmath168 where @xmath169 and @xmath170 . to break the u(1 ) symmetry spontaneously , we should turn off the source terms , i.e. , @xmath171 , then @xmath45 and @xmath46 are the vacuum expectation values of dual operators , which play the role of order parameters in the boundary field theory . so far , the holographic superconducting model with an s - wave order and a d - wave order has been constructed . we are interested in the competition between s - wave and d - wave orders . we will study the phase structure and the behaviors of the thermodynamical quantities . we emphasize the similarity and difference between the model here and the one proposed in the previous section . we take @xmath172 as an example and summarize our numerical results as follows . first , we investigate all possible phases . as the model in the previous section , except for the normal phase , there are three additional superconducting phases , the pure s - wave superconducting phase , the pure d - wave superconducting phase and the s+d coexisting phase . we plot the s - wave condensate and the d - wave condensate in figure [ sandd ] . the left plot is the s - wave condensate with @xmath3 , while the right plot is the d - wave condensate with @xmath33.,title="fig : " ] the left plot is the s - wave condensate with @xmath3 , while the right plot is the d - wave condensate with @xmath33.,title="fig : " ] we also calculate the gibbs free energy of the model to judge whether the @xmath104 coexisting phase is thermodynamically favored or not . the expression for the free energy turns out to be the same as . we show the condensation of the s+d coexisting phase in the left panel and its gibbs free energy in the right panel in figure [ condensate_2 ] . for the case with @xmath173 , the curves of the free energy for the s - wave phase and d - wave case have an intersection at a temperature , say @xmath174 . since the critical temperature of the d - wave superconducting transition is higher than the one of the s - wave case , the d - wave phase will first appear . we can see that the free energy for the d - wave phase is lower than the s - wave phase when @xmath175 , while it becomes larger than the s - wave phase when @xmath176 . one expects that there should be a transition from the d - wave phase to the s - wave phase . the let plot shows the condensation in the s+d coexisting phase . the right plot shows the differences of gibbs free energy between superconducting phases and the normal phase . here the blue line stands for the d - wave phase , the green one for the s - wave phase and the red one for the s+d coexisting phase.,title="fig : " ] the let plot shows the condensation in the s+d coexisting phase . the right plot shows the differences of gibbs free energy between superconducting phases and the normal phase . here the blue line stands for the d - wave phase , the green one for the s - wave phase and the red one for the s+d coexisting phase.,title="fig : " ] indeed , as we can see in figure [ condensate_2 ] a new phase with both s - wave order and d - wave order coexistence can appear near @xmath174 . we find that this s+d coexisting phase has the lowest free energy and is thus thermodynamically preferred to the s - wave phase and d - wave phase . in more detail , as we lower the temperature of the system , it first undergoes a phase transition from the normal phase to the pure d - wave phase at @xmath177 . then at @xmath95 , a new phase transition occurs , and the system goes into an s+d coexisting phase . at last the system undergoes the third phase transition from the s+d coexisting phase to a pure s - wave phase at @xmath96 . note that all the three phase transitions are second order . the temperature region for the s+d wave coexisting phase is very narrow , which is similar to the previous model in section [ sect : superfluid ] . the feature of the phase transitions can also be seen clearly from the charge density as the function of temperature in figure [ charge_2 ] . we find that the charge density with respect to temperature is continuous , but its derivative is discontinuous at three special points , indicating three second order phase transitions . the first transition from the normal phase to the d - wave superconducting phase occurs at the highest critical temperature . the other two transitions from the d - wave to s+d coexisting phase and from the s+d phase to the s - wave case occur inside the superconducting phase . these features are the same as those for the model in the previous section . but there is a little difference in the behavior of the total charge density for the d - wave phase . in the s - wave + bhry d - wave model , the total charge density changes monotonously with the temperature , while it behaves non - monotonous in the present model . the total charge density as a function of the temperature . the red curve is for the normal phase , while the blue one corresponds to the superconducting phase . there are three special temperatures at which the derivatives of charge density with the temperature are discontinuous.,title="fig : " ] the ratio of the superconducting charge density over the total charge density , @xmath105 , with respect to the temperature . the red curve describes the ratio @xmath105 when the system transfers from the d - wave phase to s - wave phase through the s+d coexisting phase . the green dashed blue curve is for the ratio @xmath105 of the pure s - wave phase and the blue dashed curve is the ratio for the pure d - wave phase.,title="fig : " ] the information of the phase transitions can also be revealed via the behavior of the ratio @xmath105 with respect to the temperature . from figure [ ratio_2 ] , one can see that the ratio @xmath105 also has a small kink in the region of the coexisting phase . comparing figure [ ratio_1 ] with figure [ ratio_2 ] , we see that in the former case , the green dashed curve for the pure s - wave phase intersects with the blue dashed curve for the pure d - wave phase . in contrast , the green dashed curve in figure [ ratio_2 ] is always lower than the blue dashed curve . therefore , as one lowers the temperature , the ratio @xmath105 in the s+d coexisting phase increases for the former , while it decreases for the latter . the authors of ref . @xcite investigated an s+p coexisting phase and found the decrease of the ratio @xmath105 in the coexisting phase , similar to figure [ ratio_2 ] . they suggested that it might be an experimental signal of the phase transition from a single condensate phase to a coexisting phase . nevertheless , our results uncover that the ratio @xmath105 versus temperature is model dependent . by adopting the same procedure as in section [ sect : superfluid ] , we construct the phase diagram for the model with @xmath0 and @xmath167 in the @xmath178 plane in figure [ phase_2 ] . as the s - wave + bhry d - wave model , the system also contains four kinds of phases known as the normal phase , s - wave phase , d - wave phase and s+d coexisting phase . the normal phase dominates in the high temperature region , the s - wave phase dominates in the lower temperature region with small @xmath14 below the red curve , and the d - wave phase dominates in the higher temperature zone with large @xmath14 above the blue curve . the s+d coexisting phase is favored in the area between the red and blue curves . the region for the @xmath104 coexisting phase is very narrow in the phase diagram , which indicates that the s - wave and d - wave phases generally repel each other , but they can coexist in a very small range of temperature . the @xmath14-@xmath136 phase diagram with @xmath0 and @xmath167 . we label the most thermodynamically favored phase in each part . the s+d coexisting phase exists only in a narrow region . the right plot is an enlarged version for the coexisting region in order to see this more clearly.,title="fig : " ] the @xmath14-@xmath136 phase diagram with @xmath0 and @xmath167 . we label the most thermodynamically favored phase in each part . the s+d coexisting phase exists only in a narrow region . the right plot is an enlarged version for the coexisting region in order to see this more clearly.,title="fig : " ] it is clear from the equations that the s - wave and d - wave orders now see different effective potentials . therefore , the analytical discussion in the previous section in terms of effective potential can give little useful information . we have to resort to numerical methods to find possible solutions . here we keep @xmath16 of the scalar unchanged and increase @xmath17 of the d - wave order , which lowers the critical temperature of a single d - wave condensation . we hope to find the coexisting phase with s - wave order condensed before d - wave one by increasing the mass square @xmath17 . with the same method as done in figure [ diff_mass ] , the parameter space for the @xmath104 coexisting phase where d - wave order condenses first is shown in figure [ mass_2 ] . nevertheless , from figure [ mass_2 ] we see that the value of @xmath14 ( which can indicate the appearance of @xmath104 coexisting phase ) increases when @xmath17 is increased . a larger @xmath14 in turn makes the critical temperature of d - wave condensation much more higher . the behavior here is obviously different from the one in previous model ( see figure [ diff_mass ] ) . therefore , it comes as no surprise that we do not find the @xmath104 coexisting solution for which s - wave order condenses before d - wave order . the parameter space for a @xmath104 coexisting phase where d - wave order condenses first . here we keep @xmath0 . the curves from upper to down corresponds to @xmath179 , respectively . for each @xmath17 , we give the ratio @xmath14 as a function of the critical temperature below which the d - wave order tends to vanish.,title="fig : " ] the competition between the s - wave condensate and the d - wave condensate has been studied through two holographic models . the dynamics of s - wave order dual to a complex scalar field is described by the abelian - higgs model . the dynamics of the d - wave order corresponding to a symmetric , traceless spin two tensor field , is determined by the bulk action from bhry d - wave model @xcite or ckmwy d - wave model @xcite . in our study , we did not include the direct interaction between the scalar field and the tensor field in the bulk , but , they interact with each other through the u(1 ) gauge field . note that including a direct interaction between them is equivalent to changing the effective masses of the scalar field and tensor field . based on these , we give some qualitative explanation on the completing scenario in our cases . working in the probe limit , we are left with three model parameters , i.e. , the mass square @xmath16 of scalar field , the mass square @xmath17 of tensor field and the charge ratio @xmath15 , where @xmath14 is the charge for the d - wave order and @xmath13 is for the s - wave order . without loss of generality , we have set @xmath13 to be unity in the numerical calculations . based on our analysis , there are similarity and difference between two holographic setups , i.e. , the model and the model . the common features are as follows : * the s+d coexisting phase does exist in a region of the model parameter @xmath15 . once the coexisting phase appears , it is always thermodynamically favored , compared to the pure s - wave and pure d - wave superconducting phases , which can be seen from the free energy in figure [ free ] and figure [ condensate_2 ] . * the phase transition from the coexisting phase to the phase with single order is second order , which can be seen from the charge density versus temperature in figure [ rho ] and figure [ charge_2 ] . in fact , all phase transitions are second order in these two holographic models . * one can see from figure [ phase_1 ] and figure [ phase_2 ] that the phase structure is very similar for both models . the region for the s+d coexisting phase is very narrow in the phase diagram , indicating that the s - wave and d - wave phases generally repel each other . there exist also some differences in the two models . comparing to equations and , we see that the first model exhibits an useful symmetry . taking advantage of this symmetry , one can only consider the case with @xmath36 . our numerical calculations uncover that , for suitable model parameters , as the temperature is lowered , the s - wave order condenses inside the d - wave order resulting in the coexisting phase with both orders . however , when the scalar order condenses the first one starts to disappear , and finally only the s - wave condensate is left for sufficiently low temperatures . if we change the model parameter @xmath180 , the inverse is also true : the condensate of d - wave order emerges following the condensate of s - wave order , and then the d - wave condensate finally kills the s - wave order . those two kinds of coexisting phase are one to one correspondence . and @xmath17 both non - negative , since the unitary bound requires the mass square of @xmath4 should be non - negative . ] in contrast , in the second model , we only find the first kind of the coexisting phase . what s more , for the first model , the ratio @xmath105 increases in the s+d coexisting phase as the temperature is lowered , while it decreases in the second case . this gives an obvious evidence that the ratio @xmath105 versus temperature is model dependent . the optical conductivity in the s+d coexisting phase was calculated for the s - wave+bhry d - wave model . we found a remarkable spike in the low frequency region , compared to the case for the pure d - wave superconducting phase , this is due to the additional condensation of the s - wave order in the coexisting phase . in both models , the s+d coexisting phase is narrow and one condensation tends to kill the other . this is similar to the situation of the coexisting phase with two s - wave orders @xcite as well as the case with one s - wave order and one p - wave order @xcite . this competing behavior is similar to the case shown in the condensed matter system @xcite . however , it should note that the competing scenario is model dependent . the cases in ref . @xcite and ref . @xcite are different from here . in these two cases , the condensates feed on different charge densities and the coexisting phase survives down to a low temperature . note that as found in ref . @xcite , including the back reaction of matter fields would lead to a much rich phase structure for two s - wave orders model . therefore , it will be desirable to study a consistent s+d holographic superconducting model with back reaction , although it would be a challenge in some sense due to the complexity of the spin two field theory in curved spacetime . nevertheless , we may overcome some difficulty by an effective model with a well - chosen ansatz . we will leave this for further study . l.f.l thanks c. p. herzog , a. yarom , d. marity , k. y. kim , z.y . fan and z.y . nie for various helps . this work was supported in part by the national natural science foundation of china ( no.10821504 , no.11035008 , no.11375247 , no.11205226 and no.41231066 ) , and in part by the ministry of science and technology of china under grant no.2010cb833004 . l.f.l would like to appreciate the specialized research fund for state key laboratories of cas . j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' adv . * 2 * , 231 ( 1998 ) [ int . j. theor . * 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] . s. s. gubser , i. r. klebanov and a. m. polyakov , `` gauge theory correlators from non - critical string theory , '' phys . b * 428 * , 105 ( 1998 ) [ arxiv : hep - th/9802109 ] . e. witten , `` anti - de sitter space and holography , '' adv . * 2 * , 253 ( 1998 ) [ arxiv : hep - th/9802150 ] . s. a. hartnoll , c. p. herzog and g. t. horowitz , `` building a holographic superconductor , '' phys . rev . * 101 * , 031601 ( 2008 ) [ arxiv:0803.3295 [ hep - th ] ] . s. s. gubser , `` breaking an abelian gauge symmetry near a black hole horizon , '' phys . rev . d * 78 * , 065034 ( 2008 ) [ arxiv:0801.2977 [ hep - th ] ] . s. a. hartnoll , c. p. herzog and g. t. horowitz , `` holographic superconductors , '' jhep * 0812 * , 015 ( 2008 ) [ arxiv:0810.1563 [ hep - th ] ] . s. s. gubser and s. s. pufu , `` the gravity dual of a p - wave superconductor , '' jhep * 0811 * , 033 ( 2008 ) [ arxiv:0805.2960 [ hep - th ] ] . r. -g . cai , s. he , l. li and l. -f . li , `` a holographic study on vector condensate induced by a magnetic field , '' jhep * 1312 * , 036 ( 2013 ) [ arxiv:1309.2098 [ hep - th ] ] . r. -g . cai , l. li and l. -f . li , `` a holographic p - wave superconductor model , '' jhep * 1401 * , 032 ( 2014 ) [ arxiv:1309.4877 [ hep - th ] ] . r. -g . cai , l. li , l. -f . li and y. wu , `` vector condensate and ads soliton instability induced by a magnetic field , '' jhep * 1401 * , 045 ( 2014 ) [ arxiv:1311.7578 [ hep - th ] ] . r. -g . cai , l. li , l. -f . li and r. -q . yang , `` towards complete phase diagrams of a holographic p - wave superconductor model , '' jhep * 1404 * , 016 ( 2014 ) [ arxiv:1401.3974 [ gr - qc ] ] . j. -w . chen , y. -j . kao , d. maity , w. -y . wen and c. -p . yeh , `` towards a holographic model of d - wave superconductors , '' phys . d * 81 * , 106008 ( 2010 ) [ arxiv:1003.2991 [ hep - th ] ] . f. benini , c. p. herzog , r. rahman and a. yarom , `` gauge gravity duality for d - wave superconductors : prospects and challenges , '' jhep * 1011 * , 137 ( 2010 ) [ arxiv:1007.1981 [ hep - th ] ] . j. -w . chen , y. -s . liu and d. maity , `` @xmath181 holographic superconductors , '' jhep * 1105 * , 032 ( 2011 ) [ arxiv:1103.1714 [ hep - th ] ] . k. -y . kim and m. taylor , `` holographic d - wave superconductors , '' arxiv:1304.6729 [ hep - th ] . j. zaanen , `` a modern , but way too short history of the theory of superconductivity at a high temperature , '' arxiv:1012.5461 [ cond-mat.supr-con ] . s. fujimoto , `` electron correlation and pairing states in superconductors without inversion symmetry , '' j. phys . jpn.*76 * , 051008 ( 2007 ) , [ arxiv : [ cond-mat.supr-con ] . p. basu , j. he , a. mukherjee , m. rozali and h. -h . shieh , `` competing holographic orders , '' jhep * 1010 * , 092 ( 2010 ) [ arxiv:1007.3480 [ hep - th ] ] . r. -g . cai , l. li , l. -f . li and y. -q . wang , `` competition and coexistence of order parameters in holographic multi - band superconductors , '' jhep * 1309 * , 074 ( 2013 ) [ arxiv:1307.2768 [ hep - th ] ] . z. -y . nie , r. -g . cai , x. gao and h. zeng , `` competition between the s - wave and p - wave superconductivity phases in a holographic model , '' jhep * 1311 * , 087 ( 2013 ) [ arxiv:1309.2204 [ hep - th ] ] . c. -y . huang , f. -l . lin and d. maity , `` holographic multi - band superconductor , '' phys . b * 703 * , 633 ( 2011 ) [ arxiv:1102.0977 [ hep - th ] ] . a. krikun , v. p. kirilin and a. v. sadofyev , `` holographic model of the @xmath182 multiband superconductor , '' jhep * 1307 * , 136 ( 2013 ) [ arxiv:1210.6074 [ hep - th ] ] . f. nitti , g. policastro and t. vanel , `` dressing the electron star in a holographic superconductor , '' jhep * 1310 * , 019 ( 2013 ) [ arxiv:1307.4558 [ hep - th ] ] . y. liu , k. schalm , y. -w . sun and j. zaanen , `` bose - fermi competition in holographic metals , '' jhep * 1310 * , 064 ( 2013 ) [ arxiv:1307.4572 [ hep - th ] ] . a. amoretti , a. braggio , n. maggiore , n. magnoli and d. musso , `` coexistence of two vector order parameters : a holographic model for ferromagnetic superconductivity , '' jhep * 1401 * , 054 ( 2014 ) [ arxiv:1309.5093 [ hep - th ] ] . a. donos , j. p. gauntlett and c. pantelidou , `` competing p - wave orders , '' class . grav . * 31 * , 055007 ( 2014 ) [ arxiv:1310.5741 [ hep - th ] ] . m. nishida , `` phase diagram of a holographic superconductor model with s - wave and d - wave , '' arxiv:1403.6070 [ hep - th ] . r. -g . cai , s. he , l. li and y. -l . zhang , `` holographic entanglement entropy on p - wave superconductor phase transition , '' jhep * 1207 * , 027 ( 2012 ) [ arxiv:1204.5962 [ hep - th ] ] . |
it has been argued that the quantum ( conformal ) trace anomaly could potentially provide us with a dynamical explanation of the cosmological constant problem . in this paper , however , we show by means of a semiclassical analysis that the trace anomaly does not affect the cosmological constant . we construct the effective action of the conformal anomaly for flat flrw spacetimes consisting of local quadratic geometric curvature invariants . counterterms are thus expected to influence the numerical value of the coefficients in the trace anomaly and we must therefore allow these parameters to vary . we calculate the evolution of the hubble parameter in quasi de sitter spacetime , where we restrict our hubble parameter to vary slowly in time , and in flrw spacetimes . we show dynamically that a universe consisting of matter with a constant equation of state , a cosmological constant and the quantum trace anomaly evolves either to the classical de sitter attractor or to a quantum trace anomaly driven one . when considering the trace anomaly truncated to quasi de sitter spacetime , we find a region in parameter space where the quantum attractor destabilises . when considering the exact expression of the trace anomaly , a stability analysis shows that whenever the trace anomaly driven attractor is stable , the classical de sitter attractor is unstable , and vice versa . semiclassically , the trace anomaly does not affect the classical late time de sitter attractor and hence it does not solve the cosmological constant problem . | 0803.4000 | recent observations have clearly indicated that the expansion of the universe is accelerating . according to einstein s general relativity , this can only be realised if the pressure of the dominant component of the current universe is negative . these observations have triggered a renewed interest in the cosmological constant problem ( for recent reviews , see e.g. @xcite ) . what is usually referred to as the `` old '' cosmological constant problem can be phrased as follows : why is the measured ( effective ) cosmological constant extremely close to zero ? one approach dealing with the cosmological constant problem is concerned with employing the effective field theory of gravity @xcite . lacking a full quantum theory of gravity , an effective field theory of gravity adopts the following point of view : in order to describe quantum phenomena at very large and cosmologically relevant distances , the precise physics at the shortest distance scales is irrelevant . in other words , the effective field theory of gravity is the low energy limit of quantum gravity . it combines classical general relativity with knowledge of quantum field theory in curved spacetimes @xcite . in order to describe these long distance effects accurately , one supplements the classical einstein - hilbert action with certain additional contributions . one of these additions is the trace anomaly or conformal anomaly which quantum field theories are known to exhibit @xcite . if the classical action is invariant under conformal transformations of the metric , the resulting stress - energy tensor is traceless . as an explicit example , one can easily verify that the trace of a massless , conformally coupled scalar field vanishes . in quantum field theory the stress - tensor is promoted to an operator . a careful renormalisation procedure renders its expectation value @xmath0 finite . however , inevitably , the renormalisation procedure results in general in a non - vanishing trace of the renormalised stress - energy tensor . classical conformal invariance can not be preserved at the quantum level . ever since its discovery , the trace anomaly has found many applications in various areas in physics ( see e.g. @xcite ) . an alternative approach to the backreaction problem of quantum fluctuations on the background spacetime deals with quantum fields whose spectrum is nearly flat @xcite . consequently , the spectrum in the infrared is not suppressed and is therefore expected to yield a strong backreaction on the background spacetime . examples of such fields are the minimally coupled massless scalar and the graviton . some authors stated that the trace anomaly could have effects on dark energy and the cosmological constant problem @xcite , whereas it has been argued by other authors that the trace anomaly could potentially provide us with a dynamical explanation of the cosmological constant problem @xcite . broadly speaking , the line of reasoning is as follows ( for a more in - depth review , we refer to @xcite ) . the new , conformal degree of freedom is usually parametrised by : @xmath1 according to the authors of e.g. @xcite , the trace anomaly can not be generated from a local finite term in the action , but rather stems from a non - local effective action that generates the conformal anomaly by variation with respect to the metric @xcite . it is this genuine non - locality of the trace anomaly , revealing a large distance effect of quantum physics , that is at the very foundation of its connection with the effective field theory of gravity . one then argues that the new conformal field should dynamically screen the cosmological constant , thus solving the cosmological constant problem . the proposal advocated in @xcite is very interesting and should be investigated further . before studying the effect of a new conformal degree of freedom ( [ conformaltransformation ] ) , we feel that firstly a proper complete analysis of the dynamics resulting from the effective action of the trace anomaly should be performed . this is what we pursue in this paper . according to the cosmological principle the universe is homogeneous and isotropic on the largest and cosmologically relevant scales . the cmb measurements @xcite constrain the inhomogeneities at order @xmath2 . moreover , the universe appears to be spatially flat . let us make the following observations . firstly , the cosmological principle dictates the use of the conformally flat flrw metric @xmath3 . hence , inhomogeneous fluctuations of the metric tensor and in particular of the conformal part of the metric tensor ( [ conformaltransformation ] ) are observed and expected to be small at the largest scales , comparable to the hubble radius ( also in the early universe ) . secondly , we are led to an essentially semiclassical analysis . the vacuum expectation value of the stress energy tensor resulting in the trace anomaly has been calculated semiclassically . in a semiclassical analysis quantum fluctuations backreact on the background spacetime . phase transitions aside , quantum fluctuations naturally affect the homogeneous background _ homogenously_. it is a well - known fact that quantum fluctuations can break certain symmetries present in de sitter @xcite , e.g. time translation invariance . however , we are not aware of quantum fluctuations breaking the homogeneity and isotropy of the background spacetime . finally , if quantum fluctuations compensate for or screen the cosmological constant , we must have @xmath4 . let us set : @xmath5 . stress energy conservation and metric compatibility immediately yield : @xmath6 . hence we conclude that @xmath7 must be a constant : @xmath8 . only homogeneous vacuum fluctuations can compensate the cosmological constant . moreover , @xmath9 does not break any of the symmetries of a maximally symmetric spacetime . hence , this form can not be used to study dynamical backreaction . the arguments above motivate a semiclassical approach to examining the connection between the cosmological constant and the trace anomaly . note that we do not consider a new , conformal degree of freedom ( [ conformaltransformation ] ) . hence , we do certainly not exclude any possible effect this ( inhomogeneous ) conformal degree of freedom might have on the cosmological constant . however , it is plausible that in order to address the link between the cosmological constant and the trace anomaly , a semiclassical analysis suffices . another application of the conformal anomaly can be found in what has become known as trace anomaly induced inflation : in the absence of a cosmological constant , the trace anomaly could provide us with an effective cosmological constant . originally , starobinsky @xcite realised that quantum one loop contributions of massless fields can source a de sitter stage . subsequently , the theory of trace anomaly induced inflation received significant contributions from @xcite . if one includes a cosmological constant , the theory of anomaly induced inflation is plagued by instabilities , which we will also come to address . the modified starobinsky model as advocated by @xcite , takes advantage of these instabilities to account for a graceful exit from inflation . it is argued that supersymmetry breaking changes the degrees of freedom such that it destabilises the quantum anomaly driven attractor and simultaneously stabilises the classical de sitter attractor . improving on e.g. @xcite , we incorporate matter with a constant equation of state in the einstein field equations . we argue that it is simply inconsistent not to include matter . consider the following analogy : if we examine an empty universe with a cosmological constant only , there is no dynamics and the ( 00 ) einstein equation yields @xmath10 . matter drives the dynamics and @xmath11 can only be realised with @xmath12 . in the literature , if one solves the trace of the einstein field equations in a universe with a cosmological constant and trace anomaly , one solves , however , in reality for the dynamics in a universe filled with radiation . for it is only radiation with equation of state @xmath13 that does not contribute to the trace of the einstein field equation @xmath14 . this point has not been included in other papers . we will consider matter with constant but otherwise arbitrary equation of state @xmath15 , and not just ( implicitly ) radiation with @xmath13 . in this paper we show that the cosmological constant problem can not be solved by taking account of the trace anomaly alone . the outline of this paper is as follows . in section [ tracing the einstein field equations and the trace anomaly ] we recall the basics of the conformal anomaly and discuss how to study its effect on the evolution of the universe by tracing the einstein field equations . in section [ the effective action generating the trace anomaly ] , we derive the conformal anomaly from an effective action in flat homogeneous flrw spacetimes consisting of local quadratic geometric curvature invariants . since one usually adds infinite counterterms to cancel the radiative one loop divergences , we do not see any reason why we should exclude adding a gauss - bonnet counterterm to cancel the anomaly in flat flrw spacetimes . even though this term in the effective action is formally divergent , at the level of the equation of motion it yields a finite result . hence , the coefficients multiplying the curvature invariants in the trace anomaly are not uniquely specified by the anomaly . the physical coefficient , i.e. : the parameter that can be measured , receives contributions both from the trace anomaly and from possible counterterms cancelling divergences from the underlying ( and yet unknown ) fundamental theory . this motivates varying the coupling parameters multiplying the curvature invariants in the anomaly . we can thus study all possible effects of the anomaly on the evolution of our universe . in section [ the trace anomaly in quasi de sitter spacetime ] we study the evolution of a quasi de sitter universe in the presence of matter with constant equation of state , a cosmological constant and the trace anomaly . in quasi de sitter spacetime we assume , loosely speaking , that the hubble parameter is a slowly varying function of time . effectively , we truncate the expression of the exact trace anomaly and discard higher order derivative contributions . in section [ the trace anomaly in flrw spacetimes ] we generalise our analysis and study the evolution of an flrw universe again in the presence of matter with constant equation of state , a cosmological constant and the trace anomaly . we examine the exact trace anomaly and take all higher derivative contributions into account . as the dimensionality of the phase space increases , we must carefully perform a stability analysis of the late time asymptotes . the trace anomaly or the conformal anomaly in four dimensions is in general curved spacetimes given by @xcite : @xmath16 where : [ traceanomaly2 ] @xmath17 where as usual @xmath18 is the riemann curvature tensor , @xmath19 its dual , @xmath20 the weyl tensor and @xmath21 and @xmath22 the ricci tensor and scalar , respectively . note that @xmath23 is the gauss - bonnet invariant . the general expression for the trace anomaly can also contain additional contributions if the massless conformal field is coupled to other long range gauge fields ( see e.g. @xcite ) . finally , the parameters @xmath24 , @xmath25 and @xmath26 appearing in ( [ traceanomaly1 ] ) are dimensionless quantities multiplied by @xmath27 and are given by : [ traceanomaly3 ] @xmath28 where @xmath29 , @xmath30 and @xmath31 denote the number of fields of spin 0 , 1/2 and 1 respectively ( @xmath32 ) . it is important to note that @xmath33 whereas @xmath34 in general . it turns out that the coefficient @xmath26 is regularisation dependent and is therefore not considered to be part of the true conformal anomaly . we take this into account and study the effect of @xmath26 on the stability of the solutions we are about to derive . for definiteness , we will assume that these parameters take their standard model values : @xmath35 , @xmath36 and @xmath37 . note if we were to include right - handed neutrinos , @xmath38 . one could also examine the numerical value of the coefficients ( [ traceanomaly3 ] ) for the late time universe . today s massless particle is just the photon , hence @xmath39 , @xmath40 and @xmath41 . let us specialise to flat friedmann - lematre - robertson - walker or flrw spacetimes in which the metric is given by @xmath42 where @xmath43 is the scale factor of the universe in cosmic time @xmath44 . recall that a conformal transformation leaves the weyl tensor invariant . hence , in flrw spacetimes @xmath45 . given the flrw metric one can easily verify that : [ quadraticcurvature ] @xmath46 \label{quadraticcurvaturea } \\ r_{\mu\nu } r^{\mu\nu } & = & 12 \left [ \dot{h}^{2 } + 3(\dot{h}h^{2}+h^{4})\right ] \label{quadraticcurvatureb}\\ r_{\mu\nu\kappa\lambda } r^{\mu\nu\kappa\lambda } & = & 12 \left [ \dot{h}^{2 } + 2(\dot{h}h^{2}+h^{4})\right ] \label{quadraticcurvaturec}\\ \box r & = & - 6 \left [ \dddot{h } + 7\ddot{h}h + 4 \dot{h}^{2}+12\dot{h}h^{2}\right ] \label{quadraticcurvatured } \,.\end{aligned}\ ] ] hence , the exact expression for the trace anomaly in flrw spacetimes in four dimensions reads : @xmath47 to capture the leading order dynamics we work in quasi de sitter spacetime and allow for a mildly time dependent hubble parameter : @xmath48 i.e. : we assume that @xmath49 is both small and time independent . this would truncate the trace anomaly up to terms linear in @xmath50 yielding : @xmath51 we will examine both the exact form of the trace anomaly ( [ traceanomaly4 ] ) and its truncated form ( [ traceanomaly4b ] ) . truncating the expression for the trace anomaly is motivated by the following realisation . in general backgrounds we need a non - local effective action to generate the trace anomaly in the equation of motion . the non - locality at the level of the effective action corresponds to an expansion in derivatives at the level of the equation of motion . generally , higher derivative contributions in an equation of motion have the tendency to destabilise a system unless the initial conditions are highly fine - tuned . formally , this is known as the theorem of ostrogradsky and its relevance to cosmology is outlined , for example , in @xcite . note that when @xmath52 the truncated version is exact . we discuss this further in section [ case ii : b=2b/3 ] . finally , note that although we have truncated equation ( [ traceanomaly4 ] ) to obtain ( [ traceanomaly4b ] ) , equation ( [ traceanomaly4b ] ) is still covariant . from the einstein - hilbert action : @xmath53 where : @xmath54 the einstein field equations follow as usual as : @xmath55 of which the trace can easily be verified to be : @xmath56 where @xmath57 . if one considers an empty universe with a cosmological constant , the ( 00 ) einstein field equation acts as a constraint equation for the hubble parameter and one simply finds @xmath10 as usual . however , for a non - empty universe , the ( 00 ) einstein field equation becomes a dynamical constraint . the bianchi identity for the left - hand side of equation ( [ efe1 ] ) straightforwardly results in stress - energy conservation for the right - hand side : @xmath58 because of stress - energy conservation , the ( 00 ) and the ( ij ) components of the einstein field equations are not independent . therefore , any linear combination of the ( 00 ) and ( ij ) components of the einstein field equations combined with stress - energy conservation suffice to describe the time evolution of the hubble parameter . in particular , the trace equation ( [ efetrace1 ] ) and stress - energy conservation ( [ stressenergytensor1 ] ) contain all relevant dynamics for @xmath59 . let us set @xmath60 for the quantum field in @xmath61 and require that the classical field obeys the equation of motion . note that the quantum perturbation @xmath62 does not obey this equation of motion . we expand in terms of the quantum field and construct the effective action as usual : @xmath63\right ] & = & \exp\left[i s_{\mathrm{m}}[\phi_{\mathrm{cl}}]\right ] \int \mathcal{d}\varphi \exp\left [ i \left(\int_{x } \frac{\delta s_{\mathrm{m}}}{\delta \phi_{\mathrm{cl}}(x ) } \varphi(x ) + \frac{1}{2 } \int_{x , y } \frac{\delta^{2 } s_{\mathrm{m}}}{\delta \phi_{\mathrm{cl}}(x)\phi_{\mathrm{cl}}(y ) } \varphi(x)\varphi(y ) + \mathcal{o}(\varphi^{3 } ) \right)\right ] \nonumber \\ & = & \exp\left[i s_{\mathrm{m}}[\phi_{\mathrm{cl}}]+ i \gamma_{\mathrm{q}}[\phi_{\mathrm{cl}}]\right]\,.\end{aligned}\ ] ] the first contribution to the effective action corresponds to the classical part of the action and @xmath64 $ ] is the contribution to the effective action taking account of the vacuum fluctuations . the stress - energy tensor now follows as : @xmath65 = -\frac{2}{\sqrt{-g } } \frac{\delta } { \delta g^{\mu\nu } } \left(s_{\mathrm{m}}[\phi_{\mathrm{cl } } ] + \gamma_{\mathrm{q}}[\phi_{\mathrm{cl}}]\right ) \equiv t_{\mu\nu}^{\mathrm{c } } + t_{\mu\nu}^{\mathrm{q}}\,.\ ] ] hence , there are both classical and quantum contributions to the full stress - energy tensor . classically , from the equation of motion the scalar field obeys , we have : @xmath66 hence , from ( [ stressenergytensor1 ] ) we derive : @xmath67 concluding , due to stress - energy conservation at the classical level and for the full stress - energy tensor , we have derived stress - energy conservation for the quantum contributions as well . analogously to the classical stress - energy tensor , we can symbolically write : @xmath68 . combining : [ stressenergytensor5 ] @xmath69 yields : @xmath70 = - a^{4}(t)h(t ) t_{\mathrm{q}}(t)\,.\ ] ] we thus find ( identical to @xcite ) : [ stressenergytensor6 ] @xmath71 although in general spacetimes it is not possible to perform this integral , in cosmologically relevant flrw spacetimes we can @xcite . if we consider the exact form of the trace anomaly ( [ traceanomaly4 ] ) in flat flrw spacetimes , we can easily see that @xmath72 should be of the following form : @xmath73 + b''\left [ c_{5 } \ddot{h}h + c_{6 } \dot{h}^{2}+c_{7 } \dot{h}h^{2 } \right ] \,.\ ] ] upon inserting this ansatz into equation ( [ quantumdensity1 ] ) and equating the contributions at each order , we immediately find : @xmath74 + 3b''\left [ 2\ddot{h}h -\dot{h}^{2}+6\dot{h}h^{2}\right]\,.\ ] ] because @xmath75 can be expressed in a local form , we would like to point out that this yields a local expression for the stress - energy tensor too . although quantum fluctuations of the conformal field may still act non - locally on the background spacetime , at the classical level the trace anomaly affects the spacetime only locally . again , when working in quasi de sitter spacetime where @xmath49 is both small and time independent , we can truncate this expression for the quantum density finding : @xmath76 + 18 b '' \dot{h}h^{2}\,.\ ] ] again , when @xmath77 , this analysis becomes exact . in the next sections , we solve the einstein field equations in flrw spacetimes for a universe in the presence of a ) a non - zero cosmological constant , b ) the trace anomaly as a contribution to the quantum stress - energy tensor and c ) matter with constant equation of state @xmath78 , where @xmath79 . we thus consider non - tachyonic matter only , which does not exclude @xcite , where the effect of a finite period with @xmath80 is investigated . the trace anomaly enters the einstein field equation naturally in the trace equation ( [ efetrace1 ] ) . we can thus study the effect of the trace anomaly on the evolution of a universe with a cosmological constant and matter . hence , the trace equation of the einstein field equations is given by : @xmath81 where @xmath82 now , we can employ the ( 00 ) einstein field equation from ( [ efe1 ] ) to express @xmath83 in terms of @xmath75 yielding : @xmath84\,.\ ] ] the above equation governs the dynamics for the hubble parameter that we will solve in various interesting cases . for the anomalous trace we can either take the exact expression ( [ traceanomaly4 ] ) containing higher derivatives of the hubble parameter or its truncated version ( [ traceanomaly4b ] ) . likewise , for the quantum density we can either insert the full expression ( [ quantumdensity3 ] ) or the truncated one ( [ quantumdensity4 ] ) the reason for truncating the expression for the quantum trace and density as outlined above is the realisation that higher derivative contributions in an equation of motion generally have the tendency to destabilise a system . in the literature ( see e.g. @xcite ) , one only has considered an empty universe with a cosmological term and trace anomaly . in reality , one solved for a radiation dominated universe , for only radiation ( @xmath85 ) does not contribute to the trace of the einstein field equations . we incorporate matter with constant equation of state parameter @xmath86 . independently on whether one truncates the expressions for the anomalous trace or quantum density , one can solve for the asymptotes yielding the late time behaviour . setting all time derivatives of the hubble parameter equal to zero , one can easily solve for the two late time constants : @xmath87 here , @xmath88 turns out to be the classical de sitter attractor , whereas @xmath89 is a new , quantum anomaly driven attractor . this result for the late time constants is identical to the case where matter is absent @xcite . we can write the above expression in a somewhat more convenient form by defining the dimensionless parameter @xmath90 : @xmath91 that sets the scale for the cosmological constant @xmath92 . in the current epoch , @xmath90 is extremely small , which allows us to expand ( [ asymptote1 ] ) finding ) the classical de sitter attractor , it would be more natural to denote it with the quantum corrected classical attractor . hence , the quantum attractor ( [ asymptote2b ] ) should preferably be denoted by anomaly driven attractor or planck scale attractor . we will nevertheless adopt the nomenclature existing in the literature . ] : [ asymptote2 ] @xmath93 \label{asymptote2a}\\ h_{0}^{\mathrm{a } } & = & \sqrt{\frac{-1}{16\pi g \ , b ' } -\frac{\lambda}{3 } } \label{asymptote2b}\,.\end{aligned}\ ] ] in the absence of a cosmological constant , the trace anomaly can thus provide us with an inflationary scenario which has already been appreciated by @xcite . finally , note these asymptotes are independent of @xmath26 . the authors of @xcite are correct in saying that the conformal anomaly can not be generated by a finite effective action built out of local quadratic geometric curvature invariants only . we show that the trace anomaly can be generated by an infinite effective action in flat flrw spacetimes , that consists only of local quadratic geometric curvature invariants . although infinite at the level of the effective action , we generate a finite on shell contribution . we first write the trace anomaly in conformal time @xmath94 such that the full conformal anomaly ( [ traceanomaly4 ] ) reads : @xmath95 here , dashes denote conformal time derivatives . in general , the trace of a stress - energy tensor can be written as : @xmath96 here , @xmath97 is the ( spatial ) volume . the correct effective action @xmath98 that generates the trace anomaly in spatially flat flrw spacetimes is given by : @xmath99 where @xmath100 here @xmath101 is the dimension of the spacetime and @xmath102 is an undetermined finite ( and physically irrelevant ) constant . only after variation we let @xmath103 . the numerical factors in ( [ effectiveaction1 ] ) are chosen in accordance with the trace of the einstein field equations , i.e. : the trace of the variation : @xmath104 = 0 \,,\ ] ] indeed yields : @xmath105 the effective action ( [ effectiveaction1 ] ) merits some clarifying remarks . firstly , the variation of @xmath106 straightforwardly yields the @xmath107 term in the conformal anomaly . this term is not unique for one could have equally well taken @xmath108 into account in an @xmath106 contribution and a gauss - bonnet term , that yields a vanishing surface term in four dimensions . ] . secondly , note that the local effective action depends solely on the scale factor . if one were to rewrite this effective action covariantly in terms of the full metric , the effective action would become non - local @xcite , also see @xcite . it is well - known that the effective action of the conformal factor can be written in a local form @xcite . however , we have been able to rewrite this expression in terms of the gauss - bonnet invariant multiplying an infinite constant . finally , note that the coefficient ( [ effectiveaction2 ] ) diverges in four dimensions . however , the gauss - bonnet invariant in @xmath101 dimensions reads : @xmath109 + \frac{a''}{a^{3 } } \left ( \frac{a'}{a^{2}}\right)^2\left [ 4 ( d-3)(d-2)(d-1)\right ] \,,\ ] ] which can be determined from the @xmath101 dimensional generalisations of equation ( [ quadraticcurvature ] ) . note : @xmath110 clearly , only in four dimensions the gauss - bonnet term corresponds to a total derivative that can be neglected . hence , we choose the divergent coefficient @xmath111 such that the factor @xmath112 cancels yielding a non - vanishing and finite contribution at the level of the equation of motion when @xmath113 . the contribution from @xmath102 is identically zero when we let @xmath113 . although divergent in the action , we have shown that at the level of the equation of motion the term involving @xmath111 yields a finite contribution . all physical measurements are performed on shell , hence we can not exclude a counterterm of the form @xmath114 in the effective action . this procedure can be debated . in the literature one usually prefers a finite effective action . we abandon this assumption and require a finite equation of motion only . we can not think of a physical measurement that distinguishes the two approaches and therefore we argue that one should consider and examine all possible effects the counterterms might have on the trace anomaly . in homogeneous cosmology , @xmath106 and @xmath23 are the only local quadratic geometric curvature invariants , we can express @xmath115 in terms of @xmath116 and @xmath106 . hence , @xmath106 and @xmath23 are linearly independent . note furthermore that if one investigates inhomogeneities in the universe , one measures ( statistical ) correlation functions . they are translation invariant as a consequence of the symmetry of the vacuum state and therefore respect the symmetry of the background spacetime . ] . divergences up to one loop in perturbative quantum gravity can be cancelled only by counterterms of the form @xmath106 and @xmath23 . the physical coefficients multiplying @xmath106 and @xmath23 in the renormalised one loop effective action receive contributions which one can write as : [ renormalisation1 ] @xmath117 r^{2 } \label{renormalisation1a } \\ \beta_{\mathrm{phys } } e & = & \left[\beta_{\mathrm{anom}}(m ) + \beta_{\mathrm{ct}}\right ] e\,,\end{aligned}\ ] ] where @xmath118 is the mass dependent finite contribution from the trace anomaly , where @xmath119 denotes the mass of the particles @xmath120 . it is in principle uniquely defined by the requirement that @xmath121 as @xmath122 . similarly , @xmath123 is the ( infinite ) anomalous contribution determined by the requirement that @xmath124 as @xmath122 . these requirements in principle fix @xmath123 and @xmath118 uniquely . while there is agreement in the literature on the value of @xmath118 ( also used in this paper ) , disagreement exists on the value of @xmath123 . therefore we leave it unspecified . the contributions @xmath125 and @xmath126 correspond to the parts of the counterterms that remain when eventual one loop divergences are cancelled function for a scalar field in de sitter spacetime , we refer to @xcite . it is however only the sum of these terms , yielding @xmath127 and @xmath128 , that is physical , i.e. measurable @xcite . since these parameters have not been measured , we can not simply assume that the coefficients in the trace anomaly are just given by ( [ traceanomaly3 ] ) . we should allow these coefficients to vary in order to examine the full effect of the trace anomaly on the evolution of our flrw universe . this is precisely what we pursue in the following sections . in this section we work in quasi de sitter spacetime , where we treat @xmath49 as a small and time independent constant which allows us to neglect higher order derivative contributions . the reason for discarding higher order derivative contributions is that they tend to destabilise a dynamical system , formally known as the theorem of ostrogradsky @xcite . we have to distinguish two cases separately . in the spirit of @xcite , the numerical value of the parameter @xmath26 occurring in the trace anomaly is not fixed because it is a regularisation scheme dependent parameter . generally , however , we can not exclude the presence of this term and we therefore allow it to take different values . first , we allow for an unrestricted value of @xmath26 and secondly we set @xmath77 . this case is particularly interesting as this value of @xmath26 sets the total coefficient multiplying the @xmath107 contribution in the trace anomaly to zero . we thus insert the truncated expression for the trace anomaly ( [ traceanomaly4b ] ) and the quantum density ( [ quantumdensity4 ] ) into the einstein field equation ( [ efegeneral ] ) . this yields : @xmath130\,.\ ] ] this differential equation can be solved exactly . separation of variables yields : @xmath131 where @xmath132 is conveniently defined as : @xmath133 one can perform the integral in terms of logarithms where one has to take the signs of the occurring parameters carefully into account . the integral above gives : @xmath134\nonumber \,.\end{aligned}\ ] ] note the asymptotes of this analytic solution coincide with the asymptotes obtained earlier in ( [ asymptote2 ] ) as expected . in figure [ fig : hubble1 ] , we numerically calculate the dynamics of the hubble parameter for various initial conditions . the two asymptotes divide this graph into three distinct regions that are not connected for finite time evolution . the region bounded by the two asymptotes contains initial conditions for @xmath135 such that @xmath135 grows for late times towards @xmath89 and initial conditions such that @xmath135 asymptotes to the de sitter attractor @xmath88 . in figures [ fig : epsilon1 ] and [ fig : epsilondot1 ] , we examine whether our approximation that @xmath49 is both small and a constant is valid for calculating the dynamics . in figure [ fig : epsilon1 ] one can clearly see that @xmath136 for late times . if @xmath137 , we should have @xmath138 , an assumption that is violated as depicted in figure [ fig : epsilondot1 ] . however , already for a classical cosmological constant dominated universe with matter , a similar violation occurs . furthermore , note the existence of a branching point , an initial condition for @xmath135 such that for @xmath139 the hubble parameter asymptotes to the quantum attractor @xmath89 and for @xmath140 the hubble parameter decreases to the classical de sitter attractor @xmath88 . when rewriting equation of motion ( [ efetruncated ] ) in terms of @xmath49 and noting that @xmath49 should diverge exactly at the branching point , one finds : @xmath141 we probe the dependence on scale by changing the numerical value attached to @xmath90 in equation ( [ scaledependencelambda ] ) . if we decrease @xmath90 , then also @xmath92 decreases which results in a smaller @xmath88 . also , it turns out that both of the asymptotes are already reached much faster . this improves the validity of the assumption @xmath136 , whereas the assumption @xmath138 is still seriously violated at all times . , @xmath85 , @xmath142 . [ fig : qdsinstability ] ] _ _ , @xmath85 , @xmath142 . [ fig : qdsinstability ] ] _ _ 0.1 cm , @xmath85 , @xmath142 . [ fig : qdsinstability ] ] _ _ , @xmath85 , @xmath142 . [ fig : qdsinstability ] ] _ _ let us study the analytical solution ( [ hsolution1 ] ) more closely . in particular , it is interesting to derive the high and low energy limits of this solution . of course , one naively expects in the high or low energy limit to flow towards the quantum or classical attractor , respectively . however , the analysis turns out to be somewhat more subtle . we will show that under a certain condition , the quantum anomaly driven attractor becomes unstable . although solution ( [ hsolution1 ] ) looks complicated at first glance , it simplifies when defining : [ limitdefs ] @xmath143 \label{limitdefs1 } \\ a & = & \frac{1+\frac{1}{2}\alpha ( h_{0}^{\mathrm{a}})^{2}}{h_{0}^{\mathrm{a } } } \label{limitdefs2 } \\ b & = & \frac{1+\frac{1}{2}\alpha ( h_{0}^{\mathrm{c}})^{2}}{h_{0}^{\mathrm{c } } } \label{limitdefs3 } \,.\end{aligned}\ ] ] we define the initial conditions at @xmath144 as : @xmath145 note that @xmath146 , generically . with these definitions , equation ( [ hsolution1 ] ) reduces to : @xmath147 + b \left [ \log\left(\frac{h(t)+h_{0}^{\mathrm{c}}}{\left| h(t)- h_{0}^{\mathrm{c}}\right| } \right ) - c_{2 } \right ] \,,\ ] ] in the high energy limit , we set : @xmath148 such that @xmath149 . equation ( [ hsolution2 ] ) thus modifies to : @xmath150 + b \left [ \log \left(\frac { 1 + \delta(t ) + h_{0}^{\mathrm{c}}/ h_{0}^{\mathrm{a } } } { \left| 1 + \delta(t ) - h_{0}^{\mathrm{c}}/ h_{0}^{\mathrm{a } } \right|}\right)-c_{2 } \right ] \,,\ ] ] we can expand the second logarithm making use of @xmath151 . the leading order contribution ( in @xmath152 ) is given by the denominator in the logarithm , because this term diverges as @xmath152 approaches zero . we can thus exponentiate the equation and solve for the hubble parameter : @xmath153 \,,\ ] ] where the time shift @xmath154 is given by : @xmath155 as @xmath156 , this solution converges whenever @xmath157 . this provides a stability condition on @xmath26 in terms of @xmath25 and @xmath90 . the quantum anomaly driven attractor is stable , whenever the following inequality is satisfied : @xmath158 in figure [ fig : qdsinstability ] , we have numerically calculated the evolution of the hubble parameter in a radiation dominated universe when this inequality is not satisfied . we used @xmath159 . for initial conditions above the attractor , the hubble parameter increases to even higher energies , whereas for initial conditions below the quantum attractor , the hubble parameter evolves towards the classical attractor . hence even in quasi de sitter spacetimes , physically questionable solutions occur . however , note that when @xmath160 , the specific case under consideration in subsection [ case ii : b=2b/3 ] , the above inequality is satisfied . when the above inequality is satisfied , the hubble parameter in the high energy limit decays exponentially towards the quantum anomaly driven attractor , where some frequency dependence through @xmath161 and a time shift @xmath154 can be recognised . the time shift can without observational consequences be absorbed in the initial time @xmath144 . the low energy limit reveals less surprising behaviour . here , we set : @xmath162 and @xmath163 . again we use @xmath151 in order to capture the leading order dynamics . this yields : @xmath164 \,.\ ] ] because @xmath165 , this solution converges . in the low energy limit the time shift @xmath166 is slightly different as compared to ( [ timeshift1 ] ) : @xmath167 finally , we examine the phase space flow in quasi de sitter spacetime . in figure [ fig : parametrichepsilon ] , we show a parametric plot of @xmath59 versus @xmath49 . the phase space basically consists of two lines . the phase space flow is towards either the classical or the quantum anomaly driven attractor as indicated by the arrows . note that we have chosen both attractors to be stable . furthermore , we also include the branching point ( [ branchingpoint ] ) and the classical evolution , that is , the evolution of a universe with @xmath168 . as expected , the flow is towards the classical attractor in this case . although the analysis performed above is for generic values of @xmath26 , we set it to zero in figure [ fig : parametrichepsilon ] and @xmath25 takes its standard model value . such that the @xmath107 does not contribute in the trace anomaly . qualitatively , the dynamics does not change when compared to figure [ fig : parametrichepsilon ] . again , both attractors are stable . apart from @xmath26 , the value of the parameters are identical to figure [ fig : parametrichepsilon ] . [ fig : parametrichepsilonnoboxr ] ] _ _ such that the @xmath107 does not contribute in the trace anomaly . qualitatively , the dynamics does not change when compared to figure [ fig : parametrichepsilon ] . again , both attractors are stable . apart from @xmath26 , the value of the parameters are identical to figure [ fig : parametrichepsilon ] . [ fig : parametrichepsilonnoboxr ] ] _ _ as indicated earlier , we must consider the case when @xmath77 separately because in this particular case the total coefficient in front of the @xmath107 contribution to the trace anomaly vanishes . all higher derivative contributions precisely cancel and also the @xmath170 contribution happens to cancel , such that we find ourselves immediately situated in quasi de sitter spacetime . albeit a simple case , we do take the full trace anomaly into account . the analytic solution obtained in ( [ hsolution1 ] ) still applies and moreover , it becomes exact . the branching point is still given by equation ( [ branchingpoint ] ) for which we just have to insert @xmath52 . clearly , in figure [ fig : parametrichepsilonnoboxr ] one can see that qualitatively the dynamics has not changed compared to figure [ fig : parametrichepsilon ] . the branching point has shifted somewhat to the right , and the way in which the hubble parameter approaches its two late time asymptotes differs . however , the important features of figure [ fig : parametrichepsilon ] , i.e. : two stable attractors , the occurrence of a branching point and the shape and dimension of the phase space , do not change . in which the classical attractor is stable . because of under - damping , one spirals towards the classical attractor . for clarity , we have only included the flow for two initial conditions @xmath171 and @xmath172 . the twister like structure is clearly visible . qualitatively , the phase space flow resulting from other initial conditions is identical . [ fig : phasespacefullta2 ] ] _ _ 1.0 cm in which the classical attractor is stable . because of under - damping , one spirals towards the classical attractor . for clarity , we have only included the flow for two initial conditions @xmath171 and @xmath172 . the twister like structure is clearly visible . qualitatively , the phase space flow resulting from other initial conditions is identical . [ fig : phasespacefullta2 ] , title="fig : " ] _ _ in which the classical attractor is stable . because of under - damping , one spirals towards the classical attractor . for clarity , we have only included the flow for two initial conditions @xmath171 and @xmath172 . the twister like structure is clearly visible . qualitatively , the phase space flow resulting from other initial conditions is identical . [ fig : phasespacefullta2 ] ] _ _ 0.1 cm in which the classical attractor is stable . because of under - damping , one spirals towards the classical attractor . for clarity , we have only included the flow for two initial conditions @xmath171 and @xmath172 . the twister like structure is clearly visible . qualitatively , the phase space flow resulting from other initial conditions is identical . [ fig : phasespacefullta2 ] , title="fig : " ] _ _ we turn our attention to solving the full trace equation ( [ efegeneral ] ) , where we truncate the expression neither for the anomalous trace ( [ traceanomaly4 ] ) nor for the quantum density ( [ quantumdensity3 ] ) . obviously , we can not solve this equation analytically , for it contains all higher order derivative contributions , which forces us to rely on numerical methods . firstly , we note that the asymptotes ( [ asymptote2 ] ) do not change by including the higher derivative contributions . keeping ostrogradsky s theorem in mind , we expect to incur all kinds of issues related to the stability of our system and asymptotes in particular . it is therefore essential to perform a stability analysis for small perturbations @xmath173 around both of the asymptotes . we insert : @xmath174 in equation ( [ efegeneral ] ) , where @xmath175 can either denote the classical or the quantum attractor . for the small perturbations around these attractors we make the ansatz @xmath176 $ ] . we linearise the trace equation finding the characteristic equation from which we determine the eigenvalues @xmath177 of our system : @xmath178 where @xmath179 and @xmath180 . remarkably , the solutions of this third order equation are simple : [ eigenvaluesstability ] @xmath181 where : @xmath182 clearly , when @xmath183 for @xmath184 , the corresponding attractor is stable . since we only consider non - tachyonic matter , eigenvalue ( [ eigenvaluesstabilitya ] ) is negative . however , a finite period in which @xmath185 is not excluded ( see e.g. : @xcite ) . if one were to consider other equations of state than the simple linear one @xmath186 , this statement might no longer hold @xcite . hence only for @xmath187 when @xmath188 , we could encounter a potential instability . surprisingly , the stability analysis does not depend on the equation of state @xmath86 because @xmath86 enters only through equation ( [ eigenvaluesstabilitya ] ) . the condition for instability thus reads : @xmath189 we can rewrite this equation to find : @xmath190 in the expression above , we should read the inequality @xmath191 or @xmath192 whenever @xmath193 or @xmath194 , respectively . we can now insert either the classical or quantum asymptotes previously derived in equation ( [ asymptote2 ] ) and verify which of the two above inequalities is satisfied . upon inserting the expression for the classical attractor , equation ( [ condinstab2 ] ) yields : @xmath195 where @xmath196 as before . only the first inequality @xmath191 will be satisfied . hence we conclude that the classical attractor is unstable if @xmath193 . the converse will be true if @xmath194 . likewise , equation ( [ condinstab2 ] ) for the quantum attractor after some algebra reads : @xmath197 concluding , when using @xmath198 , we unambiguously find : @xmath199 let us first of all recall that it is precisely the combination @xmath200 that multiplies the @xmath107 contribution in the trace anomaly . this calculation thus proves the statements about stability made in e.g. @xcite using the routh - hurwitz method . our proof is more general because we include a constant but otherwise arbitrary equation of state parameter @xmath201 . moreover , while the routh - hurwitz method can only guarantee stability of a solution ( when certain determinants are all strictly positive ) , it does not tell anything about instability @xcite . furthermore appreciate that the singular point in this analysis , @xmath202 , or equivalently @xmath203 , immediately directs us to the quasi de sitter spacetime analysis performed in section [ case ii : b=2b/3 ] , where all higher derivative contributions precisely cancel , rendering both attractors stable . let us compare figures [ fig : hubblefullta1 ] and [ fig : hubblefullta2 ] . in the former figure , we used @xmath204 such that @xmath193 , yielding an unstable classical attractor . however , if @xmath205 the quantum anomaly driven asymptote is not an attractor and the hubble parameter runs away to negative infinity . in the latter figure , we set @xmath206 such that @xmath194 which gives us a stable classical attractor . likewise , for initial conditions @xmath207 the de sitter solution is not an attractor and the hubble parameter rapidly blows up to positive infinity . in figure [ fig : hubblefullta2 ] , we can observe another interesting phenomenon . in this case , the classical attractor is under - damped , resulting in decaying oscillations around the de sitter attractor . in figure [ fig : hubblefullta1 ] these oscillations are not always present . the eigenvalues ( [ eigenvaluesstability ] ) develop an imaginary contribution resulting in oscillatory behaviour whenever : @xmath208 we thus find : @xmath209 the inequality @xmath192 or @xmath191 holds whenever @xmath193 or @xmath194 applies , respectively . again , we verify which of the two inequalities is actually satisfied . to study oscillations around a stable classical attractor , we should take the @xmath191 inequality ( there are no oscillations around an unstable attractor ) . we thus find : @xmath210 clearly , this inequality is always satisfied because @xmath211 . we thus conclude that whenever the classical de sitter attractor is stable , oscillations occur . furthermore , we can insert the quantum anomaly driven attractor in equation ( [ condoscill2 ] ) . now , we should use the @xmath192 inequality in order to study oscillatory behaviour around the ( stable ) quantum attractor . this yields : @xmath212 oscillatory behaviour around the quantum attractor thus occurs when : @xmath213 or because the quantum attractor has become unstable . the red region ( dark shaded ) in parameter space shows oscillatory behaviour , whereas the white region is either critically damped or over - damped . we used @xmath214 . [ fig : parmspaceoscil ] ] _ _ in figure [ fig : parmspaceoscil ] we depicted the parameter space ( the @xmath25 versus @xmath26 plane ) resulting in oscillatory behaviour around the quantum anomaly driven attractor . first of all , we should have @xmath193 yielding a stable quantum attractor . of course , one can not have oscillations around an unstable attractor . secondly , note that @xmath34 because of equation ( [ traceanomaly3b ] ) . finally , when the newly derived inequality ( [ condoscill5 ] ) is satisfied , oscillations occur . these considerations divide the phase space into three regions : a region where oscillatory behaviour occurs , another region which results in critically or over - damped behaviour and a part of phase space that is forbidden , as shown in figure [ fig : parmspaceoscil ] . let us now return to discussing the classical attractor that shows its oscillatory behaviour manifestly . we calculate the frequency of oscillations around this attractor . taking the square root of ( [ deltadef ] ) and extracting an @xmath120 we find the ( quantum corrected ) frequency for oscillations around the classical attractor : @xmath215 note that this frequency is independent on the equation of state parameter @xmath86 . the frequency of oscillations around the quantum attractor can be found analogously . let us analyse the phase space in the case of a stable quantum and classical attractor subsequently . in figure [ fig : phasespacefullta1 ] , we visualise the phase space flow for the former case parametrically in the @xmath135 , @xmath216 and @xmath217 directions . however , also @xmath218 rapidly approaches zero as time elapses . ] . the small black sphere denotes the quantum anomaly driven attractor . for the latter case , we include in figure [ fig : phasespacefullta2 ] the phase space flow for just two initial conditions for @xmath135 for clarity . the under - damped oscillatory behaviour results in the twister like structure visible in the @xmath135 , @xmath216 and @xmath217 directions . finally , we would like to point out that including all higher derivative contributions , which thus corresponds to solving the trace equation exactly , modifies the dynamics of the hubble parameter significantly . attractors that were stable in the absence of higher derivatives under certain conditions destabilise . the reason for this , clearly , is attributable to the presence of the @xmath107 term in the trace anomaly , generating these higher derivative contributions . we do not know whether or not incorporating the higher derivatives is a sensible thing to do . usually higher order derivatives tend to destabilise a system signifying that some particular solutions are not physical . therefore , in the spirit of ostrogradsky s theorem , one can question whether the analysis where higher derivative contributions are discarded is correct , or the analysis taking the full trace anomaly into account . the trace of the einstein field equations in cosmologically relevant spacetimes together with stress - energy conservation completely captures the dynamics of the hubble parameter . we have derived the trace anomaly from an effective action in spatially flat flrw spacetimes . it consists of the local quadratic geometric curvature invariants @xmath219 and the gauss - bonnet term @xmath23 . because of counterterms that are supposed to cancel divergences of the as yet unknown underlying fundamental theory , we expect the coefficients in the trace anomaly to change . the physical value of each of these coefficients receives contributions both from the anomalous trace and from these counterterms . because we do not know the physical value these parameters will take , we must allow them to vary in order to examine all possibilities . we have studied the dynamics of the hubble parameter both in quasi de sitter and in flrw spacetimes including matter , a cosmological term and the trace anomaly . in quasi de sitter spacetime , where we restrict the hubble parameter to vary slowly in time , we find that for various initial conditions @xmath135 asymptotes either to the classical de sitter attractor , or to a quantum anomaly driven attractor . we find a region in parameter space where the quantum attractor destabilises . otherwise , both attractors are stable . in flrw spacetimes we include all higher derivative contributions in the trace anomaly . we perform a stability analysis for small perturbations around the two asymptotes . for @xmath220 , the quantum attractor is stable and the classical de sitter attractor is unstable . on the contrary , for @xmath194 , the quantum attractor is unstable and the de sitter attractor becomes stable . the singular point in this analysis , @xmath221 , immediately directs us to quasi de sitter spacetime in which the dynamics is much simpler . in this case , both attractors are stable . the classical de sitter attractor always shows under - damped oscillatory behaviour and we calculate the frequency of these oscillations . we analyse the phase space of the quantum attractor and conclude there is some region in parameter space for which oscillations occur . there is no dynamical effect that influences the effective value of the cosmological constant , i.e. : the classical de sitter attractor . based on our semiclassical analysis we thus conclude that the trace anomaly does _ not _ solve the cosmological constant problem . we have studied both the truncated and the exact expression of the trace anomaly in flat flrw spacetimes . we do not know which of the two approaches is correct . keeping ostrogradsky s theorem in mind , higher derivative contributions usually have the tendency to destabilise a dynamical system . discarding these higher derivatives and studying the trace anomaly in quasi de sitter spacetime would thus seem plausible . finally , one could wonder whether the quantum anomaly driven attractor is physical . the quantum attractor is of the order of the planck mass @xmath222 , so only when matter in the early universe is sufficiently dense , @xmath223 . we then expect to evolve towards the quantum attractor . however , at these early times we also expect perturbative general relativity to break down . hence , this attractor might even not be there or it may be seriously affected by quantum fluctuations . quantum fluctuations present at that epoch might even induce tunnelling towards the regime where @xmath135 asymptotes to the classical attractor . 99 s. nobbenhuis , categorizing different approaches to the cosmological constant problem , found . * 36 * ( 2006 ) 613 [ arxiv : gr - qc/0411093 ] . s. nobbenhuis , the cosmological constant problem , an inspiration for new physics , ph.d . thesis , [ arxiv : gr - qc/0609011 ] . r. bousso , tasi lectures on the cosmological constant , [ arxiv : hep - th/0708.4231 ] . j. donoghue , the quantum theory of general relativity at low energies , helv . acta * 69 * ( 1996 ) 269 [ arxiv : gr - qc/9607039 ] . j. donoghue , perturbative dynamics of quantum general relativity , [ arxiv : gr - qc/9712070 ] . n. d. birrell and p. c. w. davies , quantum fields in curved space , cambridge monographs on mathematical physics , cambridge university press ( 1982 ) d. m. capper and m. j. duff , trace anomalies in dimensional regularization , nuovo cim . a * 23 * , 173 ( 1974 ) . s. deser , m. j. duff and c. j. isham , nonlocal conformal anomalies , nucl . b * 111 * , 45 ( 1976 ) . l. s. brown , stress tensor trace anomaly in a gravitational metric : scalar fields , phys . d * 15 * ( 1977 ) 1469 . j. s. dowker and r. critchley , the stress tensor conformal anomaly for scalar and spinor fields , phys . d * 16 * ( 1977 ) 3390 . h. s. tsao , conformal anomalies in a general background metric , phys . b * 68 * ( 1977 ) 79 . m. j. duff , observations on conformal anomalies , nucl . b * 125 * , 334 ( 1977 ) . l. s. brown and j. p. cassidy , stress tensor trace anomaly in a gravitational metric : general theory , maxwell field , phys . d * 15 * ( 1977 ) 2810 . m. j. duff , twenty years of the weyl anomaly , class . grav . * 11 * , 1387 ( 1994 ) [ arxiv : hep - th/9308075 ] . l. h. ford , quantum instability of de sitter space - time , phys . d * 31 * ( 1985 ) 710 . i. antoniadis and e. mottola , graviton fluctuations in de sitter space , j. math . * 32 * ( 1991 ) 1037 . n. c. tsamis and r. p. woodard , the structure of perturbative quantum gravity on a de sitter background , commun . * 162 * ( 1994 ) 217 . n. c. tsamis and r. p. woodard , quantum gravity slows inflation , nucl . b * 474 * ( 1996 ) 235 [ arxiv : hep - ph/9602315 ] . n. c. tsamis and r. p. woodard , one loop graviton self - energy in a locally de sitter background , phys . d * 54 * ( 1996 ) 2621 [ arxiv : hep - ph/9602317 ] . l. r. w. abramo , r. h. brandenberger and v. f. mukhanov , the energy - momentum tensor for cosmological perturbations , phys . d * 56 * ( 1997 ) 3248 [ arxiv : gr - qc/9704037 ] . f. finelli , g. marozzi , g. p. vacca and g. venturi , adiabatic regularization of the graviton stress - energy tensor in de sitter space - time , phys . d * 71 * ( 2005 ) 023522 [ arxiv : gr - qc/0407101 ] . a. bilandzic and t. prokopec , quantum radiative corrections to slow - roll inflation , phys . d * 76 * ( 2007 ) 103507 [ arxiv:0704.1905 [ astro - ph ] ] . t. janssen and t. prokopec , a graviton propagator for inflation , class . * 25 * ( 2008 ) 055007 [ arxiv:0707.3919 [ gr - qc ] ] . t. janssen and t. prokopec , implications of the graviton one - loop effective action on the dynamics of the universe , arxiv:0807.0447 [ gr - qc ] . t. janssen , s. p. miao and t. prokopec , graviton one - loop effective action and inflationary dynamics , arxiv:0807.0439 [ gr - qc ] . n. bilic , b. guberina , r. horvat , h. nikolic and h. stefancic , on cosmological implications of gravitational trace anomaly , phys . b * 657 * ( 2007 ) 232 [ arxiv : gr - qc/0707.3830 ] . r. schutzhold , small cosmological constant from the qcd trace anomaly ? , phys . * 89 * ( 2002 ) 081302 . e. t. tomboulis , dynamically adjusted cosmological constant and conformal anomalies , nucl . b * 329 * ( 1990 ) 410 . i. antoniadis and e. mottola , 4-d quantum gravity in the conformal sector , phys . d * 45 * ( 1992 ) 2013 . i. antoniadis , dynamics of the conformal factor in 4-d gravity , [ arxiv : hep - th/9211055 ] . i. antoniadis , p. o. mazur and e. mottola , fractal geometry of quantum spacetime at large scales , phys . b * 444 * ( 1998 ) 284 [ arxiv : hep - th/9808070 ] . h. salehi and y. bisabr , conformal anomaly and large scale gravitational coupling , int . j. theor . * 39 * ( 2000 ) 1241 [ arxiv : hep - th/0001095 ] . i. antoniadis , p. o. mazur and e. mottola , cosmological dark energy : prospects for a dynamical theory , new j. phys . * 9 * ( 2007 ) 11 [ arxiv : gr - qc/0612068 ] . r. j. riegert , a nonlocal action for the trace anomaly , phys . b * 134 * , 56 ( 1984 ) . e. komatsu _ et al . _ [ wmap collaboration ] , five - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological interpretation , arxiv:0803.0547 [ astro - ph ] . a. a. starobinsky , a new type of isotropic cosmological models without singularity , phys . b * 91 * ( 1980 ) 99 . s. w. hawking , t. hertog and h. s. reall , trace anomaly driven inflation , phys . d * 63 * ( 2001 ) 083504 [ arxiv : hep - th/0010232 ] . j. c. fabris , a. m. pelinson and i. l. shapiro , anomaly - induced effective action and inflation , nucl . suppl . * 95 * , 78 ( 2001 ) [ arxiv : hep - th/0011030 ] . i. l. shapiro , an overview of the anomaly - induced inflation , nucl . * 127 * ( 2004 ) 196 [ arxiv : hep - ph/0311307 ] . i. l. shapiro , effective field theory and fundamental interactions , [ arxiv : hep - th/0412115 ] . i. l. shapiro and j. sola , a modified starobinsky s model of inflation : anomaly - induced inflation , susy and graceful exit , [ arxiv : hep - ph/0210329 ] . a. m. pelinson , i. l. shapiro and f. i. takakura , on the stability of the anomaly - induced inflation , nucl . b * 648 * , 417 ( 2003 ) [ arxiv : hep - ph/0208184 ] . a. m. pelinson , i. l. shapiro and f. i. takakura , stability issues in the modified starobinsky model , nucl . suppl . * 127 * , 182 ( 2004 ) [ arxiv : hep - ph/0311308 ] . i. brevik and j. quiroga hurtado , vanishing cosmological constant in modified gauss - bonnet gravity with conformal anomaly , int . j. mod . d * 16 * ( 2007 ) 817 [ arxiv : gr - qc/0610044 ] . r. p. woodard , avoiding dark energy with 1/r modifications of gravity , lect . notes phys . * 720 * ( 2007 ) 403 [ arxiv : astro - ph/0601672 ] . m. v. fischetti , j. b. hartle and b. l. hu , quantum effects in the early universe . 1 . influence of trace anomalies on homogeneous , isotropic , classical geometries , phys . d * 20 * ( 1979 ) 1757 . v. k. onemli and r. p. woodard , quantum effects can render @xmath80 on cosmological scales , phys . d * 70 * ( 2004 ) 107301 [ arxiv : gr - qc/0406098 ] . e. o. kahya and v. k. onemli , quantum stability of a @xmath224 phase of cosmic acceleration , phys . d * 76 * ( 2007 ) 043512 [ arxiv : gr - qc/0612026 ] . i. l. shapiro , local conformal symmetry and its fate at quantum level , pos * ic2006 * ( 2006 ) 030 [ arxiv : hep - th/0610168 ] . i. l. shapiro , effective action of vacuum : semiclassical approach , class . grav . * 25 * , 103001 ( 2008 ) [ arxiv:0801.0216 [ gr - qc ] ] . p. c. w. davies , singularity avoidance and quantum conformal anomalies , phys . b * 68 * ( 1977 ) 402 . e. v. gorbar and i. l. shapiro , renormalization group and decoupling in curved space , jhep * 0302 * ( 2003 ) 021 [ arxiv : hep - ph/0210388 ] . e. v. gorbar and i. l. shapiro , renormalization group and decoupling in curved space . ii : the standard model and beyond , jhep * 0306 * ( 2003 ) 004 [ arxiv : hep - ph/0303124 ] . s. nojiri , s. d. odintsov and s. tsujikawa , properties of singularities in ( phantom ) dark energy universe , phys . d * 71 * ( 2005 ) 063004 [ arxiv : hep - th/0501025 ] . e. w. weisstein , `` routh - hurwitz theorem . '' , from mathworld a wolfram web resource , + http://mathworld.wolfram.com/routh-hurwitztheorem.html . |
we consider a model of isolation with an initial period of migration " ( iim ) , where an ancestral population instantaneously split into two descendant populations which exchanged migrants symmetrically at a constant rate for a period of time but which are now completely isolated from each other . a method of maximum likelihood estimation of the parameters of the model is implemented , for data consisting of the number of nucleotide differences between two dna sequences at each of a large number of independent loci , using the explicit analytical expressions for the likelihood obtained in wilkinson - herbots ( 2012 ) . the method is demonstrated on a large set of dna sequence data from two species of drosophila , as well as on simulated data . the method is extremely fast , returning parameter estimates in less than 1 minute for a data set consisting of the numbers of differences between pairs of sequences from 10,000s of loci , or in a small fraction of a second if all loci are trimmed to the same estimated mutation rate . it is also illustrated how the maximized likelihood can be used to quickly distinguish between competing models describing alternative evolutionary scenarios , either by comparing aic scores or by means of likelihood ratio tests . the present implementation is for a simple version of the model , but various extensions are possible and are briefly discussed . | 1511.05478 | in recent years , molecular genetic data have been used extensively to learn about the evolutionary processes that gave rise to the observed genetic variation . one important example is the use of genetic data to try to infer whether or not gene flow occurred between closely related species during or after speciation ( see for example reviews by nadachowska 2010 , pinho and hey 2010 , smadja and butlin 2011 , bird et al . 2012 and a large number of references in these papers ) . such applications typically use computer programs such as _ mdiv _ ( nielsen and wakeley 2001 ) , _ i m _ ( hey and nielsen 2004 , hey 2005 ) , _ i m a _ ( hey and nielsen 2007 ) , _ mimar _ ( becquet and przeworski 2007 ) or _ ima2 _ ( hey 2010 ) , based on the isolation with migration " ( i m ) model , which assumes that a panmictic ancestral population instantaneously split into two or more descendant populations some time in the past and that migration occurred between these descendant populations at a constant rate ever since . whilst such methods have been extensively and successfully applied to study the relationships between different populations within species , the assumption of migration continuing at a constant rate until the present is unrealistic when studying relationships between species . becquet and przeworski ( 2009 ) and strasburg and rieseberg ( 2010 ) investigated by means of simulations how robust parameter estimates ( migration rates , divergence times and effective population sizes ) , obtained by methods based on the i m model , are to violations of the i m model assumptions . becquet and przeworski ( 2009 ) found that parameter estimates obtained with _ i m _ and _ mimar _ are often biased when the assumptions of the i m model are violated , and concluded that these methods are highly sensitive to the assumption of a constant migration rate since the population split . theoretical results derived by wilkinson - herbots ( 2012 ) also suggested that estimated levels of gene flow obtained by applying an i m model to species which are now completely isolated can not simply be interpreted as average levels of gene flow over time . thus , whilst computer programs based on the i m model can be used to test for departure from a complete isolation model ( which assumes that an ancestral population instantaneously split into two or more descendant populations which have been completely isolated ever since ) , the actual parameter estimates obtained may be difficult to interpret if the i m model is not an accurate description of the history of the populations or species concerned . furthermore , in some studies the programs _ i m a _ or _ ima2 _ have also been used to estimate the times when migration events occurred , and to try to distinguish between scenarios of speciation with gene flow and scenarios of introgression , and it has recently been demonstrated that such inferences about the timing of gene flow ( obtained with programs based on the i m model ) are not valid ( strasburg and rieseberg 2011 , sousa et al . becquet and przeworski ( 2009 ) , strasburg and rieseberg ( 2011 ) and sousa et al . ( 2011 ) all suggested that more realistic models of population divergence and speciation are needed . a first step in trying to make the i m model more suitable for the study of speciation is to allow gene flow to occur during a limited period of time , followed by complete isolation of the species , and such models have been studied by teshima and tajima ( 2002 ) , innan and watanabe ( 2006 ) and wilkinson - herbots ( 2012 ) . in the latter paper , a model of isolation with an initial period of migration " ( iim ) was studied where a panmictic ancestral population gave rise to two or more descendant populations which exchanged migrants symmetrically at a constant rate for a period of time , after which they became completely isolated from each other . explicit analytical expressions were derived for the probability that two dna sequences ( from the same descendant population or from different descendant populations ) differ at @xmath0 nucleotide sites , assuming the infinite sites model of neutral mutation . it was suggested that these results may be useful for maximum likelihood estimation of the parameters of the model , if one pair of dna sequences is available at each of a large number of independent loci , and that such an ml method should be very fast as it is based on an explicit expression for the likelihood rather than on computation by means of numerical approximations or mcmc simulation . however the proposed ml estimation method had not yet been implemented and its usefulness was still to be demonstrated . in this paper we present an implementation of the ml estimation method for the parameters of the isolation with an initial period of migration " model proposed in wilkinson - herbots ( 2012 ) , and we illustrate its potential by applying it to a set of dna sequence data from _ drosophila simulans _ and _ drosophila melanogaster _ previously analysed by wang and hey ( 2010 ) and lohse et al . the parameter estimates and maximized likelihood obtained for the iim model are also compared to those under an i m model and an isolation model , and it is illustrated that the method makes it possible to distinguish between more and less plausible evolutionary scenarios , by comparing aic scores or by means of likelihood ratio tests . the implementation was done in r ( r development core team 2011 ) . the r code for the mle method described is included as supplementary material and can readily be pasted into an r document . this mle algorithm is very fast indeed : for the drosophila data mentioned above ( using the number of nucleotide differences between two dna sequences at each of approximately 30,000 loci ) , we obtained estimates of the parameters of the iim model using a desktop pc ; the program returned results instantly ( i.e. in a small fraction of a second ) if all loci had been trimmed to the same estimated mutation rate ( as proposed by lohse et al . 2011 ) , or in approximately 20 seconds if the full sequences ( and hence different mutation rates at different loci ) were used . whereas most earlier methods used large samples from a small number of loci , recent advances in dna sequencing technology and the advent of whole - genome sequencing have led to an increased interest in methods which ( like that described in this paper ) can handle data from a small number of individuals at a large number of loci ( for example , takahata 1995 ; takahata et al . 1995 ; takahata and satta 1997 ; yang 1997 , 2002 ; innan and watanabe 2006 ; wilkinson - herbots 2008 ; burgess and yang 2008 ; wang and hey 2010 ; yang 2010 ; hobolth et al . 2011 ; li and durbin 2011 ; lohse et al . 2011 ; wilkinson - herbots 2012 ; zhu and yang 2012 ; andersen et al . this type of data set has two advantages . firstly , data from even a very large number of individuals at the same locus tend to contain only little information about very old divergence or speciation events because typically the individuals ancestral lineages will have coalesced to a very small number of ancestral lineages by the time the event of interest is reached , and in such contexts a data set consisting of a small number of dna sequences at each of a large number of independent loci may be more informative ( maddison and knowles 2006 ; wang and hey 2010 ; lohse et al . 2010 , 2011 ) . secondly , considering small numbers of sequences at large numbers of independent loci is mathematically much easier and computationally much faster than working with large numbers of sequences at the same locus . in particular , explicit analytical expressions for the likelihood have been obtained for pairs or triplets of sequences for a number of demographic models ( for example , takahata et al . 1995 ; wilkinson - herbots 2008 ; lohse et al . 2011 ; wilkinson - herbots 2012 ; zhu and yang 2012 ) , which substantially speeds up computation and maximization of the likelihood . we obtain maximum likelihood estimates of the parameters of the isolation with initial migration " ( iim ) model studied in wilkinson - herbots ( 2012 ) , for the case of @xmath1 descendant populations or species . this model assumes that , time @xmath2 ago ( @xmath3 ) , a panmictic ancestral population instantaneously split into two descendant populations which subsequently exchanged migrants symmetrically at a constant rate until time @xmath4 ago ( @xmath5 ) , when they became completely isolated from each other ( see figure [ fig1 ] ) . focusing on dna sequences at a single locus that is not subject to intragenic recombination , the ancestral population is assumed to have been of constant size @xmath6 homologous sequences until the split occurred time @xmath2 ago , where @xmath7 is large . between time @xmath2 ago and time @xmath4 ago , the two descendant populations were of constant size @xmath8 sequences each and exchanged migrants at a constant rate @xmath9 , where @xmath9 is the proportion of each descendant population that was replaced by immigrants each generation . the current size of descendant population @xmath10 is @xmath11 sequences ( @xmath12 ) , and is assumed to have been constant since migration ended time @xmath4 ago . as is standard in coalescent theory and assuming that reproduction within populations follows the neutral wright - fisher model , time is measured in units of @xmath8 generations ( this also applies to the times @xmath2 and @xmath4 ) ; in practical applications where the wright - fisher model does not hold , @xmath7 is interpreted as the effective population size . the scaled " migration and mutation rates are defined as @xmath13 and @xmath14 , respectively , where @xmath15 is the mutation rate per dna sequence per generation at the locus concerned . the work described in this paper assumes that mutations are selectively neutral and follow the infinite sites model ( watterson 1975 ) . extensions to other neutral mutation models are feasible but have not yet been implemented . for this iim model , wilkinson - herbots ( 2012 ) found the probability that two homologous dna sequences differ at @xmath0 nucleotide sites : denoting by @xmath16 the number of nucleotide differences between two homologous sequences sampled from descendant populations @xmath10 and @xmath17 ( @xmath18 ) and using the subscript @xmath19 to indicate the scaled mutation rate at the locus concerned , we have for @xmath20 @xmath21 for a pair of sequences from the same descendant population , and + @xmath22 for a pair of sequences from different descendant populations , where @xmath23 with @xmath24 and where @xmath25 thus for pairwise difference data of the form @xmath26 , consisting of the number of nucleotide differences between one pair of dna sequences at each of @xmath27 independent loci , the likelihood under the iim model is given by @xmath28 where @xmath29 and @xmath30 denote the locations ( i.e. the population labels ) of the two dna sequences sampled at the @xmath31th locus , and where each factor @xmath32 is given by equation ( [ distribution0 ] ) or ( [ distribution1 ] ) as appropriate , replacing @xmath19 by the scaled mutation rate @xmath33 for the @xmath31th locus . note that the above derivation of the likelihood assumes that there is no recombination within loci and free recombination between loci . in order to jointly estimate all the parameters of the model , data from between - population sequence comparisons as well as within - population sequence comparisons from both populations should be included in the above likelihood ( where each pairwise comparison must be at a different , independent locus ) . the above explicit formula for the likelihood allows rapid computation and maximization , so that ml estimates can easily be obtained , as well as aic scores and likelihood ratios comparing the iim model with competing models such as the complete isolation " model and the symmetric isolation - with - migration " model . similar ml methods were first developed by takahata et al . ( 1995 ) for a number of different demographic models : a single population of constant size , a population undergoing an instantaneous change of size , and complete isolation models for two and for three species . innan and watanabe ( 2006 ) developed an extension of takahata et al.s mle method to a more sophisticated model of gradual population divergence than the iim model considered in this paper , but their calculation of the likelihood relies on numerical computation of the probability density function of the coalescence time using recursion equations on a series of time points ( and then numerically integrating over the coalescence time to find the probability of @xmath0 nucleotide differences ) , which can be time - consuming ; in addition , the accuracy of their recursion and likelihood calculation depend on the number of time points considered . the method described in the present paper assumes a simpler model than innan and watanabe s , but is faster because both the calculation of the pdf of the coalescence time and the integration over the coalescence time to find the probability of @xmath0 nucleotide differences have already been done , in an exact way , to give equations ( [ distribution0 ] ) and ( [ distribution1 ] ) above , leaving far less computation to be done . in order to obtain good starting values for the likelihood maximization under the iim model , our implementation first fits a complete isolation model , as the latter model gives a more tractable likelihood surface and is therefore less sensitive to the choice of starting values . the parameter estimates obtained for the complete isolation model are then used as starting values to fit the iim model ; this was found to reduce the possibility that the program might otherwise converge on a local maximum rather than on the global maximum of the likelihood under the iim model . whilst the theoretical results obtained make it possible to directly compute ml estimates of the original parameters of the iim model as described above ( @xmath34 ) , our implementation uses the following reparameterizations as this improved performance and robustness : @xmath35 ( similar to the choice of parameters in , for example , yang 2002 , and hey and nielsen 2004 ) , i.e. @xmath36 and @xmath37 represent respectively the time since the end of gene flow and the duration of the period of gene flow , measured by twice the expected number of mutations per lineage during the period concerned ; @xmath38 and @xmath39 are the population size parameters " of the current descendant populations and the ancestral population , respectively ( @xmath14 is the population size parameter of each descendant population during the migration stage of the model ) . ml estimates are obtained jointly for the parameters @xmath40 , and these can readily be converted to ml estimates of the original model parameters if required . our computer code is included as supplementary material . equations ( 1 ) and ( 2 ) rely on the assumption of symmetric migration and equal population sizes during the period of gene flow . without these assumptions , an explicit analytical formula for the likelihood becomes difficult to obtain . simulation results suggest however that our method is reasonably robust to minor violations of these assumptions ; furthermore , it is possible to extend our method to allow for asymmetric migration and unequal population sizes ( costa rj and wilkinson - herbots hm , work in progress ) . to illustrate the mle method for the iim model described above , we applied this method to the genomic data set of _ d. simulans _ and _ d. melanogaster _ compiled and analyzed by wang and hey ( 2010 ) , and reanalyzed by lohse et al . ( 2011 ) , who both used an i m model assuming a constant migration rate from the onset of speciation until the present . the data consist of alignments of 30,247 blocks of intergenic sequence of 500 bp each , from two inbred lines of _ d. simulans _ and from one inbred line each of _ d. melanogaster _ and _ d. yakuba _ , and have been pre - processed as described in wang and hey ( 2010 ) and lohse et al . we also follow these authors in using _ d. yakuba _ as an outgroup to estimate the relative mutation rate at each locus . as our method uses the number of nucleotide differences between one pair of sequences at each locus , we had to choose two of the three sequences ( which we will denote by _ d.sim_1 , _ d.sim_2 and _ d.mel _ for brevity ) at each locus . there are of course many ways in which this can be done ( for example , one could choose two sequences at random at each locus , as done by wang and hey 2010 ) . in order to use all the data , and to be able to check to what extent our results depend on our choice of sequences , we took the following approach : three ( overlapping ) pairwise data sets were formed by alternately assigning loci to the comparisons _ d.mel - d.sim_1 , _ d.mel - d.sim_2 and _ d.sim_1 - _ d.sim_2 , where data set 1 starts with _ d.mel _ - _ d.sim_1 at locus 1 ( _ d.mel _ - _ d.sim_2 at locus 2 , _ d.sim_1 - _ d.sim_2 at locus 3 , and so on ) , data set 2 uses _ _ - _ d.sim_2 at locus 1 ( _ d.sim_1 - _ d.sim_2 at locus 2 , ) , and data set 3 starts with _ d.sim_1 - _ d.sim_2 at locus 1 ( _ d.mel _ - _ d.sim_1 at locus 2 , ) . thus each sequence at each locus is used in exactly two of the three data sets , and each data set contains between - species differences at approximately 20,000 loci and within - species ( _ d . simulans _ ) differences at approximately 10,000 loci . ml estimates and estimated standard errors for the iim model parameters @xmath41 represents the current size of _ d. simulans_. we used the main wang and hey ( 2010 ) data set used also by lohse et al . ( 2011 ) , which does not include any _ d. melanogaster _ pairs and hence contains no information on the population size parameter @xmath42 corresponding to the current size of _ d. melanogaster _ ( this parameter does not appear in the likelihood of these data and thus can not be estimated here ) . ] were obtained for each of the three data sets and then averaged over the three data sets . for comparison , in addition to fitting an iim model as described above , we also obtained ml estimates assuming an iim model with @xmath43 ( or equivalently , @xmath44 , i.e. not allowing for a change of population size at the end of the migration period ) , a symmetric i m model ( which corresponds to putting @xmath45 in the iim model ) , and a complete isolation model ( which corresponds to assuming @xmath46 in addition to @xmath45 ; the size of descendant population 2 is irrelevant here ) . the results are given in table [ tab1 ] . llrrrrrrrrr + data & model & @xmath47 & @xmath48 & @xmath49 & @xmath50 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & aic + + set 1 & isolation & & 13.66 & & 5.67 & 4.69 & & -90,097.17 & 1,175.66 & 180,200.34 + & i m & & 14.79 & & 5.54 & 3.97 & 0.0209 & -89,509.34 & 270.65 & 179,026.68 + & iim with @xmath43 & 5.87 & 9.99 & & 5.47 & 3.48 & 0.1389 & -89,374.01 & 518.43 & 178,758.03 + & iim & 7.22 & 9.38 & 6.70 & 2.68 & 3.22 & 0.0888 & -89,114.80 & & 178,241.60 + & ( s.e . ) & ( 0.20 ) & ( 0.16 ) & ( 0.11 ) & ( 0.12 ) & ( 0.09 ) & ( 0.0059 ) & & & + + set 2 & isolation & & 13.57 & & 5.72 & 4.77 & & -90,339.14 & 1,185.38 & 180,684.28 + & i m & & 14.80 & & 5.58 & 3.98 & 0.0227 & -89,746.45 & 347.39 & 179,500.90 + & iim with @xmath43 & 6.27 & 9.88 & & 5.49 & 3.37 & 0.1793 & -89,572.76 & 685.95 & 179,155.51 + & iim & 7.62 & 9.43 & 6.85 & 2.31 & 3.05 & 0.0904 & -89,229.78 & & 178,471.56 + & ( s.e . ) & ( 0.17 ) & ( 0.15 ) & ( 0.11 ) & ( 0.11 ) & ( 0.09 ) & ( 0.0054 ) & & & + + set 3 & isolation & & 13.63 & & 5.59 & 4.72 & & -89,979.56 & 1,094.44 & 179,965.12 + & i m & & 14.69 & & 5.48 & 4.04 & 0.0193 & -89,432.34 & 291.86 & 178,872.68 + & iim with @xmath43 & 6.29 & 9.67 & & 5.39 & 3.46 & 0.1635 & -89,286.41 & 548.28 & 178,582.82 + & iim & 7.31 & 9.31 & 6.62 & 2.56 & 3.23 & 0.0870 & -89,012.27 & & 178,036.54 + & ( s.e . ) & ( 0.20 ) & ( 0.16 ) & ( 0.11 ) & ( 0.12 ) & ( 0.09 ) & ( 0.0058 ) & & & + + average & iim & 7.39 & 9.38 & 6.72 & 2.52 & 3.16 & 0.0887 & & & + & ( s.e . ) & ( 0.19 ) & ( 0.15 ) & ( 0.11 ) & ( 0.11 ) & ( 0.09 ) & ( 0.0057 ) & & & + + note the values of the likelihood ratio test statistic @xmath54 shown are for the comparison of the model concerned ( considered to be the null model ) against the model immediately underneath it . estimated standard errors ( numbers in brackets ) are provided for the model with the best fit , i.e. the full iim model . the bottom section of the table gives the parameter estimates under the iim model , averaged over the three data sets ; for each parameter , the average of the estimated standard errors for the three data sets is given in brackets , which serves as an estimated _ upper bound _ on the standard error of the averaged parameter estimate ( estimates of the exact standard errors would be hard to obtain as the three data sets overlap ) . [ tab1 ] fitting the full iim model took approximately 20 seconds for each of the three data sets , on a desktop pc . estimated mutation rates , and hence the estimates of the parameters @xmath55 and @xmath39 defined in ( [ pars ] ) , are averages over all the loci considered . mutation rate heterogeneity between loci was accounted for by estimating the relative mutation rate at each locus from comparison with the outgroup _ d. yakuba _ , as proposed in wang and hey ( 2010 ) ( see materials and methods " for further details ) ; for simplicity these relative mutation rates are treated as known constants ( see also yang 1997 , 2002 ) , i.e. uncertainty about the relative mutation rates is ignored . table [ tab2 ] shows the estimates ( averages over the three data sets ) converted to times in years , diploid effective population sizes , and the migration rate per generation , for each of the four models considered . lcccccc + model & @xmath56 & @xmath57 & @xmath58 & @xmath59 & @xmath60 & @xmath61 + + isolation & & 2.97my & & 6.18 m & 5.16 m & + i m & & 3.22my & & 6.04 m & 4.36 m & @xmath62 + iim with @xmath63 & 1.34my & 3.49my & & 5.95 m & 3.75 m & @xmath64 + iim & 1.61my & 3.66my & 7.33 m & 2.75 m & 3.45 m & @xmath65 + ( s.e . ) & ( 0.04my ) & ( 0.03my ) & ( 0.12 m ) & ( 0.12 m ) & ( 0.10 m ) & ( @xmath66 ) + + note the abbreviations my " and m " stand for million years " and million individuals " . the times @xmath67 and @xmath68 denote , respectively , the time since complete isolation of _ d. simulans _ and _ d. melanogaster _ ( for the iim model ) , and the time since the onset of speciation ( for all models ) , i.e. these are the times @xmath4 and @xmath2 converted into years ( see fig.1 ) . @xmath7 denotes the effective size of _ d. simulans _ in the isolation and i m models , and during the migration stage of the iim model ; @xmath69 denotes the present effective size of _ d. simulans _ in the full iim model ; @xmath70 is the effective size of the ancestral population . the estimates shown are the averages of the estimates obtained for the three ( overlapping ) data sets described in the text . for the iim model , the averaged estimated standard error is also given ( in brackets ) for each parameter ; this is an estimated upper bound on the standard error of the averaged parameter estimate . [ tab2 ] these conversions assume a generation time of 0.1 year and a 10 million year speciation time between _ d. yakuba _ and _ d. melanogaster_/_d . simulans _ ( as assumed by wang and hey 2010 , and by lohse et al . 2011 ; see also powell 1997 ) . it is seen that the iim model places the onset of speciation ( @xmath71 million years ) further back into the past than both the isolation model and the i m model , whereas the estimated time since complete isolation of the two species under the iim model ( @xmath72 million years ) is more recent than under the isolation model . under the iim model we obtain an estimated ancestral effective population size of @xmath73 million individuals , splitting into two populations containing @xmath74 million individuals each during the time interval when gene flow occurred , with _ d. simulans _ expanding to a current effective population size of @xmath75 million individuals . note that the estimated migration rate per generation ( @xmath61 ) is an order of magnitude higher under the iim model than under the i m model . on the one hand , our averaging of the estimated standard errors over the three data sets ( bottom rows of tables [ tab1 ] and [ tab2 ] ) will have given us an overestimate of the standard error of the averaged parameter estimates . on the other hand , however , it should be noted that these estimated standard errors underrepresent the true total amount of uncertainty , as they do not account for uncertainty about the relative mutation rates at the different loci ( which have been treated as known constants , whereas in practice we have estimated them from comparison with _ d. yakuba _ ) . care should be taken therefore in the interpretation of the estimated standard errors stated in tables [ tab1 ] and [ tab2 ] . table [ tab1 ] also gives the maximized loglikelihood ( @xmath53 ) for the different models fitted , for each of the three data sets . it is seen that , amongst the models being compared , the full iim model consistently gives by far the best loglikelihood value , though of course this model also has the largest number of parameters . the easiest way of comparing the fit of the different models is by using akaike s information criterion , aic , which was designed to compare competing models with different numbers of parameters ( aic scores were also used in , for example , takahata et al . 1995 , nielsen and wakeley 2001 , and carstens et al . 2009 ) . for each model ( and for the same data ) , aic is defined as @xmath76 ( akaike 1972 , 1974 ) . thus a larger maximized likelihood leads to a lower aic score , subject to a penalty for each additional model parameter . the minimum aic estimate " ( maice ) is then defined by the model ( and the maximum likelihood estimates of the model parameters ) which gives the smallest aic value amongst the competing models considered . table [ tab1 ] includes , for each data set , the aic scores of the different models considered . it is seen that for each of the three data sets , the maice is given by the full iim model . thus , out of all the models considered here , the iim model provides the best fit to the data , as measured by the aic scores . the results in table [ tab1 ] also show that an iim model with @xmath43 has a substantially better aic score than a symmetric i m model , suggesting that the improved fit of the iim model compared to the i m model is indeed due at least in part to accounting for the eventual complete isolation of the two species , and is not due solely to allowing differences between population sizes . an alternative approach is to perform a series of likelihood ratio tests for nested models ( see , for example , cox 2006 , section 6.5 on tests and model reduction " ) . focusing on any one of the three data sets in table [ tab1 ] , if we start with the full iim model and look upwards , each of the models considered reduces to the model immediately above it by fixing the value of one parameter : respectively @xmath43 , @xmath45 , and @xmath46 . each pair of neighbouring " models can then be formally compared by means of a likelihood ratio test , where the null hypothesis represents the simpler of the two models ( i.e. the one with fewer free parameters ) . if the null hypothesis is true , then the test statistic @xmath77 should approximately follow a @xmath78 distribution if the null " value of the parameter concerned is an interior point of the parameter space ( as is the case when we test @xmath43 ) , whereas we might expect @xmath54 to follow approximately a @xmath79 mixture if the null " value of the parameter concerned lies on the boundary of the parameter space ( as is the case in the tests of @xmath45 and @xmath46 ) ; using a @xmath78 distribution in the latter case is conservative ( self and liang 1987 ; see also our simulation results below ) . the value of the likelihood ratio test statistic @xmath54 is given in table [ tab1 ] for each possible null model , when evaluated against the model immediately underneath it . for each model comparison , the value of the test statistic @xmath54 is so large that the simpler model is rejected in favour of the more complex model underneath it , with , in each case , a very small p - value ( @xmath80 ) providing overwhelming evidence against the simpler model . thus these likelihood ratio tests also identify the full iim model as the most plausible amongst the different models considered . simulations were done to assess how well our method performs and how reliable the estimates obtained in the previous section are . from the three pairwise sets of drosophila sequence data considered in table [ tab1 ] , we arbitrarily selected the first one ( data set 1 " ) and mimicked this data set in our simulations . each simulated data set consists of the number of nucleotide differences between a pair of dna sequences sampled from descendant population 1 for 10,082 independent loci , and the number of differences between a pair of dna sequences taken from different descendant populations for 20,165 independent loci , generated under the full iim model with true " parameter values equal to the estimates obtained for data set 1 ( i.e. @xmath81 , @xmath82 , @xmath83 , @xmath84 , @xmath85 and @xmath86 ) and the infinite sites model of neutral mutation . these numbers of loci for both types of comparison match those in drosophila data set 1 , and so do the relative mutation rates assumed at the different loci . one hundred such data sets were generated . for each simulated data set , ml estimates of the parameters were obtained using our r program for the iim model , i.e. the method described in new approaches " . the resulting estimates are shown in figure [ fig2 ] . , @xmath82 , @xmath83 , @xmath84 , @xmath85 and @xmath86 . each simulated data set consists of the number of nucleotide differences between two dna sequences from descendant population 1 at 10,082 independent loci , and between two dna sequences from different descendant populations at 20,165 independent loci . different loci were assumed to have different mutation rates : the relative mutation rates used at the different loci match those of the drosophila data considered in this paper.,width=574 ] as one would expect , it is seen that ( at least for a large sample size ) our ml estimation procedure gives estimates centred on , and close to , the true " parameter values . in order to assess whether our method enables us to correctly select the iim model from amongst the different models considered , when the iim model is in fact the true underlying model , we also mimicked the type of analysis as was shown for data set 1 in table [ tab1 ] : for each simulated data set , we fitted an isolation model , a symmetric i m model , an iim model with @xmath43 , and a full iim model , and computed the aic scores and the values of the likelihood ratio test statistic @xmath54 . both procedures ( using aic scores or likelihood ratio tests ) correctly identified the full iim model as the best - fitting model , for each of the 100 simulated data sets . the smallest difference obtained between the aic scores of any two neighbouring models was in fact @xmath87 , with the simpler model always having the worse aic score , and the smallest difference observed between the aic scores of the full iim model and any other model was @xmath88 , indicating that the iim model could be identified with ease . for the likelihood ratio approach , comparing the value of the test statistic @xmath54 for a pair of neighbouring models with the @xmath78 distribution gave in all cases a @xmath89-value much smaller than @xmath90 , providing extremely strong evidence against the simpler of the two models compared ( the smallest value of @xmath54 obtained for any pair of neighbouring models was @xmath91 , whereas the @xmath92 quantile of the @xmath78 distribution is only @xmath93 ) . when comparing the full iim model directly with the isolation model by means of a likelihood ratio test , the smallest value of @xmath54 observed amongst the simulated data sets was @xmath94 , again giving a @xmath89-value very much smaller than @xmath90 ( using the @xmath95 distribution ) for each of the 100 simulated data sets . we also simulated 100 data sets under the isolation model , to further investigate the performance of our method in identifying the correct model , and whether false positives may be produced in particular , whether a signal of gene flow may be obtained when in reality there was no gene flow . the true " parameter values assumed in the simulations were those obtained from fitting an isolation model to drosophila data set 1 ( see table [ tab1 ] ) . two of the simulated data sets were problematic , in that only the isolation model could be fitted : due to some large values ( @xmath96 ) for the simulated numbers of nucleotide differences between pairs of sequences from different species , r was unable to evaluate the likelihood for any of the other models in the relevant part of the parameter space . for the remaining @xmath97 simulated data sets for which all models could be fitted , a likelihood ratio test using as the null distribution the naive @xmath98 distribution ( with degrees of freedom equal to the difference between the number of parameters in the two models being compared ) gave better results than did comparison of aic scores . on the basis of aic scores , an incorrect model was selected for as many as 19 of the simulated data sets : in 6 cases the i m model gave the lowest aic score , in 2 cases the iim model with @xmath43 , and in 11 cases the full iim model . however , in 6 of the 11 cases where the full iim model was selected , the estimated migration rate was @xmath99 so that this iim " model was in fact an isolation model but with an additional small change of population size . a likelihood ratio test of the isolation model against the i m model at a significance level of @xmath100 resulted in acceptance of the isolation model for all 98 data sets , regardless of whether we used @xmath78 or @xmath101 as the null distribution . at a significance level of @xmath102 , the isolation model was rejected for 4 of the 98 simulated data sets if we used @xmath78 , and for 6 of the simulated data sets if the mixed @xmath98 distribution was used . figure [ fig3 ] against the @xmath78 distribution ( left ) , and against the mixture @xmath79 ( right ) . on the vertical axis we have plotted the sample quantiles of the values of @xmath54 observed when testing the isolation model against the i m model for @xmath103 data sets simulated from the isolation model ( 2 data sets have been omitted as r was unable to fit the i m model - please see the main text for further details ) . the line @xmath104 is also shown for ease of comparison.,title="fig:",width=317 ] against the @xmath78 distribution ( left ) , and against the mixture @xmath79 ( right ) . on the vertical axis we have plotted the sample quantiles of the values of @xmath54 observed when testing the isolation model against the i m model for @xmath103 data sets simulated from the isolation model ( 2 data sets have been omitted as r was unable to fit the i m model - please see the main text for further details ) . the line @xmath104 is also shown for ease of comparison.,title="fig:",width=317 ] shows that the use of @xmath78 as the null distribution is indeed conservative , and suggests that this may be preferable to using the mixed @xmath98 distribution . similarly , at a significance level of @xmath100 , a likelihood ratio test of the isolation model against the iim model with @xmath43 led to acceptance of the isolation model for all 98 data sets , whether we used @xmath105 or the mixture @xmath106 as the null distribution . at a significance level of @xmath102 , the use of @xmath105 resulted in rejection of the isolation model for 1 of the 98 data sets , whilst the use of @xmath106 as the null distribution led to rejection of the isolation model for 5 of the data sets . the qq - plots shown in figure [ fig4 ] confirm again that the use of @xmath105 in this case is conservative . against the @xmath105 distribution ( left ) , and against the mixture @xmath107 ( right ) . on the vertical axis we have plotted the sample quantiles of the values of @xmath54 observed when testing the isolation model against the iim model with @xmath43 for @xmath103 data sets simulated from the isolation model ( 2 data sets have been omitted as r was unable to fit the iim model with @xmath43 ) . the line @xmath104 is also shown for ease of comparison.,title="fig:",width=317 ] against the @xmath105 distribution ( left ) , and against the mixture @xmath107 ( right ) . on the vertical axis we have plotted the sample quantiles of the values of @xmath54 observed when testing the isolation model against the iim model with @xmath43 for @xmath103 data sets simulated from the isolation model ( 2 data sets have been omitted as r was unable to fit the iim model with @xmath43 ) . the line @xmath104 is also shown for ease of comparison.,title="fig:",width=317 ] similarly , the qq - plots in figure [ fig5 ] indicate that the use of the @xmath95 distribution is conservative when testing the isolation model against the full iim model , and that this may be preferable to the use of a mixed @xmath98 distribution . a likelihood ratio test at a significance level of @xmath100 led to acceptance of the isolation model for all 98 data sets if a @xmath95 distribution was used , and resulted in rejection of the isolation model for 1 of the 98 data sets when using the mixed @xmath108 distribution . at a significance level of @xmath102 , using the @xmath95 distribution resulted in rejection of the isolation model for 4 of the 98 data sets ( in 2 of these 4 cases the estimated migration rate was @xmath99 , reducing the iim model to an isolation model with an additional slight change of population size ) , whereas the use of the mixed @xmath98 distribution led to rejection of the isolation model for 12 of the 98 data sets ( in 6 of these 12 cases , the estimated migration rate was @xmath99 ) . against the @xmath95 distribution ( left ) , and against the mixture @xmath109 ( right ) . on the vertical axis we have plotted the sample quantiles of the values of @xmath54 obtained when testing the isolation model against the full iim model for @xmath103 data sets simulated from the isolation model ( 2 data sets have been omitted as r was unable to fit the iim model ) . the line @xmath104 is also shown for ease of comparison.,title="fig:",width=317 ] against the @xmath95 distribution ( left ) , and against the mixture @xmath109 ( right ) . on the vertical axis we have plotted the sample quantiles of the values of @xmath54 obtained when testing the isolation model against the full iim model for @xmath103 data sets simulated from the isolation model ( 2 data sets have been omitted as r was unable to fit the iim model ) . the line @xmath104 is also shown for ease of comparison.,title="fig:",width=317 ] in this paper we have presented a very fast method to obtain ml estimates and to distinguish between different evolutionary scenarios , using nucleotide difference data from pairs of sequences at a large number of independent loci . the iim model considered allows for an initial period of gene flow between two diverging populations or species , followed by a period of complete isolation , and is more appropriate in the context of speciation than the i m model commonly used in the literature ( which assumes that gene flow continues at a constant rate all the way until the present time ) . we have illustrated the speed and power of our method by applying it to a large data set from two related species of drosophila , and to simulated data . fitting an iim model to a data set of approximately 30,000 loci ( with an average length of about 400 bp and varying mutation rates ) from two species of drosophila took approximately 20 seconds on a desktop pc ; this time included fitting an isolation model first in order to obtain good starting values for the parameters . moreover , the results made it possible to distinguish between more and less plausible models ( representing alternative evolutionary scenarios ) with ease , and identified the iim model as providing the best fit amongst the models considered . the drosophila data set studied in this paper was previously analysed by wang and hey ( 2010 ) , who compiled the data , and by lohse et al . both fitted i m models to these data , assuming that gene flow occurred at a constant rate from the time of separation of the two species until the present time . in tables [ tab1 ] and [ tab2 ] , alongside our results for the iim model , we also included parameter estimates for a symmetric i m model , for the sake of comparison . as expected , our i m results are in close agreement with those of lohse et al . ( 2011 ) , who fitted two versions of the i m model , one with symmetric migration and one with migration in only one direction ( _ d . simulans _ to _ d. melanogaster _ forward in time , motivated by wang and hey s results ) . they found that their estimated migration rate for the symmetric model is approximately half that obtained for the model with migration in one direction , i.e. both models gave the same estimated total number of migrants per generation between the two species in the two directions combined , whereas their estimates of the other parameters are virtually identical for the two models ( see p.23 of their supporting information ) . our i m results are also in good agreement with those of wang and hey ( 2010 ) , even though they fitted a more general version of the i m model ( allowing for unequal sizes of the two species and unequal migration rates in the two directions ) and assumed the jukes - cantor model of mutation , whereas our results and those of lohse et al . ( 2011 ) assume the infinite sites model for its mathematical simplicity . indeed , lohse et al . ( 2011 ) comment on how little difference the assumption of the infinite sites model makes to the i m results for the pairwise drosophila data , compared to wang and hey s results based on the jukes - cantor model . of more interest is the comparison of these authors i m results with those obtained for our iim model . we note firstly that their estimates of 3.04 million years ( wang and hey 2010 ) or 2.98 million years ( lohse et al . 2011 ) for the _ simulans - melanogaster _ speciation time , and also our i m estimate of 3.22 million years , fall in between our iim estimates of 3.66 million years for the time since the onset of speciation ( @xmath68 ) and 1.61 million years for the time since complete isolation of these two species ( @xmath67 ) . secondly , under the i m model the amount of gene flow between the two species ( in the two directions combined ) was estimated at 0.0134 migrant gene copies per generation ( wang and hey 2010 ) , 0.0255 migrant gene copies per generation ( lohse et al . 2011 , asymmetric i m model ) , 0.0256 ( lohse et al . 2011 , symmetric i m model ) , or 0.0210 migrant gene copies per generation ( our i m results in table [ tab1 ] , averaged over the three data sets ) . these estimates are considerably smaller than the corresponding estimate of 0.0887 migrant gene copies per generation during the period of migration under our iim model . lohse et al . ( 2011 ) also considered a trimmed version of the wang and hey ( 2010 ) drosophila data , where they shortened all loci to a fixed number of nucleotide differences between _ d. melanogaster _ and the _ d. yakuba _ outgroup , so that the estimated mutation rate for all loci was equal . this assumption that all loci have the same mutation rate massively speeds up the computation of the likelihood , since in this case the probability of @xmath0 differences needs to be calculated only once for each observed value of @xmath0 , rather than having to calculate this probability over and over again for different mutation rates at different loci . further to this idea , we also implemented a simplified , much faster , version of our mle method to be used for data sets where all loci have the same mutation rate ( this r code is also included as supplementary material ) . applying this version of our iim program to the trimmed drosophila data ( still approximately 30,000 loci ) gave virtually instant results on a desktop pc , i.e. the computing time was @xmath110 second . however , the substantial shortening of loci did result in loss of information , and the distinction between more and less plausible models was not always as clear anymore , as the differences between the values of the maximized loglikelihood for the different models ( and hence the values of the likelihood ratio test statistic @xmath54 , and the differences between the aic scores of different models ) were not as large as they were for the full sequence data . nevertheless , the full iim model could still be identified as giving the best fit amongst the models considered in this paper , both on the basis of aic scores and by means of likelihood ratio tests . table [ tab3 ] gives the estimates obtained by fitting the iim model to the trimmed data ; the conversion to times in years , diploid effective population sizes and migration rate per generation was done using a generation time of 0.1 year and a 10 million year speciation time between _ d. yakuba _ and \{_d . melanogaster _ , _ d. simulans _ } , as before . the results are broadly in agreement with the corresponding estimates for the full sequence data ( see table [ tab2 ] , bottom row ) . lcccccc + model & @xmath56 & @xmath57 & @xmath58 & @xmath59 & @xmath60 & @xmath61 + + iim & 1.52my & 3.67my & 6.04 m & 3.94 m & 3.82 m & @xmath111 + ( s.e . ) & ( 0.16my ) & ( 0.10my ) & ( 0.11 m ) & ( 0.33 m ) & ( 0.18 m ) & ( @xmath112 ) + + note the abbreviations my " and m " stand for million years " and million individuals " . the times @xmath67 and @xmath68 denote , respectively , the time since complete isolation of _ d. simulans _ and _ d. melanogaster _ , and the time since the onset of speciation , i.e. these are the times @xmath4 and @xmath2 converted into years ( see fig.1 ) . the effective population sizes @xmath113 , @xmath7 and @xmath114 refer respectively to the present _ d. simulans _ population , each population during the migration stage of the model , and the ancestral population . the estimates shown have been averaged over the three overlapping data sets described in the text . the averaged estimated standard error is also given ( in brackets ) for each parameter ; this is an estimated upper bound on the standard error of the averaged parameter estimate . [ tab3 ] the purpose of our analysis of the drosophila data was merely to illustrate the potential of our method , rather than to draw any firm conclusions about the evolutionary history of the particular species concerned . whilst the iim model provides the best fit to the drosophila data amongst the different models considered in this paper , it is likely that more realistic models can be constructed which provide yet a better fit than the version of the iim model considered here . our assumption of equal population sizes and symmetric migration during the period of gene flow in the model is obviously an unrealistic oversimplification , but was made for the sake of mathematical tractability and computational speed . an extension of our method to a more general iim model allowing for unequal migration rates and unequal population sizes during the migration stage of the model is possible and is being developed ( costa rj and wilkinson - herbots hm , work in progress ) . similarly , the current implementation of our method assumes the infinite sites model of neutral mutation because of its mathematical ease . extensions of our method to other neutral mutation models ( for example , the jukes - cantor model ) can be done but have not yet been implemented . it should also be feasible to extend our method to incorporate more than two species . whilst our method explicitly allows for mutation rate heterogeneity between loci ( whether due to variation in sequence length or variation in the mutation rate per site , or both ) , our current implementation assumes that accurate estimates of the relative mutation rates of the different loci are available ( our implementation treats the relative mutation rates as known constants ) . this limitation will become less important as more and more genome sequences become available , allowing more accurate estimation of the relative mutation rates . for our analysis of the drosophila data presented in this paper , however , the relative mutation rates of the different loci were estimated by comparison of \{_d . simulans _ , _ d. melanogaster _ } with the outgroup _ d. yakuba _ , as was done also by wang and hey ( 2010 ) and by lohse et al . ( 2011 ) . using more than one outgroup sequence ( if available ) to estimate the relative mutation rate at each locus should improve the accuracy . it may also be possible to adapt our method to incorporate uncertainty about the relative mutation rates by modelling mutation rate variation as a random variable and integrating out over it ( see yang 1997 , or the all - rate " method proposed by wang and hey 2010 ) . the drosophila data considered in this paper are the _ d. melanogaster - d. simulans _ divergence data compiled and analyzed by wang and hey ( 2010 ) ; we used the subset of the data that was studied also by lohse et al . the data consist of alignments of 30,247 segments of intergenic sequence of length 500 bp each , from two inbred lines of _ d. simulans _ and one inbred line of _ d. melanogaster _ , and from an inbred line of _ d. yakuba _ for use as an outgroup . the data have been pre - processed as described in wang and hey ( 2010 ) and lohse et al . ( 2011 ) ; the version of the data used is that labelled wangheyraw " in the supporting information of lohse et al . as our method uses pairwise nucleotide differences , subsets of the data were formed by selecting one pair of sequences from each locus ( i.e. _ d.mel _ - _ d.sim_1 , _ d.mel _ - _ d.sim_2 , or _ d.sim_1 - _ d.sim_2 ) . three such subsets were formed as described in the section on results " ; these three data sets overlap as each sequence at each locus is used in two of the three data sets . parameter estimates and estimated standard errors were obtained for each of these three data sets separately and then averaged over the three data sets . this gives too large estimates of the standard errors of the averaged parameter estimates ; however , due to the overlap between the data sets , more accurate estimated standard errors would be difficult to obtain . following wang and hey ( 2010 ) and lohse et al . ( 2011 ) , mutation rate heterogeneity between loci was accounted for by comparing the _ d. melanogaster _ and _ d. simulans _ sequences with the outgroup _ d. yakuba_. calculating the outgroup divergence @xmath115 between \{_d . simulans _ , _ d. melanogaster _ } and _ d. yakuba _ at locus @xmath31 as a weighted average over the available sequences , assigning 25% weight to each of the _ simulans _ sequences and 50% weight to the single _ melanogaster _ sequence , the relative mutation rate @xmath116 of locus @xmath31 was estimated as @xmath117 , where @xmath118 is the average of the @xmath115 over all loci . the scaled mutation rate at locus @xmath31 is then given by @xmath119 , where @xmath19 is the average scaled mutation rate over all the loci considered . the relative mutation rates @xmath116 are treated as fixed constants in the likelihood maximization ( see also yang 1997 , 2002 ) . estimates of the parameters of the iim model were obtained by maximizing the likelihood , given by equation ( [ likelihood ] ) , using the reparameterization ( [ pars ] ) . in order to obtain reasonable starting values for the likelihood maximization under the iim model , it can be helpful to first fit an isolation model , as the latter has a more tractable likelihood surface . estimated standard errors are obtained from the inverted hessian matrix . the computer code was written in r and is included as supplementary material . table [ tab1 ] shows the results obtained for the three sets of drosophila data . to convert the parameter estimates into readily interpretable units , we followed wang and hey ( 2010 ) and lohse et al . ( 2011 ) in using a generation time of @xmath120 year and a 10 million year speciation time between _ d. yakuba _ and \{_d . simulans _ , _ d. melanogaster _ } ( see also powell 1997 ) , which gives us an estimate of @xmath121 for the mutation rate per locus per generation , averaged over all the loci included in the analysis . the converted estimates given in table [ tab2 ] were then calculated according to the following equations : @xmath122 for the times in years since complete isolation and since the onset of speciation , respectively ; @xmath123 for the effective size ( number of diploid individuals ) of either population during the migration stage of the model , and similarly for all other population sizes ; and @xmath124 for the migration rate per generation . estimated standard errors for the converted estimates were obtained by re - running a reparameterized version of the program , in terms of @xmath125 and @xmath126 instead of @xmath37 and @xmath127 , using the ml estimates already obtained as starting values , allowing easy computation of the hessian matrix . as before , estimated standard errors were computed separately for each of the three data sets and then averaged . following lohse et al . ( 2011 ) we also analyzed a trimmed version of the wang and hey ( 2010 ) drosophila data , where each locus was cut after 16 nucleotide differences between _ d. melanogaster _ and _ d. yakuba _ , and the remainder of the locus was ignored ; the 2,090 loci which had fewer than 16 differences between _ d. melanogaster _ and _ d. yakuba _ were omitted altogether , leaving 28,157 trimmed loci for analysis . this trimming amounts to a shortening of the loci by roughly a factor of 3 on average . again , three ( overlapping ) subsets of the data were formed , each containing one pair of sequences from each locus ( i.e. _ d.mel _ - _ d.sim_1 , _ d.mel _ - _ d.sim_2 , or _ d.sim_1 - _ d.sim_2 ) as described above and in the section on results " . parameter estimates and estimated standard errors were obtained for each of these three data sets separately and then averaged over the three data sets . we used a simplified version of our r code written specifically for data sets where all loci have the same estimated mutation rate ; this simplified code is also included as supplementary material . the likelihood calculation uses the frequencies of the observed numbers of pairwise differences , which is much faster than the corresponding code for the case where different loci have different mutation rates : for the trimmed drosophila data , instead of evaluating 28,157 terms in the loglikelihood ( one term for each locus ) , only about 50 different terms need to be calculated ( corresponding to the different values of the number of pairwise differences observed , within and between species ) and multiplied by their frequencies , which hugely speeds up the likelihood computation and maximization . table 3 shows the results obtained under the iim model , after conversion into conventional units as explained above , using a mutation rate of @xmath128 per locus per generation for the trimmed data ( see also lohse et al 2011 ) . simulated pairwise difference data were generated by first simulating the coalescence times of pairs of sequences and then superimposing neutral mutation under the infinite sites model ; our r code is included as supplementary material . for pairs of sequences from the same species , our simulation algorithm uses the shortcut " provided by equations ( 8) and ( 9 ) in wilkinson - herbots ( 2012 ) , exploiting the fact that the distribution of the coalescence time of a pair of sequences sampled from the same species in the iim model is a mixture of two piecewise exponential " distributions ( with change points " @xmath4 and @xmath2 ) , eliminating the need to explicitly simulate migration events . i would like to express my sincere thanks to yong wang , jody hey , konrad lohse and nick barton for the use of their drosophila data . thanks are also due to paul northrop and rex galbraith for some helpful tips on programming in r , and to ziheng yang for some valuable discussions . this work was supported by the engineering and physical sciences research council via an institutional sponsorship award to university college london ( grant number ep / k503459/1 ) . the r code which fits an iim model and an isolation model to pairwise difference data from a large number of independent loci is supplied as supplementary material . the r code for simulating pairwise difference data under the iim model is also provided . : : akaike h. 1972 . information theory and an extension of the maximum likelihood principle . in : petrov bn , csaki f , editors . 2nd int . information theory , supp . to problems of control and information theory . p. 267 - 281 . : : akaike h. 1974 . a new look at the statistical model identification . ieee transactions on automatic control ac-19:716 - 723 . : : andersen ln , mailund t , hobolth a. 2014 . efficient computation in the i m model . j math biol . 68:1423 - 51 . : : becquet c , przeworski m. 2007 . a new approach to estimate parameters of speciation models with application to apes . genome res . 17:1505 - 1519 . : : becquet c , przeworski m. 2009 . learning about modes of speciation by computational approaches . evolution 63:2547 - 2562 . : : bird ce , fernandez - silva i , skillings dj , toonen rj . 2012 . sympatric speciation in the post modern synthesis " era of evolutionary biology . evol biol . 39:158 - 180 . : : burgess r , yang z. 2008 . estimation of hominoid ancestral population sizes under bayesian coalescent models incorporating mutation rate variation and sequencing errors . mol biol evol . 25:1979 - 1994 . : : carstens bc , stoute hn , reid nm . an information - theoretical approach to phylogeography . mol ecol 18:4270 - 4282 . : : cox dr . 2006 : principles of statistical inference . cambridge university press . : : hey j. 2005 . on the number of new world founders : a population genetic portrait of the peopling of the americas . plos biol . 3:965 - 975 . : : hey j. 2010 . isolation with migration models for more than two populations . mol biol evol 27:905 - 920 . : : hey j , nielsen r. 2004 . multilocus methods for estimating population sizes , migration rates and divergence time , with applications to the divergence of drosophila pseudoobscura and d. persimilis . genetics 167:747 - 760 . : : hey j , nielsen r. 2007 . integration within the felsenstein equation for improved markov chain monte carlo methods in population genetics . proc natl acad sci usa . 104:2785 - 2790 . : : hobolth a , andersen ln , mailund t. 2011 . on computing the coalescence time density in an isolation - with - migration model with few samples . genetics 187:1241 - 1243 . : : innan h , watanabe h. 2006 . the effect of gene flow on the coalescent time in the human - chimpanzee ancestral population . mol biol evol . 23:1040 - 1047 . : : li h , durbin r. 2011 . inference of human population history from individual whole - genome sequences . nature 475(7357):493 - 496 . : : lohse k , sharanowski b , stone gn . 2010 . quantifying the pleistocene history of the oak gall parasitoid _ cecidostiba fungosa _ using twenty intron loci . evolution 64:2664 - 2681 . : : lohse k , harrison rj , barton nh . a general method for calculating likelihoods under the coalescent process . genetics 189:977 - 987 . : : maddison wp , knowles ll . 2006 . inferring phylogeny despite incomplete lineage sorting . . 55:21 - 30 . : : nadachowska k. 2010 . divergence with gene flow - the amphibian perspective . herpetological journal 20:7 - 15 . : : nielsen r , wakeley j. 2001 . distinguishing migration from isolation : a markov chain monte carlo approach . genetics 158:885 - 896 . : : pinho c , hey j. 2010 . divergence with gene flow : models and data . annu rev ecol evol syst . 41:215 - 230 . : : powell jr . progress and prospects in evolutionary biology : the drosophila model . oxford university press . : : r development core team . 2011 . r : a language and environment for statistical computing . r foundation for statistical computing , vienna , austria . url http://www.r - project.org/. : : self sg , liang ky . 1987 . asymptotic properties of maximum likelihood estimators and likelihood ratio tests under non - standard conditions . j am stat assoc . 82:605 - 610 . : : smadja cm , butlin rk . a framework for comparing processes of speciation in the presence of gene flow . 20:5123 - 5140 . : : sousa vc , grelaud a , hey j. 2011 . on the nonidentifiability of migration time estimates in isolation with migration models . molecular ecology 20:3956 - 3962 . : : strasburg jl , rieseberg lh . 2010 . how robust are isolation with migration " analyses to violations of the i m model ? a simulation study . mol biol evol . 27:297 - 310 . : : strasburg jl , rieseberg lh . 2011 . interpreting the estimated timing of migration events between hybridizing species . molecular ecology 20:2353 - 2366 . : : takahata n. 1995 . a genetic perspective on the origin and history of humans . annu rev ecol syst . 26:343 - 372 . : : takahata n , satta y , klein j. 1995 . divergence time and population size in the lineage leading to modern humans . theor popul biol . 48:198 - 221 . : : takahata n , satta y. 1997 . evolution of the primate lineage leading to modern humans : phylogenetic and demographic inferences from dna sequences . proc natl acad sci usa . 94:4811 - 4815 . : : teshima km , tajima f. 2002 . the effect of migration during the divergence . theor popul biol . 62:81 - 95 . : : wang y , hey j. 2010 . estimating divergence parameters with small samples from a large number of loci . genetics 184:363 - 379 . : : watterson ga . 1975 . on the number of segregating sites in genetical models without recombination . theor popul biol . 7:256 - 276 . : : wilkinson - herbots hm . the distribution of the coalescence time and the number of pairwise nucleotide differences in the `` isolation with migration '' model . theor popul biol . 73:277 - 288 . : : wilkinson - herbots hm . the distribution of the coalescence time and the number of pairwise nucleotide differences in a model of population divergence or speciation with an initial period of gene flow . theor popul biol . 82:92 - 108 . : : yang z. 1997 . on the estimation of ancestral population sizes of modern humans . genet res camb . 69:111 - 116 . : : yang z. 2002 . likelihood and bayes estimation of ancestral population sizes in hominoids using data from multiple loci . genetics 162:1811 - 1823 . : : yang z. 2010 . a likelihood ratio test of speciation with gene flow using genomic sequence data . genome biol evol . 2:200 - 211 . : : zhu t , yang z. 2012 . maximum likelihood implementation of an isolation - with - migration model with three species for testing speciation with gene flow . mol biol evol . 29:3131 - 3142 . |
we present and discuss the results of new multiple - scattering radiative transfer calculations for three representative types of galactic environments , filled with either homogeneous or two - phase clumpy dust distributions . extinction and scattering properties for two types of interstellar dust , similar to those found in the average diffuse medium of the milky way galaxy ( mw ) and the bar of the small magellanic cloud ( smc ) , are considered . the wavelength coverage extends from 1000 to 30,000 , with particular emphasis on the rest - frame uv . this makes these models especially applicable to starburst galaxies and lyman - break galaxy samples . the examination of the models concentrates on the study of uv / visual / near - ir reddening effects , the wavelength dependence of attenuation , and on the changes that arise from the transition from homogeneous to clumpy dust distributions in different star / dust geometries . embedded dust , especially when clumpy , leads to saturation at fairly low reddening values with correspondingly gray attenuation functions . this makes the assessment of the attenuation of the far - uv flux from starburst galaxies difficult , if only uv / visual / near - ir data are available . existing data for uv - selected starburst galaxies indicate a range of uv attenuation factors of 0150 . our models reproduce the `` calzetti attenuation law '' , provided one adopts smc - type dust and a clumpy shell - type dust distribution surrounding the starbursts . the average far - uv attenuation factor for the calzetti sample is found to be 7.4 . the only relatively reliable measure for the uv attenuation factor for single galaxies was found in the ratio of the integrated far - ir flux to the far - uv flux , measured near 1600 , requiring the measurement of the entire spectral energy distribution of galaxies . | astro-ph9907342 | the presence of dust in galaxies has a profound impact on observable characteristics such as their apparent uv / optical / near - ir luminosity , their spectral energy distribution , their recognizable level of star - formation activity , and their morphology . in addition , dust directly affects and often controls the physical conditions within the interstellar medium of galaxies and plays a major role in interstellar chemistry by providing the sites for the formation of interstellar molecular hydrogen and by regulating the interstellar radiation field . quantitative estimates of the dust mass responsible for the observed effects are needed in order to yield gas - to - dust ratios in galaxies of different types , essential if we are to understand the chemical evolution of galaxies over cosmological time scales . radiative transfer simulations often provide the most direct way to arrive at a quantitative assessment of the ways that lead to observed dust effects , provided such simulations are sufficiently realistic to apply to actual physical systems . all such simulations involve approximations . historically , these have ranged from the adoption of absorbing screen geometries ( @xcite , 1975 ) in analogy to the study of extinction of starlight in our galaxy , to the consideration of various geometries with embedded dust , including the effects of multiple scattering ( witt , thronson , & capuano 1992 , hereafter @xcite ; @xcite ) . according to @xcite , a galaxy may be looked upon as consisting of environments which are characterized by certain geometrical relations between the attenuating dust and the illuminating stars , for example , starburst regions may be approximated by regions occupied by luminous stars , surrounded on the outside by the remnants of molecular clouds from which the stars originated . quiescent parts of dusty disk galaxies may resemble systems in which stars and dust are fairly uniformly mixed , or , if one considers the same case at longer wavelengths where the dominant sources might be k - giants , a case where the scale height of sources is much larger than the scale height of the dust which is concentrated toward the center or central plane . as shown by @xcite , equal amounts of dust in these different configurations produce very different reddening and attenuation effects . for the analysis of effects resulting from changes in the inclination of disk galaxies , disk models with embedded dust have been developed , either in plane - parallel or doubly exponential geometry ( e.g. @xcite ; @xcite ; @xcite ; @xcite ) . all of the models mentioned so far consistently illustrate the fact that a certain quantity of embedded dust produces much smaller attenuation effects than a foreground screen made from the same amount of dust , and that the inclusion of multiple scattering reduces these effects even further by returning the scattered radiation to the optical radiation field , albeit with an altered spectral and directional distribution . an approximation in common among all the models mentioned above is the homogeneous nature of their assumed dust distributions , which are either constant or of a monotonically varying continuous variety . in the first paper in the present series ( witt & gordon 1996 , hereafter @xcite ) , we conducted an extensive exploration of the effects of clumpiness on the transport of radiation through a spherical system containing scattering dust and a single central source this study , together with the results of earlier work on transfer through clumpy media ( e.g. @xcite ; @xcite ; @xcite ) clearly revealed that the clumpy multi - phase structure of the interstellar medium ( ism ) leads to a significant reduction in the opacity produced by a given distribution and mass of interstellar dust . in addition , @xcite showed that the relationship between the effective optical depth of a clumpy distribution and that of an equivalent homogeneous distribution is non - linear ; this leads to an effective attenuation as a function of wavelength that differs from the opacity law intrinsic to the dust in the system . these conclusions have been further amplified in a recent investigation of the radiative transfer in the clumpy environment of young stellar objects by wolf , fischer , & pfau ( 1998 ) and general investigation of clumpy radiative transfer by vrosi & dwek ( 1999 ) . since clumpiness is a reality of the ism wherever it can be observed with sufficient spatial resolution , we are introducing it into the transfer of radiation in such galactic environments as investigated in @xcite now . a preliminary version of some aspects of this work appeared in gordon , calzetti , & witt ( 1997 ) . in addition to adding models with a clumpy two - phase structure , we have improved several other aspects of the original @xcite study . our knowledge of dust scattering properties in the uv has greatly improved since 1992 , and with some confidence we are now modeling the transfer for 15 ( out of a total of 25 ) bandpasses in the wavelength range from 1000 to 3000 , compared to two in @xcite . this improved uv coverage makes the current models particularly well suited for the analysis of starburst galaxies , both nearby and at high redshift , since these systems emit much of their stellar energy at uv wavelengths . furthermore , most previous multi - wavelength radiative transfer studies have assumed opacity functions and scattering properties representative of average milky way ( mw ) dust . even within the mw dust characteristics vary from one environment to another ( e.g. @xcite ) and already in the local group of galaxies there exists a considerable variety of opacity laws ( @xcite ; @xcite ; @xcite ) . in starburst galaxies , in particular , it has been shown ( @xcite ) that the opacity function intrinsic to the dust present there is much more similar to dust found typically in the bar of the small magellanic cloud ( smc ) . we therefore model all our galactic environments with both the average mw and smc bar dust . other radiative transfer studies which have included both mw and smc bar type dust are those detailed by gordon et al . ( 1997 ) , takagi , arimoto , & vansevicius ( 1999 ) , and ferrara et al . the models presented in the present paper are particularly well suited to analyze the reddening effects and the corresponding uv attenuation in lyman - break galaxies and galaxies seen in the hubble deep fields ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite , @xcite ) . they are equally well usable for studying opacity effects in nearby dusty galaxies ( e.g. @xcite ; @xcite ; @xcite ; @xcite ) . the principal results from the current work are an extensive set of tables , accessible in the on - line version of this issue of the astrophysical journal or available by contacting the authors . we will present the details of the models in the following section 2 ; selected results will be summarized in graphical form in section 3 ; we will discuss the implications of these results for currently ongoing galaxy studies in section 4 , to be followed by a summary in section 5 . in order to model the dust radiative transfer in complex systems such as galaxies , we use a radiative transfer model ( named dirty ) based on monte carlo techniques ( @xcite ) . use of such a model allows for arbitrary distributions of both the emitting sources ( stars ) and absorbing and scattering material ( dust ) . the full description of dirty can be found in gordon et al . ( 1999b ) . the inputs to dirty are the stellar distribution , the dust distribution , and the dust grain properties . with monte carlo techniques , individual photons are emitted isotropically from stars according to the stellar distribution and followed through a dust distribution . each photon s interaction with the dust is parameterized by the dust optical depth ( @xmath0 ) , albedo , and scattering phase function . the optical depth determines where the photon interacts , the albedo gives the probability the photon is scattered from a dust grain , and the scattering phase function gives the angle at which the photon scatters . as was done in @xcite , we have chosen to model spherical galactic environments instead of entire galaxies . a galactic environment describes the distribution of the stars and gas . in this paper we concentrate on three galactic environments : cloudy , dusty , and shell . these three geometries span the range of star / dust distributions and are pictorially represented in figure [ fig_geometries ] . the cloudy geometry has dust extending to 0.69 of the system radius and stars extending to the system radius . the dusty geometry has both dust and stars extending to the system radius . this geometry represents a uniform mixture of stars and dust . the shell geometry has stars extending to 0.3 of the system radius and dust extending from 0.3 to 1 of the system radius . this geometry was named nuclei in @xcite . these three geometries characterize the _ global _ star and dust distribution . the _ local _ distribution of the dust can be either homogeneous or clumpy . for this paper , we define the clumpy distribution ( using the definitions of @xcite ) to have a filling factor ( _ ff _ ) of 0.15 , a low - density to high - density ratio ( @xmath1 ) of 0.01 , and a system size characterized by a cube divided into @xmath2 cubical bins ( @xmath3 ) . a forthcoming third paper in this series will present the radiative transfer through doubly exponential disk / bulge galaxies with clumpy ism . we have chosen to compute models using both milky way ( mw ) and small magellanic cloud ( smc ) dust . by using these two different types of dust , we are spanning the range of _ known _ dust extinction curves . for monte carlo radiative transfer , we need to know the dependence of the dust s optical depth ( @xmath0 ) , albedo ( @xmath4 ) , and scattering phase function on wavelength . in this work , we used the @xcite phase function which is parameterized by the phase function weighted average of the cosine of the scattering angle ( @xmath5 ) . for the mw , we used dust which is associated with the diffuse interstellar medium and has @xmath6 . the wavelength dependence of @xmath0 was determined from the parameterization of milky way extinction curves by cardelli , clayton , & mathis ( 1991 , hereafter @xcite ) . the @xmath4 and @xmath7 values were determined by comparing empirical determinations of these numbers to a dust grain model which reproduces the @xmath8 extinction curve ( @xcite ) . figure [ fig_dust_a_g]a shows the empirical determinations , dust grain model values , and the numbers adopted in this paper . in determining the albedo values to use for this paper , we visually averaged the empirical determinations and then shifted the average to agree with the dust grain model values . the shift was done because most of the empirical determinations were done for reflection nebulae which usually have higher values of @xmath9 than 3.1 . the dust grain model predicts decreasing albedo s for decreasing @xmath9 s . this is also consistent with recent work on the ultraviolet diffuse galactic light ( @xcite ) . when we averaged the albedo values , we also made sure that @xmath10 ( @xmath11 ) was smooth across the 2175 bump as it is an absorption feature ( @xcite ) . figure [ fig_dust_q]a plots @xmath12 , @xmath10 , and @xmath13 for the mw albedo values used in this paper . the @xmath7 values shown in fig . 2b were taken as averages of the empirical determinations , as the @xmath7 values from the ultraviolet diffuse galactic light work are not significantly different from those of work on reflection nebulae . also , the dust grain model still has problems in that it predicts the 2175 bump as partly a scattering feature , contrary to observational evidence . the @xmath0 , @xmath4 , and @xmath7 values adopted for mw dust in this paper are given in table [ tab_dust_prop ] . for the smc , we used dust which is associated with the star forming bar of the smc ( @xcite ) . this dust produces extinction curves which are quite different from dust extinction curves in the mw ( @xcite ) or the large magellanic cloud ( @xcite ) . it lacks a 2175 bump and is roughly linear with @xmath14 . this type of dust seems to be associated with intense star forming regions ( @xcite ; @xcite ; @xcite ) . the @xmath0 values were taken from an average of the three known extinction curves in the smc bar ( @xcite ) . there is no empirical work on the @xmath4 and @xmath7 values in the smc . as there are also still problems with dust grain models reproducing the mw @xmath4 and @xmath7 values , we have chosen to start with the mw @xmath4 values and then modify them to reflect the lack of a 2175 bump . this time , we required that the @xmath10 and @xmath13 values did not show any features across the 2175 region . figure [ fig_dust_q]b displays @xmath12 , @xmath15 , and @xmath13 for the smc albedo values used in this paper . the @xmath0 , @xmath4 , and @xmath7 values adopted for smc dust in this paper are given in table [ tab_dust_prop ] . ccccccc 1000.0 & 5.238 & 0.320 & 0.800 & 9.675 & 0.400 & 0.800 1142.9 & 3.918 & 0.409 & 0.783 & 7.440 & 0.449 & 0.783 1285.7 & 3.182 & 0.481 & 0.767 & 6.068 & 0.473 & 0.767 1428.6 & 2.780 & 0.526 & 0.756 & 5.167 & 0.494 & 0.756 1571.4 & 2.584 & 0.542 & 0.745 & 4.536 & 0.508 & 0.745 1714.3 & 2.509 & 0.536 & 0.736 & 4.074 & 0.524 & 0.736 1857.1 & 2.561 & 0.503 & 0.727 & 3.700 & 0.529 & 0.727 2000.0 & 2.843 & 0.432 & 0.720 & 3.379 & 0.528 & 0.720 2142.9 & 3.190 & 0.371 & 0.712 & 3.101 & 0.523 & 0.712 2285.7 & 2.910 & 0.389 & 0.707 & 2.857 & 0.520 & 0.707 2428.6 & 2.472 & 0.437 & 0.702 & 2.642 & 0.516 & 0.702 2571.4 & 2.194 & 0.470 & 0.697 & 2.452 & 0.511 & 0.697 2714.3 & 2.022 & 0.486 & 0.691 & 2.282 & 0.505 & 0.691 2857.1 & 1.905 & 0.499 & 0.685 & 2.133 & 0.513 & 0.685 3000.0 & 1.818 & 0.506 & 0.678 & 2.031 & 0.515 & 0.678 3776.8 & 1.527 & 0.498 & 0.646 & 1.610 & 0.498 & 0.646 4754.7 & 1.199 & 0.502 & 0.624 & 1.221 & 0.494 & 0.624 5985.8 & 0.909 & 0.491 & 0.597 & 0.880 & 0.489 & 0.597 7535.8 & 0.667 & 0.481 & 0.563 & 0.630 & 0.484 & 0.563 9487.0 & 0.440 & 0.500 & 0.545 & 0.430 & 0.493 & 0.545 11943.5 & 0.304 & 0.473 & 0.533 & 0.272 & 0.475 & 0.533 15036.0 & 0.210 & 0.457 & 0.511 & 0.166 & 0.465 & 0.511 18929.2 & 0.145 & 0.448 & 0.480 & 0.111 & 0.439 & 0.480 23830.6 & 0.100 & 0.424 & 0.445 & 0.075 & 0.417 & 0.445 30001.0 & 0.069 & 0.400 & 0.420 & 0.033 & 0.400 & 0.420 the results of our radiative transfer calculations are contained in tables 2 , 3 , and 4 , accessible in full in the online edition of this issue or available directly from the authors . table 2 contains the data for the dusty galactic environment , table 3 those for the shell galactic environment , and table 4 those for the cloudy environment . only the beginning of table 2 is shown in this paper in order to illustrate the type of data available in the full set of tables . each model is identified by its geometry ( dusty , shell , cloudy ) , the intrinsic opacity function ( mw , smc ) , the radial extinction optical depth from the center to the edge of the dust environment at v ( @xmath16 0.25 , 0.5 , ... , 50 ) , assuming a constant density homogeneous distribution , and the type of structure ( homogeneous , clumpy ) in the dust region . each individual model contains calculations for 25 wavelengths , ranging from 1000 to 30000 , shown in column 1 . column 2 identifies the optical depths for extinction at each wavelength , derived from the applicable intrinsic opacity function . columns 3 , 4 , 5 , and 6 list , respectively , the attenuation optical depth ( @xmath17 ) , the scattered flux fraction , the direct attenuated stellar flux fraction , and the total escaping flux fraction , containing both attenuated stellar as well as scattered radiation , for the homogeneous case . columns 7 to 10 repeat this information for the equivalent clumpy case . all fluxes are referred to unity at each wavelength , with unit flux corresponding to no attenuation , and in cases with clumpy dust distributions , they are averages over all directions . for a given geometry , the value of the v - band optical depth may be regarded as a measure of the dust mass present in that geometry . the attenuation optical depth ( column 3 and 7 ) is defined as @xmath18 and thus measures the absorption occurring in the system . this is a rather useful measurement since in galactic environments viewed from afar both the attenuated stellar flux as well as the scattered light are combined in the escaping total flux observed . note that this definition of the attenuation optical depth differs from that of the effective optical depth ( @xmath19 ) defined in @xcite , which refers to the attenuated , non - scattered stellar flux only . the @xmath17 values are always smaller than the radial optical depth listed in column 2 , because the dust albedo is always non - zero and the stars are distributed throughout extended volumes . also , the attenuation optical depth for the clumpy cases never exceeds that of the corresponding homogeneous cases ( @xcite ) . it needs to be reemphasized that the clumpy case treated here is for one specific set of clumpiness parameters , namely a filling factor ff = 0.15 , an interclump medium to clump density ratio @xmath1 = 0.01 , and a system size divided into @xmath2 cubical bins . the variations of transfer characteristics expected when any of these parameters are changed within reasonable ranges were explored in @xcite , and they are briefly discussed in section 3.5 . rc|cccc|cccc & & 1000 & 1.309 & 0.589 & 1.07e-01 & 4.48e-01 & 5.55e-01 & 0.517 & 8.84e-02 & 5.08e-01 & 5.96e-01 1142 & 0.979 & 0.402 & 1.37e-01 & 5.32e-01 & 6.69e-01 & 0.366 & 1.16e-01 & 5.77e-01 & 6.93e-01 1285 & 0.795 & 0.295 & 1.54e-01 & 5.90e-01 & 7.45e-01 & 0.274 & 1.34e-01 & 6.26e-01 & 7.60e-01 1428 & 0.695 & 0.238 & 1.62e-01 & 6.26e-01 & 7.88e-01 & 0.225 & 1.42e-01 & 6.56e-01 & 7.99e-01 1571 & 0.646 & 0.215 & 1.61e-01 & 6.45e-01 & 8.06e-01 & 0.204 & 1.43e-01 & 6.72e-01 & 8.15e-01 1714 & 0.627 & 0.212 & 1.56e-01 & 6.53e-01 & 8.09e-01 & 0.201 & 1.39e-01 & 6.78e-01 & 8.17e-01 1857 & 0.640 & 0.230 & 1.47e-01 & 6.48e-01 & 7.94e-01 & 0.218 & 1.30e-01 & 6.74e-01 & 8.04e-01 2000 & 0.711 & 0.289 & 1.29e-01 & 6.20e-01 & 7.49e-01 & 0.270 & 1.13e-01 & 6.51e-01 & 7.64e-01 2142 & 0.798 & 0.353 & 1.13e-01 & 5.90e-01 & 7.03e-01 & 0.325 & 9.76e-02 & 6.25e-01 & 7.23e-01 2285 & 0.728 & 0.315 & 1.15e-01 & 6.14e-01 & 7.30e-01 & 0.293 & 1.00e-01 & 6.46e-01 & 7.46e-01 2428 & 0.618 & 0.251 & 1.22e-01 & 6.56e-01 & 7.78e-01 & 0.236 & 1.08e-01 & 6.82e-01 & 7.90e-01 2571 & 0.549 & 0.211 & 1.24e-01 & 6.86e-01 & 8.10e-01 & 0.200 & 1.11e-01 & 7.07e-01 & 8.18e-01 2714 & 0.506 & 0.190 & 1.23e-01 & 7.04e-01 & 8.27e-01 & 0.181 & 1.11e-01 & 7.24e-01 & 8.35e-01 2857 & 0.476 & 0.175 & 1.22e-01 & 7.17e-01 & 8.39e-01 & 0.167 & 1.10e-01 & 7.36e-01 & 8.46e-01 3000 & 0.454 & 0.165 & 1.20e-01 & 7.28e-01 & 8.48e-01 & 0.158 & 1.09e-01 & 7.45e-01 & 8.54e-01 3776 & 0.382 & 0.141 & 1.05e-01 & 7.63e-01 & 8.68e-01 & 0.136 & 9.61e-02 & 7.77e-01 & 8.73e-01 4754 & 0.300 & 0.111 & 8.85e-02 & 8.07e-01 & 8.95e-01 & 0.107 & 8.24e-02 & 8.16e-01 & 8.99e-01 5985 & 0.227 & 0.086 & 6.94e-02 & 8.48e-01 & 9.18e-01 & 0.084 & 6.56e-02 & 8.54e-01 & 9.20e-01 7535 & 0.167 & 0.065 & 5.24e-02 & 8.85e-01 & 9.37e-01 & 0.063 & 5.00e-02 & 8.89e-01 & 9.39e-01 9487 & 0.110 & 0.041 & 3.77e-02 & 9.22e-01 & 9.60e-01 & 0.040 & 3.65e-02 & 9.24e-01 & 9.60e-01 11943 & 0.076 & 0.030 & 2.53e-02 & 9.45e-01 & 9.70e-01 & 0.029 & 2.47e-02 & 9.46e-01 & 9.71e-01 15036 & 0.052 & 0.021 & 1.72e-02 & 9.62e-01 & 9.79e-01 & 0.021 & 1.69e-02 & 9.62e-01 & 9.79e-01 18929 & 0.036 & 0.015 & 1.18e-02 & 9.73e-01 & 9.85e-01 & 0.015 & 1.16e-02 & 9.74e-01 & 9.86e-01 23830 & 0.025 & 0.011 & 7.80e-03 & 9.82e-01 & 9.89e-01 & 0.011 & 7.68e-03 & 9.82e-01 & 9.89e-01 30001 & 0.017 & 0.008 & 5.11e-03 & 9.87e-01 & 9.92e-01 & 0.008 & 5.03e-03 & 9.87e-01 & 9.92e-01 & & 1000 & 2.419 & 0.887 & 1.29e-01 & 2.83e-01 & 4.12e-01 & 0.733 & 1.04e-01 & 3.76e-01 & 4.80e-01 1142 & 1.860 & 0.671 & 1.62e-01 & 3.50e-01 & 5.11e-01 & 0.578 & 1.32e-01 & 4.29e-01 & 5.61e-01 1285 & 1.517 & 0.543 & 1.75e-01 & 4.06e-01 & 5.81e-01 & 0.481 & 1.45e-01 & 4.74e-01 & 6.18e-01 1428 & 1.292 & 0.455 & 1.83e-01 & 4.51e-01 & 6.35e-01 & 0.409 & 1.53e-01 & 5.11e-01 & 6.64e-01 1571 & 1.134 & 0.393 & 1.85e-01 & 4.89e-01 & 6.75e-01 & 0.359 & 1.57e-01 & 5.42e-01 & 6.98e-01 1714 & 1.018 & 0.345 & 1.87e-01 & 5.21e-01 & 7.08e-01 & 0.318 & 1.60e-01 & 5.68e-01 & 7.28e-01 1857 & 0.925 & 0.313 & 1.83e-01 & 5.48e-01 & 7.32e-01 & 0.290 & 1.58e-01 & 5.91e-01 & 7.49e-01 2000 & 0.845 & 0.287 & 1.77e-01 & 5.74e-01 & 7.50e-01 & 0.268 & 1.53e-01 & 6.11e-01 & 7.65e-01 2142 & 0.775 & 0.267 & 1.68e-01 & 5.97e-01 & 7.65e-01 & 0.250 & 1.47e-01 & 6.32e-01 & 7.79e-01 2285 & 0.714 & 0.249 & 1.61e-01 & 6.19e-01 & 7.80e-01 & 0.234 & 1.41e-01 & 6.50e-01 & 7.91e-01 2428 & 0.660 & 0.232 & 1.53e-01 & 6.40e-01 & 7.93e-01 & 0.219 & 1.36e-01 & 6.68e-01 & 8.03e-01 2571 & 0.613 & 0.218 & 1.46e-01 & 6.58e-01 & 8.04e-01 & 0.206 & 1.29e-01 & 6.84e-01 & 8.14e-01 2714 & 0.571 & 0.206 & 1.38e-01 & 6.76e-01 & 8.14e-01 & 0.196 & 1.23e-01 & 6.99e-01 & 8.22e-01 2857 & 0.533 & 0.190 & 1.35e-01 & 6.92e-01 & 8.27e-01 & 0.181 & 1.21e-01 & 7.13e-01 & 8.35e-01 3000 & 0.508 & 0.181 & 1.31e-01 & 7.03e-01 & 8.35e-01 & 0.173 & 1.19e-01 & 7.23e-01 & 8.41e-01 3776 & 0.403 & 0.149 & 1.09e-01 & 7.53e-01 & 8.62e-01 & 0.143 & 9.93e-02 & 7.68e-01 & 8.67e-01 4754 & 0.305 & 0.115 & 8.82e-02 & 8.04e-01 & 8.92e-01 & 0.111 & 8.19e-02 & 8.13e-01 & 8.95e-01 5985 & 0.220 & 0.084 & 6.73e-02 & 8.52e-01 & 9.20e-01 & 0.081 & 6.36e-02 & 8.58e-01 & 9.22e-01 7535 & 0.157 & 0.061 & 5.02e-02 & 8.91e-01 & 9.41e-01 & 0.059 & 4.81e-02 & 8.94e-01 & 9.42e-01 9487 & 0.108 & 0.041 & 3.64e-02 & 9.24e-01 & 9.60e-01 & 0.040 & 3.53e-02 & 9.26e-01 & 9.61e-01 11943 & 0.068 & 0.027 & 2.29e-02 & 9.51e-01 & 9.74e-01 & 0.026 & 2.24e-02 & 9.52e-01 & 9.74e-01 15036 & 0.041 & 0.017 & 1.40e-02 & 9.69e-01 & 9.83e-01 & 0.016 & 1.37e-02 & 9.70e-01 & 9.84e-01 18929 & 0.028 & 0.012 & 8.95e-03 & 9.80e-01 & 9.89e-01 & 0.012 & 8.79e-03 & 9.80e-01 & 9.89e-01 23830 & 0.019 & 0.008 & 5.78e-03 & 9.86e-01 & 9.92e-01 & 0.008 & 5.70e-03 & 9.86e-01 & 9.92e-01 30001 & 0.008 & 0.003 & 2.47e-03 & 9.94e-01 & 9.97e-01 & 0.003 & 2.43e-03 & 9.94e-01 & 9.97e-01 reddening of a spectral energy distribution compared to that expected from spectroscopic indicators is usually taken as the first sign of the presence of dust in a galactic system . color - color diagrams derived from multi - color photometry of galaxies frequently exhibit reddening _ arrows _ which purport to show the direction in which data points move in the color - color plane in response to the presence of a given column density of dust . implicit in such displays is the assumption that the reddening is proportional to the dust column . it was shown by @xcite that the relationships between reddenings and dust columns is usually quite non - linear and highly dependent upon details of the geometry and the scattering properties of the dust . here we show the additional effects due to the presence of a clumpy structure in the attenuating medium . in fig . [ fig_ebv_tauv]a we show the traditional reddening parameter @xmath20 as a function of the homogeneous radial optical depth @xmath21 ; in fig . [ fig_ebv_tauv]b the congested region near the origin of fig . [ fig_ebv_tauv]a is enlarged . several interesting points can be made . the most dramatic effect arising from the transition from a homogeneous structure to a two - phase clumpy structure occurs in the shell geometry . the homogeneous shell is similar to a spherical homogeneous screen , except that flux is scattered back into the beam in the shell geometry . as expected for a screen , the predicted reddening for the homogeneous shell geometry increases linearly with increasing dust column density . the scaling factor is less than that for a screen due to the blue color of scattered radiation included in the measurement . consequently , the ratios @xmath22 for the homogeneous shell models are 0.20 for the mw case , 0.24 for the smc case . this should be compared to the standard ratio of 0.35 for average galactic extinction . the transition to a two - phase clumpy dust structure in the shell leads to a reddening function which quickly becomes non - linear and saturates near @xmath23 for mw dust , near @xmath24 for smc dust . this initial saturation is caused by the fact that individual clumps become optically thick while the low - density interclump medium initially contributes little to the reddening . at @xmath25 , the clumpy shell geometry exhibits a gradual increase in the reddening again , now solely due to the growing attenuation by the interclump medium . note the interesting parallel to the behavior of a curve - of - growth in astronomical spectroscopy , where the increase of absorption in the core of a line is followed by saturation and subsequent further growth due to absorption in the damping wings of the line . the dusty geometry exhibits a permanent saturation at larger optical depths both in the clumpy and in the homogeneous states , as was shown already by @xcite . the transition to a clumpy structure makes the saturation occur at roughly half the amount of reddening in @xmath20 than was found in the homogeneous case . the saturation in the homogeneous case is a result of the fact that radiation emerges primarily from the outermost regions within optical depths one to two , regardless of how much dust is present further inside the system . this is fundamentally unchanged by adopting a clumpy structure , except that light now escapes preferentially through low - optical depth lines - of - sight between clumps , resulting in an overall lower amount of reddening . the level of @xmath20 at which realistic dusty structures saturate is highly significant . it has been noted with great emphasis by dickinson ( 1997 ) , sawicki and yee ( 1998 ) , and steidel et al . ( 1999 ) , among others , that the @xmath20 color excesses of lyman - break galaxies crowd around values of @xmath26 , with hardly any exhibiting color excesses greater than 0.35 . this result is fully consistent with our model results . all models except the homogeneous shell models saturate at reddening values @xmath20 less than 0.3 , a value which is breached by the clumpy shell models only for @xmath27 . this suggests that the highly idealized homogeneous shell model , which is basically the historical screen model with scattering , does not apply to these lyman - break galaxies , as was argued forcefully by @xcite already . however , the result also means that the actual dust column densities and the resulting attenuations are quite unconstrained by these color excess measurements . the fact that these galaxies were selected on the basis of their rest - frame - uv emission characteristics does argue , however , for moderate total attenuations . the saturation of color excesses occurs for higher column densities of dust when wavelengths longer than the v - band are involved . in fig . [ fig_evk_ebv ] , this is illustrated in the case of a @xmath28 versus @xmath20 color - color plot . here , models with increasing optical depths are plotted as series of connected symbols , starting with @xmath29 at the origin . the separation of symbols from one another along a given reddening line is a measure of how each color excess scales with optical depth . not only is the reddening line not a simple _ arrow _ ( with the exception of the homogeneous shell model ) , the detailed shape of the reddening line depends on the dust / star geometry and the structure of the dusty medium . of particular interest again is the clumpy shell geometry , which shows the initial saturation in @xmath20 when clumps become optically thick in the v - band , followed by a saturation in @xmath28 at much higher radial dust column density when the same clumps reach the optically thick state in the k - band . finally , after clumps are sufficiently thick to block light in all bands , the interclump medium begins to dominate the reddening process . where and when this occurs depends on the ratio of densities in the clump and the interclump medium . in a real physical system where the differences between different phases are rather more gradual than the fixed ratio assumed in our models , the transition shown in fig . 5 between the reddening caused by the clumps and that caused by the interclump medium will also occur in a less abrupt manner . in fig . [ fig_ext_mw ] , we present the predicted attenuation optical depths ( see equation 1 ) as a function of inverse wavelength for mw dust for three cases , the optically thin case for @xmath30 ( top ) , the moderately optically thick case @xmath31 ( middle ) , and the very optically thick case @xmath32 ( bottom ) . also shown in each panel is the average galactic extinction for the same dust column densities . while the left - hand panels show the absolute values of the respective optical depths , the corresponding right - hand panels present the same data normalized at v , in order to facilitate the comparisons of the changes in the wavelength dependence of each function . the dusty , shell , and cloudy models are contrasted in both their homogeneous and their clumpy forms . fig . [ fig_ext_smc ] presents the analogous set of data for smc dust . insert boxes are included to show the behavior of those curves which go out of bounds vertically on the main graphs . + + + + several significant conclusions can be drawn from the illustrated cases . in all instances , attenuation optical depths for the three geometries are less than the interstellar extinction produced by the same column density of dust . this difference is due to the return of scattered radiation in all cases , exclusively so in the case of the shell geometry , and also due to lower attenuation in cases of stars only lightly embedded in the dust distribution , which is happening in the cloudy and dusty geometries . the descending order from shell , to dusty , to cloudy geometry can thus be understood . similarly , clumpy structures containing the same amounts of dust as corresponding homogeneous structures always provide less attenuation than the homogeneous ones at all wavelengths . more importantly , and this difference is particularly prominent in the shell geometry , the clumpy structures provide a substantially grayer wavelength dependence for the attenuation than would have been expected from adopting the original extinction curve . the trend toward gray attenuation increases strongly with increasing column density , because an increasing fraction of the total attenuation is caused by clumps which are optically thick shortward of the v - band . for this same reason , the relative strength of the 2175 uv extinction feature , prominent in the mw dust , is greatly reduced in clumpy structures . this is in contrast to the case of the homogeneous shell model , which exhibits a 2175 feature even stronger than that in the interstellar extinction ( fig . [ fig_ext_mw ] , rhs ) . this is due to the fact that the albedo of mw dust displays a minimum coinciding with this feature , leading to return less scattered light here than at wavelengths immediately longward or shortward of the feature . similarly , the declining dust albedo shortward of 1500 causes the predicted attenuation curve for the homogeneous shell model curves to be steeper in the far uv than the interstellar extinction curve . it is important to recognize that the slope of the attenuation functions in the uv for the more realistic embedded and clumpy geometries is dependent on the total dust column density . the use of a single attenuation function , for example the widely used `` calzetti attenuation law '' ( @xcite ) , for the dereddening and flux correction of individual starburst galaxies is therefore not justified in the absence of information about the dust column density . a comparison of the `` calzetti attenuation law '' with predictions from our models will be discussed in more detail in section 4 . + + the integrated light of dusty galaxies consists of a certain fraction of scattered light , with a spectrum similar to that of the integrated star light . this fraction depends on wavelength , geometry , total dust column density , and dust scattering properties , as well as viewing direction in the case of disk galaxies . in the latter , the relative fraction of scattered light is largest in directions perpendicular to the disk , this being the direction in which the escape probability is highest for both stellar and particularly for scattered radiation . given the spherical geometry of our models , we are able to predict the direction - averaged scattered light components . these are shown in fig . [ fig_scat_flux ] as fractions of the attenuated stellar flux for three characteristic visual radial optical depths ( 0.5 , 1.5 , 4.5 ) , and for three geometries with both mw and smc dust , both clumpy and homogeneous . as expected from observations of reflection nebulae with embedded stars ( @xcite , 1993 ) , the scattered light fraction for homogeneous shell models easily exceeds unity and rises strongly with increasing optical depth , more as a result of the increasing attenuation of the starlight , less as a result of increasing scattered light flux . we view the embedded dusty , the clumpy shell models , and the cloudy models to be representative of actual geometries of galaxies . for these , the scattered light fraction appears to be fairly constant across the uv / visible spectrum and is likely between a quarter and one half of the total integrated light . in the case of the mw dust , the depression near @xmath33 is due to the albedo minimum associated with the 2175 extinction bump , and this feature is actually seen in the spectrum of the diffuse ( i.e.scattered ) galactic light ( @xcite ) . the transition from homogeneous to clumpy structures is associated with a substantial reduction of the fractional scattered light flux . this is due to two causes : direct starlight escapes more easily from the system , mainly through gaps in the clump distribution , while the scattered light flux is diminished , because clumps are mainly illuminated externally with the scattered light being directed preferentially into the clump interior where final absorption is more likely . in clumpy media , clumps in effect become the scattering centers instead of individual dust grains . as shown in @xcite , the effective albedo of clumps is always lower than the input dust albedo . overall , the fractional amounts of scattered light for mw dust and for smc dust are about the same for identical dust column densities . we have emphasized that the models involving clumpy dust distributions presented in the paper are based upon one particular set of clumpiness parameters ( @xmath34 , @xmath35 , @xmath3 ) . this set was found by @xcite to be closest to that representing the structured ism in the galaxy . detailed results on how the radiative transfer is affected when the clumpiness parameters are varied can be found in @xcite as well . here we summarize the principal effects . the density ratio @xmath1 and the filling factor @xmath36 together control the contributions to the total opacity made by the interclump medium and the clumps . our case of @xmath36 = 0.15 and @xmath1 = 0.01 is intermediate to the case of @xmath1 near unity , where the interclump medium controls the system opacity , with results very similar to the homogeneous case ( @xmath37 ) and the case of @xmath38 , where the opacity is controlled by the blocking effect of the clumps and is then proportional to the filling factor . as the filling factor increases , the appearance of the system will approach that of the homogeneous case . the grayness of the attenuation ( section 3.3 ) increases as @xmath1 decreases for the same amount of dust . also , the ratio of scattered light to direct , partially attenuated stellar light ( fig . 8) decreases as the ratio @xmath1 decreases , both as a result of the growing probability of stellar light escaping without any interaction and as a consequence of the lower scattering ability of a medium dominated by increasingly optically thick clumps . finally , for a given amount of dust , the overall opacity of the medium will increase as the number of clumps ( n ) increases . in this section we discuss which of our models with respect to geometry , optical depth , and dust type might best be used for the analysis of observations of different kinds of galaxies . in particular , we concentrate on dust attenuation studies of nearby starburst galaxies and lyman break galaxies . + + starburst galaxies exhibit a wide range of dustiness as evidenced by large variations in the uv spectral index @xmath39 , ratios of recombination line strengths , and ratios of far - ir to uv luminosities . the uv spectral index @xmath39 was defined by calzetti et al . ( 1994 ) through the relation @xmath40 in the range @xmath41 . by comparing the spectral energy distributions ( sed ) of more highly reddened starbursts with those of barely reddened starbursts , calzetti et al . ( 1994 ) derived the observed wavelength dependence of the dust attenuation in these systems . the attenuation function takes into account the return of scattered light into the integrated light of a galaxy and also reflects radiative transfer effects arising in the three - dimensional distributions of stars , gas , and dust in these galaxies . this observed attenuation function , further refined in calzetti ( 1997 ) , is referred to in the literature as the `` calzetti attenuation law '' or `` calzetti law '' . this attenuation due to internal dust must be clearly distinguished from the wavelength dependence of interstellar extinction , which is the attenuation due to absorption _ and _ scattering by a screen of dust located between a distant source and an observer . the principal characteristic of the `` calzetti law '' is the absence of a 2175 feature . gordon et al . ( 1997 ) showed that the `` calzetti law '' can be reproduced , if the radiative transfer in starburst galaxies is occurring in a medium with intrinsic dust properties closely resembling that of smc dust . also , the large range of observed reddenings , as measured by variations in the spectral index @xmath39 , appears to require efficient attenuation geometries such as provided by our shell or dusty geometries . in fig . [ fig_ext_cal ] we compare attenuation values derived from starburst galaxies by calzetti ( 1997 ) with our model attenuation functions . in order to facilitate this comparison , models and observations are renormalized to @xmath42 . in order to transform the values in table 5 of calzetti ( 1997 ) , we have used the result of calzetti ( 1997 ) that the stellar continuum suffers 40% the reddening that the gas suffers , i.e. @xmath43 . the data point capable of discriminating between different models most clearly is the one near @xmath44 , derived from the far - uv slope in the reddened seds of starbursts . no model with mw dust matches this data point . in addition , the lack of a 2175 feature in the observations compared to the prediction of a 2175 feature in the models with mw dust argues strongly against the presence of mw dust in starburst galaxies . this will be discussed further in section [ sec_bump ] . among smc dust models , the same point does discriminate between dust column densities and clearly favors the clumpy shell model for @xmath45 . as is apparent from fig . [ fig_ext_smc ] , such a model predicts an attenuation optical depth of about @xmath46 at @xmath44 , and , according to fig . [ fig_ebv_tauv ] , a color excess @xmath47 . an attenuation optical depth of 2.0 implies that the flux at @xmath44 has been reduced by a factor 7.4 on average in this uv - selected sample of starburst galaxies . we need to re - emphasize that there is no universal attenuation function and that the attenuation factor of 7.4 corresponds simply to the average of the sample of uv / optically - selected starbursts studied by calzetti et al . our average attenuation factor of 7.4 compares well with the luminosity - weighted mean absorption factor of 5.4 derived by meurer et al . ( 1999 ) for the same sample . we discuss the uv flux reduction factor and the uncertainties associated with its determination further in section [ sec_uv_atten ] . it came as an initial surprise during the iras mission that far - ir ultraluminous and by inference , very dust - rich , galaxies in most instances exhibited optical colors which were barely distinguishable from those of normal galaxies of similar population type . @xcite showed that this was due to the fact that , given embedded dust geometries , the optical colors were determined by the least attenuated stars , while the far - ir flux was due to the absorption of light from stars which contribute little or nothing to the optical flux . the present calculations were extended to dust column densities corresponding to @xmath48 instead of 15 in the case of @xcite . in fig . [ fig_ratio ] , we show the resulting ratios of the far - ir flux , integrated over all frequencies at which dust emission occurs , to the integrated uv / optical / near - ir stellar flux . we reproduce the most extreme observed ratios near 100 with the clumpy shell model of @xmath48 with a maximum reddening near 0.8 in @xmath20 and only slightly lower ratios with the homogeneous dusty models with a reddening less than 0.3 in @xmath20 . only the homogeneous shell , as expected for a simple screen , produces very large amounts of reddening of @xmath49 and @xmath50 for @xmath48 . the fact that most ir luminous galaxies appear to be relatively blue seems to indicate that clumpy dust structures prevail in most of them . this is also confirmed by the correlation between the ratio @xmath51 and the spectral index @xmath52 for different geometries and dust types , shown in fig . [ fig_uvratio ] . a major aim of studies of high - z starbursts ( or lyman - break galaxies ) is to determine the epoch of peak star formation in the history of the universe , if such a peak indeed exists . the outcome of such an investigation depends critically upon the corrections applied for dust attenuation in the rest - frame uv continuum spectrum of the observed galaxies , given that these systems contain substantial amounts of internal dust ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . in addition , the upcoming nasa galex mission is to produce about 100,000 uv galaxy spectra in order to probe the global history of star formation , and the observed uv spectral slope is be be related uniquely to the attenuation factor by which the uv flux has been reduced by dust internal to the galaxies . our figures [ fig_ext_mw ] and [ fig_ext_smc ] already indicated some of the problems to be expected in such an approach . the attenuation functions are geometry dependent and they become increasingly gray as the dust column density increases . this means that a heavily attenuated starburst may exhibit a steep uv slope in its sed , resembling that of a barely reddened galaxy , but show a greatly reduced uv flux , corresponding to a system with little star forming activity . the chances for underestimating the star formation rate in early epochs of the universe are therefore severe , if dust is abundant generally in the early galaxies . fortunately , our models provide an excellent set of tools with which to evaluate different approaches to the problem of determining the far - uv attenuation factor . in fig . [ fig_a1600]a we illustrate the computed attenuation factor at 1600 as a function of @xmath20 found for different model geometries and dust types . as anticipated above , the observation that most observed values of @xmath20 for lyman - break galaxies fall into the narrow range of 0.2 to 0.3 barely constrains the attenuation factor to the range 4 to 200 . this is not very useful as a constraint . in fig . [ fig_a1600]b we investigate the use of the spectral index @xmath39 as a tool to evaluate the attenuation factor . our radiative transfer models do not assume a specific energy distribution for the source spectrum . hence , we predict only the change in the spectral index @xmath39 which occurs as a result of the wavelength dependent attenuation . the galex program advocates this particular approach by suggesting that the change in the uv slope of the galaxies sed can be converted into information about the applicable attenuation factor . meurer et al . ( 1999 ) have taken the additional step of calibrating the correlation between the spectral index @xmath39 and the uv - attenuation , derived for nearby uv - selected starburst galaxies . in observed starburst spectra , this change in spectral index ranges from zero to 3 ( meurer et al.1997 ) . our fig . [ fig_a1600]b reveals that correlations between the spectral index change and the uv attenuation factor are extremely model dependent and even more dependent on the type of dust present in a given system . the calibration proposed by meurer et al . ( 1999 ) is shown in fig . [ fig_a1600]b as a solid black line . for this to be valid universally , one must assume that smc - type dust is present in all starburst galaxies and that shell - type geometries with various degrees of clumpiness are the prevailing star / dust geometries . even when restricting considerations to smc - type dust , the range of observed values of @xmath52 is consistent with attenuation factors in the range 0150 . as is evident from fig . [ fig_a1600]b , mw - type dust is not expected to produce positive values of @xmath52 . this is contrary to observations in starbursts , where such positive changes in the uv spectral index are seen ( meurer et al . however , mw - type dust will reveal its presence by the appearance of a 2175 feature . this has been seen in the heavily obscured starburst nucleus of m33 ( @xcite ) , which indicates that smc - type dust is not a universal characteristic of all starburst environments . however , the preference for smc - type dust in starburst galaxies is affirmed , when we compare values of @xmath53 flux ratios and corresponding values of @xmath52 from observations ( meurer et al . 1997 ) with our models in fig . [ fig_uvratio ] . for this calculation we assumed that the unattenuated source sed was given by a 1 or a 1000 myr constant star formation sed of fioc & rocca - volmerange ( 1999 ) . the observed data points clearly follow the trend set by the model for clumpy shell configurations with smc dust . this agreement would become even more convincing , were we to present models with different ratios of clump to interclump density , different clump size , and different filling factors , which is , however , beyond the scope of this present discussion . the demonstrated model dependence of the ratio of fir to f160bw flux suggests that the observed scatter in the data points is most likely real , as was concluded by meurer et al . ( 1999 ) also . the spread in the fir to f160bw ratio for a given value of @xmath52 corresponds to a factor of 10 uncertainty in the corresponding uv attenuation factor . we conclude , therfore , that the uv spectral slope is not a reliable indicator of the dust attenuation in a given galaxy . it is usable only , if a large sample of starburst galaxies with relatively uniform characteristics is being considered at once . in fig . [ fig_ratio_1600]a , we plot the ratio of dust - emitted ir flux to the integrated uv nir flux against the attenuation factor at 1600 . the model calculations assume two starbursts of different ages , one a 1 million year old constant star formation sed , the other a 1 billion year old constant star formation sed . we note that even this ratio does not yield a unique relation to the uv attenuation factor . in particular , clumpiness in a given dust distribution will reduce the attenuation at 1600 for a given value of the ratio of dust - emitted ir flux to the integrated uv nir flux , because clumpiness is associated with a much grayer attenuation function . this is particularly evident in shell - type geometries . variations in the age of the starburst and the associated changes in the uv nir sed affect the ratio as well , leading to a significant model dependence of the predicted attenuation factor at 1600 . by contrast , if we plot the ratio of integrated fir flux to the measured flux at 1600 against the attenuation factor at 1600 , as shown in fig . [ fig_ratio_1600]b , the model and age dependence practically vanish . we conclude , therefore , that the uv attenuation factor can be evaluated reliably for individual galaxies only , if a complete spectral survey of the fir emission of the galaxy is carried out and a flux measurement in the far - uv is obtained . gordon et al . ( 1997 ) showed that the absence of the 2175 extinction bump in the attenuation function of starbursts can not be explained by radiative transfer effects but requires the actual absence of the absorber giving rise to this feature . this is despite the fact that some weakening of the 2175 feature was seen in the attenuation curves ( fig . [ fig_ext_mw ] ) for embedded geometries involving mw dust and further weakening occurred when clumpiness was included . as was found by gordon et al . ( 1997 ) , the 2175 bump can be suppressed using radiative transfer effects , but at the expense of a gray extinction throughout the uv . thus , the existence of a substantial slope in the calzetti ( 1997 ) attenuation curve ( fig . [ fig_ext_cal ] ) implies that the dust in starburst galaxies must lack the absorbers which produce the 2175 bump . the dust in starbursts most closely resembles that found in the smc ( @xcite ) . recent modeling of this dust by zubko ( 1999 ) shows that such dust lacks the component normally introduced to produce the 2175 feature ( graphite ) , and that the overall size distribution of smc grains is shifted toward smaller sizes compared to mw dust . we find the color index ( @xmath54 ) to be particularly sensitive for distinguishing between the presence of smc dust and mw dust , one without the 2175 feature and the other with . in fig . [ fig_f170wf218w_bv ] , we show a color - color diagram of starburst data , which are compared to the colors of unreddened starburst models ( open symbols ) and to reddening lines produced by our models . these reddening lines are attached to different unreddened models to show them distinctly ; they may be shifted horizontally to fit the data . several facts emerge clearly : all starburst data appear on or below the line of unreddened burst models , i.e. they are reddened in the ( @xmath54 ) index . only the reddening lines for smc dust fall below the unreddened bursts , while all mw models fall above the unreddened bursts , where no data points are found . both the amount of reddening in the ( @xmath54 ) index and the distribution of data points in the color plane can best be fitted with a clumpy shell model with smc dust . we take this ( as did gordon et al . 1997 ) to be one of the strongest indications that nearby starbursts contain smc type dust and that the prevailing geometry is that of a clumpy shell model . this sensitive test might be applied to the rest - frame uv seds of high - redshift galaxies in general in order to test for the presence of the type dust ( mw or smc ) . we have presented new multiple - scattering radiative transfer calculations for three representative types of galactic environments and two dust types , as found within the local group in the form of the average mw and smc type dust . both homogeneous and two - phase clumpy dust distributions have been explored over a large range of dust column densities , measured by @xmath55 . the full numerical details of the models are made available to the community via the authors , the authors www site , and eventually various national / international data centers . the environments examined include dusty , a fully mixed spherical distribution of dust and embedded stars ; shell , a dust - free star filled region , surrounded by a dust shell ; and cloudy , a mixed dust - star region surrounded by stars in a dust - free region on the outside . the models cover the wavelength range 1000 to 30000 with exceptionally close coverage of the rest - frame uv . this aspect makes these models particularly applicable to uv - selected starburst galaxies and lyman - break galaxies . we examined the models with respect to predicted reddening effects , attenuation functions , and the contribution of scattered light to the total flux as a function of wavelength . all model types except the homogeneous shell model exhibit reddening saturation effects . in particular , models that are likely representatives for starburst galaxies show values of @xmath20 not in excess of 0.3 , despite a large range of dust column density . reddening `` vectors '' in color - color diagrams are complex non - linear functions depending on the prevailing dust type and structure , as well as on the geometry . attenuation functions , which measure the wavelength - dependent reduction in the total stellar and scattered flux from a dusty galaxy , exhibit a trend toward increasing grayness with increasing dust column density . this trend is stronger in clumpy dust distributions than in homogeneous ones . no justification was found for the use of a universal attenuation function for the analysis of a large sample of galaxies . in particular , we found that the widely - employed `` calzetti law '' is most closely reproduced by our clumpy shell model with smc - type dust and a dust column density corresponding to @xmath31 . such a case corresponds to an attenuation optical depth in the uv of @xmath56 or a uv attenuation correction factor of 7.4 . the calzetti law is valid for statistical samples which have similar dust / star geometries as found in uv - selected starbursts and which have dust column densities within narrow limits . under no circumstances should it be used for the attenuation correction for single galaxies . when investigating various approaches to determining this attenuation factor in general , we found that the use of the color excess @xmath20 and the uv spectral index @xmath39 lead to results that depend strongly on dust type , dust distribution , and source - dust geometry . only the measurement of the @xmath57 flux ratio promises reasonable certainty for the determination of the uv attenuation correction factor in individual galaxies . this reaffirms the need for observations of starburst galaxies and lyman - break galaxies over the entire range of rest - wavelengths , particularly the fir , in addition to the uv . we further showed that observations can clearly discriminate between the presence of mw - type dust and smc - type dust through the use of a @xmath58 vs. @xmath59 color - color diagram . when applied to existing data on uv - selected nearby starburst galaxies , a consistent prevalence of smc dust is found . we gratefully acknowledge support from nasa grants nag 5 - 3367 and nagw-3168 to the university of toledo . constructive conversations with daniela calzetti , geoffrey clayton , and karl misselt are also gratefully acknowledged . berlind , a. a. , quillen , a. c. , pogge , r. w. , & sellgren , k. 1997 , , 114 , 107 bianchi , l. , clayton , g. c. , bohlin , r. c. , hutchings , j. b. , & massey , p. 1996 , , 471 , 203 boiss , p. 1990 , , 228 , 483 bruzual , a. g. , magris , g. , & calvet , n. 1988 , , 333 , 673 byun , y. i. freeman , k. c. , & kylafis , n. d. 1994 , 432 , 114 calzetti , d. 1997 , , 113,162 calzetti , d. , bohlin , r. c. , gordon , k. d. , witt , a. n. , bianchi , l. 1995 , , 446 , l97 calzetti , d. , kinney , a. l. , & storchi - bergmann , t. 1994 , , 429 , 582 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 [ ccm ] corradi , r. l. m. , beckman , j. e. , & simonneau , e. 1996 , , 282 , 1005 di bartolomeo , a. , barbaro , g. , & perinotto , m. 1996 , in new light on galaxy evolution , iau symp . 171 , ed . r. bender and r. l. davies , kluwer acad . 363 dickinson , m. 1997 , in the hubble deep field , stsci symp . , eds . m. livio , s. m. fall , & p. madau , ( in press ) ferrara , a. , bianchi , s. , cimatti , a. , & giovanardi , c. 1999 , , in press ( astro - ph/9903078 ) fioc , m. & rocca - volmerange , b. , 1999 , in preparation fitzpatrick , e. l. 1999 , , 111,63 gordon , k. d. , calzetti , d. , & witt , a. n. 1997 , , 487 , 625 gordon , k. d. & clayton , g. c. 1998 , , 500 , 816 gordon , k. d. , witt , a. n. , carruthers , g. r. , christensen , s. a. , & dohne , b. c. 1994 , , 432 , 641 gordon , k. d. , hanson , m. m. , clayton , g. c. , rieke , g. h. , & misselt , k. a. 1999a , , 519 , 165 gordon , k. d. , witt , a. n. , & clayton , g. c. 1999b , in preparation heckman , t. m. 1998 , in the birth of galaxies : xth recontres de blois , ed . b. guiderdoni , in press henyey , l.g . & greenstein , j.l . 1941 , , 93 , 70 hobson , m. p. & scheuer , p. a. g. 1993 , , 264 , 145 holmberg , e. 1958 , medd . ser . 2 , no . 6 holmberg , e. 1975 , stars and stellar systems ix , univ . of chicago press , 123 hurwitz , m. 1994 , apj , 433 , 149 kim , s .- h . , martin , p. g. , & hendry , p. d. 1994 , , 422 , 164 kuchinski , l. e. , terndrup , d. m. , gordon , k. d. , & witt , a. n. 1998 , , 115 , 1438 larson , r. b. & tinsley , b. m. 1978 , , 219 , 46 lehtinen , k. & mattila , k. 1996 , , 309 , 570 lillie , c. f. , & witt a. n. 1976 , , 208 , 64 meurer , g. r. , heckman , t. m. , lehnert , m. d. , leitherer , c. , & lowenthal , j. 1997 , , 114 , 54 meurer , g. r. , heckman , t. m. , & calzetti , d. 1999 , , in press ( astro - ph/9903054 ) misselt , k. a. , clayton , g. c. , & gordon , k. d. 1999 , , 515 , 128 morgan , d. h. , nandy , k. , & thompson g. i. 1976 , , 177 , 531 natta , a. & panagia , n. 1984 , , 287 , 228 sawicki , m. & yee , h. k. c. 1998 , , 115 , 1329 sawicki , m. j. , lin , h. , & yee , h. k. c. 1997 , , 113 , 1 smith , d. a. , herter , t. , & haynes , m. p. 1998 , , 494 , 150 steidel , c. c. et al . 1999 , , in press ( astro - ph/9811399 ) takagi , t. , arimoto , n. , & vanseviius , v. 1999 , , in press ( astro - ph/9902219 ) toller , g. n. 1981 , phd thesis , state univ . of new york at stony brook vrosi , f. , & dwek , e. 1999 , , submitted witt , a. n. 1977 , , 35 , 1 witt , a. n. , friedmann , b. c. , sasseen , t. p. 1997 , , 481 , 809 witt , a. n. & gordon , k. d. 1996 , , 463 , 681 [ wg96 ] witt , a. n. & lillie , c. f. 1973 , , 25 , 397 witt , a. n. , lindell , r. s. , block , d. l. , & evans , r. 1994 , apj , 427 , 227 witt , a. n. , oliveri , m. v. , & schild , r. e. 1990 , , 99 , 888 witt , a. n. , petersohn , j. k. , bohlin , r. c. , oconnell , r. w. , roberts , m. s. , smith , a. m. , & stecher , t. p. 1992a , , 395 , l5 witt , a. n. , petersohn , j. k. , holberg , j. b. , murthy , j. , dring , a. , & henry , r. c. 1993 , , 410 , 714 witt , a. n. , thronson , h. a. , & capuano , j. m. 1992b , , 393 , 611 [ wtc ] witt , a. n. , walker , g. a. h. , bohlin , r. c. , & stecher , t. p. 1982 , , 261 , 492 wolf , s. , fischer , o. , & pfau , w. 1998 , , 340 , 103 zubko , v. g. 1999 , , in press |
we analyze a candidate theory for the strange metal near optimal hole - doping in the cuprate superconductors . the theory contains a quantum phase transition between metals with large and small fermi surfaces of spinless fermions carrying the electromagnetic charge of the electron , but the transition does not directly involve any broken global symmetries . the two metals have emergent su(2 ) and u(1 ) gauge fields respectively , and the transition is driven by the condensation of a real higgs field , carrying a finite lattice momentum and an adjoint su(2 ) gauge charge . this higgs field measures the local antiferromagnetic correlations in a ` rotating reference frame ' . we propose a global phase diagram around this higgs transition , and describe its relationship to a variety of recent experiments on the cuprate superconductors . | 1412.1086 | several recent experiments @xcite have provided strong evidence for a dramatic change in the nature of the low temperature electronic state of the hole - doped cuprate superconductors near optimal doping ( @xmath0 ) . moreover , zero field photoemission experiments carried out in the normal state have seen evidence for a ` large ' fermi - surface for @xmath1 , consistent with the overall luttinger count @xcite , and disconnected fermi ` arcs ' near the nodal regions for @xmath2 @xcite . at high fields , quantum oscillations also reveal a ` large ' fermi - surface for @xmath1 @xcite , but a closed electron - like fermi - surface with an area that constitutes a small fraction of the entire brillouin - zone for @xmath2 @xcite . it is therefore quite natural to associate the transition with decreasing @xmath3 at @xmath0 with the loss of a ` large ' fermi - surface and the simultaneous opening of a pseudogap . there has also been significant experimental progress @xcite in understanding the structure of the density - wave ordering at lower doping , which is likely responsible for the reconstructed electron - like fermi - surface seen in quantum oscillation experiments @xcite . in this paper we will use these advances to motivate and develop a previously proposed model @xcite for the physics of the strange metal near optimal doping . we argue that the rich phenomenology observed in the underdoped cuprates is primarily driven by a transition between non - fermi liquid metals with large and small fermi surfaces which does not directly involve any broken global symmetry . all states with broken symmetry observed at low temperatures and low doping are not part of the critical field theory @xcite , but are derived as low energy instabilities of the parent small fermi surface phase . this diminished role for broken symmetries is consistent with absence of any observed order with a significant correlation length at higher temperatures . we will also construct a global phase diagram to describe the many phases and crossovers around the strange metal . a quantum phase transition which does not involve broken symmetries is necessarily associated with a _ topological _ change in the character of the ground state wavefunction . emergent gauge fields are a powerful method of describing this topological structure , and they remain applicable also to the gapless metallic phases of interest to us here . given the fundamental connection between emergent gauge fields and the size of the fermi surface , which was established in ref . using oshikawa s method @xcite , we are naturally led to a quantum phase transition in which there is a change in the structure of the deconfined gauge excitations . indeed , this describes a higgs transition in a metal , such as that discussed in ref . . this argument is a general motivation for higgs criticality near optimal doping in the cuprates , which applies beyond the specific model considered here . we emphasize that we are using the traditional particle - physics terminology in which a `` higgs transition '' describes the breaking of a local gauge invariance . we are not referring to the longitudinal mode of a broken global symmetry , which has also been labeled `` higgs '' in condensed matter contexts @xcite . the primary new motivation for the model of ref . arises from our recent work @xcite analyzing the @xmath4-form factor density waves observed in scanning tunnelling microscopy @xcite and x - ray experiments @xcite . in this work @xcite , we argued that such density waves arise most naturally as an instability of a metallic higher temperature pseudogap state with small fermi surfaces described as a @xcite ` fractionalized fermi liquid ' ( fl * ) ; other works with related ideas on the pseudogap are refs . . specifically , we used a theory of the fl * involving a background u(1 ) spin liquid with bosonic spinons @xcite : it is therefore convenient to dub this metallic state for the pseudogap as a u(1)-fl*. these results are also easily extended to a @xmath5 spin liquid , and we will consider this case in appendix [ app : spiral ] . the presence of a small fermi surface without symmetry breaking requires topological order and emergent gauge fields @xcite , and so also a higgs transition to the large fermi surfaces at larger doping : here we provide a natural embedding of a fl * theory into such a transition , and we expect similar approaches are possible for other possible topological orders in the underdoped regime . we now consider the evolution of the u(1)-fl * , and its small electronic fermi surfaces , to the conventional ` large ' fermi surface fermi liquid state at large doping . there is an existing conventional theory of the transformation from small to large fermi surfaces driven by the disappearance of antiferromagnetic order . this is a transition between two fermi liquids , and the vicinity of the transition is described by the hertz - millis theory @xcite and its field - theoretic extensions @xcite , as shown in fig . [ fig : phase1 ] . . note that the u(1)-fl * has a ` small ' fermi surface of electrons due to the presence of topological order , while phase a above has a ` small ' fermi surface of electrons because of translational symmetry breaking.,width=432 ] here , we describe a detour from this direct route @xcite in which two new non - fermi liquid phases appear between the conventional phases of hertz - millis theory . the detour is described by a su(2 ) gauge theory , and the transition from small to large fermi surfaces is now a higgs transition without any local order parameter , in which the emergent gauge structure describing the topological order in the ground state changes from u(1 ) to su(2 ) . the higgs field of this transition is a measure of the local antiferromagnetic correlations in a rotating reference frame to be introduced below in eq . ( [ r ] ) . note that the higgs transition in fig . [ fig : phase1 ] is between metallic states which we denote as ` algebraic charge liquids ' ( acl ) . the small and large fermi surfaces in the acls are those of spinless fermions which carry the electromagnetic charge of the electron . for the u(1 ) acl , a bound state forms between the spinless fermions and a spin @xmath6 boson @xcite , leading to small fermi surfaces of fermionic quasiparticles carrying the same quantum numbers as the electron in the u(1)-fl * : so photoemission will detect a small fermi surface of electrons in the u(1)-fl*. we anticipate that similar effects are also present in the su(2 ) acl metal : there is a large density of states of thermally excited @xmath6 bosons at low energy , so that the photoemission spectral function reflects the large fermi surface of the spinless fermions . we also note that although the higgs field plays a central role in our phase diagram , its direct experimental detection will be difficult . it is overdamped via its coupling to the fermi surfaces , and gauge invariance prohibits any experimental probe from coupling linearly to it . nevertheless , we will see below that it has significant experimental consequences via its strong effect on the fermionic spectrum . we will present details of this theory starting from a microscopic model in section [ su2 ] , but first , in section [ sec : prelim ] , we shall describe some key aspects using our proposed phase diagram in fig . [ phasecup ] . let us begin with a simplified picture of the optimal doping strange metal with a large fermi surface . we consider a model of electrons @xmath7 on the sites @xmath8 of a square lattice , with @xmath9 a su(2 ) spin index . we transform the electrons to a rotating reference frame using a su(2 ) rotation @xmath10 and ( spinless-)fermions @xmath11 with @xmath12 , ( c c_i + c_i ) = r_i ( c _ i,+ + _ i,- ) , [ r ] where @xmath13 . note that this representation immediately introduces a su(2 ) gauge invariance ( distinct from the global su(2 ) spin rotation ) ( c _ i,- ) u_i ( c _ i,+ + _ i,- ) , r_i r_i u_i^ , [ gauge ] under which the original electronic operators remain invariant , @xmath14 ; here @xmath15 is a su(2 ) gauge - transformation acting on the @xmath16 index . so the @xmath17 fermions are su(2 ) gauge fundamentals , they carry the physical electromagnetic global u(1 ) charge , but they do not carry the su(2 ) spin of the electron . the density of the @xmath17 is the same as that of the electrons . such a rotating reference frame perspective was used in the early work by shraiman and siggia on lightly - doped antiferromagnets @xcite , and the importance of its gauge structure was clarified in ref . . the strange metal is obtained by forming a large fermi surface state of the @xmath17 fermions , while @xmath10 fluctuate isotropically over all su(2 ) rotations with a moderate correlation length . this description suggests a simple trial wavefunction for this strange metal . begin with a large fermi surface ( lfs ) state of free @xmath17 fermions : _ inside lfs , p= _ p^ ( ) | 0 . expand this out in position space , insert the inverse of eq . ( [ r ] ) to write the wavefunction in terms of @xmath18 and the physical electrons @xmath19 , and finally average over @xmath18 , to obtain _ i dr_i w [ \{r_j } ] _ inside lfs , p= | 0 , where @xmath20 is a variational weight - function of the @xmath10 , invariant under global spin rotations . for @xmath21 , we have a zero correlation length for @xmath10 , and we obtain a wavefunction for the @xmath19 involving only empty and doubly - occupied sites . with non - trivial @xmath20 , the correlation length of @xmath18 increases , we also build in spin singlet pairs of @xmath19 electrons on nearby sites . comparing to the gutzwiller - projected trial states commonly used for the underdoped cuprates @xcite , this wavefunction includes the possibility of doubly - occupied sites and assigns different complex weights to the off - site singlet pairs . for a more precise and complete description of the strange metal , which accounts for the gauge structure in eq . ( [ r ] ) , we must turn to a quantum effective action for the @xmath17 which necessarily includes an emergent su(2 ) gauge field . in the terminology of ref . , such a theory of spinless , gapless fermions coupled to an emergent gauge field is an ` algebraic charge liquid ' ( acl ) , and hence we have labeled the strange metal as su(2 ) acl in fig . [ phasecup ] . this name implies that the su(2 ) gauge symmetry is unbroken ( _ i.e. _ not ` higgsed ' ) , and in such a situation the @xmath17 fermions have a large fermi surface with a shape similar to that of the electron fermi surface in fermi liquid state at large doping . now let us consider the transition to the u(1 ) acl in fig . [ phasecup ] . this is described by the condensation of a real higgs field @xmath22 , where @xmath23 indicates that the higgs field transforms as a su(2 ) adjoint . as we will see below in eq . ( [ hn ] ) , this higgs field is a measure of the local antiferromagnetic order in the rotating reference frame defined by @xmath18 ( see also ref . for an illuminating analogy ) . the condensation of the higgs field breaks the gauge symmetry from su(2 ) to u(1 ) and reconstructs the @xmath17 fermi surface from large to small . it is this higgs transition which describes the optimal doping qcp in fig . [ phasecup ] , and analyzing its structure is the main purpose of the present paper . in the case where @xmath22 is complex , the higgs phase can break the gauge symmetry down to @xmath5 , and we consider this case in appendix [ app : spiral ] . the shraiman - siggia analyses @xcite of doped antiferromagnets were effectively within such a higgs - condensed regime , and this obscured the gauge structure of their formulation @xcite . let us also note from fig . [ phasecup ] that the u(1 ) acl is the parent of the u(1)-fl*. this was discussed in refs . , and will be reviewed below : the u(1)-fl * arises by the formation of bound states between the spinons , @xmath18 , and @xmath24 fermions around the small fermi surface . we expect that a similar phenomenon also happens at low @xmath25 in the su(2 ) acl at lower temperatures , so that the photoemission largely reflects the structure of the large fermi surface of the su(2 ) acl . the phase diagram in fig . [ phasecup ] is meant to be schematic ; determining the exact nature of the various crossover and phase - transition lines is beyond the scope of this work . the higgs transition is present only at @xmath26 , and there is only a crossover at @xmath27 ( shown as the dashed green line in fig . [ phasecup ] ) . also , we have assumed that the energy scales associated with the acl / fl * and @xmath28bdw vanish with the same power law as a function of @xmath29 at the qcp . theoretically speaking , other possibilities @xcite are also allowed . we now specify the imaginary time lagrangian of the optimal doping qcp in fig . [ phasecup ] , and its vicinity . for now , the lagrangian will not include the @xmath18 bosons : we assume that @xmath18 fluctuations are short - ranged , but the associated spin - gap in the su(2 ) acl phase of fig . [ phasecup ] is small because of proximity to the multi - critical point m in fig . [ fig : phase1 ] ; we will include the @xmath18 contributions in section [ su2 ] . then we have , _ = _ + _ h + _ y . [ lqcp ] the first term describes a large fermi surface of @xmath24 fermions minimally coupled to a su(2 ) gauge field @xmath30 : _= _ i _ i , p^_i , p + _ i , jt_ij^_i , p_pp_j , p , where @xmath31 are the fermion hopping parameters , @xmath32 are the spatial co - ordinates of the sites , @xmath33 is the chemical potential , and @xmath34 are pauli matrices acting on the su(2 ) gauge indices . the higgs lagrangian is denoted @xmath35 , and it has a form familiar from its particle - physics incarnations , _h = ( _ h^a-2i_abca^b_h^c)^2 + ( h^a-2i_abc^b h^c)^2 + ( h^a)^2 + [ ( h^a)^2]^2 . [ lh ] the higgs potential is determined by the parameters @xmath36 and @xmath37 , and transition across the qcp is controlled by the variation in @xmath36 . as usual , for negative @xmath36 , the higgs field condenses , and this breaks the gauge symmetry from su(2 ) to u(1 ) ; and for positive @xmath36 , the higgs field is gapped , and then the su(2 ) gauge symmetry remains unbroken . finally , we have the yukawa coupling in @xmath38 . as in particle - physics , this is a trilinear coupling between the higgs field and the fermions , but now it has a different spatial structure : _ y = - h_i^a e^i _ i _ i , p^^a_pp _ i , p , [ yukawa ] where @xmath39 is the antiferromagnetic wavevector . this spatial structure indicates that @xmath22 transforms non - trivially under lattice translations : h^a h^a e^i ; [ higgstrans ] note that this is permitted because @xmath40 is real for all spacings @xmath41 . the transformation in eq . ( [ higgstrans ] ) arises from the role of the higgs field as a measure of the antiferromagnetic correlations in a rotating reference frame . in the presence of the higgs condensate , this yukawa coupling reconstructs the @xmath24 fermi surface from large to small , and the @xmath42 factor is crucial in the structure of this reconstruction . while in the particle physics context the higgs condensate gives the fermions a mass gap , here the fermions acquire a gap only on certain portions of the large fermi surface , and a small fermi surface of gapless fermions remains . we note that the effective gauge theory will also acquire a yang - mills term for the su(2 ) gauge field @xmath43 when high energy degrees of freedom are integrated out . as is well known in theories of emergent gauge fields , such a term helps stabilize deconfined phases of the type considered here . we do not write this term out explicitly here , but will include its contributions in section [ sec : landaudamping ] , and specifically in the @xmath44 term in eq . ( [ eq : lsu2 ] ) . the body of our paper will describe a field theoretic analysis of the non - fermi liquid properties of @xmath45 . this combines recent progress in the theories of fermi surfaces coupled to order parameters @xcite and gauge fields @xcite . here we mention one notable result on the electrical resistivity in the quantum - critical region of the higgs transition . as in recent work @xcite on other quantum critical points of metals , we consider the situation in which there is a strong momentum bottleneck _ i.e. _ there is rapid exchange of momentum between the fermionic and bosonic degrees of freedom , and the resistivity is determined by the rate of loss of momentum . in particular , it is possible for the resistivity to be dominated by the scattering of neutral bosonic degrees of freedom , rather than that of charged fermionic excitations near the fermi surface . in our model , we argue that an important source for momentum decay is the coupling of the higgs field to disorder _ = v ( ) ^2 , [ ldis ] where @xmath46 is quenched gaussian random variable with v ( ) = 0 ; v ( ) v ( ^)=v_0 ^ 2 ^ 2(-^ ) , where the double angular brackets indicate an average over quenched disorder . comparing with eq . ( [ lh ] ) , we see that @xmath46 can be viewed as a random local variation in the value of @xmath36 , the tuning parameter which determines the position of the qcp . we will show that the analysis of the contribution of @xmath47 to the resistivity closely parallels the computation in ref . for the spin - density - wave quantum critical point . and as in ref . , we find a resistivity for weak disorder which is proportional to @xmath48 , ( t)~v_0 ^ 2 t^2(+1-z)/z , [ dct ] where @xmath49 is the scaling dimension of the @xmath50 operator , @xmath51 is the correlation length exponent and @xmath52 is the dynamical exponent . as we will see in section [ sec : higgscritical ] , this predicts a linear - in-@xmath25 resistivity for the leading order values of the exponents . the outline for the rest of our paper is as follows . in section [ su2 ] , we arrive at the above gauge - theoretic description starting from the theory of a metal with fluctuating antiferromagnetism and discuss the mean - field phase diagram as a function of the relevant tuning parameters . in section [ qcp ] , we describe the properties of the qcp using a low - energy description of the fermi - surface coupled to a gauge - field and the critical fluctuations of the higgs field . finally in section [ conc ] , we discuss the relation of our proposed phase - diagram to the actual phase - diagram in the hole - doped cuprates . appendix [ app : spiral ] contains the extension to spiral order and @xmath5 gauge theory , while technical details are in appendix [ diags ] . we summarize the derivation in ref . of the su(2 ) gauge theory , starting from a model of electrons on the square lattice coupled to the fluctuations of collinear antiferromagnetism at the wavevector @xmath53 . the case of collinear antiferromagnetism at other wavevectors was also considered in ref . , and we treat spiral antiferromagnets at incommensurate wavevectors in appendix [ app : spiral ] . we begin with a model of electrons coupled to the quantum fluctuations of antiferromagnetism represented by the unit vector @xmath54 , with @xmath55 and @xmath56 . the lagrangian is given by &=&_f + _n + _fn , + _f&=&_i c_i^c_j , + _n&= & , + _fn&=&- _ i e^i_i n_i c_i^^ _ c_i . [ model ] in the above @xmath57 measures the strength of quantum fluctuations associated with the orientation of @xmath58 , @xmath59 is an @xmath60 spin - fermion coupling and @xmath61 is a characteristic spin - wave velocity . now we insert the parametrization in eq . ( [ r ] ) into eq . ( [ model ] ) and proceed to derive an effective theory for @xmath62 and @xmath18 . the formulation of the latter theory is aided by the introduction of a su(2 ) gauge connection @xmath30 . as is familiar in many discussions of emergent gauge fields in correlated electron systems , this gauge field arises after decoupling hopping terms via an auxiliary field ; here we skip these intermediate steps , and simply write down appropriate hopping terms for the @xmath62 and @xmath18 which are made gauge - invariant by suitable insertions of the gauge connection . with the parameterization in eq . ( [ r ] ) we notice that the coupling @xmath63 in eq . ( [ model ] ) maps precisely onto the yukawa coupling in eq . ( [ yukawa ] ) with h_i^a ^a_pp = n_i r^_i p ^ _ r_i p , and so we _ define _ the higgs field @xmath64 by h_i^a n_i [ ^r_i^a r_i^ ] . [ hn ] this identifies @xmath22 as the antiferromagnetic order in the rotating reference frame defined by eq . ( [ r ] ) . an important property of this definition is that the field @xmath22 is _ invariant _ under a global su(2 ) spin rotation @xmath65 , which rotates the direction of the physical electron spin and of the antiferromagnetic order , ( c c_i + c_i ) v ( c c_i + c_i ) , r_i v r_i . [ spin ] note that the su(2 ) spin rotation is a left multiplication of @xmath18 above , while the su(2 ) gauge transformation in eq . ( [ gauge ] ) is a right multiplication of @xmath18 . with these properties , eq . ( [ hn ] ) implies that @xmath22 transforms as a vector under the su(2 ) gauge transformation in eq . ( [ gauge ] ) . we have now assembled all the steps taken after substituting eq . ( [ r ] ) into eq . ( [ model ] ) . the lagrangian of the resulting gauge theory is then obtained as _ + _r , [ su2l ] where @xmath66 was described below eq . ( [ lqcp ] ) in section [ sec : ft ] , and @xmath67 is the lagrangian for @xmath18 . the structure of the latter is determined by the transformations of @xmath18 in eqs . ( [ gauge ] ) and eq . ( [ spin ] ) . so we have _r = . this completes our derivation of the su(2 ) gauge theory . it is useful here to collect the transformations of the fields under the su(2 ) gauge transformation , the global su(2 ) spin rotation , and electromagnetic u(1 ) charge , as summarized in table [ tab ] . .summary of the transformations of the fields under the various gauge - transformations . su(2 ) representations of spin @xmath68 are labelled by their dimension of @xmath69 . the u(1 ) column contains the charge under the u(1 ) gauge field . [ cols="^,^,^,^",options="header " , ] finally , we can make contact with other approaches by expressing @xmath18 as r_i =( cc z_i & -z_i^ * + z_i & z_i^ * ) , [ rz ] with @xmath70 , but this parameterization will not be useful to us . consider the situation in the higgs phase , where the field @xmath22 is condensed . then we are free to choose a gauge in which the higgs condensate is @xmath71 . in such a condensate , after inverting the relation in eq . ( [ hn ] ) we find n_i & = & h_i^a [ ^r_i^a r_i^ ] + & = & z_i ^^ _ z_i . [ e1 ] the last relationship is the familiar connection between the o(3 ) and cp@xmath72 variables , but note that it holds here only within the phase where the higgs field is condensed _ i.e. _ in the u(1 ) acl . we now describe the phases of @xmath73 obtained in a simple mean field theory @xcite in which we allow condensates of the bosonic field @xmath18 and @xmath22 . these phases are obtained by varying the tuning parameters @xmath36 and @xmath57 , and were shown in fig . [ fig : phase1 ] ; in fig . [ fig : phase2 ] , we label the phases by their condensates . the phases are : * the higgs phase , labelled as ( a ) in figs . [ fig : phase1],[fig : phase2 ] , where both su(2)@xmath74 and su(2)@xmath75 are broken , leading to @xmath76 . the gauge - excitations , @xmath77 , are gapped here . this phase describes the _ afm - metal _ where the large fermi - surface gets reconstructed into hole ( and electron ) pockets due to condensation of @xmath78 , the n@xmath79el order parameter . * the su(2 ) confining phase , labelled as ( b ) in figs . [ fig : phase1],[fig : phase2 ] . note that the su(2)@xmath74 here remains unbroken . we have @xmath80 , which is necessary to preserve spin - rotation invariance since @xmath81 from eq . ( [ hn ] ) . this is the usual _ fermi liquid _ phase , with a large fermi - surface . * the higgs phase , labelled as ( c ) in figs . [ fig : phase1],[fig : phase2 ] , where the su(2)@xmath75 is broken , but the su(2)@xmath74 remains unbroken , leading to @xmath82 . by recalling the physical interpretation of the fields , this amounts to a locally well developed amplitude of the afm , without any long - range orientational order . we can choose @xmath83 by carrying out a gauge - transformation , which immediately implies that a u(1 ) subgroup of the su(2)@xmath75 remains unbroken , so that the @xmath84 photon remains gapless . thus this phase describes a _ u(1 ) algebraic charge liquid _ , or , the _ holon - metal _ @xcite . however , due to the locally well developed afm order , the fermi - surface is reconstructed into @xmath17 holon pockets that are minimally coupled to a u(1 ) gauge - field . + as a function of temperature , there could be a continuous crossover from a u(1 ) acl to a u(1 ) fl@xmath85 ( or a holon - hole " metal ) , where some of the holons ( @xmath86 ) start forming bound states with the gapped spinons ( @xmath87 ) @xcite . * the final phase ( d ) in figs . [ fig : phase1],[fig : phase2 ] has the full symmetry , with none of the fields condensed : @xmath88 . instead of the above u(1 ) acl , where only @xmath84 was gapless , in this phase there are a triplet of gapless su(2 ) photons coupled to a large fermi - surface . this phase can be described as a _ su(2 ) algebraic charge liquid_. formally , this phase a spin gap , but we assume that @xmath25 is greater than the gap in the metallic regions of fig . [ phasecup ] because of proximity to the point m in fig . [ fig : phase1 ] . at low enough @xmath25 , this phase is unstable to superconductivity @xcite . we should emphasize that the above mean - field analysis has been rudimentary ; _ e.g. _ we can not rule out the possibility that higher order couplings could induce first - order transitions , that could even eliminate an intermediate phase . the next section shall present the theory for the interplay between the fluctuations of the gauge and higgs fields , within a low - energy field - theoretic formulation . we are interested in studying the properties of the qcp between the su(2 ) acl and the u(1 ) acl . at the qcp , @xmath89 , the entire fermi - surface is coupled to the transverse fluctuations of a su(2 ) gauge field . there have been studies in the particle physics literature of fermi surfaces coupled to non - abelian gauge fields @xcite ; however these have been restricted to spatial dimension @xmath90 , where a rpa analysis gives almost the complete answer . in spatial dimension @xmath91 of interest to us here , we shall follow the approach taken for abelian gauge theories @xcite which uses a patch decomposition of the fermi surface . the same approach transfers easily to the non - abelian case ; indeed because of the landau damping of the gauge bosons , there is little difference between the abelian and non - abelian cases @xcite , as will also be clear from our analysis in section [ sec : landaudamping ] . apart from their coupling to a su(2 ) gauge field , the fermionic @xmath17 particles are also coupled to a quantum critical higgs field . this coupling is strongest at 8 ` hot spots ' around the fermi surface , and in section [ sec : higgscritical ] we shall be able to use the methods developed from the case of a spin - density - wave transition of fermi liquids @xcite some of the details of the computations appear in appendix [ diags ] . here we describe the low energy theory of the su(2 ) acl , away from the higgs condensation at the qcp . we need only consider a su(2 ) gauge field coupled to the large fermi surface of the @xmath17 fermions . as in the u(1 ) case @xcite , we can make a patch decomposition of the fermi surface , and treat antipodal pairs of patches separately . for a single pair of antipodal patches , we have the fermions @xmath92 ( see fig . [ fs ] ) , with @xmath93 the patch index , and @xmath94 the usual su(2 ) gauge index . this is coupled to the transverse components of the su(2 ) gauge field , @xmath95 . l&=&l_f+l_+l _ , + l_f&=&_+p^(_-i_x-_y^2)_+p + _ -p^(_+i_x-_y^2)_-p , + l_&= & ( _ y a^a_x)^2 , + l_&= & a^a_x ( _ + p^ ^a_pp _ + p - ^_-p ^a_pp _ -p ) [ eq : lsu2 ] let us review the one - loop renormalization of the gauge and fermionic matter fields . we start by looking at the self - energy of the gauge - field due to the particle - hole bubble ( fig . [ gse]a ) . we have , ^_0(q)=2 _ s g_s^0 ( ) g_s^0(+q ) , [ gfs ] where @xmath96 and the bare fermionic propagator is given by , g_s^0()=. the final result is of the form , the uv cutoff , drops out . the factor that appears in general is of the form @xmath97 , where @xmath4 is the number of space - dimensions . ] , ^_0(q)=c_b , c_b=. [ pigf ] the computations are summarized in appendix [ gsea ] . computing the fermionic self - energy due to the bosonic - propagator dressed with the rpa level polarization bubble ( fig . [ gse]b ) leads to , _ s , pp(k)&=&-_p^a _ p^a d ( ) g_s^0(k- ) , + & = & -3 _ pp d ( ) g_s^0(k- ) , where @xmath98 is the gauge - field propagator , d^-1()=(c_b + _ y^2 ) . we then obtain , _ s(k)&=&- , + _ s(k)&=&-ic_f ( k_)|k_|^2/3 , c_f=2 ( ) ^2/3.[sig1 ] this self - energy contribution is larger than the bare @xmath99 term at low energies . therefore , upon including the rpa contribution into the fermionic propagator , we have , g_s()= , [ fg ] which is the well known result for the quasiparticles being damped all along the fermi - surface . now we consider the qcp at which the higgs boson condensed from the non - fermi liquid su(2 ) acl state described in the previous subsection . across this higgs transition from the su(2 ) acl to the u(1 ) acl , the fermi - surface gets reconstructed this is controlled by the real higgs field , @xmath22 , which carries lattice momentum , @xmath100 . by the same arguments used for the onset of spin - density - wave order in a fermi liquid @xcite , the low energy physics of the qcp is dominated by the vicinity of the hot - spots : these are points on the fermi surface which are connected by @xmath101 ( see fig . [ fs ] ) . the computation for the present non - fermi liquid fermi surface proceeds just as for the fermi liquid case , by linearizing the bare dispersion for the fermions around the hot spots : l&=&l_hs+l_h+l_fh , + l_hs&=&_1p^m(_- i^m_1)_1p^m + _ 2p^m(_- i^m_2)_2p^m , + l_fh&=&h^a(_1p^m^a_pp_2p^m + _ 2p^m^a_pp_1p^m ) , where @xmath102 already appeared in eq . ( [ lh ] ) ; @xmath103 is the hot - spot pair index ( fig . [ fs ] ) . let us first look at the one - loop self energy of the @xmath22 field ( fig . [ hse]a ) . this is given by , ^h(q)=2_m , [ pih ] where we now use the non - fermi liquid fermion green s function renormalized by the gauge field fluctuations , as discussed in section [ sec : landaudamping ] : g_^m()=. [ gf1 ] note the @xmath104 scaling of the fermion self energy , which allows us to drop the bare frequency dependence from above ( @xmath105 ) . + upon including contributions from all pairs of hot - spots , we obtain ( see appendix [ apph ] ) , ^h(q)=^h(q=0)+|q_| , = , [ pih1 ] where @xmath106 is the number of pairs of hot spots . note that the @xmath107 dependence has completely dropped out and the above result is precisely the expression that we would have obtained if we had started with the bare fermion green s functions ( or , any anomalous power @xmath108 ) . this result is not surprising it just reproduces the landau - damped " form of the propagator for @xmath22 . as we know , the only requirement for the appearance of landau - damping is the existence of particle - hole excitations around the fermi - surface in the limit of @xmath109 . in the general case , this always leads to @xmath110 for a bosonic order - parameter coupled to a fermion - bilinear . when the order parameter itself carries a finite momentum @xmath101 , as is the case here , then the denominator in the damping term gets cut off and leads to @xmath111 . equipped with the above expression , let us now compute the self - energy of the fermions in the vicinity of the hot - spots ( fig . [ hse]b ) , _ 1,pp(p)&=&_p^a_p^a g_2(p- ) ( ) , + & = & 3_pp g_2(p- ) ( ) , [ fhse ] where the propagator tuned to the critical point ( @xmath89 ) is given by : @xmath112 . we are interested in the singular power law frequency dependence of the self - energy at the hot - spots . for future use , it is useful to express the green s function in eq . ( [ gf1 ] ) in the more general form g_^m()= , [ gf2 ] where the exponent @xmath113 from the coupling to the su(2 ) gauge field . upon evaluating the momentum integrals , the self - energy ( for @xmath114 ) becomes ( see appendix [ fse ] ) , _ 1(p_)=3i ^-1 ( ) , [ sef ] which correctly reproduces @xmath115 . furthermore , note that if @xmath116 , i.e. if the anomalous self - energy contribution were to be absent , then the above reduces to the well known form @xcite _ 1(p _ ) = 3i = - ( p _ ) , [ z2 ] reproducing the @xmath117 result . let us now proceed to evaluate the expression in the presence of a finite @xmath118 . rescaling @xmath119 leads to , & & _ 1(p _ ) = ( p _ ) |p_|^1- + & & ^-1 ( ) , [ scself ] where the dimensionless parameter , @xmath120 . an asymptotic analysis of eq . ( [ scself ] ) shows that _ 1 ( |p_| 0 ) ~-i ( p _ ) , so the low energy singularity of the self - energy is _ independent _ of the non - fermi liquid exponent @xmath121 in the fermion green s function , and has the same value as in the spin density wave case without the gauge field . this is the key observation of the present subsection . to estimate the co - efficient , we can use a self - consistent approach in which we use a self energy in eq . ( [ gf2 ] ) with @xmath122 . this self - energy arises from the coupling to the higgs field , and is always dominant over the one obtained from the gauge field with @xmath123 . assembling all the constraints , the final expression takes the following @xmath117 form _ 1(p_)= ( c ) ( p _ ) , where the function @xmath124 is defined in appendix [ ints ] and @xmath125 . so we reach our main conclusion that , in both the fermionic and bosonic sectors , the low energy physics of the higgs qcp is essentially identical to that of the spin - density - wave onset transition in a fermi liquid . and the basic reason for this is simple . the hotspot theory has dynamic critical exponent @xmath117 , while the singularities arising from the su(2 ) gauge field coupling around the fermi surface have @xmath104 . at a given length scale , the contributions with the larger @xmath52 dominate because they have a lower energy . hence the higgs criticality of a non - fermi liquid maps onto the spin density wave criticality of a fermi liquid . with this conclusion in hand , we can now directly apply the results of ref . on the dc resistivity to the higgs qcp . the approach of ref . requires that there is quasiparticle breakdown around the entire fermi surface , and the fermionic excitations rapidly equilibriate with all the bosonic modes . while this was only marginally true for the spin - density - wave quantum critical point considered in ref . , it is easily satisfied for the higgs qcp being considered here : the su(2 ) gauge field makes the entire fermi surface `` hot '' , while the higgs field fluctuations induce additional fermion damping at the hot spots on the fermi surface . as in the previous case , it is possible for disorder to couple to the square of the higgs field because such an operator is gauge - invariant , as we noted in eq . ( [ ldis ] ) . and the corresponding contribution to the resistivity is in eq . ( [ dct ] ) . for the exponents @xmath91 , @xmath117 , and @xmath126 presented above , this yields a linear - in - temperature results @xmath127 . the primary goal of this paper has been to propose a candidate theory for the quantum phase transition near optimal doping in the cuprates . we analyzed the qcp between metals with ` large ' and ` small ' fermi - surfaces , which did not involve any broken global symmetries , but instead involved a higgs transition between metals with emergent su(2 ) and u(1 ) gauge fields . the higgs field acts as a measure of the local antiferromagnetic order in the rotating reference frame defined by eq . ( [ r ] ) . as we discussed in sections [ sec : intro ] and [ sec : prelim ] , the symmetry broken phases observed in the underdoped cuprates arise as low temperature instabilities of the ` small ' fermi - surface metal . the underlying qcp we studied was between two metals ( the u(1 ) acl and the su(2 ) acl ) in which the fermi surface excitations are coupled to emergent gauge fields , and so there are no landau quasiparticles . however , electron - like quasiparticles do re - emerge around a small fermi in the u(1)-fl * , and we will discuss similar features around the large fermi surface in the su(2 ) acl below . the reconstruction to the ` small ' fermi - surface in the acl phases is driven by the condensation of the higgs field , and the higgs critical point has additional singular structure in the vicinity of the hot - spots " . the higgs criticality has associated with it an interplay of both @xmath104 physics on the whole fermi - surface , and @xmath117 physics in the vicinity of the hot - spots . we showed that near the higgs qcp the @xmath117 physics dominates , and hence many critical properties map onto the previously studied problem of the onset of spin density wave order in a fermi liquid @xcite . let us now conclude with a discussion of the relationship of our proposed phase diagram in fig . [ phasecup ] to the experimentally obtained phase diagram in the non - la - based cuprates . the @xmath28sc and @xmath28bdw both arise as instabilities of the u(1 ) fl * , as has been discussed in refs . ( the su(2 ) acl is also unstable to superconductivity @xcite ) . the high temperature pseudogap phase for @xmath128 is a u(1 ) acl or more appropriately a _ holon - hole _ metal @xcite . there is a crossover to the u(1 ) fl * phase at @xmath129 , where all the holons have formed bound - states with the spinons . an important feature of the u(1 ) fl * phase is that its transport and photoemission signatures are mostly identical to those of a fermi liquid . the primary difference from a landau fermi liquid is that the volume enclosed by the fermi surface is proportional to the density of holes , @xmath3 , and not to the luttinger density @xmath130 . the u(1 ) fl * phase also has an emergent u(1 ) gauge field , as required by the topological arguments in ref . , but the fermi surface quasiparticles are gauge neutral . the recent remarkable observation of fermi liquid transport properties in the pseudogap phase of hg1201 @xcite below @xmath131 , with some possibly non - fermi liquid behavior between @xmath131 and @xmath132 , can therefore be viewed as strong support for the existence of a fl * derived out of a parent acl . in particular , the alternative ` fluctuating order ' picture of the pseudogap does not naturally lead to such temperature dependent crossovers from non - fermi liquid to fermi liquid regimes . for the la - based cuprates , there is a larger doping regime with magnetic order , overlapping with the regime of charge order . this can be accommodated in our phase diagram @xcite by moving the full red arrow in fig . [ fig : phase1 ] just to the other side of the point m , and allowing for incommensurate order as in appendix [ app : spiral ] . an important challenge for future experiments is to detect direct experimental signatures of the complete small fermi surface of the proposed fl * phase . we presume that it is the small quasiparticle residue on the ` back side ' of the small fermi surface @xcite which is responsible for the arc - like features in the photoemission spectrum @xcite . therefore , we need a probe which does not involve adding or removing an electron from the sample , and so is not sensitive to the quasiparticle residue . possibilities are friedel oscillations , the kohn anomaly , or ultrasonic attenuation . within our proposed phase diagram in fig . [ phasecup ] , the strange metal phase is to be viewed as a su(2 ) acl at the higgs critical point , and proximate to the multicritical point m to ensure the spin gap is smaller than @xmath25 . the dc transport properties of this phase are controlled by the coupling of the gauge - invariant square of the higgs field to _ long _ wavelength disorder , following an analysis of ref . for the spin density wave critical point . such a coupling leads to a linear - in - temperature resistivity . also , as emphasized in ref . , the residual resistivity is proportional to _ short _ wavelength disorder which can scatter fermions across the fermi surface . so there is no direct relationship between the residual resistivity and the slope of the linear resistivity . it would be interesting in future work to explore the role of intrinsic umklapp scattering events in the transport properties of such strange metals in the strong coupling regime . the electron spectral function in the su(2 ) acl is a convolution of the spectra of the @xmath24 fermions and the @xmath18 bosons . as in the computation in ref . , we assume the @xmath18 spectrum is thermally overdamped ( because of the proximity to m ) , and the electron spectral function primarily reflects the @xmath24 spectrum ; we also expect precursors of the bound state formation between the @xmath24 and the @xmath18 to enhance the @xmath24 features in the electron spectrum @xcite , just as in the u(1)-fl*. then the electron spectral functions should have an anisotropic structure around the fermi surface , with the weaker gauge field - induced damping in the nodal region , and the stronger higgs field - induced damping in the anti - nodal region . also note that while the higgs field coupling does show up in the resistivity as discussed above , the gauge fields coupling has a weaker effect on transport . this is because gauge - invariance prevents a non - derivative coupling between the gauge field and perturbations that violate momentum conversation . an important open question is whether this rich theoretical structure can be made consistent with the complex experimental features of the conductivity and magnetotransport in the strange metal @xcite . our linear-@xmath25 resistivity is proportional to disorder , as in the previous model in ref . . however , because the disorder couples to the higgs field , the relevant disorder is long - wavelength . this is in contrast to short wavelength disorder , which can lead to efficient large momentum scattering of fermions around the fermi surface . modifying the coefficient of the resistivity therefore requires modifying long - wavelength disorder , and this may be difficult to do because of the intrinsic disorder from the dopant ions . inducing short - wavelength disorder , by including _ e.g. _ zn impurities , may not be effective in modifying the co - efficient of the linear-@xmath25 resistivity . these features can act as tests of our proposed mechanism for the resistivity of the strange metal @xcite . finally , we note from figs . [ phasecup ] and [ fig : phase1],[fig : phase2 ] that the su(2 ) acl survives for an extended region beyond the higgs qcp . this implies strange metal behavior over a finite range of doping as @xmath133 , and not only at a single qcp . transport measurements @xcite in magnetic fields which have suppressed superconductivity appear to be consistent with such a non - fermi liquid phase . we thank seamus davis and martin greven for useful discussions . is supported by the harvard - gsas merit fellowship . this research was supported by the nsf under grant dmr-1360789 , the templeton foundation , and muri grant w911nf-14 - 1 - 0003 from aro . research at perimeter institute is supported by the government of canada through industry canada and by the province of ontario through the ministry of research and innovation . here we generalize the theory in eq . ( [ model ] ) to the case of spiral spin order at an incommensurate wavevector @xmath101 . in this case the antiferromagnetic order is not characterized by a single unit vector @xmath58 , but by two orthogonal unit vectors @xmath134 and @xmath135 which obey n_1 ^ 2 = n_2 ^2 = 1 , n_1 n_2 = 0 . [ triad ] the spin - fermion coupling to the electrons in eq . ( [ model ] ) is replaced by _fn = - _ i c_i^^ _ c_i , after the change of variables in eq . ( [ r ] ) , this leads to the yukawa coupling _y = - ( h_i^a e^i _ i + h_i^a e^-i _ i ) _ i , p^^a_pp _ i , p , [ yukawac ] where , in contrast to eq . ( [ yukawa ] ) , the higgs field @xmath22 is now _ complex _ and is defined by h^a ( n_1 + i n_2 ) [ ^r_i^a r_i^ ] , [ hnc ] generalizing eq . ( [ hn ] ) it is now also clear from eq . ( [ yukawac ] ) that under translation by a distance @xmath41 , the higgs field transforms as in eq . ( [ higgstrans ] ) , where @xmath136 can now be complex . the structure of su(2 ) gauge theory with a complex higgs field remains essentially the same as for the real higgs discussed in the body of the paper , with one important distinction . the quartic term in eq . ( [ lh ] ) is replaced by two terms u_1 [ h^a h^a]^2 + u_2 [ h^a]^2 [ h^b]^2 , [ u12 ] and the presence of spiral order requires that @xmath137 . then in the higgs phase , the minimum energy condensate can always be oriented so that h^a = ( 1,i , 0 ) . such a higgs condensate breaks the su(2 ) gauge symmetry all the way down to @xmath5 . and using eq . ( [ rz ] ) , the analog of the relationship in eq . ( [ e1 ] ) for the orientation of the spiral order is n_1 + i n_2 & = & h^a [ ^r^a r^ ] + & = & - _ z_^ _ z _ . [ e2 ] this co - incides with the conventional representation @xcite of the spiral orientation in terms of spinons @xmath138 , and it can verified that the values in eq . ( [ e2 ] ) obey eq . ( [ triad ] ) . for the case with @xmath139 in eq . ( [ u12 ] ) , the higgs condensate is instead a su(2 ) rotation of h^a = e^i ( 1,0 , 0 ) , where @xmath140 is an arbitrary phase . this corresponds to incommensurate collinear spin order @xcite . since we are interested in the singular structure of @xmath141 in eq . ( [ gfs ] ) , we shall evaluate the integral over @xmath142 first , followed by @xmath143 . therefore , ^_0(q)&= & 2 + - , + & = & + - , + & = & . this leads to the expression for @xmath141 in eq . ( [ pigf ] ) . it is not hard to see that the only non - zero contribution comes from the imaginary parts of both the terms . then , ^h_m=1(q)= . upon carrying out the @xmath147 integrals , this becomes , ^h_m=1(q)=-[(_+q _ ) ( _ ) + q - q ] . this directly leads to the expression for @xmath148 in eq . ( [ pih1 ] ) . the fermionic self - energy due to the higgs field fluctuations ( eq . ( [ fhse ] ) ) becomes , _ let us now change coordinates such that @xmath149 and @xmath150 is the component along the fermi - surface of @xmath151 . then , _ 1(p)=-3 . it is straightforward to carry out the integral over @xmath150 , which gives , _ let us now study the form of the self - energy at the hot - spot , @xmath114 , and extract the @xmath152 dependence . we can symmetrize the above form then to give , _ 1(p_)=3i _ carrying out the integral over @xmath153 ( see appendix [ ints ] ) leads to the expression in eq . ( [ sef ] ) . m. hashimoto _ et al . _ , direct spectroscopic evidence for phase competition between the pseudogap and superconductivity in bi@xmath160sr@xmath160cacu@xmath160o@xmath161 , http://arxiv.org/abs/1405.5199[arxiv:1405.5199 ] . et al . _ , fermi surface and quasiparticle excitations of overdoped tl@xmath160ba@xmath160cuo@xmath162 , phys . * 95 * , 077001 ( 2005 ) ; http://arxiv.org/abs/cond-mat/0503117[arxiv:cond-mat/0503117 ] . d. c. peets , j. d. f. mottershead , b. wu , i. s. elfimov , ruixing liang , w. n. hardy , d. a. bonn , m. raudsepp , n. j. c. ingle and a. damascelli , tl@xmath160ba@xmath160cuo@xmath162 brings spectroscopic probes deep into the overdoped regime of the high - t@xmath163 cuprates , new j. of phys . * 9 * , 28 ( 2007 ) ; http://arxiv.org/abs/cond-mat/0609250[arxiv:cond-mat/0609250 ] . b. vignolle , a. carrington , r. a. cooper , m. m. j. french , a. p. mackenzie , c. jaudet , d. vignolles , cyril proust and n. e. hussey , quantum oscillations in an overdoped high - t@xmath163 superconductor , nature * 455 * , 952 ( 2008 ) . n. doiron - leyraud , c. proust , d. leboeuf , j. levallois , jean - baptiste bonnemaison , ruixing liang , d. a. bonn , w. n. hardy and louis taillefer , quantum oscillations and the fermi surface in an underdoped high - t@xmath163 superconductor , nature * 447 * , 565 ( 2007 ) ; http://arxiv.org/abs/0801.1281[arxiv:0801.1281 ] . t. wu , h. mayaffre , s. kramer , m. horvatic , c. berthier , w. n. hardy , ruixing liang , d. a. bonn and m .- h . julien , magnetic - field - induced charge - stripe order in the high - temperature superconductor yba@xmath160cu@xmath164o@xmath165 , nature * 477 * , 191 - 194 ( 2011 ) ; http://arxiv.org/abs/1109.2011[arxiv:1109.2011 ] . et al . _ , distinct charge orders in the planes and chains of ortho - iii - ordered yba@xmath160cu@xmath164o@xmath162 superconductors identified by resonant elastic x - ray scattering , phys . rev . * 109 * , 167001 ( 2012 ) ; http://arxiv.org/abs/1207.3667[arxiv:1207.3667 ] . et al . _ , direct observation of competition between superconductivity and charge density wave order in yba@xmath160cu@xmath164o@xmath167 , nat . * 8 * , 871 - 876 ( 2012 ) ; http://arxiv.org/abs/1206.4333[arxiv:1206.4333 ] . k. fujita , m. h. hamidian _ et al . _ , direct phase - sensitive identification of a @xmath4-form factor density wave in underdoped cuprates , proc . sci . * 111 * , e3026-e3032 ( 2014 ) ; http://arxiv.org/abs/1404.0362[arxiv:1404.0362 ] . s. e. sebastian , n. harrison , and g. g. lonzarich , towards resolution of the fermi surface in underdoped high - t@xmath163 superconductors , rep . . phys . * 75 * , 102501 ( 2012 ) ; http://arxiv.org/abs/1112.1373[arxiv:1112.1373 ] . a. allais , d. chowdhury and s. sachdev , connecting high - field quantum oscillations to zero - field electron spectral functions in the underdoped cuprates , nat . * 5 * , 5771 ( 2014 ) ; http://arxiv.org/abs/1406.0503[arxiv:1406.0503 ] . l. balents , l. bartosch , a. burkov , s. sachdev , and k. sengupta , putting competing orders in their place near the mott transition , phys . rev b * 71 * , 144508 ( 2005 ) ; http://arxiv.org/abs/cond-mat/0408329[arxiv:cond-mat/0408329 ] . t. senthil , m. vojta , and s. sachdev , weak magnetism and non - fermi liquids near heavy - fermion critical points , phys . b * 69 * , 035111 ( 2004 ) ; http://arxiv.org/abs/cond-mat/0305193[arxiv:cond-mat/0305193 ] . m. ferrero , p. s. cornaglia , l. de leo , o. parcollet , g. kotliar , a. georges , valence bond dynamical mean - field theory of doped mott insulators with nodal / antinodal differentiation , europhys . lett . * 85 * , 57009 ( 2009 ) ; http://arxiv.org/abs/0806.4383[arxiv:0806.4383 ] . jia - wei mei , s. kawasaki , guo - qing zheng , zheng - yu weng , and x .- wen , luttinger - volume violating fermi liquid in the pseudogap phase of the cuprate superconductors , phys . b * 85 * , 134519 ( 2012 ) ; http://arxiv.org/abs/1109.0406[arxiv:1109.0406 ] . long zhang and jia - wei mei , charge order instability in doped resonating valence bond state and magnetic orbits from reconstructed fermi surface in underdoped cuprates : a phenomenological synthesis , http://arxiv.org/abs/1408.6592[arxiv:1408.6592 ] . r. k. kaul , a. kolezhuk , m. levin , s. sachdev , and t. senthil , hole dynamics in an antiferromagnet across a deconfined quantum critical point , phys . b * 75 * , 235122 ( 2007 ) ; http://arxiv.org/abs/cond-mat/0702119[arxiv:cond-mat/0702119 ] . m. punk and s. sachdev , fermi surface reconstruction in hole - doped @xmath171-@xmath172 models without long - range antiferromagnetic order , phys . b * 85 * , 195123 ( 2012 ) ; http://arxiv.org/abs/1202.4023[arxiv:1202.4023 ] . ar . abanov and a. v. chubukov , anomalous scaling at the quantum critical point in itinerant antiferromagnets , phys . * 93 * , 255702 ( 2004 ) http://arxiv.org/abs/cond-mat/0409601[arxiv:cond-mat/0409601 ] . m. a. metlitski and s. sachdev , quantum phase transitions of metals in two spatial dimensions : ii . spin density wave order , phys . b * 82 * , 075128 ( 2010 ) ; http://arxiv.org/abs/1005.1288[arxiv:1005.1288 ] . e. g. moon and s. sachdev , competition between spin density wave order and superconductivity in the underdoped cuprates , phys . b * 80 * , 035117 ( 2009 ) ; http://arxiv.org/abs/0905.2608[arxiv:0905.2608 ] . p. w. anderson , p. a. lee , m. randeria , t. m. rice , n. trivedi , and f. c. zhang , the physics behind high - temperature superconducting cuprates : the `` plain vanilla '' version of rvb , j phys . condens . matter * 16 * , r755-r769 ( 2004 ) ; http://arxiv.org/abs/cond-mat/0311467[arxiv:cond-mat/0311467 ] . m. a. metlitski , and s. sachdev , quantum phase transitions of metals in two spatial dimensions : i. ising - nematic order , phys . rev . b * 82 * , 075127 ( 2010 ) ; http://arxiv.org/abs/1001.1153[arxiv:1001.1153 ] . s. a. hartnoll , r. mahajan , m. punk and s. sachdev , transport near the ising - nematic quantum critical point of metals in two dimensions , phys . rev . b * 89 * 155130 ( 2014 ) ; http://arxiv.org/abs/1401.7012[arxiv:1401.7012 ] . s. i. mirzaei _ _ , spectroscopic evidence for fermi liquid - like energy and temperature dependence of the relaxation rate in the pseudogap phase of the cuprates , proc . sci . * 110 * , 5774 ( 2013 ) ; http://arxiv.org/abs/1207.6704[arxiv:1207.6704 ] . m. k. chan _ , in - plane magnetoresistance obeys kohler s rule in the pseudogap phase of cuprate superconductors , phys . lett . * 113 * , 177005 ( 2014 ) ; http://arxiv.org/abs/1402.4472[arxiv:1402.4472 ] . yang , j. d. rameau , z .- h . pan , g. d. gu , p. d. johnson , h. claus , d. g. hinks and t. e. kidd , reconstructed fermi surface of underdoped bi@xmath160sr@xmath160cacu@xmath160o@xmath173 cuprate superconductors , phys . rev . lett . * 107 * , 047003 ( 2011 ) ; http://arxiv.org/abs/1008.3121[arxiv:1008.3121 ] . s. a. hartnoll , p. k. kovtun , m. mller , and s. sachdev , theory of the nernst effect near quantum phase transitions in condensed matter , and in dyonic black holes , phys . b * 76 * , 144502 ( 2007 ) ; http://arxiv.org/abs/0706.3215[arxiv:0706.3215 ] . f. rullier - albenque , h. alloul , f. balakirev and c. proust , disorder , metal - insulator crossover and phase diagram in high-@xmath174 cuprates , europhys . lett . * 81 * , 37008 ( 2008 ) ; http://arxiv.org/abs/0710.3737[arxiv:0710.3737 ] . |
we have used the very large array to image the isolated `` superthin '' galaxy ugc 7321 in the line with a spatial resolution of 16 and a spectral resolution of 24 khz ( 5.2 ) . we have reached a sensitivity of ( 0.36 0.40 ) mjy / beam per channel , which correspond to a column density of ( 8 9) atoms ( ) . ugc 7321 has a gas - rich disk with d and ( d is the distance to ugc 7321 in units of 10 mpc , the value adopted in this paper ) , and no detectable radio continuum emission ( mjy ) . the global distribution of ugc 7321 is rather symmetric and extends to.5 times the optical radius ( at atoms ) . an `` integral sign '' warp is observed in the disk , commencing near the edge of the stellar distribution , and twisting back toward the equatorial plane in the outermost regions . in addition , the position - velocity diagram suggests the presence of a bar or inner arm within from the center . the rotation curve of ugc 7321 is slowly rising ; it reaches its asymptotic velocity of at from the center ( about 0.9 optical radii ) and declines near the edge of the disk . the ratio of the inferred dynamical mass to the mass in gas and stars is d , implying that ugc 7321 is a highly dark - matter dominated galaxy . 0.6truein 0.2truein | astro-ph0303275 | `` superthin '' galaxies appear `` needle - like '' on the sky because of their edge - on orientation , highly flattened stellar disks ( @xmath1810 ) and ( near ) absence of a discernable spheroidal component ( vorontsov - velyaminov 1967 ) . a spectrophotometric study showed them to have gas - rich , optically diffuse disks with little internal absorption , as well as low emission - line intensity ratios and slowly rising rotation curves ( goad & roberts 1981 ) . therefore , many of the so - called superthins are late - type , low surface brightness ( lsb ) disk galaxies seen at high inclination ( matthews , gallagher , & van driel 1999 ; dalcanton & bernstein 2000 ; see also gerritsen & de blok 1999 ) . they seem to be among the least evolved disk galaxies in the local universe , having undergone only minimal dynamical heating , star formation , and angular momentum transport ( bergvall & rnnback 1995 ; matthews et al . 1999 ; dalcanton & bernstein 2000 ) . in addition , their orientation and simple structure make superthins ideal laboratories for probing how galaxy disk structure and morphology are influenced by internal and external processes . observational and theoretical arguments suggest that disk structure and thickness are affected by environment ( tth & ostriker 1992 ; odewahn 1994 ; reshetnikov & combes 1997 ; schwarzkopf & dettmar 2001 ) and imply that a disk must remain isolated in order to remain `` superthin . '' this is often observed , as most galaxies with @xmath19 are devoid of close neighbors ( karachentsev 1999 ; matthews & van driel 2000 ) . however , some exceptions do exist ( duc et al . 2000 ; hibbard et al . 2001 ; matthews & uson 2002 ) . since superthin galaxies are rich in neutral hydrogen ( ) gas ( giovanelli , avera , & karachentsev 1997 ; matthews & van driel 2000 ) , imaging observations of the ( 21-cm ) line emission can be used to test whether these objects have truly remained isolated throughout their lifetimes . dynamical timescales of the outer regions of the gaseous disks of galaxies are longer than those of their stellar disks ; hence signatures of past perturbations and interactions such as warps and gas asymmetries can be preserved in the distribution , even when none are obvious in the stellar disk ( rix & zaritsky 1995 ) . in addition , high - sensitivity observations can uncover the presence of previously unseen gas - rich companions which ( if present ) might also play a role in the evolution of these disks . we can gain further insight into the nature of superthin galaxies by comparing images with data obtained at other wavelengths . for example , measurements of the thickness of the stellar and gaseous disks of superthins are of considerable interest in understanding the evolution of these galaxies . while the stellar scale - heights of superthin disks are typically found to be quite small ( @xmath20200 pc , matthews 2000 ; but see also van der kruit et al . 2001 ) , this does not necessarily mean that their gas disks are also unusually thin , since gaseous and stellar disks are affected by different heating processes . comparison of the structure and scale heights of the gas disks of superthins with those of other types of galaxies should constrain how the dynamically coldest disks are stabilized , and what mechanisms are responsible for heating and supporting their ism component , especially in the absence of significant energy input from massive star formation ( sellwood & balbus 1999 ; snchez - salcedo 2001 ) . since the stellar disks of these galaxies remain so thin , internal as well as external heating processes must be suppressed in them , including ` firehose ' instabilities ( fridman & polyachenko 1984 ) , bar formation ( mihos et al . 1997 ) , and the development of spiral arms ( noguchi 1987 ) . in addition , their stability requires the presence of a massive dark halo ( efstathiou , lake and negroponte 1982 ) . in fact , superthins should be amongst the most dark - matter dominated of disk galaxies ( zasov , makarov , & mikhailova 1991 ; gerritsen & de blok 1999 ) . however , constraints on the dark matter content of edge - on superthin disks have been derived for only a few objects such as ngc 4244 ( olling 1996 ) . existing optical rotation curve studies of superthins typically show a slow rise throughout much of the stellar disk ( e.g. , goad & roberts 1981 ; makarov et al . 1997 ) ; but only observations permit the derivation of the extended outer disk rotation curves which are necessary to constrain the mass models and dark matter halo parameters . for superthins , such analyses are not complicated by the presence of a bulge component , and the mass contributions of stellar disks are expected to be low , making them particularly well - suited for probing the shape of the distribution of dark matter in these galaxies ( olling 1995 ) . in spite of the unique information that can be gleaned from detailed studies of edge - on spiral galaxies , only a small number of these have been imaged in the line with high sensitivity and high spatial and spectral resolution ( e.g. , rupen 1991 ; olling 1996 ; swaters , sancisi , & van der hulst 1997 ; garca - ruiz , sancisi , & kuijken 2002 ) . moreover , with the exception of olling ( 1996 ) , the most sensitive imaging observations of edge - on galaxies to date have concentrated on massive , luminous systems of mid - to - early hubble type rather than on moderate - to - low mass and/or lsb systems like the superthins . this paper is the first in a series describing new high - sensitivity imaging observations of superthin galaxies with moderate optical luminosities ( @xmath21 ) , moderate masses ( @xmath22100 ) , and highly flattened disks ( @xmath2310 ) . the galaxies that we have observed are undergoing differing levels of perturbation from a companion , ranging from being isolated to being strongly perturbed , and show differing levels of current star formation . in this paper , we present our study of ugc 7321 , an isolated superthin galaxy which offers a unique perspective on the ism structure of an lsb disk viewed edge - on . by comparing its properties to those of other lsb and edge - on spiral galaxies , and ultimately , to the results to be presented in our following papers , we aim to address many of the questions discussed above i.e . , to gain clues on the formation histories of pure disk galaxies and probe how environmental factors , interactions , and internal processes influence the structure , evolution , and star formation histories of late - type disk galaxies . ugc 7321 is a highly inclined ( @xmath24 ) sd spiral galaxy . its distance is somewhat uncertain , with estimates ranging from @xmath105 mpc from a kinematical model of the local supercluster ( tully , shaya , & pierce 1992 ) to 14.9 mpc using the @xmath25-band tully - fisher relation ( garca - ruiz et al . one of us ( ldm ) has used the _ hubble space telescope _ to obtain images of ugc 7321 in collaboration with j. s. gallagher ( wisconsin ) for a different project . they estimate that ugc 7321 is located at a distance in the range of ( 7 13 ) mpc based on the brightest resolved stars . in this paper , we adopt a distance of 10 mpc for the derivation of physical parameters , and indicate the dependence of our results with distance explicitly by using the symbol d@xmath8 to denote distance in units of 10 mpc . we summarize the optical and near - infrared properties of ugc 7321 in table 1 . the scale height of the disk of ugc 7321 is one of the smallest ever reported for a galaxy disk ( matthews 2000 ) . ugc 7321 is also unique among the well - studied edge - on spirals because it has an extremely diffuse stellar disk with only minimal patchy dust obscuration and very low internal absorption . correspondingly , ugc 7321 has very low far - infrared and co luminosities ( matthews & gao 2001 ) . ugc 7321 is a very -rich galaxy with a high @xmath6 ratio . indeed , haynes et al . ( 1998 ) obtained a carefully calibrated ( 5% ) spectrum of ugc 7321 with the green bank 43 m telescope , which yielded an integrated line flux density of 44.51 jy km s@xmath26 . matthews et al . ( 1999 ) made a 7-pointing map with the nanay decimetric radio telescope and showed that the gas extends beyond 1.2 times the optical radius . we compare these single - dish observations with our new vla measurements in more detail in section [ global ] . it has also been included in the recent study of warps in edge - on galaxies of garca - ruiz et al . ( 2002 ) who used observations made for the whisp survey . we observed ugc 7321 with the very large array ( vla ) in c configuration during 2000 may 26 ( hereafter day 1 ) and 2000 may 30 ( day 2 ) . since early 1998 , the c configuration is a modified version of the original c configuration ( 3 km maximum baseline ) in which the fifth antenna of the north arm is placed at the center of the array thus improving its sensitivity to extended emission . indeed , it allowed us to `` recover '' ( nearly ) all of the `` single - dish '' signal from ugc 7321 ( see section[global ] ) . the observations were obtained during two 8-hour ( late - afternoon to mostly nighttime ) observing sessions . to obtain the best possible velocity resolution over the frequency range of interest ( defined from the single - dish observations ) , we observed the field centered at the optical position of the galaxy with spectral mode ` 4 ' using two pairs of 127 channels with on - line hanning smoothing ; this produced two pairs of 63 nearly - independent , contiguous channels of width 24.4 khz ( @xmath105.2 ) . for both observing sessions , the first `` if pair '' ( consisting of separate right and left circular polarizations ) was tuned to the known heliocentric velocity of ugc 7321 ( 407 ) , while the second `` if pair '' was tuned to a central velocity of 650 ( on day 1 ) , and 300 ( on day 2 ) in order to search for possible dwarf companions or high - velocity gas in the neighborhood of ugc 7321 ( see section [ companions ] ) . observations of the point source 1221 + 282 ( 5 minute `` scans '' each ) were interspersed with observations of ugc 7321 ( 30 minute scans ) . the strong source 1331 + 305 ( 3c286 ) was observed twice each day ( 10 minute scans ) in order to obtain good flux calibration . most of the data on ugc 7321 ( 720 minutes ) were obtained at elevations above 40@xmath27 . details appear in table 2 . because our chosen phase - calibrator ( 1221 + 282 ) turned out to be considerably weaker ( 0.86 jy ) than the value given in the vla calibrator list ( 1.8 jy ) we used a non - conventional path for the data reduction . we generated a pseudo - continuum database by vector - averaging the data for channels 4 through 60 at each time stamp , but only used this file to find and edit corrupted data ( mostly due to interference and cross - talk at small spacings as well as a few malfunctioning correlators ) . we determined the overall shape of the bandpass as well as the zeroth - order amplitude and phase calibration in one operation using the strong primary calibrator 1331 + 305 ( 3c286 , which was observed at an elevation intermediate to those corresponding to the observations of ugc 7321 ) . we adopted a flux of 14.7363 jy at a frequency of 1418.36 mhz , appropriately scaled to the other two frequency settings , as indicated in the vla calibration manual . this bandpass calibration was applied to all data , which provided the amplitude scale of each frequency channel as well as an initial calibration of the instrumental phases . as is well known , phase errors increase at the band - edges and prevent an accurate calibration of the affected channels . we found that channels 13 and 6163 could not be calibrated to the accuracy of the 57 central channels . therefore , we have not used those six edge - channels in the analysis that follows . next , we made a ( deconvolved ) continuum image of the phase - calibrator using the spectral data from channels 4 through 60 with each visibility gridded at the correct `` uv - cell . '' this image was used for a subsequent `` global '' amplitude and phase ( self ) calibration ( i.e. , one amplitude and phase correction per antenna per scan which was applied to all frequency channels ) in which we constrained the average of the amplitudes to remain constant ( a global re - scaling by the factor 0.996 ) and where the averaging time was that of a `` scan . '' these corrections were subsequently applied to the ugc 7321 visibilities after linear 2-point interpolation ( in time ) of amplitudes and phases . these steps completed the calibration of the observations of day 1 , as we obtained two sets of spectral images ( `` image cubes '' ) that were free of any noticeable artifacts . in addition , the spectrum of the strongest continuum source in the ugc 7321 field ( j1217 + 2239 , with an observed flux density of 64.5 mjy ) , located at an angular distance of @xmath28 , did not show any systematic effects and was well - fitted with a constant over both frequency settings . on the second day , the vla was not quite as stable as on the first day and we found it necessary to improve on the calibration of the instrumental antenna - based passbands . we followed the operations described above with a secondary calibration of the passband determined with the data obtained on j1221 + 282 ( the phase calibrator ) . because of the low flux density of this calibrator ( 0.86 jy ) and the confusion in the surrounding area ( @xmath100.3 jy ) we could only afford to determine one ( complex ) correction per antenna and per channel for the whole observing run , which was obtained by comparing the visibilities channel - by - channel with the theoretical values obtained from the continuum image of j1221 + 282 obtained as described above . these corrections were subsequently applied to the ugc7321 data and new image cubes were made . this step was sufficient for the second frequency setting used on this day , but not so for the first frequency setting , as the spectrum through the aforementioned continuum source , j1217 + 2239 , showed significant curvature . therefore , we were forced to evaluate scan - based corrections to the spectral passband for the main frequency - setting on this day using the observations of j1221 + 282 ( the phase calibrator ) . however , given the low signal - to - noise ratio per channel of these observations , we could only determine ( scan based ) global corrections to the passband in the form of 5-th order chebyshev polynomials using channels 4 through 60 by comparing the so - corrected data to the theoretical visibility function derived from the continuum image of the j1221 + 282 field determined in the previous step . these corrections were once again applied to the ugc7321 data using linear - interpolation between pairs of observations of 1221 + 282 . the image cube determined after this calibration was free of defects , and the source j1217 + 2239 discussed above was well fit with a constant . we have performed a number of tests on the continuum images of our amplitude and phase calibrators and estimate that the uncertainty in the amplitude scale of the images that we present in this paper is about 1% . we used the line - free channels in the second frequency setting ( hereafter if2 ) of each observation to compute an image of the continuum emission in the vicinity of ugc7321 . because there were 47 such channels on day 1 and only 23 on day 2 , we proceeded in the following manner . we first made a multi - field ( 20 fields ) image with the data from channels 450 from day 1 using channel - based `` 3-d '' gridding and a robustness factor @xmath29 , which offers an optimal compromise between noise and resolution when imaging vla observations obtained with full uv - coverage ( briggs 1995 ) . this yielded a continuum image computed over an effective bandwidth of @xmath101.15 mhz . these images were used to generate phase corrections through self - calibration of the data with a 5 minute averaging interval . next we subtracted the theoretical visibilities derived from the images on a channel - by - channel basis and subsequently determined the frequency - averaged residual visibilities . the increased sensitivity of these pseudo - continuum visibilities allowed us to recognize some residual low - level contamination which was mostly due to solar interference ( in the data obtained in the late - afternoon ) as well as cross - talk on short spacings . these data were discarded ( about 11% of the visibilities ) prior to restoring the ( theoretical ) contribution of the radio sources determined above . the same procedure was applied to the line - free channels ( 3860 ) in if2 of day 2 , except that here a 10 minute averaging time was used for the phase - only self - calibration because of the reduced effective bandwidth ( @xmath30 mhz ) . again , about 10% of the visibilities were found to show low - level contamination and discarded . the resulting pseudo - continuum visibilities were combined and a new set of final images ( 20 fields ) was made using a @xmath31 cell - size . the images have a nearly circular synthesized - beam with @xmath32 and an rms noise level of @xmath33jy / beam ( essentially the theoretical noise level ) , and are devoid of any noticeable artifacts . the total deconvolved flux is 199 mjy . it is customary to remove the continuum emission from the visibilities prior to imaging the line emission in the individual channels . this is often accomplished with a visibility - based subtraction of the continuum in which a first - order polynomial is fit to the real and imaginary components of each visibility , and the residuals are used to make the channel images ( cornwell , uson , & haddad 1992 ) . in our case , there are only four line - free channels at each end of if1 ( channels 47 and 5760 ) and the limited number of degrees of freedom in the visibility - based subtraction of the continuum results in an artificial lowering of the noise in the images corresponding to those channels ( to @xmath1075% of the theoretical level ) . therefore , we have followed an alternate path to compute the channel images , whereby we have made images of the total ( line plus continuum ) emission in each channel and subsequently determined the line emission as discussed below . we made twelve image cubes using the combined data from both of our observing dates for if1 . only the first image cube ( which is centered on ugc 7321 ) contained emission ; the remaining eleven image cubes were centered on neighboring continuum sources in order to image the total continuum flux in the field properly . we used a @xmath34-cellsize and a robustness factor @xmath35 in order to minimize the noise level while still yielding a reasonable synthesized - beam which was nearly circular with @xmath36 at position angle pa@xmath37 . the deconvolution was deep , to 0.35 mjy beam@xmath26 ( @xmath38 ) and was followed by a restoration of spurious ( isolated ) components if the absolute value of the flux inside a six - pixel radius was below 0.7 mjy beam@xmath26 . the peak deconvolved flux ( line plus continuum , uncorrected for primary - beam attenuation ) was 480 mjy at 1417.922 mhz . the rms noise was 0.350.40 mjy beam@xmath26 . this procedure resulted in images that were free of artifacts and did not show a negative bowl surrounding the emission . we made the corresponding sets of twelve image cubes with the data from if2 for each one of the two observing dates . because the noise level was 40% higher than for the combined data from if1 , we stopped the deconvolution and `` filtered '' the spurious components at correspondingly scaled levels . next we combined the three image cubes corresponding to the central field at the three observed frequency settings into a global cube with 125 channels in the following way : channels 1 through 47 correspond to channels 4 through 50 of if2 from day 1 . because the if settings used on day 1 resulted in overlapping channels that were reasonably aligned ( a misalignment of only @xmath104% ) channels 48 through 57 are a weighted average of the images in channels 51 through 60 of if2 ( day 1 ) and channels 4 through 13 of if1 ( both days ) respectively . most of the weight is attached to the second image in each pair ( created from the combined data from both days ) . this might seem unnecessary and even undesirable , as corresponding channels observed through both ifs ( on day 1 ) view the same sky - photons , but the vla backend introduces some noise into the observation which can be reduced somewhat through this average . care was taken that the weights accurately reflected the loss of independence of the images averaged . channels 58 through 101 of the global cube were channels 14 through 57 of if1 ( obtained from the combined observations of both days ) ; while channels 102 through 104 resulted from the weighted average of channels 58 through 60 of if1 and 37 through 39 of if2 , day 2 , respectively , with weights determined as above . because the frequency alignment of the channels was somewhat poorer on day 2 ( a misalignment of @xmath1025% ) here we have only combined line - free channels , so that the small blurring ( in frequency ) is of no consequence . finally , channels 105 through 125 of the global cube correspond to channels 40 through 60 of if2 of day 2 . we have performed a number of tests on this combined image cube . figure [ fig : contspectrum ] shows the spectrum at the position of the continuum source j1217 + 2239 in the combined , global 125-channel cube . it is well fit by a constant @xmath39 mjy with a chi - square of 123.9 with 124 degrees of freedom . this is an indication that the bandpass correction has been successful in flattening the spectrum and that the amplitude and phase calibrations on the four different if - day combinations are consistent to within the noise level reached . it also shows that the deconvolution has been successful in eliminating the bowl around the emission that is quite apparent in the raw images and that any residual deconvolution bias is fairly consistent from channel to channel . we have examined the statistics of the image cube to check for artifacts . the cube is remarkably clean . for example , figure [ fig : imeannoise ] shows a histogram of the values in planes 48104 which correspond to if1 but with the region of emission from ugc 7321 excluded ( the central @xmath40 for planes 52100 ) . the histogram is as expected from a statistical distribution with the corresponding number of degrees of freedom ( values are correlated within any synthesized - beam area ) . even more remarkable is the comparison with the histogram in figure [ fig : imeansignal ] , which shows the distribution of values in the central @xmath40 for planes 52100 that had been excluded from figure [ fig : imeannoise ] . the signal is clearly noticeable , but it is also clear that the negative values show the distribution that is expected in the absence of any deconvolution artifacts , such as bowls . we have determined the continuum emission from this cube in the image - plane by linear fitting of a first - order polynomial to each line - of - sight ( pixel ) in the cube ( cornwell et al . 1992 ) after blanking the region that contains the emission from ugc 7321 . because of the finite spacial distribution of the emission , this procedure results in a larger number of `` line - free '' channels per line - of - sight than if the subtraction is performed in the visibility domain and the procedure does not alter the noise characteristics of the channel images in a detectable way . this continuum - subtracted cube has been used for the analysis of the emission reported in this paper . finally , we have made a similar 125-channel image cube following the procedure just described but with a robustness parameter @xmath41 . this is closer to uniform weigthing and results in a somewhat sharper synthesized beam that is again nearly circular ( fwhm@xmath42 at pa @xmath43 , with a @xmath34 cellsize ) . the weighting results in noise levels that are about 40% larger that the corresponding ones for the @xmath35 cube and the deconvolution limits and filtering levels were correspondingly scaled . the region around ugc 7321 contains only relatively weak radio continuum sources ( figure [ fig : continuum ] ) . the brightest source within the primary beam , j1217 + 2239 , is located @xmath44 to the nw of ugc 7321 and has an integrated flux of @xmath45 mjy ( after correcting for the attenuation due to the primary beam ) , in good agreement with the value @xmath46 mjy listed in the nrao vla sky survey ( hereafter nvss ; condon et al . 1998 ) . it is clear from figure [ fig : continuum ] that we have failed to detect continuum emission from ugc 7321 itself . the peak value within the ( optical ) disk of the galaxy is only @xmath47 mjy beam@xmath26 ( formal error ) at 12@xmath48 17@xmath49 33.8@xmath50 + @xmath51(j2000 ) . however , because of the uncertainty introduced by the deconvolution of the sidelobes from j1217 + 2239 , the actual uncertainty at this position is somewhat larger . ugc 7321 was only weakly detected in the fir by _ iras _ in the 60@xmath52 m and 100@xmath52 m bands . upon examining the _ iras _ 60@xmath52 m and 100@xmath52 m survey scans from the nasa / ipac infrared science archive , we found ugc 7321 to be consistent with a point source in the fir with fwhm@xmath53 . iras _ faint - source catalog indicates flux densities of 0.344 jy at 60@xmath52 m ( @xmath54 ) and 0.964 jy at 100@xmath52 m ( @xmath55 ) which yields a fir flux of @xmath56 w m@xmath57 ( using eq . 14 in condon 1992 ) and a predicted 1.4 ghz continuum flux density of @xmath103.1 mjy ( eq . 15 in condon 1992 ) . to place a meaningful limit on the radio continuum emission from ugc 7321 , we have summed the flux in our continuum image ( section [ imagcont ] ) within a @xmath58 box centered at the coordinates quoted above , which yields a flux density of @xmath59 mjy given the low significance of the _ iras _ detections , it is perhaps not surprising that the galaxy is not detected in the radio continuum . in addition , some of the fir luminosity could be due to a contribution from cirrus emission as well as from a weak embedded nucleus . indeed , matthews et al . ( 1999 ) found a rather red ( @xmath601.5 ) , compact feature inside the _ iras _ box . moreover , a low - mass galaxy such as ugc 7321 might be unable to confine cosmic rays with sufficient efficiency and would thus fall short of the prediction from the fir - radio correlation . using the relations given by condon ( 1992 ) , the low fir emission from ugc 7321 yields a rather small estimate of its high - mass star formation rate of @xmath100.006/year ( for @xmath61 ) . such a rate is more than one order of magnitude smaller than the mean value estimated for all ugc galaxies of type scd - im ( see roberts & haynes 1994 ) , confirming that ugc 7321 is a bona fide low surface - brightness spiral galaxy . figure [ fig : channelmaps ] shows the images corresponding to the channels that contain detectable emission , flanked with one extra channel on each side , overlayed on an @xmath62-band image of ugc 7321 . a number of features of the disk and velocity field of ugc 7321 can be seen in these channel images . for example , the channels covering the velocity ranges 293.4 319.2 and 479.4 510.4 , clearly show warping of the outer gas layers . in addition , in most of the channel images , the spread in @xmath63 of the emission increases with distance from the disk center , which indicates flaring of the gas layer . however , the gas in the most extreme velocity channels is confined to a narrower distribution than at small and intermediate velocities , contrary to the expectation for a pure flaring model ( see also swaters et al . 1997 ) . it is interesting that in the two channels representing the velocity extrema of the observed emission ( i.e. , corresponding to velocities of 283.0 and 531.1 ) , we see gas that is not located at the edge of the disk , but rather at smaller galactocentric radii , near the edge of the stellar disk . this is indicative of a falling rotation curve in the outskirts of ugc 7321 ( see section [ rotcurve ] below ) . in order to estimate the total content of ugc 7321 and derive a global rotation profile , we measured the total flux in each of the channel images from our global , @xmath35 , cube , after discarding low values ( those with absolute value @xmath64 ) and restricting the sum to the pixels corresponding to the galaxy . the corresponding pixels were determined from an initial selection consisting of those pixels which were above the @xmath65 level ( where the galaxy is quite distinct ) and expanding such areas by successive surrounding bands 3-pixels wide ( @xmath66 ) until the enclosed flux converged . for each channel , the selected areas were summed after correction for the attenuation of the primary - beam . we estimated the error in the total flux in each channel from the rms noise in that channel , taking into account the correlation introduced by the synthesized - beam and the correction due to the attenuation of the primary beam . the resulting errors are quite small ( @xmath67 mjy per channel ) . the resulting global profile is shown in figure [ fig : global ] . also shown are the profiles obtained with the nrao 43 m telescope ( haynes et al . 1998 ) and with the nanay telescope ( matthews et al . the three profiles match extremely well given the 5% calibration uncertainty ( and statistical correction for primary beam attenuation ) of the green bank data , and the estimated calibration uncertainty of the seven - pointing nanayspectrum . the total flux derived from the present data is @xmath68 jy km s@xmath26 , where the error comes from a statistical contribution of 0.11 jy km s@xmath26 , an estimated systematic contribution of 0.36 jy km s@xmath26 and a global calibration uncertainty of @xmath69% . comparison with the values measured with the 43 m ( 44.51 jy km s@xmath26 ) and the nanay ( 47.6@xmath70jy km s@xmath26 with 15% calibration uncertainty ) telescopes indicates that the measurements reported in this paper have likely recovered most of the flux . the measured brightness temperatures peak at a value of t@xmath71 k at the center of the galaxy and at the systemic velocity . the somewhat better resolution image made using a robustness factor r@xmath72 ( fwhm@xmath73 ) yields a corresponding peak brightness temperature t@xmath74 k. this value is interestingly close to those measured in the milky way ( burton 1970 ) suggesting that the spectrum of ugc 7321 might be affected by self - absorption . we have examined cuts along the major and minor axes of the emission at the systemic frequency ( which contains the peak values ) . the latter is well fit by a gaussian of half - width @xmath75 , with the data falling short of the gaussian fit at the center position by @xmath76% , indicating only mild self - absorption . neglecting this leads to an estimate of the total mass of ugc 7321 of @xmath77 d@xmath5 . from the global profile , we measure velocity widths at 20% and 50% of the peak maximum of 234.3 and 219.8 , respectively . the heliocentric recessional velocity ( optical definition ) determined as the average of the 20%-level values of the global profile is 406.8 ( table 3 ) . these values are all in agreement with previous single - dish measurements . the sub - images containing extracted as discussed in the previous section have been added to obtain the total intensity map of ugc 7321 which is shown in figure [ fig : totintmap ] , overplotted on a continuum - subtracted h@xmath78 $ ] image of the galaxy . the overall morphology of ugc 7321 is typical of normal , edge - on spirals . it shows a relatively smooth , unperturbed , symmetric distribution , and we see no obvious enhancements in the emission corresponding to the locations of the h@xmath78 $ ] emission . at a resolution of @xmath79 , the peak intensity observed within the galaxy corresponds to @xmath80 @xmath2 , while the maximum observed extent of the disk along the major axis is @xmath12@xmath13 @xmath14 at @xmath81 @xmath2 , or roughly 1.5 times the ( d@xmath82 ) diameter of the stellar disk . at @xmath83 @xmath2 , we measure an diameter of @xmath84@xmath13@xmath85 . the slight depression in the intensity distribution visible near the optical center of ugc 7321 is possibly due to the mild self - absorption discussed above . such features are frequently seen in images of edge - on spirals . in addition , small filaments appear to be protruding vertically from the disk to higher latitudes . these features are discussed further in section [ pvz ] . one particularly interesting feature of the total intensity map of ugc 7321 is the `` integral sign '' warp visible in the disk . the warp commences near the edge of the stellar disk , reaching a peak amplitude of @xmath86 on the east side of the disk and @xmath87 on the west ( i.e. , roughly 4% of the disk radius ) , before the gas twists back toward the midplane in the outermost regions . these features can be seen clearly in the warp curve shown in figure [ fig : warpcurve ] , where we have plotted the mean deviation of the gas from the disk midplane as a function of radius , measured by fitting single gaussians to intensity profiles perpendicular to the disk . warps with features similar to those of ugc 7321 have now been observed in the disk of the milky way ( burton 1988 ) and numerous edge - on spiral galaxies ( sancisi 1976 ; briggs 1990 ; garca - ruiz et al . gaseous warps usually originate at the edge of the stellar distribution , which is to be expected because it would seem easier to pull material out from the midplane in regions where the self - gravity is lowest , and because once material becomes displaced from the midplane , dissipational and dynamical - friction forces will cause it to settle back into equilibrium more quickly within the stellar disk than outside of it ( e.g. , bottema 1996 ) . however , the outer stellar disk of ugc 7321 is extraordinarily diffuse ( matthews et al . 1999 ) , implying that its self - gravity should be far less significant than in the outer regions of a galaxy like the milky way . therefore , the commencement of the warp near the edge of the stellar disk ( with essentially no hint of warping being visible in the stars ) seems somewhat surprising . a number of researchers have shown that warps are more commonly seen in galaxies that have companions ( e.g. , reshetnikov & combes 1998 ) , implying that interactions play a role in triggering them . however , ugc 7321 appears to be an unusually isolated galaxy , with no close companions detectable in the optical bands , and no gas - rich companions with @xmath88 ( see section [ companions ] ) . the nearest optically catalogued neighbors to ugc 7321 are ugc 7236 ( type i m ) , at @xmath89=945 and a projected distance of @xmath90 ( 0.34 d@xmath8 mpc ) , and ngc 4204 ( type sbdm ) , at @xmath91861 and a projected distance of @xmath92 ( again 0.34 d@xmath8 mpc ) . assuming a typical group speed of 200 , the last encounter of ugc 7321 with one of its neighbors would have taken place at least @xmath93 years ago , even if they were separated only by their ( minimal ) projected distance . is it plausible that an encounter with one of these systems could have triggered the warp in ugc 7321 ? a consequence of tidal interactions will be to cause heating ( thickening ) of galaxy disks ( e.g. , reshetnikov & combes 1997 ) . although the stellar disk of ugc 7321 is ( dynamically ) much colder than those of typical spiral galaxies , matthews ( 2000 ) has shown that ugc 7321 does show distinct vertical color gradients and a multi - component stellar disk structure , consistent with a disk that has been mildly heated . using statistical arguments , reshetnikov & combes ( 1997 ) suggested that the typical `` cooling '' time for a disk mildly perturbed by an interaction should be @xmath1010@xmath94 years . hofner & sparke ( 1994 ) , who treat warps as discrete bending modes in a disk embedded in a dark halo potential , have developed a formalism to estimate the age of a warp . in ugc 7321 , the rotation curve flattens at about @xmath95 from the center ( the hofner - sparke @xmath96 ) , the warp commences at about @xmath97 from the center of the disk , the peak rotational velocity ( @xmath98 ) is about 110 , the disk scale length ( @xmath99 ) is @xmath100 kpc and the disk mass ( in gas and stars ) is @xmath101 . in this latter relation , the factor of 1.34 corrects the gas mass for the contribution of helium and @xmath102 is the stellar mass - to - light - ratio . we assume that the contributions of molecular and ionized gas are negligible . using the models of bell & de jong ( 2001 ) , we adopt a stellar mass - to - light ratio @xmath102=1.0 based on ugc 7321 s @xmath103 color of 0.97 . then , equation 7 of hofner & sparke ( 1994 ) leads to an estimate of the age of the warp of @xmath1010@xmath104 years . thus , our estimate does not exclude the possibility that a mild encounter between ugc 7321 and one of its nearest catalogued dwarf neighbors could have occurred @xmath105 years ago and triggered its warp ( as well as some mild dynamical heating of its disk ) . assuming that the timescale for the cooling of the disk is @xmath1010@xmath94 years as suggested by reshetnikov & combes ( 1997 ) , such a scenario would have allowed sufficient time for subsequent star formation ( and possible gas accretion ) to re - establish the very thin young stellar disk that is observed in ugc7321 . ultimately , self - consistent numerically calculations ( including the gas and the stars , as well as a `` live '' dark matter halo ) will be needed in order to gauge whether the amplitude and line of nodes of the warp of ugc 7321 are consistent with its triggering by the passage of one of its dwarf neighbors , and whether a thin , moderate - mass disk like ugc 7321 could have survived such an encounter relatively unscathed . we have computed a map of the first moment of the emission ( representing the velocity field ) of ugc 7321 from our @xmath35 spectral image cube using the standard `` cutoff '' technique ( rots & shane 1975 ) . a 2@xmath106 clipping was applied to the absolute values in the image cube after spatially smoothing the data with a gaussian kernel of @xmath107 and hanning - smoothing the data in velocity . in order to maintain a uniform cutoff with respect to the noise , these operations were performed on images uncorrected for primary beam attenuation . the spatial and velocity smoothing were used for the purpose of deciding which pixels to discard ; the moment map was made with the full resolution of the input cube . because of the biases that clipping of any sort can introduce ( bosma 1981 ) , we used the moment-1 map ( shown in figure [ fig : mom1 ] ) only to obtain a qualitative impression of the global velocity field of ugc 7321 . overall , the velocity field of ugc 7321 is fairly smooth and normal , although some minor irregularities are visible . in the innermost regions , the isovelocity contours exhibit the characteristic `` v '' shape of a differentially rotating disk , with only a slight asymmetry in the shape of these contours near the eastern and western ends of the disk . the symmetry of these contours about the midplane shows that the ( kinematic ) major axis of ugc 7321 does not appear to undergo any significant twisting or precession . although the velocity field of ugc 7321 is fairly regular , a mild asymmetry between the receeding and the approaching sides is noticeable about half - way through the extent of the optical disk . in addition , at galactocentric radii @xmath10875@xmath0 , the isovelocity contours begin to exhibit some curvature at the edges that can be attributed to the warping of the disk . finally , at the far edges of the disk , we see the most pronounced deviations from purely regular rotation . on the approaching side , the isovelocity contours exhibit their greatest distortion as the terminal velocity of the disk is neared , while the receding side of the disk shows three irregular patches of higher velocity material superimposed on the constant - velocity outer disk region . in figure [ fig : pv ] we show position - velocity ( p - v ) diagrams for ugc 7321 along the major axis of the disk ( center panel ) , as well as at heights of @xmath109 , above and below the midplane . these diagrams were made from unsmoothed data with no averaging applied along the vertical direction . the overall shape of the p - v diagrams of ugc 7321 is that of a `` scaled - down '' giant spiral ( e.g. , m31 ; brinks & shane 1984 ) ; although it shows a smaller fraction of its emission near the terminal velocity ( along the `` flat '' part of the p - v diagram ) than in most giant spirals , and the peak rotational velocity of ugc 7321 is only about half of that of a system like m31 . a large range of gas velocities is observed at all positions across the disk , particularly in the central regions . while similar features are commonly seen in the p - v diagrams of more luminous spirals , many low - mass spiral galaxies with luminosities and rotational velocities similar to those of ugc 7321 exhibit slowly rising , solid - body rotation curves with a much narrower spread of velocities at any given radius [ e.g. , ic 2233 ( matthews & uson 2002 ) ; ngc 55 ( puche , carignan , & wainscoat 1991 ) ; ngc 4395 ( swaters et al . 1999 ) ] . one interesting feature of the major axis p - v profile of ugc 7321 is its `` figure-8 '' shape , which is clearly visible in figure [ fig : pv ] . m31 exhibits a similar signature in its major axis p - v curve , and brinks & shane ( 1984 ) demonstrated that this results from the warping and flaring of its disk . as discussed above , the disk of ugc 7321 is also warped , and the twisting of its gas disk is likely to be the origin of the `` figure-8 '' shape seen in this galaxy . this can be seen in figure [ fig : mom1 ] , where the isovelocity contours are found to be predominantly v - shaped in the inner disk regions , but at @xmath110 they begin to twist near the edges . it is also interesting to note that the central panel of figure [ fig : pv ] suggests the presence of a ( small ) bar or inner arm that extends from about @xmath111 to @xmath112 and manifests itself mainly in the decrease of the intensity at @xmath113 , @xmath114 ) . indeed , in a barred galaxy viewed edge - on , the lack of available orbits for gas near the corotation radius manifests itself as a perturbation in the major - axis p - v diagram ( kuijken & merrifield 1995 ) . in addition , the optical h - band major - axis profile shows two distinct `` shoulders '' at @xmath115 where the light exceeds that of a pure exponential disk ( matthews et al . 1999 , their figure 3 ) . given its edge - on orientation it might be difficult to distinguish between a small bar and an inner arm or even confirm their reality . however , should further observations confirm that ugc 7321 is in fact barred , they would challenge the conventional view that bar formation is strongly suppressed in galaxies with low surface densities and large dark matter contents ( mihos et al . 1997 ) . in figure [ fig : zpv ] we display p - v diagrams extracted at five different cuts parallel to the rotational axis of ugc 7321 : along the minor axis ( @xmath116 ) , at @xmath117 , and at @xmath118 . in all of these p - v profiles , we identify features at @xmath119 levels that appear to be filaments extending out of the plane of ugc 7321 to distances up to @xmath120 ( @xmath102.9 d@xmath8 kpc ) . in addition to these features , other apparent asymmetries are visible in the minor axis p - v profiles when the two sides of the disk are compared . these trends are reminiscent of features reported in the giant , edge - on sbc galaxy ngc 891 by swaters et al . ( 1997 ) , which they attributed to the presence of an halo produced by a galactic fountain . as discussed by these authors , detailed modelling is required in order to demonstrate conclusively that the apparent high - latitude extensions are not simply manifestations of a warp or of flaring of the gas disk . an analysis of the high - latitude gas will be reported elsewhere ( matthews and wood 2002 ) . commonly used methods of converting p - v data into a true disk rotation curve , such as fitting tilted ring models to the velocity field , extracting major axis cuts along the velocity field , or fitting single , intensity - weighted gaussians to one - dimensional p - v slices , break down for edge - on and nearly edge - on galaxies as a consequence of projection effects . indeed , if applied to highly inclined galaxies , these methods will result in a rotation curve which is artificially shallow in the inner regions ( sancisi & allen 1979 ) . therefore , we have used a technique that was first introduced to galactic rotation studies by shane & bieger - smith ( 1966 ) which involves the determination of the terminal velocity at a given position , @xmath121 , from one - dimensional p - v cuts at various radii using an `` equivalent rectangular measure '' based on the location of the extreme velocity features : @xmath122 where @xmath123 is the peak flux density in the profile , @xmath124 is its corresponding velocity , @xmath125 is the channel separation , and @xmath126 is the velocity at which the wing of the extreme velocity edge of the line profile drops to zero . the last term in the equation is a correction that accounts for gas turbulence and was introduced by burton & gordon ( 1978 ) who also noted that this method is rather insensitive to noise . we make the further approximation that @xmath106 is everywhere equal to 7 the average of the range of @xmath127(5 9 ) observed in normal spiral and quiescent dwarf galaxies ( van der kruit & shostak 1984 ; lo et al . 1993 ; olling 1996 ) and which we also find to be consistent with our data based on gaussian fits to the extreme velocity envelopes of our one - dimensional p - v cuts . we have ignored any additional corrections for asymmetric drift in ugc 7321 , as such corrections for the gaseous components of disks are typically on the order of only a few i.e . small compared with the uncertainties of applying such corrections ( de blok & bosma 2002 ) . the resulting rotation curve is shown in figure [ fig : rotcurve ] . we have indicated on this figure the location of the edges of the stellar disk . the statistical errors of these measurements have a median value of 0.7 , with only two points at the end of the receding side and three points at the end of the approaching side having significantly larger statistical errors of up to 6 . however , the uncertainties are likely to be dominated by systematic errors . we have compared the values that we derived using the `` equivalent rectangular measure '' to those obtained from fitting `` half - gaussian '' functions to the outer part of the profiles in the flat region of the rotation curve and estimate systematic errors of order @xmath67 . we see that the rotation curve of ugc 7321 flattens on both sides of the disk , reaching a peak rotational velocity of @xmath10110 , but that it rises rather slowly , not reaching @xmath128 until about half - way through the disk . this is in contrast to high surface brightness spirals , whose rotation curves tend to rise rapidly to their asymptotic value , ( e.g. , casertano & van gorkom 1991 ) . it is clear from figure [ fig : rotcurve ] that the rotation curve of ugc 7321 does not exhibit any strong asymmetries or lopsidedness , particularly in the inner regions . it does however show a slight ( @xmath1010% ) difference between the peak amplitudes of the approaching and receding sides of the disk , and the rotation curve is observed to extend about 1 kpc further on the approaching side ( see also figure [ fig : pv ] . figure [ fig : rotcurve ] shows that near the edge of its disk , the rotation of ugc 7321 appears to decline , decreasing by @xmath109% from its peak value . evidence for this decline can also be seen in the major axis p - v plot in figure [ fig : pv ] and the channel maps in figure [ fig : channelmaps ] . a declining rotation curve was also reported for the thin , edge - on spiral ngc 4244 by olling ( 1996 ) . two possible interpretations of this falling rotation curve in ugc 7321 are that either the dark matter halo of ugc 7321 is truncated near the edge of the disk , or that we are seeing a projection effect caused by the warping of ugc 7321 . both interpretations are consistent with the data . whenever a slowly rising rotation curve is inferred from aperture synthesis measurements , it is of some concern whether the apparent slow rise could be an artifact of beam smearing ( begeman 1989 ) . indeed , some of the slowly rising rotation curves reported for lsb galaxies in the literature have been attributed to this effect ( swaters , madore , & trewhella 2000 ; van den bosch et al . these rotation curves can not be used to constrain the shape of dark matter halos , since the inner region is the most critical for distinguishing between the cuspy halos predicted by cold dark matter ( cdm ) models and dark matter halo models with lower density cores . in order to gauge the possible effects of beam smearing in the rotation curve of ugc 7321 , we have performed two tests . firstly , we produced a new deconvolved image cube of ugc 7321 using a robustness value of @xmath129 as described in section [ imagline ] , which yielded a spatial resolution of @xmath130 , 25% narrower than that of our initial ( @xmath35 ) derivation . the rotation curve obtained following the steps described above is shown as a dotted line on figure [ fig : rotcurve ] . aside from the uncertainties caused by the @xmath131% higher noise level in the higher resolution data , the agreement between the two rotation curves is excellent . a second check is provided by the comparison of the rotation curve of ugc 7321 with the rotation curve derived by goad & roberts ( 1981 ) from optical ( h@xmath132 ) emission line spectroscopy ( shown with ` + ' signs on figure [ fig : rotcurve ] ) . the optical data have much higher spatial resolution than the data and hence do not suffer from beam smearing . given the uncertainties introduced by the more irregular distribution of the regions compared with that of the gas , the agreement between the and h@xmath132 rotation curves is quite good , even in the innermost regions . indeed , maximum deviations are @xmath205 ( our spectral resolution ) . as shown by matthews & wood ( 2001 ) , the effects of internal extinction on the optical rotation curve should be negligible in ugc 7321 . we conclude that the effects of beam smearing are insignificant in our derived rotation curve of ugc 7321 . assuming a spherical mass model for ugc 7321 , the dynamical mass of ugc 7321 within radius @xmath133 is simply @xmath134 . therefore , the mass contained within the radius of the last measured point of the rotation curve ( taken to correspond to @xmath135 d@xmath8 kpc , @xmath136 ) , is @xmath137 d@xmath8 . this implies ratios of the total mass to the galaxy s @xmath25-band luminosity and mass , respectively , of @xmath138 d@xmath139 and @xmath140 d@xmath139 . these ratios are 78 times larger and 3 times larger , respectively , than the median values found by roberts & haynes ( 1994 ) for scd sd galaxies , indicating a large dark matter content in ugc 7321 . assuming a stellar mass - to - light ratio in the range @xmath141 1.4 , consistent with the @xmath142 color of ugc 7321 , and a correction to the mass of 1.34 for the contribution of helium , the ratio of the dynamical mass to the visible mass in ugc 7321 is in the range ( 1113 ) d@xmath139 , which is again rather large . we have begun detailed modelling of the underlying distribution of dark matter in ugc 7321 . preliminary results indicate a slight preference for a pseudo - isothermal sphere over the cuspy distributions suggested by cdm numerical simulations ( navarro , frenk , & white 1996 ) . in all cases , ugc 7321 is highly dark - matter dominated at all radii beyond @xmath143 d@xmath8 kpc . because of the high inclination of ugc 7321 , the column densities inferred directly from the total - intensity map ( figure [ fig : totintmap ] ) represent only projected values rather than the true gas surface densities in the disk . deprojected gas surface densities are of interest for assessing the star - formation efficiency of ugc 7321 and for comparing its distribution to that of other less inclined late - type and lsb spirals . in order to estimate the deprojected surface density in ugc 7321 as a function of radius , we have followed an iterative modelling procedure using the formalism developed by irwin and her collaborators ( the cubit software , irwin & seaquist 1991 ; irwin 1994 ) . this is a non - linear least - squares fitting program that operates on a three - dimensional ( ) spectral - line image cube and is capable of simultaneously determining the underlying density distribution and velocity field . we adopted the observed rotation curve for ugc 7321 ( see section [ rotcurve ] ) and used it to determine the density distribution . from the fits , we found that we were unable to reproduce the density distribution of ugc 7321 with a single - component disk model either gaussian or exponential in @xmath133 and @xmath63 . neglecting the warping and flaring of the gas , we were able to find a reasonable fit to the data using a two - component model : ( 1 ) a ring of gaussian cross - section with a central hole of radius @xmath144 , a scale - height ( @xmath106 ) of @xmath145 , a scale - length of @xmath146 , and peak density of 0.19 atoms @xmath147 ; ( 2 ) a gaussian with scale - height of @xmath148 , a scale - length of @xmath149 , and a central density of 0.06 atoms @xmath147 . fixing these parameters , we used cubit to compute a face - on rendition of the galaxy and `` observed '' the intensity distribution along its major axis . subsequently , we used this model profile in order to scale the observed major axis intensity profile of ugc 7321 to a deprojected ( face - on ) value which is shown in figure [ fig : majorint ] . this profile is in good agreement with the one derived for ugc 7321 by garca - ruiz et al . ( 2002 ) from westerbork data using an abel inversion technique . figure [ fig : majorint ] shows that the peak surface density in ugc 7321 is only @xmath10 5.8 pc@xmath57 , somewhat smaller than the mean value of 10 pc@xmath57 found in normal surface brightness sd galaxies by cayatte et al . ( 1994 ) or broeils & van woerden ( 1994 ) . moreover , it appears that at all radii , the gas surface density in ugc 7321 lies below that required for efficient star formation based on the dynamical criterion of kennicutt ( 1989 ) . while this star formation criterion should only be considered an approximation ( ugc 7321 is clearly forming some stars ) , this result is nonetheless consistent with the very low star formation rate implied by its low ( optical ) surface brightness disk , weak co emission , and low radio continuum and far - infrared fluxes . although the gas surface densities observed in ugc 7321 are very low compared with those of normal spirals , they are however rather similar to the typical ( subcritical ) surface densities seen in other late - type , lsb spiral galaxies ( van der hulst et al . the deprojected functions indicate a drop in the radial surface density distribution of ugc 7321 at @xmath15012 kpc ; a step of only one synthesized beam ( 0.8 kpc ) leads to a drop from a deprojected surface density of @xmath151 atoms @xmath2 ( @xmath152 pc@xmath57 ) to @xmath153 atoms @xmath2 ( @xmath154 pc@xmath57 , @xmath155 ) . we have searched for companions to ugc 7321 within a @xmath156 field and over the full velocity range covered by our observations . visual inspection of the image cube shows no recognizable companions . in addition , we have done a `` matched - filter '' automated search using a set of gaussian functions in frequency as our kernel ( uson , bagri , & cornwell 1991 ) . for @xmath157 in the range ( 24.4244 ) khz ( 110 channels ) we search for an expected signal of the form : @xmath158 where the kernel has the form @xmath159,\ ] ] where @xmath160 and @xmath161 are the edges of the observed band , and the norm @xmath162 is @xmath163 the noise is weighted by the same kernel and is approximately equal to @xmath164 which is the value that corresponds to the case in which all image planes have similar noise . the method is surprisingly efficient at finding signals even if their intrinsic shape is not gaussian . indeed , it is a good discriminant of faint extensions to the disk of ugc 7321 as they fade into the noise . when we exclude the galaxy delineated as discussed in section [ global ] , a search done with a kernel of width equal to one channel yields a set of amplitudes , @xmath165 , which are nearly gaussian distributed , with tails extending up to @xmath166 . this is reasonably close to what is expected from statistics only as the image cube has @xmath167 degrees of freedom ( so that tails up to @xmath168 are expected if the noise is gaussian distributed ) . in addition , spurious `` absorption '' at the @xmath169 level is found against the source j1217 + 2239 discussed in section [ cont ] . we have traced this systematic to small positional offsets between the three image cubes ( corresponding to the three observational settings ) that we have combined to generate the 125 channel image cube . these offsets are a consequence of residual phase noise left behind in the calibration . because companions are likely to show emission with width of up to @xmath1040 , we have repeated our search with kernels of width 210 channels . even though the smoothing introduced by the kernel decreases the number of degrees of freedom , the algorithm `` finds '' somewhat more extended distributions with tails that reach @xmath170 , especially when using kernels with a width of a few ( 37 ) channels ( in addition , the line of sight to j1217 + 2239 again shows spurious signals due to the image jitter just described ) . this is likely due to a lack of independence between neighboring channels in excess of the @xmath104% expected from hanning - smoothing , which could be introduced by low - level errors in the vla backend electronics , as well as by errors in the correction of the spectral bandpass during the data reduction due to insufficient sensitivity ( see also section [ imagline ] ) . we have verified that these tails have not been introduced because of our decision to image the line emission without prior subtraction of the continuum emission ( section [ imagline ] ) by reprocessing the data after performing the customary visibility - based subtraction of the continuum emission ( cornwell et al . the resulting spectral images agree with the ones described above within the noise ; although they have a higher `` zero - point '' uncertainty due to the small number of line - free channels in if1 . in addition , the rms noise is not uniform , being significantly lower for the four line - free channels at each end that are used to define the continuum . in addition , the matched filter finds a `` signal '' at a level of @xmath10@xmath171 @xmath172 mjy ) ; which appears in a single synthesized - beam area with a width of about six channels . we are unable to provide any corroborating evidence that would lead us to believe that this small blip represents real emission . moreover , given that the tail of negative values of @xmath165 reaches down to @xmath10@xmath173 for this kernel , we prefer to treat it at this time as spurious excess noise . we have added three standard deviations to its fiducial value and corrected it for the attenuation of the primary beam of the vla antennas in order to derive a robust upper - limit to the mass of the fiducial companions to ugc 7321 located within a radius of 12 minutes of arc ( 36 d@xmath8 kpc ) of @xmath174 d@xmath5 . therefore , we find no close companion that might be responsible for the warp of ugc 7321 ( section [ warp ] ) . even allowing for a significant dark matter component in any such companions to ugc 7321 , their total masses would still be below 1% of the inferred total mass of ugc 7321 ( section [ dynmass ] ) . such a mass is likely to be too small to excite a warp via tidal forcing ( hunter & toomre 1969 ; kuijken & garca - ruiz 2001 ) , thus requiring either the warp of ugc 7321 to be quite old ( in order to have allowed the perturber sufficient time move outside our search region ) , or that the warp was excited by some other mechanism . we have presented high - sensitivity images of the isolated , edge - on , ` superthin ' spiral galaxy ugc 7321 with a resolution of 24.4 khz and a total bandwidth of 3.052 mhz ( 125 channels ) . in the optical , ugc 7321 exhibits a diffuse stellar disk with little dust obscuration , an extremely small global scale height , and no discernible spheroid component . these features indicate that ugc 7321 is a late - type , lsb spiral seen at high inclination . ugc 7321 thus allows us to explore both the radial and vertical structure of the ism of an ordinary lsb galaxy whose disk has not been recently perturbed . ugc 7321 shows similar global characteristics to other late - type , lsb spirals : it has a gas - rich disk with = @xmath77 d@xmath175 , and an inferred ratio of @xmath6 = 1.0(in solar units ) , which is at the high end of the observed values for spiral galaxies . assuming a stellar mass - to - light ratio similar to the sun ( based on its global @xmath103 color ) ugc 7321 contains a similar fraction of its visible mass in gas as in stars , suggesting that it has been a relatively inefficent star former . ugc 7321 is undetected in the 21-cm radio continuum ( @xmath9mjy ) , which is consistent with its status as a low star formation rate system . assuming a spherical mass distribution , the ratio of the dynamical mass to the total visible mass ( total gas+stars ) is @xmath176 , implying that ugc 7321 is a highly dark - matter dominated galaxy . to first order , the total- distribution of ugc 7321 is regular and symmetric with a slight depression near its center , and extends to approximately 1.5 times the stellar radius . the deprojected surface density is low ( peak value @xmath177 pc @xmath57 ) , which is consistent with other lsb spirals . the outer disk of ugc 7321 displays mild flaring in addition to an `` integral sign '' warp that commences near the edge of the stellar disk . the warped material bends back toward the equatorial plane in the outermost regions . the origin of this warp is still not understood , as ugc 7321 appears to be an isolated galaxy . the nearest known companions to ugc 7321 are two dwarf galaxies , both with projected distances of 0.34 d@xmath8 mpc and @xmath178 which would have had their last encounter with ugc 7321 @xmath93 years ago . in addition , the observations described in this paper rule out the presence of any close gas - rich companions with masses greater than 2.2@xmath179 ( @xmath180 ) within @xmath181 ( 36 d@xmath8 kpc ) . the disk of ugc 7321 appears thicker and more complex than what a single , cold layer confined to the midplane would display . however , the projection effects expected from warping and flaring of the gas layer have to be modelled in order to untangle the three - dimensional structure of the galaxy , which is complicated by its near - edge - on orientation . in addition , we detect filaments of gas extending to distances of up to @xmath1060@xmath0 ( 2.9 d@xmath8 kpc ) from the mid - plane in the total intensity images and in the position - velocity cuts along the minor axis of the disk , which suggests the presence of high - latitude gas . the position - velocity plot along the major axis of ugc 7321 displays a `` figure-8 '' pattern that we interpret as a manifestation of the warping of the disk . in addition , the diagram reveals the presence of a small bar or inner arm . overall ugc 7321 has a rather regular velocity field with only minor perturbations . similarly , the rotation curve of ugc 7321 exhibits only a mild asymmetry about the two sides of the disk . the rotation curve rises rather slowly and linearly throughout the stellar disk , indicating a low central matter density ; this effect can not be attributed to the finite spatial resolution of our observations . a peak rotational velocity of @xmath10110 is reached near 0.9 optical radii , and a slight decline in the rotation curve occurs in the last few measured points . two possible interpretations of this decline are that the dark matter halo of ugc 7321 is truncated near the edge of the disk or that we are seeing a projection effect due to the warping of the layer . @xmath132 ( j2000.0 ) & 12 17 33.8 @xmath182 ( j2000.0 ) & + 22 32 25 hubble type & sd iv distance ( mpc ) & 7 13 p. a. ( deg ) & @xmath183 inclination ( deg ) & @xmath184 @xmath185 & 10.3 @xmath186 ( arcmin ) & 5.6 @xmath186 ( kpc ) & 16.3 d@xmath8 @xmath187 ( kpc ) & @xmath188 d@xmath8 @xmath189 ( pc ) & @xmath190 d@xmath8 |
we present a harmonic expansion of the observed line - of - sight velocity field as a method to recover and investigate spiral structures in the nuclear regions of galaxies . application to the emission - line velocity field within the circumnuclear starforming ring of ngc1097 , obtained with the gmos - ifu spectrograph , reveals a _ three_-arm spiral in the non - circular motions , which agrees with a _ two_-arm dust spiral in the surface brightness . this nuclear spiral is consistent with a weak perturbation in the gravitational potential due to a two - arm logarithmic spiral , with a pitch angle of derived directly from the harmonic expansion of the velocity field . next , we use a simple spiral perturbation model to constrain the fraction of the measured non - circular motions that is due to radial inflow . we combine the resulting inflow velocity with the gas density in the spiral arms , inferred from emission line ratios , to estimate the mass inflow rate as a function of radius , which reaches aboutmyr at a distance ofpc from the center . this corresponds to a fraction of about of the eddington mass accretion rate onto the central black hole in ngc1097 , and is fully consistent with previous mass accretion models fitted to the observed spectral energy distribution in the nucleus of this liner / seyfert1 galaxy . we conclude that the line - of - sight velocity not only can provide a cleaner view of nuclear spirals than the associated dust , but that the presented method also allows the quantitative study of these possibly important links in fueling the centers of galaxies , including providing a handle on the mass inflow rate as a function of radius . | 0905.3556 | gas transport to the centers of galaxies is still mainly an unsolved problem ( e.g. * ? ? ? since most of this gas is residing in a rotating disk well beyond the center , it is essentially a problem of angular momentum transport . proposed transport mechanisms range from galactic interactions and bars , to nuclear bars and spirals , to stellar mass loss and disruption near the central black hole ( bh ) . this range in scales also closely represents a range in decreasing mass inflow rates , which in turn might be correlated with activity in the galactic nucleus ( agn ) , ranging from quasar , seyfert , liner to quiescent galaxies . however , not only is it challenging to ( observationally ) establish fueling mechanisms down to a few parsec from the center , time delays between changes in the mass inflow rate and the onset of nuclear activity further complicate linking them . it is also likely that multiple fueling mechanisms are important and act together . large - scale bars are efficient at transporting gas inward ( e.g. * ? ? ? * ) , but the presence of an inner lindblad resonance ( ilr ) will cause the gas to pile up in a nuclear ring , often clearly visible due to the intense star formation ( see e.g. figure [ fig : images ] ) . the gas might continue further inward through nested bars ( e.g. * ? ? ? * ; * ? ? ? * ) , although dynamical constraints on a double - barred system may prohibit inflow down to the center @xcite . moreover , observational evidence of nuclear bars is scarce ( e.g. * ? ? ? * ; * ? ? ? * ) , but should be interpreted with care since the prominent dust lanes in the main bar might be absent in secondary bars ( e.g. * ? ? ? on the other hand , nuclear spirals seem to be commonly observed , both in active and quiescent galaxies ( e.g. * ? ? ? * ) , ranging from flocculent to grand - design nuclear spirals ( e.g. * ? ? ? whereas the former are suggested to form by acoustic instabilities ( e.g. * ? ? ? * ) , the grand - design nuclear spirals are thought to be the result of gas density waves ( e.g. * ? ? ? * ) or shocks ( e.g. * ? ? ? * ; * ? ? ? * ) induced by the non - axisymmetric gravitational potential of a large - scale bar . the latter two might be connected in the sense that the bar - driven spiral shocks trigger the gas density waves throughout the disk @xcite , which in turn seem to be necessary for the nuclear spiral to be long - lived @xcite . the inward extended gas inflow through a nuclear spiral not only depends on the torque of the large - scale bar , but also the gas having high enough sound speed so as to loose angular momentum ( e.g. * ? ? ? * ; * ? ? ? * ) , as well as the presence of a central mass concentration such as a super - massive bh to overcome a closer - in ( inner ) ilr ( e.g. * ? ? ? * ; * ? ? ? the deviations in the gas density due to nuclear spirals are typically only a few per cent @xcite , which makes direct imaging very difficult . instead , the obscuration due to dust thought to be associated with the gas overdensities is often employed , but when the extinction is small or the dust not well - mixed with the gas , it can lead to unclear or even missed detections of nuclear spirals . at the same time , nuclear spirals induce non - circular motions in the gas , with resulting deviations in the observed velocity field that can be a significant fraction of the underlying circular velocity . if the gas is ionized , the kinematics might be inferred from emission lines , which in general can be detected more easily and at higher spatial resolution in the optical , than can be tracers of the molecular or atomic gas such as co and hi at radio wavelengths . moreover , while imaging typically only yields a detection of the nuclear spiral , the observed non - circular motions in combination with the gas ( over)density inferred from the simultaneously measured ( line ) fluxes might be used to derive an estimate of the gas mass in / out - flow rate . an alternative approach to get a handle on the gas flow rates is to compute the gravitational torques from the observed surface brightness using a mass - to - light ratio conversion calibrated against ( circular ) velocity measurements @xcite , or against stellar population models fitted to color measurements @xcite . in these cases , various assumptions are made , most importantly that stellar light is a clean ( once corrected for dust obscuration ) and direct ( mass follows light ) tracer of the underlying gravitational potential . even so , this approach shows that gravitational torques due to non - axisymmetric structures are very efficient at transporting gas inward while overcoming dynamical barriers such as the corotation resonance . however , when the resulting predicted mass inflow rates as functions of radius are compared with the non - circular motions in the observed gas velocity fields the correlation between both is often not evident @xcite . the main difficulty with the non - circular motions is to identify the fraction that is due to pure radial flow . gas on closed elliptic orbits not only contributes to non - circular motions ( or elliptic streaming ) in the azimuthal direction but also in the radial direction @xcite . however , since closed elliptic orbits are applicable strictly only to collisionless ( stellar ) orbits , we expect that as a result of shocks and other dissipational effects the angular momentum of the gas will change , leading to in / out - flow . while the measured non - circular motions are often taken to be directly representative of the radial flow velocities ( e.g. * ? ? ? * ) , only a fraction of them are expected to be truly radial in / out - flows . the goal of this paper is two - fold : ( i ) to show that harmonic expansion of gas velocity fields provides a clean way to detect nuclear spirals and to estimate their pitch angles ; ( ii ) to use simple perturbation models to constrain the fraction of measured non - circular motions that is due to radial flow , and to estimate the corresponding mass inflow rate . in section [ sec : harmanalysis ] we present the harmonic analysis , and in section [ sec : ngc1097 ] we apply it to the observed emission - line velocity field within the circumnuclear starforming ring of ngc1097 . we find that the three - arm spiral structure in the non - circular motions is consistent with a weak two - arm spiral perturbation of the gravitational potential , and use the model derived in appendix [ app : gasorbits ] to constrain the inflow rate as function of distance from the center of ngc1097 . in section [ sec : discussion ] we discuss the corresponding spiral distortion in the surface brightness , and possible additional non - circular motion contributions to the velocity field . finally , we link the estimated mass inflow rate to the accretion onto the central bh of ngc109 . we summarize and draw our conclusions in section [ sec : summaryandconcl ] . maps of the surface brightness ( sb ) , line - of - sight velocity ( @xmath7 ) , velocity dispersion ( @xmath8 ) , and higher order velocity moments ( often expressed in term of gauss - hermite moments @xmath9 , @xmath10 , ) of nearby galaxies generally show organized , periodic features , which can be suitably studied by means of a harmonic expansion . we start by dividing a map into a number of elliptic annuli , each with a different semi - major axis length @xmath11 , but we assume all have the same flattening @xmath12 and share the same angle @xmath13 and center @xmath14 , so that in terms of cartesian coordinates on the map ( i.e. , on the plane of the sky ) @xmath15 if the map is aligned such that @xmath16 is pointing north and @xmath17 pointing west , then @xmath18 , where @xmath19 is the common ( observational ) definition of the position angle of the major axis of the galaxy , measured from north through east . next , we extract the profiles along each of the annuli and describe them by a finite number of harmonic terms @xmath20 , @xmath21 there are various ways to obtain the set of best - fit ellipses , but given the highly non - linear nature of the optimization for the best - fit parameters , this is commonly done in a stepwise , iterative process . we follow a similar approach as described in @xcite based on the usually adapted procedure in photometry @xcite which we briefly summarize . for ( early - type ) galaxies , the maps of sb , @xmath8 and _ even _ higher - order velocity moments ( @xmath10 , @xmath22 , ) each are to first order well approximated by a function that is constant along ( concentric , equally flattened and orientated ) ellipses , and only varies as function of the semi - major axis length of these ellipses . similarly , @xmath7 and _ odd _ higher - order velocity moments ( @xmath9 , @xmath23 , ) each are well described by a similar function , but one that has an additional cosine variation along the ellipses . in these cases , the harmonic expansion in equation truncates after the third - order term ( @xmath24 ) . the parameters of the best - sampling ellipses for even velocity moments , can then be obtained by minimizing @xmath25 , and without @xmath26 for odd velocity moments . we first perform the minimization for a grid of ( fixed ) @xmath12 and @xmath13 values , where we might use external constraints such as the inclination @xmath27 and measured position angle @xmath19 together with an initial estimate of the center @xmath14 . next , starting from the best - fit grid pair , we optimize for all four parameters . we sample the semi - major axis lengths as @xmath28 $ ] for @xmath29 , with the initial @xmath30 depending on the spatial resolution ; for the geometric increase factor we take @xmath31 . we achieve this sampling , together with uniform sampling in @xmath32 , by bilinear interpolation of the observed map . when fitting the observed line - of - sight velocity field shown in figure [ fig : hardec_maps ] , we assume all motions are within the equatorial plane . we divide all harmonic terms by @xmath33 to take into account the projection effect . for further details , including error estimates , see @xcite . once we have obtained the set of best - fitting ellipses , we fit the harmonic expansion of equation to each of the corresponding profiles to obtain the harmonic terms @xmath34 and @xmath35 ( up to higher order @xmath36 ) , as functions of @xmath11 . the difference with the above expansion up to and including @xmath24 , reveals the deviations from the latter smooth model : for example , boxiness and diskiness in the sb map , and non - circular motion in the residual @xmath7 map . in the case of a spiral structure as observed in ngc1097 , both through the dust in the sb and in the residual @xmath7 map @xcite , the deviations are more naturally described as a ( radially varying ) offset in the angle @xmath32 , than through a complex combination of variations in the amplitudes @xmath34 and @xmath35 . henceforth , we rewrite the harmonic expansion of equation ( in a mathematically equivalent expression ) as @xmath37\right),\ ] ] where the amplitudes @xmath38 and phase shifts @xmath39 are related to the coefficients @xmath34 and @xmath35 ( @xmath40 ) by @xmath41 for example , the @xmath7 map of ngc1097 in figure [ fig : hardec_maps ] reveals , after subtracting the best - fit circular motion @xmath42 ( with @xmath43 , since we set @xmath44 ) , what seems to be a three - arm spiral structure , @xmath45)$ ] . here , @xmath46 traces the spiral arms as function of radius @xmath11 . an axisymmetric gravitational potential in a frame that co - rotates with a weak perturbation of harmonic number @xmath47 can be written in polar coordinates as @xmath48\right),\ ] ] where @xmath49 is the phase of the perturbation . through poisson s equation it follows that the corresponding surface mass density exhibits a harmonic @xmath47 distortion . to derive the line - of - sight velocity , we follow @xcite . as described in appendix [ app : gasorbits ] , we extend their collisionless analysis by including radial damping in the equations of motion to take into account the dissipative nature of gas . we assume that the gas moves on closed loop orbits in the equatorial plane , which we observe at an inclination @xmath27 away from its normal and at an ( azimuthal ) angle @xmath50 . given a point @xmath51 in the equatorial plane , the projection of the azimuthal and radial velocity onto the corresponding line - of - sight yields @xmath52,\ ] ] where @xmath53 is zero on the line of nodes ( see also fig . 1 of * ? ? ? to first order the solutions of the equations of motion in the perturbed gravitational potential of equation yield @xmath54 , \\ \label{eq : solvphi } v_\phi(r,\psi ) & = & v_c(r ) \left [ 1 + c_\phi \cos m\psi + s_\phi \sin m\psi\right],\end{aligned}\ ] ] where @xmath55 is the circular velocity and @xmath56 , @xmath57 , @xmath58 and @xmath59 are functions of @xmath11 given in appendix [ app : gasorbits ] . substituting these solutions into equation , we obtain @xmath60 with @xmath61 and @xmath62 , and @xmath63 the circular velocity in projection . we thus find , as concluded before by @xcite and already qualitatively inferred by @xcite that , if the gravitational potential has a perturbation of harmonic number @xmath47 , the line - of - sight velocity field contains an @xmath64 and an @xmath65 harmonic term . how loosely or tightly wound a spiral is can be quantified via its pitch angle , @xmath66 , which , at a given radius , @xmath11 , measures the angle between the tangent of the spiral arm and a circle with radius @xmath11 in the plane of the disk . inverting the phase - shift , @xmath67 , of a spiral , we can parameterize an arm of the spiral as @xmath68 one can then show that the pitch angle @xmath66 is given by @xmath69 which is positive ( negative ) if the spiral curves anti - clockwise ( clockwise ) with increasing radius . the smaller the pitch angle , the more tightly the spiral is wound , with @xmath70 a circle , while @xmath71 corresponds to a straight line . a specific case that is often encountered in nature is that of a logarithmic spiral @xmath72 with constants @xmath73 and @xmath74 . from equation , we find the well - known property that the logarithmic spiral has a constant pitch angle @xmath75 . in the case of an @xmath76-spiral in the line - of - sight velocity field , the corresponding pitch angle , @xmath77 , follows directly from the phase - shift , @xmath78 , in the harmonic expansion in equation . the pitch angles @xmath79 and @xmath80 generally take different values and bracket the pitch angle @xmath81 of the @xmath47-spiral perturbation in the gravitational potential that caused them . to show this , we start from equations and in appendix [ app : gasorbits ] and rewrite the coefficients @xmath82 and @xmath83 as @xmath84 so that after substitution into equation we obtain @xmath85 where @xmath86 . this links the phase - shift @xmath87 of the @xmath47-spiral perturbation in the gravitational potential with the phase - shifts @xmath78 of the @xmath76-spirals in the line - of - sight velocity field . the corresponding pitch angles are related as @xmath88 in general , @xmath89 and @xmath90 depend in a rather complex way on the gravitational potential , but as we show in appendix [ app : gasorbits ] they possess some generic properties . first , the amplitude @xmath91 is larger ( smaller ) than the amplitude @xmath92 outside ( inside ) the corotation radius @xmath93 , and equal to it at @xmath93 . this implies a transition in the line - of - sight velocity field at @xmath93 , going from the @xmath94-spiral dominating outside @xmath93 to the @xmath95-spiral dominating inside @xmath93 ( see also figure [ fig : model_pl ] ) . this is also concluded by @xcite for the collisionless case and earlier by @xcite for the less general case of a tightly wound spiral in the linear density - wave theory ( see also * ? ? ? second , @xmath90 typically varies much less with radius than the spiral phase - shifts , so that @xmath96 in equation is relatively small . as a result , we can estimate the pitch angle @xmath81 of the @xmath47-spiral perturbation in the gravitational potential from the pitch angles @xmath79 and/or @xmath80 of the @xmath95-spiral and @xmath94-spiral in the observed line - of - sight velocity field , without constructing a full dynamical model ( see also the last two panels of figure [ fig : model_pl ] ) . nuclear spirals in principle provide a mechanism to transport gas from kpc scales , where it often stalls inside a nuclear ring , into the center of the galaxy . still , as mentioned in section [ sec : intro ] , if a nuclear spiral ( or a nuclear bar ) is due to gas moving on closed elliptic orbits this results in non - circular motions , also referred to as elliptic streaming , but not necessarily in net inflow towards and/or outflow away from the center . however , unlike stars , gas is not collisionless and its orbits interact and ex - change angular momentum leading to net radial flows ( e.g. * ? ? ? henceforth , in the analytic models in appendix [ app : gasorbits ] , we assume a weak perturbation in the gravitational potential causing gas to deviate from circular onto elliptic orbits , while taking into account its dissipative nature via radial damping . these analytic spiral models are an extension of the analytic bar models introduced by @xcite , who showed that they describe well the gas behavior seen in hydrodynamical simulations . the radial damping causes the gas to lose / gain angular momentum inside / outside the corotation radius , which nicely matches the angular momentum transfer due to the torque from the bar potential in numerical simulations . the amount of radial damping , controlled through the dimensionless parameter is the same as the dimensionless parameter @xmath97 in @xcite , while we define @xmath98 . ] , @xmath99 , thus provides a handle on the amount of net radial flow that is needed to explain the observed non - circular motions in addition to elliptic streaming . in this way , a rather straightforward estimate of the net radial flow velocity , @xmath100 , can be obtained by comparing the radial velocity @xmath101 ( equation [ eq : appsolvr ] ) of the analytic model that includes radial damping ( @xmath102 ) with the analytic model without radial damping ( @xmath103 ) . at a given radius @xmath11 , the maximum radial velocity is given by @xmath104 here , @xmath105 and @xmath106 only depend on the weak perturbation as given in equation , while @xmath107 are functions of the angular frequency @xmath108 and the epicycle frequency @xmath109 of the axisymmetric gravitational potential , as well as the harmonic number @xmath47 and pattern speed @xmath110 of the weak perturbation . since without radial damping ( @xmath111 ) all radial motion is due to elliptic streaming , we subscribe a fraction @xmath112 of the radial velocity to elliptic streaming , leaving as an estimate of the radial flow velocity @xmath113 even though the analytic models neglect possible non - linear effects , they can capture most features of observed non - circular motions ( as we show next in the case of ngc1097 ) , and at the same time provide an estimate of the fraction of the observed non - circular motions that is due to net radial flow in addition to elliptic streaming . note that radial flow here does not mean that the gas is following pure radial orbits with zero angular momentum , which would contribute only to the harmonic term @xmath114 ( e.g. * ? ? ? * ) . instead , the gas is expected to gradually spiral inward / outward as it has both azimuthal and radial velocity components . in case the angular momentum loss / gain is driven by a weak gravitational potential perturbation with harmonic number @xmath47 , this results in a contribution to @xmath82 and @xmath83 . we apply the above harmonic analysis to the observed emission - line velocity field within the circumnuclear starforming ring of ngc1097 . we recover in the non - circular motions a spiral structure and infer its pitch angle directly from the harmonic components . next , we use a spiral perturbation model to constrain the radial inflow velocity and combine this with the gas density in the nuclear spiral to estimate the mass inflow rate as function of distance from the center of ngc1097 . ngc1097 ( eso32558 ) is a nearby ( distance 14.5mpc , so @xmath115pc ) liner / seyfert1 host with a strong , @xmath116kpc long , bar and a @xmath117kpc in radius circumnuclear starforming ring ( figure [ fig : images ] ) . in this sb galaxy dust can be traced within the large - scale spiral arms out to a ( outer lindblad resonance , olr ) radius of @xmath118kpc , in prominent lanes along the bar , and continuing within the nuclear ring as a spiral structure down to @xmath119pc from the center @xcite . non - circular motions associated with this nuclear spiral structure @xcite , indicate a possible mechanism to drive gas from kpc scales down to a few pc from the center , where a double - peaked broad h@xmath120 emission profile @xcite indicates the presence of a super - massive bh . @xcite describe in detail the observations and reduction of the two - dimensional spectroscopy of ngc1097 obtained with the gmos - ifuon the gemini south telescope ( gs-2004b - q-25 , pi : storchi - bergmann ) . three pointings within the nuclear ring region ( figure [ fig : images ] ) provided 1500 individual spectra covering 56007000 at a velocity resolution of 85kms@xmath4 and with a spatial sampling of 01 . in the first panel of figure [ fig : hardec_maps ] , we present the [ nii]emission - line velocity field covering the inner @xmath121kpc . a systemic velocity of @xmath122kms@xmath4 ( equivalent to a constant @xmath123 term ) has been subtracted . for an inclination @xmath124@xmath1 ( flattening @xmath125 ; * ) , and position angle of @xmath126@xmath1 , the other panels show the results of applying the harmonic expansion of section [ sec : expansion ] . from left to right : the best - fit circular motion @xmath127 , the remaining non - circular motion , and the harmonic reconstruction from the sum of the latter two . in figure [ fig : hardec_profiles ] , the first two panels show the adopted position angle and systemic velocity . the next four panels show the coefficients @xmath34 and @xmath35 ( in kms@xmath4 ) for the first three harmonic terms , as function of the ( deprojected ) radius @xmath11 . the bottom three panels , show the phase shifts @xmath39 ( in degrees ) for the alternative formulation of the harmonic expansion given in equation , as functions of the ( natural ) logarithm of @xmath11 . the apparent three - arm spiral structure in the non - circular motion ( third panel in figure [ fig : hardec_maps ] ) is confirmed by the significant amplitudes of @xmath128 and @xmath129 , and in particular by the the smooth variation of the corresponding phase - shift @xmath130 with radius . excluding the uncertain measurements within @xmath131 ( vertical dotted line ) and the single measurement at the edge of the map , the remaining measurements indicated by solid circles show a clear linear relation between @xmath132 and @xmath133 . this is consistent with a logarithmic spiral defined in equation , with fitted @xmath134 and @xmath135 . the slope provides a direct and robust measurement of the pitch angle of @xmath136@xmath1 . this logarithmic spiral , overplotted in the third panel of figure [ fig : hardec_maps ] , indeed closely traces the three - arm spiral structure in the non - circular motions ; possible reasons for the deviations are discussed below in section [ sec : discussion ] . the positive sign of the pitch angle indicates the nuclear spiral is curved anti - clockwise with increasing radius , equivalent to the spiral arms extending outwards of the large - scale bar , as can be seen in the left panel of figure [ fig : images ] . since the global rotation in ngc1097 is clockwise from the observed velocity field , and the north - east is the `` far side '' from being more obscured , it follows that both the large - scale and nuclear spirals are trailing . the above three - arm spiral structure in the non - circular motions is consistent with a perturbation in the gravitational potential due to a @xmath137 harmonic spiral ( see also section [ sec : potentialpertubation ] ) . from equation it then follows that the pitch angle @xmath138 of this two - arm spiral perturbation in the gravitational potential follows from the measured pitch angle @xmath139 of the three - arm spiral in the velocity field as @xmath140 . we show below that , as expected , the later term is small , so that for a measured @xmath141@xmath1 , we calculate @xmath142@xmath1 . as illustrated in figure [ fig : model_pl ] of appendix [ app : gasorbits ] , the true value of @xmath138 might differ slightly depending on the details of the gravitational potential perturbation . we construct a model for the nuclear spiral in ngc1097 to get a handle on the latter difference as well as to constrain the fraction of the observed non - circular motions that is due to net radial inflow in addition to elliptical streaming . we use the analytic solutions of appendix [ app : gasorbits ] for gaseous orbits in an axisymmetric gravitational potential , @xmath143 , that is weakly perturbed by a logarithmic @xmath137 spiral with pitch angle @xmath138 . note that these analytic spiral models are based on linearized equations of motion under the epicycle approximation , but are not restricted to tightly wound spirals ( e.g. * ? ? ? * ; * ? ? ? * ) . in this way , we show below that the observed loosely wound nuclear spiral in ngc1097 is still consistent with a density wave , and not necessarily driven by shocks as suggested by @xcite . we adopt the power - law model @xcite with axisymmetric gravitational potential @xmath144 & \text{$\beta \ne 0 $ } , \\ v_0 ^ 2 \ , \frac{1}{2 } \left [ 1 - \ln\left(1+\frac{r^2}{r_c^2}\right ) \right ] & \text{$\beta = 0 $ } , \end{cases}\ ] ] and corresponding circular velocity @xmath145 so that @xmath146 at the core radius @xmath147 . the parameter @xmath148 controls the logarithmic gradient of the rotation curve at large radii : @xmath149 rising , @xmath150 flat , and @xmath151 falling . the three gravitational potential parameters @xmath152 , @xmath147 and @xmath148 are set by comparing the corresponding circular velocity @xmath153 with the measured radial profile of @xmath26 , taking into account the non - circular contribution due to the @xmath137 spiral perturbation . in the third panel of figure [ fig : hardec_profiles ] , the black dashed curve shows @xmath153 for @xmath154kms@xmath4 , @xmath155 , and @xmath156 . except for the three values within @xmath157 , this simple power - law model is a good representation of the @xmath26 measurements , once the non - circular contribution derived below is added , as indicated by the blue dashed curve . the advantage of such a simple analytic representation of the gravitational potential is that it makes all subsequent calculations concerning the perturbation very convenient . for the amplitude of the gravitational potential perturbation we assume @xmath158 , with constant strength @xmath159 , while the phase - shift is given by @xmath160 . here , @xmath161 from the above fit to the @xmath162 harmonic terms in the line - of - sight velocity field , while the corresponding approximation @xmath163@xmath1 , is taken as the initial value for the pitch angle . the three additional free parameters are the pattern speed , @xmath110 , of the spiral perturbation , the azimuthal viewing angle , @xmath50 , and finally the amount of radial damping , @xmath99 . equations and then provide predictions for the non - circular motion contribution in terms of the harmonic coefficients @xmath26 , @xmath114 , @xmath128 , and @xmath129 , which we compare with the corresponding measured radial profiles for ngc1097 in figure [ fig : hardec_profiles ] . the flattening of the hi rotation curve in the outer parts of ngc1097 @xcite implies a nearly constant circular velocity @xmath164kms@xmath4 . with the angular and epicycle frequencies approximately given by @xmath165 , the olr being at @xmath166kpc yields a pattern speed @xmath167kms@xmath4kpc@xmath4 the latter places the corotation radius , @xmath168kpc , or @xmath169% beyond the extent of the large - scale bar , consistent with numerical simulations of `` fast bars '' ( e.g. * ? ? ? * ; * ? ? ? * ) , and measured pattern speeds in similar galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? assuming that the two - arm nuclear spiral as a gas density wave is being driven by the large - scale bar @xcite , we adopt the same value for the pattern speed of the perturbation . since we consider the harmonic coefficients well within the corotation radius , they are not sensitive to , and hence do not constrain @xmath110 . on the contrary , a small change in the azimuthal viewing angle already causes a significant radial offset in the harmonic coefficients , so that we need @xmath170 . next , too little radial damping results in too small amplitudes of the coefficients @xmath128 and @xmath129 with respect to @xmath114 and the non - circular contribution to @xmath26 . specifically , since @xmath114 and @xmath129 are of similar amplitude and shape over most of the radial range , significant radial damping with @xmath171 is needed . as described in section [ sec : radialflowvelocity ] and discussed below in section [ sec : massinflowrate ] , this implies that net radial flow makes up most of the intrinsic radial velocity . matching the amplitudes of the harmonics terms yields a strength of the gravitational potential perturbation of @xmath172 . finally , values for the pitch angle of the spiral perturbation that are in the range of @xmath142@xmath1 , approximated above from the pitch angle @xmath139 of the three - arm spiral in the velocity field , yield predictions for the harmonic coefficients that are consistent with the measured harmonic coefficients . the effect of the additional term @xmath173 is indeed small , so that @xmath174 provides a robust measurement of @xmath138 . this is further illustrated in figure [ fig : hardec_profiles ] , where in addition to the measured harmonic coefficients , we show with dashed curves the predictions of the above two - arm spiral perturbation model with pitch angle @xmath163@xmath1 , azimuthal viewing angle @xmath175@xmath1 , and radial damping parameter @xmath176 . both @xmath177 and @xmath178 , shown as dot - dashed curves in respectively the bottom - left and bottom - right panel , indeed vary only little with radius . similarly , the dashed curve in the bottom - right panel shows the predicted relation between @xmath132 and @xmath133 while taking into account @xmath179 , which is nearly indistinguishable from the fitted solid line . the predicted harmonic terms match well the @xmath128 and @xmath129 measurements in the middle - right panel , and rather well the @xmath114 measurements in the middle - left panel , except for those at larger radii . however , the latter measurements are quite uncertain , and might well be disturbed by the nuclear ring . the blue dashed curve in the top - right panel is the spiral model prediction for @xmath26 , which apart from the innermost measurements , nicely traces the measured rotation curve . due to the significant non - circular motion contribution the latter is different from the circular velocity of the ( power - law ) axisymmetric gravitational potential shown as the black dashed curve . the corresponding @xmath180 black dot - dashed curve is still well above the pattern speed @xmath181kms@xmath4kpc@xmath4 ( solid horizontal line ) at radius @xmath182 , which places the ilr well beyond the nuclear ring radius of @xmath117kpc . if indeed the radius of nuclear rings is set by the location of the ilr ( e.g. * ? ? ? * ) , this indicates that the nuclear ring in ngc1097 has migrated inward . inward migration has also been suggested for the nuclear ring in ngc4314 @xcite , has been seen in hydrodynamic simulations ( e.g. * ? ? ? * ; * ? ? ? * ) , and might be a consequence of shepherding of the gas ring by star clusters that formed in it @xcite . we found that a spiral perturbation model with only elliptic streaming , in which gas is moving on closed elliptic orbits without radial damping , can not explain the observed non - circular motions in ngc1097 . in section [ sec : radialflowvelocity ] , we showed that the amount of radial damping required provides an estimate of the radial flow velocity @xmath100 in terms of the radial velocity @xmath101 of the spiral model which matches the observed harmonic terms . since in the case of ngc1097 a large amount of radial damping @xmath171 is needed , the ratio @xmath183 with the dashed curves , the corresponding radial inflow velocity @xmath184 is shown in the top - left panel of figure [ fig : mass_inflow_rate ] with a solid curve . even though nearly identical to the radial velocity @xmath101 , it still is only a fraction of the total non - circular motion ( dashed curve ) , which also includes the azimuthal velocity . to turn this radial inflow velocity into a constraint on the mass inflow rate , we furthermore need to know the gas density in the nuclear spiral , as well as the geometry or the area through which the gas is flowing . the flux ratio of the emission - line doublet included in the spectral range of the gmos - ifu observations allows us to constrain the mean electron density @xmath185 over the observed field . taking all values together yields an average flux ratio of @xmath186 , which , adopting the prescription by @xcite and using the atomic parameters compiled by @xcite and @xcite , corresponds to @xmath187@xmath188assuming a mean electron temperature of @xmath189k . even though the flux ratio map might indicate a spiral structure similar to that in the non - circular motions , the signal - to - noise is not high enough to quantify the gas ( over)density in the spiral arms from it . instead , we apply the same elliptic annuli used to extract the harmonic coefficients to the flux ratio map . from the distribution within each annulus , we derive @xmath185 as a function of radius , resulting in the dashed curve in the top - right panel of figure [ fig : mass_inflow_rate ] ( divided by 10 for illustrative purposes ) . next , we assume that the width of the distribution , indicated by the solid curve , is driven by the density wave contrast , and hence yields an estimate of the electron ( over)density @xmath190 within the spiral arms . assuming a @xmath191% lower ( higher ) electron temperate of @xmath192k , leads to an increase ( decrease ) in the electron densities by a factor of about 20% . in all cases , the density contrast is of the order of @xmath193% , in agreement with hydrodynamic simulations ( e.g. * ? ? ? * ; * ? ? ? * ) and @xmath194-band imaging presented by @xcite . using the proton mass and a factor 1.36 to account for the presence of helium , we convert @xmath190 to a gas mass overdensity @xmath195 . for the geometry we assume in agreement with the spiral perturbation model that both the radial flow velocity and gas mass overdensity , at each radius @xmath11 in a disk with scale height @xmath196 , vary as a sinusoidal function in azimuthal angle @xmath32 . integrating over the @xmath32 values for which the radial flow velocity is positive , the mass inflow rate then reduces to @xmath197 for gas of density @xmath195 flowing in the @xmath47 spiral arms towards the center at a velocity @xmath184 . to estimate the scale height @xmath196 , we start from a marginally stable disk with toomre s ( @xcite ) @xmath198 , and substitute in @xmath199 a constant sound speed @xmath200kms@xmath4 , the epicycle frequency @xmath201 of the power - law axisymmetric gravitational potential , and @xmath202 , with @xmath203 inferred from the average electron density @xmath185 . the dashed curve in the bottom - left panel of figure [ fig : mass_inflow_rate ] shows @xmath204 as function of radius . clearly , towards the center this leads to an unrealistically high scale height , so that we constrain @xmath204 to be not larger than the relative width @xmath205 of a spiral arm , resulting in the solid curve . finally , in the bottom - right panel of figure [ fig : mass_inflow_rate ] , we present the mass inflow rate @xmath206 ( in m@xmath3yr@xmath4 ) as function of radius . we have argued for a two - arm spiral perturbation in the gravitational potential as the source of the three - arm spiral structure in the velocity field of ngc1097 . we verify the implied two - arm spiral distortion in the surface brightness , and discuss possible causes for contributions from additional harmonic terms . finally , we link the mass inflow rate to the accretion onto the central bh . an @xmath137 spiral perturbation in the gravitational potential that would explain the three - arm spiral structure in the velocity field of ngc1097 , implies an @xmath137 spiral distortion in the surface mass density and hence a two - arm spiral structure in the surface brightness . even if the latter distortion is too weak to measure directly , it might show itself through obscuration by correlated dust . indeed , structure maps of ngc1097 ( e.g. * ? ? ? * ; * ? ? ? * ) show spiral - like features , but they are not evidently a two - arm nuclear spiral . @xcite note in high - resolution ` vlt`/naco infrared images a central spiral with a three - arm symmetry , although one of the three arms does not seem to continue towards the nuclear ring , but instead splits into a number of spiral filaments . on the other hand , @xcite use wavelet analysis on a ` hst`/wfpc2 image to reveal a clear two - arm spiral structure , but additional spiral features can not be ruled out . in the right panel of figure [ fig : images ] , we reproduce the wavelet map from ( * ? ? ? * their fig . 1 ) on top of a structure map from @xcite . the latter structure map is based on the richardson - lucy image restoration technique @xcite , using a multi - step convolution of the ` hst`/acs high resolution camera fr656n image with a two - dimensional psf model constructed using tiny tim @xcite . the solid magenta curves show the two - arm spiral with pitch angle @xmath163@xmath1 , which we predict based on the three - arm spiral in the velocity field ( section [ sec : spiralmodel ] ) . the two open arms trace well the spiral structures in both the wavelet and structure map , except closer to the nuclear ring when , at least in the northern part , a more tightly wound spiral seems needed . the dashed magenta curves indicate the two additional arms in case of an @xmath207 spiral perturbation with the same pitch angle . such higher - order _ even _ harmonic terms can result from non - linear coupling of modes , and although smaller in amplitude then the @xmath137 spiral perturbation might give rise to possibly additional spiral features . furthermore , a spiral perturbation driven by the large - scale bar is not necessarily restricted to an @xmath137 harmonic term by @xmath208 with @xmath209 , creates besides @xmath137 also higher - order even harmonic terms . ] . still , as long as the perturbation is bi - symmetric all resulting harmonic terms are even . only in case of an asymmetric perturbation , or lopsidedness , are contributions of @xmath210 or higher - order odd harmonic terms to the surface brightness possible . since the stars make up nearly all the mass in the center of ngc1097 , such a deviation from bi - symmetry , e.g. due to an eccentric nuclear disk as in m31 @xcite , or a dwarf elliptical companion , should reveal itself in the stellar kinematics . however , the central stellar kinematics of ngc1097 do not show such deviations nor any ( spiral ) structure in the non - circular motions @xcite , in support of the nuclear spiral being a perturbation in the gas . instead , @xcite propose that a difference in strength between the two principal shocks in the large - scale bar give rise to the lopsidedness . this only works in their scenario in which the shocks , and not the bar itself , create the nuclear spiral . however , while the shocks might be the trigger , it seems that density waves in the gas , driven by the bi - symmetric bar , are necessary for the nuclear spiral to be long - lived @xcite . alternatively , the typically asymmetric ( projected ) dust distribution might mimic odd harmonic terms in the surface brightness , and possible even give rise to even harmonic terms in the velocity field , as we discuss next . the coefficients @xmath34 and @xmath35 , that provide the phase - shift @xmath211 , may be `` contaminated '' by additional contributions to the non - circular motion , but in general they are not expected to result in a smooth variation of the phase - shift with radius , as seen for a spiral structure . an exception is the @xmath210 harmonic term , since the measured coefficient @xmath26 also incorporates the circular velocity as @xmath212 , which in general dominates over the non - circular motions @xmath213 and @xmath114 . this results in a phase - shift @xmath214 that is everywhere close to zero , and due to the degeneracy in @xmath26 there is little hope of constraining a two - arm spiral perturbation from its contribution to the @xmath210 harmonic term in the velocity field . however , we can predict its pitch angle @xmath215 and phase - shift @xmath216 from the ( logarithmic ) spiral inferred from the contribution to the @xmath217 harmonic term : @xmath218 and @xmath219 . as a result , we might use @xmath220 as an estimate of the non - circular contribution to @xmath26 . the effect is shown in the top - right panel of figure [ fig : hardec_profiles ] , where the black dashed curve is @xmath221 of the ( power - law ) axisymmetric potential , which after taking into account @xmath213 yields the blue dashed curve that matches the measured @xmath26 coefficients indicated by the blue diamonds . this provides a novel way to correct for non - circular motions , which otherwise might , for example , lead to an underestimation of the ( inner ) slope of the mass distribution ( e.g. * ? ? ? figure [ fig : hardec_profiles ] shows that , besides @xmath210 and @xmath217 harmonic terms in the velocity field expected from a @xmath137 perturbation in the gravitational potential , the non - circular motions also seem to contain an @xmath137 harmonic term . there are several effects that might ( partly ) cause this additional contribution . as shown by @xcite ( @xcite , their equation 7 ) , an error in the ( kinematic ) center results in a spurious @xmath222 and @xmath137 contribution to the line - of - sight velocity as @xmath223,\end{aligned}\ ] ] we see that the effect on @xmath224 and @xmath225 vanishes if @xmath226 , i.e. , when the circular velocity curve increases linearly with radius , @xmath227 . as in most galaxies , the latter is also the case in the inner region of ngc1097 , but still @xmath224 is significantly positive within the central @xmath228 . not surprisingly , varying the kinematic center ( in the process of finding the best - fit set of ellipses as described in section [ sec : expansion ] ) does not remove the @xmath137 harmonic contribution , but in contrast makes the overall fit worse . this makes a significant effect due to an incorrect center unlikely . based on the analysis in section [ sec : potentialpertubation ] , we expect a similar contribution of both @xmath222 and @xmath137 harmonic terms to the velocity field from an @xmath210 distortion of an axisymmetric distribution . as mentioned above in section [ sec : spiralmodel ] , it is unlikely that the gravitational potential itself is lopsided since the stars that dominate in mass do not show any such signature . also , the nuclear spiral itself is expected to be bi - symmetric if it is indeed a gas density wave driven by the large - scale bar . still , the phase - shift @xmath229 in the bottom - middle panel of figure [ fig : hardec_profiles ] seems to vary smoothly , and even close to linearly as a function of @xmath133 within the central @xmath230 where prominent dust features are present . moreover , the amplitude of the slope is similar to the linear relation of @xmath132 versus @xmath133 in the bottom - right panel , but with negative instead of positive sign . this is consistent with a logarithmic spiral with the same pitch angle but orientated clockwise instead of anti - clockwise , i.e. , leading instead of trailing . in principle , both leading and trailing nuclear spirals can exist , as shown by @xcite , but in their simulations the smaller leading spiral dissolves before the long - lived larger trailing spiral fully emerges . alternatively , the spiral - like contribution to the @xmath137 harmonic term , appearing like a `` negative image '' , might result from an asymmetric dust obscuration , mimicking a lopsided distortion . the often prominent dust lanes along the leading edges of bars in galaxies are associated with shocks in the gas streaming along the length of the bar ( e.g. * ? ? ? * ) , which in turn lead to velocity jumps across the dust lanes ( e.g. * ? ? ? moreover , numerical models of dust lanes ( e.g. * ? ? ? * ) as well as analytical models of diffuse disks ( e.g. * ? ? ? * ) show that dust extinction can have a significant effect on the velocity along the line - of - sight . henceforth , we expect the dust and possible shocks associated with the spiral features to distort the velocity field , but the modeling required to understand the specific effects on the non - circular motion is beyond the scope of this paper . the eddington accretion rate onto a central black hole @xmath231 adopting @xmath232 for the radiative efficiency , and a mass @xmath233m@xmath3 for the central black hole in ngc1097 based on the measured central stellar velocity dispersion of @xmath234kms@xmath4 @xcite substituted in the @xmath235 relation @xcite yields @xmath236m@xmath3yr@xmath4 . this value is indicated in the bottom - right panel of figure [ fig : mass_inflow_rate ] by the upper dashed horizontal line . the lower dotted horizontal line at @xmath237 is the approximate transition from a seyfert1 to a liner active galactic nucleus ( see for a review * ? ? ? * ) , and thought to be the rate below which mass accretion becomes radiatively inefficient ( see for reviews * ? ? ? * ; * ? ? ? whereas ngc1097 is typically classified as a liner galaxy , monitoring of the nucleus reveals evolution in its activity up into the seyfert1 regime @xcite . @xcite find that the observed optical to x - ray spectral energy distribution in the nucleus of ngc1097 is consistent with an inner radiatively inefficient accretion flow plus outer standard thin disk , with a mass accretion rate of @xmath238 . these observational constraints are fully consistent with our estimate of the mass inflow rate , down to @xmath239m@xmath3yr@xmath4(@xmath240 ) at a distance @xmath241pc ( @xmath242 ) from the center , where the gas kinematics are still accurately measured and well described by the two - arm spiral perturbation model . our mass inflow rate is significantly lower than @xmath243m@xmath3yr@xmath4 estimated by @xcite at a distance @xmath244pc . in the latter estimate , a non - circular motion of @xmath191kms@xmath4 is adopted for the inflow velocity , whereas we find from our spiral model that even though the contribution of the radial inflow dominates over elliptic streaming , it is still only a fraction ( @xmath245kms@xmath4 at @xmath246pc ) of the non - circular motions . moreover , the estimated electron density of @xmath247@xmath188 is similar to the average density we derive directly from emission line ratios , but more than an order of magnitude higher than the expected ( over)density in the spiral arms ( @xmath248@xmath188 at @xmath246pc ) . finally , instead of assuming a fixed opening angle of @xmath249@xmath1 , we incorporate the geometry from the spiral model , and allow for the scale height to vary , in order for the disk to remain marginally stable . in this way , we obtain an estimate of the mass inflow rate fully consistent with the expected mass accretion rate onto the central bh in ngc1097 , without having to invoke an unclear filling factor ( e.g. * ? ? ? even so , the mass inflow rate might be even further constrained , in particular by measuring emission line ratios at higher signal - to - noise to obtain a more accurate estimate of the ( over)density , as well as using other emission lines to break the degeneracy with temperature . even though the ratios of emission lines constrain the total gas density in the nuclear spiral , most likely only part of the gas is ionized . on the other hand , the high amount of radial damping needed for the spiral model to match the observed non - circular motions ( section [ sec : spiralmodel ] ) , indicates strong dissipation , possibly due to shocks that in turn can ionize a significant fraction of the gas . at the same time , it implies that this `` hot '' gas does not necessarily show the same kinematics as the underlying `` cold '' gas , as traced for example by hi and co measurements at radio wavelengths , like those presented recently @xcite . in the latter study , the projected molecular gas distribution in the center of ngc1097 shows after subtracting elliptical isophotes , a clear two - arm spiral structure ( their fig . 3 ) , consistent with a @xmath137 spiral perturbation in the gravitational potential . however , the corresponding velocity field after subtraction of an axisymmetric disk model ( their fig . 5 ) does not reveal a clear three - arm spiral structure , as seen in the non - circular motions of the ionized gas ( figure [ fig : hardec_maps ] ) . it is possible that the kinematics of the ionized gas better trace the nuclear spiral , whereas the corresponding three - arm spiral is weaker in the non - circular motions of the cold gas , and perhaps more contaminated by a @xmath137 contribution through asymmetric dust obscuration . we presented harmonic expansion of the line - of - sight velocity field as a suitable method to identify and quantify possible structures in the non - circular motions , including nuclear spirals . we showed that a weak perturbation in the gravitational potential of harmonic number @xmath47 , causes the surface brightness to also exhibit @xmath47 distortion , but leads to @xmath64 and @xmath65 harmonic terms in the velocity field . in the case of a @xmath47-arm spiral perturbation in the gravitational potential , we found that the corresponding @xmath94-arm and @xmath95-arm spirals in the velocity field are respectively less and more tightly wound , with pitch angles approximately related as @xmath250 . in appendix [ app : gasorbits ] , we derived an analytic perturbation model , which allows for a simple estimate of the fraction of the measured non - circular motions that is due to radial flow . we applied this method to the emission - line velocity field within the circumnuclear starforming ring of ngc1097 , obtained with the gmos - ifu spectrograph . the resulting non - circular motions reveal a _ three_-arm spiral structure , and from the corresponding @xmath217 harmonic terms , we measure a pitch angle @xmath136@xmath1 . we linked the @xmath210 and @xmath217 harmonic terms in the velocity field to a weak perturbation of the gravitational potential due to a two - arm nuclear spiral with an inferred pitch angle @xmath251@xmath1 . we showed that this agrees with a two - arm dust spiral in ( optical ) images of the center of ngc1097 , although additional spiral structure might be present as a result of higher - order even harmonic terms . we argued that it is unlikely that the presence of also @xmath137 harmonic terms in the velocity field is due to lopsidedness in the gravitational potential nor in the nuclear spiral itself , which we expect to be due to long - lived density waves in the gas driven by the bi - symmetric large - scale bar . instead , we postulated that an asymmetric dust obscuration mimics a lopsided distortion , and gives rise to the additional even harmonic terms in the velocity field . to match the measured odd harmonic terms in the velocity field of ngc1097 , a spiral perturbation model with a large amount of radial damping is required . this indicates both strong dissipation and , that , in addition to elliptic streaming , a significant fraction of the non - circular motions is due to radial inflow . we combined the inferred radial inflow velocity with the gas density in , and the geometry of , the spiral arms to estimate the mass inflow rate as function of radius . we calculated the gas density from the variation in the flux ratio of the emission - line doublet , and we incorporated in the geometry the scale - height assuming a marginally stable disk . the resulting mass inflow rate decreases to @xmath252m@xmath3yr@xmath4 at a distance of @xmath241pc from the center . we showed that this corresponds to @xmath253 , where the latter eddington accretion rate is onto a black hole in ngc1097 with a mass @xmath233m@xmath3 based on the central stellar velocity dispersion . this rate is fully consistent in general with the active galactic nucleus in ngc1097 varying between liner and seyfert1 , and in particular with previous mass accretion models fitted to the observed spectral energy distribution in its nucleus . we conclude that the line - of - sight velocity can provide not only a cleaner view on nuclear spirals than does the associated dust , but that the presented method also allows one to quantitatively study these possibly important links in fueling the centers of galaxies , including a handle on the mass inflow rate as a function of radius . it is a pleasure to thank paul wiita for discussions and a critical reading of the manuscript . gvdv acknowledges support provided by nasa through hubble fellowship grant hst - hf-01202.01-a awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa , under contract nas 5 - 26555 . kf is supported by the swedish research council ( vetenskapsrdet ) , and gratefully acknowledges the hospitality of the institute for advanced study , to which a visit contributed greatly in completing this project . in figure [ fig : images ] , the ` vlt`/vimos images ( left panel ) have been taken and pre - processed by european southern observatory ( eso ) paranal science operation astronomers , with additional image processing by henri boffin ( eso ) , and the ` vlt`/naco images ( middle panel ) is based on research published in the october issue of astronomical journal , vol . 130 , p. 1472 . we derive analytic solutions of gaseous orbits in a weakly perturbed gravitational potential ( for modeling of a strong perturbation see e.g. * ? ? ? we follow the treatment of @xcite ( @xcite , p. 146 - 148 ) for a weak harmonic perturbation . like @xcite , we include an additional phase - shift @xmath87 as in equation to accommodate for a spiral perturbation . since gas at the same spatial location has the same velocity , we look for closed loop orbits that do not intersect themselves . however , to take into account the dissipative nature of gas , we include , like @xcite , radial damping , so that the first - order equation of motion becomes @xmath254,\ ] ] where @xmath99 controls the amount of radial damping . the subscript zero refers to zeroth - order with constant @xmath255 and @xmath256 , with @xmath110 the pattern speed of the perturbation . we have introduced the epicycle frequency , @xmath257 , and angular frequency , @xmath258 , with @xmath259 the logarithmic slope of the circular velocity , @xmath260 , of the axisymmetric part of the gravitational potential . furthermore , @xmath261 , \quad b = \frac{\phi_m}{r^2 } m \cot\zeta_m , \ ] ] where we have introduced @xmath262 , and @xmath81 is the pitch angle as defined in equation . solving equation we find that the solution for closed loop orbits is @xmath263 , \\ \label{eq : appsolphi } \phi & = & \phi_0 + [ 2 ( a \sin\eta - b \cos\eta ) - \xi \sin\eta ] / [ m(1-\omega ) ] \\ \label{eq : appsolvr } v_r & = & v_c \left\ { m ( 1-\omega ) [ a \sin\eta - b \cos\eta ] \right\ } , \\ \label{eq : appsolvphi } v_\phi & = & v_c \left\ { 1 + ( 1+\alpha ) [ ( a-\xi ) \cos\eta + b \sin\eta ) \right\ } , \ ] ] with @xmath120 appearing in the last line because of a first - order conversion from guiding center @xmath264 to a point @xmath265 in the observed velocity field . moreover , @xmath266 where we have defined @xmath267 and @xmath98 . the orbit solutions have a singularity at the corotation radius where @xmath268 ( @xmath269 ) , because the adopted epicycle approximation breaks down . without radial damping ( @xmath270 ) , the ( collisionless ) orbit solutions also have singularities when @xmath271 , i.e. , at the lindblad resonances given by @xmath272 . next , we assume to first order @xmath273 and replace it by @xmath274 which is zero along the line of nodes . we also define @xmath275 , so that @xmath276 $ ] . this allows us to recast the above expressions for @xmath101 and @xmath277 in multiple angles of @xmath32 as in equations and , with @xmath278 , \\ \label{eq : appsolsr } s_r & = & m(1-\omega ) [ a\cos(m\varphi_m ) - b\sin(m\varphi_m ) ] , \\ \label{eq : appsolcphi } c_\phi & = & ( 1+\alpha)(a - \xi ) \cos(m\varphi_m ) - ( 1+\alpha ) b \sin(m\varphi_m ) . \\ \label{eq : appsolsphi } s_\phi & = & ( 1+\alpha)(a - \xi ) \sin(m\varphi_m ) + ( 1+\alpha ) b \cos(m\varphi_m ) , \ ] ] substitution in equation results in the expression for the line - of - sight velocity in equation , where the harmonic coefficients are given by @xmath279 with @xmath280 \ , a - ( 1+\alpha ) \xi \ , \right\ } , \qquad b_\pm = \frac12 { v_\star}\ , [ ( 1+\alpha ) \mp m(1-\omega ) ] \ , b.\ ] ] substituting @xmath281 and @xmath282 , with @xmath283 we arrive at the form given in equation . in figure [ fig : model_pl ] , we present an example of a weakly perturbed axisymmetric logarithmic potential defined in equation , with @xmath284kms@xmath4 , @xmath285kpc and @xmath150 . the amplitude of the perturbation is a factor @xmath286 times the axisymmetric logarithmic potential , while the angular dependence is due to a logarithmic spiral defined in equation , with pitch angle @xmath287@xmath1 . this results in non - circular motions which contribute a 1-arm and 3-arm spiral structure to the observed line - of - sight velocity field as shown in figure [ fig : model_pl ] for an adopted inclination of @xmath288@xmath1 . in general , @xmath89 and @xmath90 in equation above depend in a rather complicated way on the gravitational potential . however , in case of the linear spiral density - wave theory @xcite the expressions reduce to @xmath289 , \qquad \tan\theta_{m\pm1 } = \pm \cot\zeta_m,\ ] ] where the constant amplitude @xmath290 measures the strength of a _ tightly _ wound spiral without radial damping . in this case , @xmath291 and @xmath270 , so that @xmath292 , and @xmath293 in equation , and hence @xmath294 and @xmath295 . the expression for @xmath296 in equation is proportional to the reduced expression for @xmath89 in equation , since by substituting @xmath297 and @xmath298 , the term in square brackets in equation is proportional to @xmath299 $ ] in equation . , r. a. & dufour , r. j. 1994 , in astronomical society of the pacific conference series , vol . 61 , astronomical data analysis software and systems iii , ed . d. r. crabtree , r. j. hanisch , & j. barnes , 327 , s. , gebhardt , k. , bender , r. , bower , g. , dressler , a. , faber , s. m. , filippenko , a. v. , green , r. , grillmair , c. , ho , l. c. , kormendy , j. , lauer , t. r. , magorrian , j. , pinkney , j. , & richstone , d. 2002 , , 574 , 740 |
we develop a rigorous quantum mechanical theory for collisions of polyatomic molecular radicals with s - state atoms in the presence of an external magnetic field . the theory is based on a fully uncoupled space - fixed basis set representation of the multichannel scattering wavefunction . explicit expressions are presented for the matrix elements of the scattering hamiltonian for spin-1/2 and spin-1 polyatomic molecular radicals interacting with structureless targets . the theory is applied to calculate the cross sections and thermal rate constants for spin relaxation in low - temperature collisions of the prototypical organic molecule methylene [ ch with he atoms . to this end , two highly accurate three - dimensional potential energy surfaces ( pess ) of the he - ch complex are developed using the state - of - the - art ccsd(t ) method and large basis sets . both pess exhibit shallow minima and are weakly anisotropic . our calculations show that spin relaxation in collisions of ch , chd , and cd molecules with he atoms occurs at a much slower rate than elastic scattering over a large range of temperatures ( 1k 1 k ) and magnetic fields ( 0.01 1 t ) , suggesting excellent prospects for cryogenic helium buffer - gas cooling of ground - state-ch molecules in a magnetic trap . furthermore , we find that-ch undergoes collision - induced spin relaxation much more slowly than-ch , which indicates that magnetic trapping can be used to separate nuclear spin isomers of open - shell polyatomic molecules . | 1209.3071 | cooling and trapping polyatomic molecules is a promising research direction within the rapidly expanding field of cold molecular gases @xcite . because of their rich internal structure and coupled vibrational degrees of freedom , polyatomic molecules offer new research opportunities at the interface between chemistry , physics , combustion @xcite , atmospheric chemistry @xcite , and astrochemistry @xcite . studies of intramolecular energy redistribution in complex organic molecules yield insight into a broad range of phenomena , ranging from biochemical reactions involved in visual transduction @xcite to quantum decoherence in mesoscopic systems @xcite . unlike diatomic molecules , polyatomic molecules composed of four ( or more ) atoms can exist in the form of left and right - handed isomers and high - precision spectroscopy of cold molecular ensembles can provide insight into parity violation @xcite and hund s paradox @xcite . internal cooling increases the population of low - lying rovibrational states and hence the molecular response to external field perturbations , while cooling of external ( translational ) degrees of freedom enhances the molecule - field interaction time . high - precision spectroscopy of cold polyatomic molecules may thus enable ultrasensitive molecular detection and separation of different conformers @xcite as well as the development of novel time and frequency standards @xcite . while indirect cooling techniques such as photo - association and magneto - association are highly efficient at producing ultracold alkali - dimers @xcite , they can not be extended in a straightforward manner to cool polyatomic molecules due to the lack of efficient laser cooling methods for all but a handful of atomic species @xcite . direct cooling techniques such as sympathetic cooling @xcite , velocity filtering @xcite , rotating nozzle slowing @xcite , and stark deceleration @xcite are free from these difficulties , as demonstrated by recent experiments on the production of cold ensembles of fully deuterated ammonia @xcite , benzonitrile @xcite , formaldehyde @xcite , and naphthalene @xcite . slow beams of polyatomic molecules thus produced have already been used , in combination with electromagnetic traps , to explore the dynamics of oh + nd@xmath6 @xcite and li + sf@xmath7 @xcite collisions at low temperatures . however , quantum collision dynamics of polyatomic molecules in the presence of external electromagnetic fields remains poorly understood , necessitating the development of theoretical tools for _ ab initio _ modeling of these pioneering experiments as well as for the realization of efficient strategies for controlled molecular cooling @xcite . theoretical studies of polyatomic molecule collisions in the absence of external fields were pioneered by green @xcite , who developed a formal quantum theory of collisions involving asymmetric tops , and carried out scattering calculations on various problems of astrophysical interest , including he + hcn , he + h@xmath2o , and he + nh@xmath6 @xcite . many authors employed green s formalism to carry out accurate scattering calculations of rotational cooling rates for interstellar molecules colliding with atomic and molecular species based on accurate _ ab initio _ interaction potentials @xcite . several theoretical studies have used a combination of high - level _ ab initio _ and quantum scattering calculations to explore the dynamics of collision - induced rovibrational relaxation of methylene ( ch@xmath8 ) and methyl ( ch@xmath6 ) radicals in room - temperature he buffer gas @xcite , for which accurate rate measurements have recently become available @xcite . a number of related theoretical studies have focused on field - free collisions of h@xmath2o , co@xmath2 , and benzene molecules with he atoms at temperatures below 1 k @xcite . these calculations were motivated by the ongoing experiments on sympathetic cooling of optically decelerated @xcite polyatomic molecules with rare gas atoms @xcite . in particular , evidence was reported for slow rotational energy transfer in cold collisions of highly rotationally excited co@xmath2 molecules with he atoms @xcite . a recent classical trajectory study of he - naphthalene scattering @xcite has shown that the unexpected lack of cluster formation observed in buffer - gas cooling experiments on naphthalene @xcite can be attributed to the short lifetime of the he - naphthalene complex . we have recently presented a rigorous quantum scattering methodology for numerical calculations of collisional properties of polyatomic molecular radicals with @xmath9-state atoms in the presence of an external magnetic field @xcite . the method is based on the fully uncoupled space - fixed representation of the scattering wave function originally introduced by volpi and bohn @xcite and krems and dalgarno @xcite to investigate collisions of diatomic molecules in a magnetic field . in ref . we studied the quantum dynamics of spin relaxation in collisions of spin-1/2 and spin-1 polyatomic molecular radicals induced by collisions with he atoms . we found that many of these polyatomic molecules have favorable collisional properties ( that is , large ratios of elastic to inelastic collisions rates ) with he atoms , making them amenable to cryogenic buffer - gas cooling and possibly sympathetic cooling with @xmath9-state atoms in a magnetic trap @xcite . here , we present a full account of the theory and computational methodology outlined in our previous work @xcite . we begin with the derivation of coupled - channel quantum scattering equations for a polyatomic molecule colliding with an @xmath9-state atom in the presence of an external magnetic field ( sec . iib ) . in sec . iii , we use the newly developed scattering methodology to address the question of collisional stability of methylene radicals in cold @xmath10he gas in the presence of an external magnetic field . after a brief outline of the _ ab initio _ techniques employed to obtain the adiabatic pes for the he - ch@xmath2 complex ( sec . iiia ) , we present and discuss the main results for elastic and inelastic cross sections and thermal rate constants for spin relaxation in he + ch@xmath2 collisions ( sec . we then analyze the scaling properties of the inelastic cross sections and discuss the mechanism of spin relaxation of different nuclear spin isomers ( sec . iiic ) and isotopomers ( sec . iiid ) of ch@xmath2 . section iv summarizes the main results of this work . atomic units are used throughout unless stated otherwise . in this section , we will outline our approach to the solution of the atom polyatomic molecule collision problem in the presence of an external magnetic field . we first describe our _ ab initio _ calculations of the potential energy surfaces ( pess ) of the he - ch@xmath2 complex . in sections iib and iic , we generalize this work to collisions of polyatomic molecules ( asymmetric tops ) with @xmath9-state atoms . we present the derivation of the matrix elements required to parametrize the close - coupling ( cc ) scattering equations for collisions of polyatomic molecular radicals with @xmath9-state atoms . ii c presents a brief description of our numerical calculations . the hamiltonian for an open - shell polyatomic molecule colliding with a @xmath11 atom in the presence of an external magnetic field may be written as @xmath12 where @xmath3 is the reduced mass of the collision complex , @xmath13 is the atom - molecule separation vector of length @xmath14 , @xmath15 is the orbital angular momentum for the collision , @xmath16 is the anisotropic potential describing the atom - molecule interaction , and @xmath17 is the asymptotic hamiltonian describing the internal structure of the molecule and its interaction with external fields . in this work , we focus on collisions at low incident kinetic energies ( 10@xmath18 100 k ) , much lower than the typical energy scale for intramolecular electronic and vibrational excitations ( typically , hundreds and thousands of kelvin ) so we can assume that the molecule remains in its ground electronic and vibrational states during the collision . thus we can use the following effective hamiltonian @xcite to describe the rotational and fine structure of a polyatomic molecular radical like ch@xmath2 @xcite : @xmath19 the rotational part of this hamiltonian is that of a rigid asymmetric top with rotational constants @xmath20 , @xmath21 , and @xmath22 : @xmath23 , where the operators @xmath24 ( @xmath25 ) yield the molecule - fixed ( mf ) components of rotational angular momentum @xmath26 . the fine - structure hamiltonians @xmath27 and @xmath28 describe the spin - rotation and spin - spin interactions in polyatomic molecules with non - zero total spin @xmath9 ( radicals ) . the term @xmath29 describes the centrifugal distortion of rotational levels and the term @xmath30 describes the interaction of the molecular spin with an external magnetic field . here , we choose the mf axes to correspond to the symmetry axis of the molecule , as illustrated in fig . [ fig : mf ] . this choice is the same as adopted by hutson in his theoretical study of ar + h@xmath2o collisions @xcite but differs from the recent work of dagdigian and co - workers , where the principal axis of the triatomic is used to define the @xmath31 axis @xcite . the ch@xmath2 molecule lies in the @xmath32 plane , with the @xmath31-axis bisecting the hch angle , and the @xmath33-axis perpendicular to the molecular plane . with this choice , the he - ch@xmath2 interaction potential is symmetric under the reflection in the @xmath32 plane . the rotational part of the effective hamiltonian ( [ hmol ] ) becomes @xmath34 and the centrifugal distortion hamiltonian is ( up to fourth order ) @xcite @xmath35 the spin - rotation interaction is given by @xcite @xmath36 where @xmath37 is the electron spin , @xmath38 are the spin - rotation constants . ( [ hsr_bf ] ) gives the spin - rotation interaction in terms of the mf angular momentum operators @xmath26 and @xmath37 . the corresponding expression in terms of sf angular momentum operators is derived in appendix a. the spin - spin interaction may be written as @xmath39 \\ + \frac{3e - d}{\sqrt{6 } } \mathcal{d}^2_{p,0}(\hat{\omega } ) \biggr { ] } [ \hat{\bm{s}}\otimes\hat{\bm{s}}]^{(2)}_p,\ ] ] where @xmath40 are the wigner @xmath41-functions , @xmath42 are the euler angles which specify the orientation of mf axes in the space - fixed ( sf ) coordinate frame , and @xmath41 and @xmath43 are the zero - field - splitting parameters @xcite . equation ( [ hss_bf ] ) may be derived from the standard textbook expression for the spin - spin interaction in the mf frame @xcite as described in appendix a. note that the spin - spin interaction is only different from zero in molecular radicals bearing more than one unpaired electron ( @xmath44 ) . the interaction of the polyatomic molecule with an external magnetic field is described by the hamiltonian @xcite @xmath45 where @xmath46 is the magnetic field vector , @xmath47 is the bohr magneton , and @xmath48 is the @xmath49-tensor . following ref . @xcite , we take @xmath50 where @xmath51 is the unit matrix , so eq . ( [ hext ] ) reduces to a form commonly used for diatomic @xmath52 molecules @xmath53 where @xmath54 yields the projection of @xmath37 on the magnetic field axis . the rotational and nuclear spin zeeman interactions are much smaller than the @xmath30 term , and we neglect them in this work . table ii lists the molecular constants used in this work to parametrize the hamiltonian ( [ hmol ] ) for ch@xmath55 . to solve the quantum scattering problem for the hamiltonian ( [ h ] ) , we expand the wave function of the collision complex in direct products of internal basis functions @xmath56 and partial waves of relative motion @xmath57 in the space - fixed ( sf ) coordinate frame @xmath58 describe the rotational and spin degrees of freedom of the polyatomic molecule , @xmath59 is a symmetric - top eigenfunction , @xmath60 ( @xmath61 ) is the projection of @xmath26 on the sf ( mf ) quantization axes , and @xmath62 is the sf projection of @xmath37 . we note that the projection @xmath63 of the total angular momentum on the sf quantization axis is rigorously conserved for collisions in a magnetic field . as in the atom - diatom case considered previously @xcite this conservation can be exploited to factorize the collision problem into smaller problems which can be solved independently for each @xmath64 . using the expansion ( [ expansion ] ) in combination with the time - independent schrdinger equation for hamiltonian ( [ h ] ) leads to a system of close - coupling ( cc ) equations @xmath65 f_{\alpha lm_l}(r ) = 2 \mu \sum_{\alpha',l',m_l ' } \langle \alpha | \langle l m_l | v(r,\theta,\phi ) + \hat{h}_\text{mol}| \alpha ' \rangle | l ' m_l'\rangle f_{\alpha ' l'm_l'}(r).\ ] ] the matrix elements of the asymptotic hamiltonian @xmath17 and of the atom - molecule interaction potential @xmath66 are evaluated as described below . as in the atom - diatom case @xcite , the matrix of the asymptotic hamiltonian is not diagonal in the fully uncoupled sf representation . in order to properly apply the scattering boundary conditions , we introduce an asymptotic basis @xcite @xmath67 where @xmath68 are the asymptotic eigenvalues , which define scattering channels in the presence of an external magnetic field . the coefficients @xmath69 in eq . ( [ basis_gamma ] ) are independent of @xmath70 and @xmath71 and form the matrix @xmath72 , which satisfies @xmath73 , where @xmath74 is the diagonal matrix of eigenvalues @xmath68 . the asymptotic behavior of the solutions to cc equations ( [ cc ] ) in the asymptotic basis @xcite defines the elements of the scattering @xmath9-matrix @xmath75 as in the atom - diatom case @xcite . the cross sections for the collision - induced transition @xmath76 in the polyatomic molecule may be expressed through the @xmath9-matrix elements as @xmath77 where @xmath78 is the wavevector for channel @xmath79 , @xmath80 is the total energy , and @xmath81 is the collision energy . the matrix elements entering the cc equations ( [ cc ] ) can be expressed analytically via molecular constants and expansion coefficients of the atom - molecule interaction potential . in this section , we present the expressions for these matrix elements in the fully uncoupled sf basis of eq . ( [ expansion ] ) . we start with the asymptotic hamiltonian of the polyatomic molecule ( [ hmol ] ) . first , we note that since the asymptotic hamiltonian ( [ hmol ] ) is independent of the atom - molecule orientation , its matrix elements are independent of @xmath70 and @xmath71 : @xmath82 the matrix element on the right - hand side can be written as the sum of rotational ( [ hrot ] ) , centrifugal distortion ( [ hcd ] ) , fine - structure ( [ hsr_bf ] ) , ( [ hss_bf ] ) and magnetic field - dependent terms ( [ hext2 ] ) given by eq . ( [ hmol ] ) . in order to evaluate the matrix elements of @xmath83 , we express the mf angular momentum operators @xmath84 and @xmath85 in terms of the ladder operators @xmath86 and use the well - known expressions for their matrix elements @xcite to obtain @xmath87 \delta_{k_n k_n ' } + bk_n^2\delta_{k_nk_n ' } \biggr{]},\end{gathered}\ ] ] where @xmath88^{1/2}.\ ] ] the matrix elements of the centrifugal distortion term ( [ hcd ] ) can be evaluated following the same procedure . explicit expressions for the matrix elements of @xmath29 are presented in appendix b. a triatomic molecule with two identical h nuclei ( like ch@xmath2 ) can exist in two different nuclear spin modifications , ortho ( @xmath89 ) and para ( @xmath90 ) . nuclear spin selection rules strongly suppress collision - induced transitions between the energy levels of different nuclear spin isomers @xcite . because the hamiltonian is invariant to interchange of the two h atoms , @xmath89-ch@xmath2 and @xmath90-ch@xmath2 can be considered separate molecular species @xcite . we obtain the energy level spectrum of each nuclear spin isomer by diagonalizing the hamiltonian ( [ hmol ] ) using a restricted basis of symmetric - top eigenfunctions @xmath91 with @xmath92 ( for @xmath89-ch@xmath2 ) and @xmath93 ( for @xmath90-ch@xmath2 ) . in order to evaluate the matrix elements of the spin - rotation interaction ( [ hsr_bf ] ) , we rewrite it in the form @xmath94}\hat{s}_{+1 } \\ & + \frac{1}{2 } \biggl { [ } ( \gamma_x-\gamma_y ) \hat{n}_{-1 } - ( \gamma_x+\gamma_y ) \hat{n}_{+1 } \biggr{]}\hat{s}_{-1 } + \gamma_z \hat{n}_0 \hat{s}_0,\end{aligned}\ ] ] where @xmath95 and @xmath96 are the spherical tensor components of @xmath97 and @xmath98 in the mf frame . rewriting the mf spin operators in terms of their sf counterparts @xcite ( see eq . ( [ a2 ] ) in appendix a ) , evaluating the integral over three wigner @xmath41-functions , and rearranging the result , we obtain @xmath99^{1/2}[(2s+1)s(s+1)]^{1/2 } \\ \times ( -1)^{m_n'-k_n'+s - m_s } \mathcal{m}_{nk_n;n'k_n ' } \sum_q { \left(\begin{array}{ccc } s & 1 & s\\ -m_s & q & m_s ' \end{array } \right ) } { \left(\begin{array}{ccc } n & 1 & n'\\ m_n & q & -m_n ' \end{array } \right)},\end{gathered}\ ] ] where @xmath100 } \\ \notag - \frac{1}{2\sqrt{2 } } \biggl { [ } & \lambda_+ ( n',k_n ' ) ( \gamma_x-\gamma_y ) { \left(\begin{array}{ccc } n & 1 & n'\\ k_n & -1 & -k_n'-1 \end{array } \right ) } \\ \notag + & \lambda_-(n',k_n ' ) ( \gamma_x+\gamma_y ) { \left(\begin{array}{ccc } n & 1 & n'\\ k_n & -1 & -k_n'+1 \end{array } \right ) } \biggr { ] } + \gamma_z k_{n } ' { \left(\begin{array}{ccc } n & 1 & n'\\ k_n & 0 & -k_n ' \end{array } \right)},\end{aligned}\ ] ] and the quantities in parentheses denote 3-@xmath101 symbols . an alternative expression for the matrix elements of the spin - rotation interaction may be obtained by directly evaluating the matrix elements of the tensor product in eq . ( [ hsr_bf ] ) in the sf frame ( see appendix a ) . the resulting expression is rather cumbersome , and we do not present it here . we used this expression to cross - check eq . ( [ me_hsr_2 ] ) and verified that both expressions yield the same results . the matrix elements of the spin - spin interaction ( [ hss_bf ] ) in the fully uncoupled basis factorize into products of matrix elements of the form @xmath102^{(2)}_p |s m_s'\rangle.\ ] ] the first matrix element reduces to a product of two 3-@xmath101 symbols @xcite and the second matrix element is given by eq . ( 19 ) of ref . @xcite . combining these results , we obtain ( neglecting constant @xmath9-dependent terms ) @xmath103^{1/2 } ( -1)^{m_n'-k_n'+s - m_s } \\ \times \biggl { [ } \frac{1}{2}(e+d ) \biggl { [ } { \left(\begin{array}{ccc } n & 2 & n'\\ k_n & 2 & k_n ' \end{array } \right ) } + { \left(\begin{array}{ccc } n & 2 & n'\\ k_n & -2 & -k_n ' \end{array } \right ) } \biggr { ] } + \frac{3e - d}{\sqrt{6 } } { \left(\begin{array}{ccc } n & 2 & n'\\ k_n & 0 & -k_n ' \end{array } \right ) } \biggr { ] } \\ \times \sqrt{5 } [ ( 2s+1)s(s+1 ) ] { \left\{\begin{array}{ccc } 1 & 1 & 2\\ s & s & s \end{array } \right\ } } \sum_q { \left(\begin{array}{ccc } n & 2 & n'\\ m_n & q & -m_n ' \end{array } \right ) } { \left(\begin{array}{ccc } s & 2 & s\\ -m_s & q & m_s ' \end{array } \right)}.\end{gathered}\ ] ] substituting @xmath104 in eq . ( [ me_hss ] ) , we recover the expression for the matrix element of the spin - spin interaction derived previously for linear @xmath52 molecules @xcite . in order to evaluate the matrix elements of the atom - molecule interaction potential in the fully uncoupled sf basis ( [ expansion ] ) , we expand the potential in renormalized spherical harmonics @xcite @xmath105,\ ] ] where @xmath106 and @xmath107 are the spherical polar angles of @xmath108 in the mf frame ( see fig . [ fig : mf ] ) and @xmath109^{1/2}y_{\lambda\mu}(\theta,\phi)$ ] . the expansion ( [ vbf ] ) is valid for any polyatomic molecule interacting with an @xmath9-state atom provided that @xmath66 is symmetric under reflection in the @xmath32 plane , i.e. , @xmath110 . in our case , because of the @xmath111 symmetry of ch@xmath2 , the potential is also symmetric under reflection in the @xmath112 plane , so we have @xmath113 . this condition implies that the expansion coefficients @xmath114 vanish for odd @xmath3 . we note that this property is a consequence of our using a specific mf system shown in fig . [ fig : mf ] . as mentioned above , dagdigian and co - workers chose their mf @xmath31-axis to be perpendicular to the c@xmath2 axis of ch@xmath2 @xcite . as a result , their @xmath114 coefficients are not the same as defined in this work , and they do not vanish for odd @xmath3 @xcite . because our sf basis functions do not depend explicitly on @xmath106 and @xmath107 , we transform the expansion ( [ vbf ] ) to the sf frame . transforming the spherical harmonics to the sf system , we obtain @xcite @xmath115 c_{\lambda m_\lambda}(\hat{r}).\ ] ] while this expansion is convenient for obtaining analytical expressions for matrix elements , it involves five angular variables , and is thus not well suited for computational purposes . a convenient way of evaluating the @xmath114 coefficients is outlined in appendix c. making use of standard expressions for the integrals involving products of three @xmath41-functions and spherical harmonics @xcite , we obtain the final result for the matrix element of the interaction potential @xmath116^{1/2 } \times ( -1)^{m_n'-k_n'+m_l } \\ \times \sum_{\lambda= 0}^{\lambda_\text{max } } \sum_{\mu \ge 0}^\lambda v_{\lambda\mu } ( r ) \frac{1}{1+\delta_{\mu 0 } } \sum_{m_\lambda } { \left(\begin{array}{ccc } n & \lambda & n'\\ m_n & m_\lambda & -m_n ' \end{array } \right ) } { \left(\begin{array}{ccc } l & \lambda & l'\\ -m_l & m_\lambda & m_l ' \end{array } \right ) } { \left(\begin{array}{ccc } l & \lambda & l'\\ 0 & 0 & 0 \end{array } \right ) } \\ \biggl { [ } { \left(\begin{array}{ccc } n & \lambda & n'\\ k_n & \mu & -k_n ' \end{array } \right ) } + ( -1)^\mu { \left(\begin{array}{ccc } n & \lambda & n'\\ k_n & -\mu & -k_n ' \end{array } \right ) } \biggr{]}. \end{gathered}\ ] ] setting @xmath104 reduces this expression to the familiar legendre expansion of the atom - diatomic molecule interaction potential @xcite . as in the atom - diatom case @xcite , the interaction potential @xmath16 does not couple basis functions with different @xmath62 . the coefficients @xmath117 were evaluated as described in appendix c using a @xmath118 direct - product gauss - legendre quadrature grid in @xmath106 and @xmath107 . the expansion ( [ me_v ] ) was truncated at @xmath119 and we verified that increasing @xmath120 does not affect the numerical results . the system of cc equations ( [ cc ] ) was solved numerically using the log - derivative algorithm @xcite on a grid of @xmath14 extending from @xmath121 to @xmath122 with a grid step of @xmath123 for collision energies larger than 0.1 @xmath124 . for ultralow collision energies ( @xmath125 @xmath124 ) , @xmath126 was increased to @xmath127 ( for @xmath128 @xmath129 @xmath124 ) and 160.04@xmath130 ( @xmath131 @xmath132 @xmath124 ) . to obtain the cross sections converged to within @xmath13310% , the coupled - channel basis set included @xmath134 rotational states of ch@xmath2 augmented with 5 ( for @xmath135 @xmath124 ) or 6 ( for @xmath136 @xmath124 ) partial waves . the same basis set parameters were used for all ch@xmath2 isotopomers . in this work , we use two different _ ab initio _ potential energy surfaces ( pess ) for the he - ch@xmath2 van der waals complex : pes a and pes b. the former ( pes a ) was utilized in our previous theoretical work on he + ch@xmath2 scattering in a magnetic field @xcite . a more recent , high - quality pes b developed by ma _ @xcite has been used in quantum calculations of rotational relaxation in he + ch@xmath2 collisions at room temperature . pes a was calculated using a uccsd(t ) method ( coupled - cluster including single and double excitations along with a perturbative correction for triple excitations ) @xcite and a large cc - pvtz basis set @xcite . for fitting purposes , the pes is described by an analytic function that sums over six pairs of interaction potentials,@xcite @xmath137 with the damping function @xmath138 the six pairs refer to @xmath139 , where g@xmath140 are three pseudo - atoms . their mf coordinates relative to c are fixed at ( 0,1,0)@xmath130 for g@xmath141 , ( 0,-1,0)@xmath130 for g@xmath2 , and ( 0,0,1)@xmath130 for g@xmath6 . the pseudo - atoms are used to take the many - body interaction potentials into account . the parameters in eq . ( [ eq : pes ] ) are determined by fitting to 985 ccsd(t)/cc - pvtz points spanning the range of @xmath142\,a_0 $ ] , @xmath143 $ ] , and @xmath144 $ ] with a fixed geometry of ch@xmath2 at its triplet equilibrium structure . the potential energy surface is well fitted to a root - mean - square error of 0.8 @xmath124 . the parameters obtained are given in table [ tab : pes ] . figures [ fig : pesba1 ] [ fig : pesba3 ] ( left panels ) show a typical long - range potential energy surface for the he@xmath145ch@xmath2 complex , where the ch@xmath2 molecule is fixed at its ccsd(t ) optimized geometries , @xmath146 and @xmath147 . it displays two equivalent global minima located at @xmath148 with a well depth of 9.7@xmath124 . owing to the shallow minima , the interaction potentials between the two collision partners are weakly anisotropic . this feature of the potential energy surface is also consistent with the small permanent electric dipole moment of 0.2293 atomic units ( @xmath149d ) calculated with ccsd(t)/cc - pvtz and nearly isotropic quadrupole moments [ ( 6.0344 , 4.3788 , 5.6653 ) atomic units ] of ch@xmath2 . a detailed description of the _ ab initio _ approach and the topology of pes b has been presented elsewhere @xcite . spin - restricted coupled - cluster [ rccsd(t ) ] calculations @xcite were carried out on a 4-dimensional grid consisting of the position ( in spherical polar coordinates ) of the he atom with respect to the center - or - mass of the ch@xmath55 molecule for a grid of values of the ch@xmath2 bond angle . an aug - cc - pvqz @xcite basis set was used with a set of bond functions @xcite added at the mid - point of @xmath14 , the vector pointing to the he atom from the center of mass of ch@xmath2 . the 3-dimensional pes used in the scattering calculations was determined by averaging the 4-dimensional potential over the bending angle weighted by the square of the @xmath150 bending vibrational wave function . figures [ fig : pesba1 ] [ fig : pesba3 ] show that pes a is considerably less deep , more repulsive at short range , and less anisotropic than pes b. the increased depth of the potential arises from a better recovery of the correlation energy , which is known to be responsible for dispersion interactions , due to the inclusion of diffuse atom - centered functions as well as bond functions @xcite for a weakly - bound complex . the calculations on which pes a is based are deficient in that the van der waals attraction is significantly less well described . we also note that because the van der waals well is deeper in pes b , the repulsive core of the pes the angle - dependent hard sphere radius is significantly smaller in extent than the predictions of pes a. as will be shown below ( sec . iiib ) , the significant difference in the two pess has a strong effect on both the elastic and inelastic he + ch@xmath55 collisions at low temperatures . figure [ fig : zeeman ] shows the zeeman energy levels of _ ortho _ and _ para _ nuclear spin isomers of ch@xmath2 . the ground rotational state of @xmath89-ch@xmath2 ( @xmath151 ) is split by magnetic fields into three zeeman sublevels with @xmath152 , 0 , and 1 . here , we adopt the conventional @xmath153 notation for the rotational states of an asymmetric top , where @xmath154 ( @xmath155 ) are the projections of @xmath26 on the mf quantization axis in the oblate ( or prolate ) symmetric top limits @xcite . the ground rotational state of @xmath90-ch@xmath2 ( @xmath156 ) has @xmath157 @xcite and splits into nine zeeman sublevels correlating with the @xmath158 , @xmath159 , and @xmath160 fine - structure levels in the limit of zero magnetic field , where @xmath161 is the total angular momentum of ch@xmath2 excluding nuclear spin . here , we are interested in collisions of ch@xmath2 molecules in their magnetically trappable , maximally spin - stretched zeeman levels @xmath162 ( for @xmath89-ch@xmath2 ) and @xmath163 ( for @xmath90-ch@xmath2 ) , where @xmath164 is the projection of @xmath101 on the magnetic field axis . it is clear from fig . [ fig : zeeman ] that the zeeman energy level patterns of @xmath89-ch@xmath2 and @xmath90-ch@xmath2 are very different . as will be shown below , this difference leads to dramatic variations in the low - temperature collisional properties of the spin isomers . in this section , we focus on collisions of @xmath89-ch@xmath2 molecules and omit the prefix @xmath89 for brevity . figure [ xs_potab ] shows the elastic and inelastic cross sections for @xmath10he + ch@xmath165 collisions @xcite as functions of collision energy calculated for @xmath166 t , the field strength typically used in buffer - gas cooling and magnetic trapping experiments with @xmath52 molecules @xcite . we calculate the inelastic spin relaxation cross section by summing the state - to - state cross sections for a given incident collision channel ( dashed lines in fig . [ fig : zeeman ] ) over all energetically accessible zeeman levels of ch@xmath2 ( full lines in fig . [ fig : zeeman ] ) . both the elastic and inelastic cross sections are very sensitive to the interaction pess : the @xmath167-wave limit of the elastic cross section calculated using pes a is @xmath16810 times larger than the limit for pes b. overall , because calculations on which pes a is based predict a larger hard - sphere radius , we anticipate that the @xmath167-wave elastic cross section , which is a measure of the size of the system , will be larger . the cross sections calculated using pes a are dominated by a pronounced resonance feature at @xmath169 k with a very broad shoulder extending down to @xmath170 k , whereas the results for pes b display two resonance peaks above @xmath171 k. the presence of the broad resonance leads to a large difference between the thermally averaged ratios of elastic to inelastic collision rates calculated with potentials a and b ( see the inset of fig . [ xs_potab ] ) . since pes b has a more pronounced long - range attractive component , we anticipate that any elastic resonances at low energy predicted by the two pes s will be significantly different . figure [ xs_b ] shows the cross sections for inelastic spin relaxation in he + ch@xmath2 collisions as functions of collision energy calculated for several values of the magnetic field . at very low collision energies , we observe the @xmath167-wave threshold behavior @xmath172 typical of two - body inelastic processes in the limit of zero collision energy @xcite . the elastic cross section is independent of the field and becomes a constant at @xmath173 100 mk . at moderately low collision energies ( above 0.5 k ) , the cross sections exhibit resonant oscillations due to the excited rotational states ( @xmath174 ) of the collision complex ( shape resonances ) which decay by tunneling through the centrifugal barrier @xcite . a crossover between the single @xmath167-wave and multiple partial - wave regimes occurs at a collision energy of 1 mk at low magnetic fields ( @xmath175 t ) and shifts to higher collision energies with increasing field . a similar trend manifests itself in cold collisions of @xmath52 molecules @xcite . figure [ xs_b ] shows that the magnitude of the inelastic cross section in the @xmath167-wave limit increases dramatically with the field strength . as in previous theoretical work on collisions of @xmath176-state molecules , this can be explained by the following argument @xcite . the height of the centrifugal barrier in the outgoing ( @xmath177 ) collision channels becomes smaller with increasing field , making it easier for collision products to escape over the barrier , thereby enhancing the inelastic collision rate @xcite . in the inset of fig . [ xs_b ] , we plot the key figure of merit for sympathetic cooling experiments the ratio of the thermal rate constants for elastic scattering and inelastic scattering ( spin relaxation ) in he - ch@xmath2 collisions : @xmath178 the rates in eq . ( [ gamma ] ) are obtained as functions of temperature by averaging the cross sections shown in fig . [ xs_b ] over a maxwell - boltzmann velocity distribution . we observe that @xmath179 over the range of magnetic fields typically used in cryogenic cooling and magnetic trapping experiments ( 0.01 1 t ) . an empirical rule states that @xmath180 is required to ensure efficient cryogenic he buffer - gas cooling of a molecule @xcite , so the results shown in fig . [ xs_b ] indicate that @xmath90-ch@xmath2 is an ideal candidate for cryogenic cooling and magnetic trapping experiments @xcite . as shown in the inset of fig . [ xs_potab ] , the temperature profiles of @xmath79 calculated for pess a and b differ substantially , with @xmath181 at @xmath182 k due to the broad scattering resonance that occurs on pes a ( see the inset of fig . [ xs_potab ] ) . we emphasize , however , that despite these quantitative differences , both pes a and pes b calculations predict @xmath183 , so our main qualitative conclusion as to the suitability of ch@xmath2 for sympathetic cooling experiments is independent of the choice of the pes . so far we have been focusing on collisions involving a single nuclear spin isomer of ch@xmath2 ( _ ortho_-ch@xmath2 ) . however , as pointed out in sec . ii , the rotational and zeeman structures of @xmath89- and @xmath90-ch@xmath2 are very different , so it is of interest to compare the collisional properties of the nuclear spin isomers . figure [ xs_op ] shows the elastic and inelastic cross sections for @xmath89-ch@xmath2 and @xmath90-ch@xmath2 in he as functions of collision energy . the inelastic cross section calculated for @xmath90-ch@xmath2 is about four orders of magnitude larger than that for @xmath89-ch@xmath2 over the range of collision energies from 10@xmath184 to @xmath185 k. in the multiple partial wave regime ( @xmath186 k ) the difference becomes somewhat smaller , but never falls below three orders of magnitude . in order to understand this remarkable disparity in low - temperature collisional properties of different nuclear spin isomers of ch@xmath2 , we note that in our calculations , the incident collision channels for both @xmath4 and @xmath5-ch@xmath2 are the maximally spin - stretched zeeman states ( shown in fig . [ fig : zeeman ] by dashed lines ) . these are the zeeman states in which the molecules reside when confined in a permanent magnetic trap @xcite . however , the @xmath187 manifolds to which these zeeman states belong are different for @xmath4 and @xmath5 nuclear spin isomers : as shown in the left panel of fig . [ fig : zeeman ] , the lowest maximally spin - stretched zeeman state of @xmath89-ch@xmath2 belongs to the @xmath151 rotational manifold , while that for @xmath90-ch@xmath2 belongs to the @xmath156 rotational manifold @xcite . the spin - spin interaction ( [ hss_bf ] ) couples the ground @xmath151 and the second excited @xmath188 rotational manifolds of @xmath89-ch@xmath2 . as a result , both @xmath189 and @xmath62 lose the status of good quantum numbers , and the wavefunction of the lowest magnetically trappable zeeman state becomes latexmath:[\[\label{initial_state_o } where the tildes over @xmath189 and @xmath62 indicate the approximate nature of these quantum numbers . a first - order perturbation theory estimate gives @xmath191 where the basis functions are the products of asymmetric - top eigenfunctions @xmath192 times the spin functions @xmath193 , and we have neglected the dependence of the matrix element of the spin - spin interaction ( [ hss_bf ] ) on the zero - field - splitting parameter @xmath43 assuming @xmath194 , which is justified for ch@xmath55 where @xmath195 . in eq . ( [ beta0 ] ) @xcite @xmath196^{1/2}\ ] ] is the energy gap between the ground and the second excited rotational states . in @xmath89-ch@xmath2 the gap ( 47 @xmath124 ) is much larger than the spin - spin coupling @xmath197 @xmath124 , so the admixture of the @xmath188 rotational state into the ground state is on the order of a few percent ( depending on @xmath198 ) . in the first born approximation , the cross section for spin relaxation out of the lowest magnetically trappable state of ch@xmath2 ( @xmath199 ) is proportional to the square of the matrix element between the molecular spin states involved in the transition . in particular , the cross section for relaxation from the @xmath200 state in the lowest rotational manifold of @xmath89-ch@xmath2 may be written as @xcite @xmath201 . using the basis functions given by eq . ( [ initial_state_o ] ) , we find for the @xmath202 transition to the leading order @xmath203 this result shows that spin relaxation in spin-1 molecular radicals is a first - order process mediated by the combined action of the spin - spin interaction and the anisotropy of the interaction potential ( the second term in eq . [ born2 ] ) . note that both these interactions are equally important : without the spin - spin interaction there would be no admixture of the spin - down state in the initial spin - up state ( @xmath204 implies @xmath205 in eq . [ beta0 ] ) , and without the potential anisotropy there would be no coupling between the different @xmath189 states . we further note that the expression ( [ born2 ] ) contains the factor @xmath206 for light molecules with large rotational splittings ( [ deltae02 ] ) . thus , collision - induced spin relaxation in such molecules should be slower than rotational relaxation , which is induced by the anisotropic part of the interaction potential _ without _ the small factor @xmath207 . these considerations explain the small magnitude of the calculated inelastic cross sections for @xmath89-ch@xmath2 shown in fig . [ xs_b ] . a closely related result for collision - induced spin relaxation in linear @xmath52 molecules was first derived theoretically by krems and co - workers @xcite and then observed experimentally using he buffer - gas cooled nh(@xmath52 ) molecules in a magnetic trap at 600 mk @xcite . the wavefunction for the lowest magnetically trappable zeeman state of @xmath90-ch@xmath2 may be written as latexmath:[\[\label{initial_state_p } unlike the ground rotational state of @xmath89-ch@xmath2 the spin - rotation and spin - spin interactions couple the different @xmath60 and @xmath62 sublevels _ within _ the @xmath156 rotational manifold of @xmath90-ch@xmath2 . thus , the mixing coefficients @xmath209 , and @xmath210 in eq . ( [ initial_state_p ] ) are all of similar magnitude , and the cross section for spin relaxation in @xmath90-ch@xmath2 collisions is given by an expression similar to eq . ( [ born2 ] ) _ without _ the small factor @xmath207 ( assuming the validity of born approximation ) . thus the mechanism of spin relaxation in he + @xmath90-ch@xmath2 collisions is similar to that of rotational relaxation or rotational depolarization with the corresponding cross sections ranging from 1 to 100 @xmath211 ( ref . this lack of suppression is responsible for the large magnitude of spin relaxation cross sections for he + @xmath90-ch@xmath2 shown in fig . [ xs_op ] , explaining the large differences between the collisional properties of different nuclear spin isomers of ch@xmath2 in low - temperature collisions with he atoms . a simple qualitative explanation for this difference is as follows : inelastic changes in the spin - multiplets correspond to a reorientation of @xmath212 followed by a recoupling with @xmath213 , which is unaffected by the intermolecular potential which has no magnetic terms . in the case of the lowest low - field - seeking zeeman state of @xmath89-ch@xmath2 ( @xmath214 ) , the nonzero @xmath215 component is only a small admixture in the ground - state wavefunction ( see eq . [ initial_state_o ] ) so reorientation occurs at a much slower rate . in the previous section , we have established that the cross section of spin relaxation in ground - state @xmath89-ch@xmath2 collisions scales as @xmath216 with the splitting @xmath217 between the ground @xmath151 and the second excited @xmath188 rotational levels of @xmath89-ch@xmath2 . this scaling law is based on a simple first born approximation , which does not take into account the effects of higher - order potential couplings and shape resonances ( see fig . [ xs_op ] ) on collision dynamics . it is therefore desirable to test the accuracy of the scaling prediction , which can be performed by changing the rotational constant ( and hence the energy gap @xmath217 ) of ch@xmath2 . to this aim , we have performed quantum scattering calculations for partially and fully deuterated isotopomers of ch@xmath2 using the same pes b for all the isotopomers . the rotational and zero - field - splitting constants of ch@xmath2 , chd , and cd@xmath2 used in these calculations are listed in table ii @xcite . figure [ xs_isotopes ] compares the cross sections for spin relaxation in ch@xmath2 , chd , and cd@xmath2 induced by collisions with he at @xmath166 t. remarkably , the cross sections exhibit a nearly identical dependence on collision energy in the few - partial - wave regime ( @xmath218 mk ) . this result points to the validity of the inverse quadratic scaling of spin relaxation cross sections with the rotational splitting ( [ born2 ] ) . table iii lists the ratios of the cross sections for chd and cd@xmath2 to those of ch@xmath2 calculated at two different collision energies . the ratios are seen to increase with a decreasing rotational splitting @xmath217 . the deviation from the @xmath219 scaling does not exceed 10% , demonstrating the araccuracy of the scaling prediction ( [ born2 ] ) which is quite remarkable given its simplicity . in the multiple partial wave regime , the inelastic cross sections in fig . [ xs_isotopes ] tend to be dominated by shape resonances , which occur due to the centrifugal barriers in the incoming and outgoing collision channels @xcite . these resonances lead to an enhancement of scattering cross sections that is correlated with the magnitude of the rotational constant of the metastable collision complex . the properties of this collision complex are determined not only by the rotational structure of the molecule , but also by the reduced mass and interaction potential of the he - molecule complex , leading to pronounced deviations from the simple @xmath216 scaling near scattering resonances . this effect is clearly observed in fig . [ xs_isotopes ] : in contrast with the scaling prediction , the inelastic cross section for ch@xmath2 can _ exceed _ that of cd@xmath2 at selected collision energies . we note that similar deviations from the @xmath216 scaling were observed experimentally for @xmath10he + nh ( but not @xmath220he - nh ) collisions at 600 k @xcite , demonstrating the important role of shape resonances in atom - molecule collisions at sub - kelvin temperatures . we have developed a theoretical methodology for exact quantum mechanical calculations of polyatomic molecule collisions in the presence of an external magnetic field . the theoretical description is based on a fully uncoupled sf basis set expansion to represent the scattering wave function . a system of coupled - channel equations for the radial part of the wave function is derived and parametrized by the matrix elements of the molecular hamiltonian and of the atom - molecule interaction pes . using ch@xmath221 as a general example of a non - linear , asymmetric - top polyatomic molecule , we present the expressions for the matrix elements of rotational , centrifugal distortion , spin - rotation , and spin - spin interactions in terms of the relevant spectroscopic constants @xcite ( sec . ii and appendices a and b ) . the matrix elements of the atom - molecule interaction are evaluated by expanding the pes in angular basis functions , and we outline the details of this procedure in sec . ii and appendix c. we have applied our theoretical framework to explore the dynamics of spin relaxation of ground - state ch@xmath221 molecules induced by collisions with he atoms in a magnetic field . the scattering calculations were based on two different _ ab initio _ pess developed independently of each other : pes a @xcite and pes b @xcite . we found that , even though the results obtained for pes a and b exhibit large quantitative differences , the ratio of elastic to inelastic collision rates for he + @xmath89-ch@xmath2 collisions greatly exceeds 100 over a wide range of temperatures and magnetic fields regardless of the pess used to describe the interaction of he with ch@xmath2 . the magnitude of this ratio is thus relatively insensitive to the details of the long - range part of the potential , which is described quite differently by the two pess , as well as to the overall `` size '' of the potential . the large magnitude of the ratio strongly suggests the possibility of cryogenic buffer - gas cooling and magnetic trapping of @xmath89-ch@xmath2 . our calculations demonstrate that different nuclear spin isomers of ch@xmath2 undergo collision - induced spin relaxation on widely different timescales . as shown in fig . [ xs_op ] , the cross section for he - induced spin relaxation in @xmath90-ch@xmath2 is more than three orders of magnitude larger than that for @xmath89-ch@xmath2 . these differences arise because the lowest @xmath222 rotational state , which is more immune to spin relaxation due to the absence of direct intramultiplet spin - rotation and spin - spin couplings , is part of the rotational spectrum of @xmath89-ch@xmath2 , but not of @xmath90-ch@xmath2 . the results shown in fig . [ xs_op ] thus suggest that it may be possible to isolate the @xmath4 nuclear spin isomer of ch@xmath2 by magnetic trapping of the naturally occurring @xmath4-@xmath5 mixture . separation of nuclear spin isomers is an active field of research @xcite , and our results suggest that such separation will occur automatically when a cold @xmath4-@xmath5 mixture is confined in a magnetic trap in the presence of a cryogenic he gas . using the first born approximation , we have shown that the cross sections for spin relaxation in the @xmath222 rotational state scale as @xmath207 with the spin - spin coupling @xmath41 and the rotational splitting @xmath217 . this result is fully analogous to the previously derived scaling law for diatomic @xmath52 molecules @xcite . we verified that the deviation of the scaling predictions from exact cc calculations does not exceed 10% in the @xmath167-wave regime ( fig . [ xs_isotopes ] and table iii ) . the scaling prediction becomes less accurate with increasing collision energy , as the collision system enters the multiple partial wave regime ( fig . [ xs_isotopes ] ) and shape resonances occur . the results of our analysis bear important implications for collisional cooling of polyatomic molecules , opening up the possibility of cryogenic buffer - gas cooling of ch@xmath221 and possibly other @xcite open - shell polyatomic molecules in a magnetic trap . experimental realization of this possibility may open up new directions of research in the study of reaction intermediates @xcite , large molecule spectroscopy and photochemistry , and may even enable external field control of complex organic reactions @xcite . moreover , since the magnitudes of the elastic - to - inelastic ratios for he + ch@xmath2 plotted in fig . [ xs_op ] are similar to those measured and calculated for he + nh(@xmath52 ) @xcite , it is reasonable to assume that collisional properties of ch@xmath2 molecules with atoms other than he will be similar to those of the isoelectronic nh@xmath223 molecule @xcite . if so , it may be possible to use sympathetic cooling with laser - cooled alkaline earth or spin - polarized nitrogen atoms @xcite to create ultracold ensembles of ch@xmath2 , chd , cd@xmath2 , and possibly other spin-1 and spin-1/2 polyatomic molecules @xcite . a more detailed theoretical study would be needed to fully explore this possibility , and could be done using the formalism presented in this work . we thank gregory e. hall , john m. doyle , and paul j. dagdigian for discussions . this work was supported by the doe office of basic energy science and nsf grants to the harvard - mit cua and itamp at harvard university and the smithsonian astrophysical observatory as well as to mha at the university of maryland ( grants che-0848110 and che-1213332 ) . here , we present the derivation of eq . ( [ hss_bf ] ) from the standard mf expression for the spin - spin interaction in polyatomic molecules @xcite @xmath224 where the subscripts @xmath225 , and @xmath31 refer to mf cartesian components of the spin operators . the mf and sf spherical tensor components of @xmath37 are related by @xcite @xmath226 here and below , the index @xmath227 ( @xmath90 ) is used to denote the mf ( sf ) components of @xmath37 @xcite . a product of two mf spherical tensor operators can thus be expanded as @xmath228 where we have used a clebsch - gordan expansion for the product of two @xmath41-functions @xcite and the quantities in square brackets are clebsch - gordan coefficients . interchanging the order of summation , we obtain @xmath229 where @xmath230 . the last sum in this expression is , by definition , a spherical tensor product of @xmath37 with itself @xcite , so eq . ( [ a4 ] ) takes the form @xmath231^{(k)}_p.\ ] ] using the standard relations between the cartesian and spherical tensor components of angular momentum operators @xcite , we get three important particular cases of eq . ( [ a5 ] ) @xcite : @xmath232^{(k)}_p , \\ \notag \hat{s}_\pm^2 & = 2\hat{s}_{\pm 1}^2 = 2 \sum_{k=0,2 } { \begin{bmatrix } { 1 } & { 1 } & { k } \\ { \pm1 } & { \pm 1 } & { \pm 2 } \\ \end{bmatrix } } \sum_p \mathcal{d}^k_{p,\pm 2}(\hat{\omega } ) [ \hat{\bm{s } } \otimes \hat{\bm{s}}]^{(k)}_p , \\ \hat{s_\pm}\hat{s}_\mp & = -2\hat{s}^{(1)}_{\pm1 } s^{(1)}_{\mp1 } = - 2 \sum_{k=0,2 } { \begin{bmatrix } { 1 } & { 1 } & { k } \\ { \pm1 } & { \mp 1 } & { 0 } \\ \end{bmatrix } } \sum_p \mathcal{d}^k_{p0}(\hat{\omega } ) [ \hat{\bm{s } } \otimes \hat{\bm{s}}]^{(k)}_p.\end{aligned}\ ] ] by combining these expressions with the definitions @xmath233 and @xmath234 and neglecting the terms proportional to @xmath235 , we obtain eq . ( [ hss_bf ] ) of the main text . we note that the following sf tensor expression for the spin - rotation interaction ( [ hsr_bf ] ) @xmath236 + { \textstyle \frac{1}{\sqrt{6 } } } ( 2\gamma_z-\gamma_x-\gamma_y ) \mathcal{d}^2_{p,0}(\hat{\omega } ) \biggr { ] } [ \hat{\bm{n } } \otimes \hat{\bm{s}}]^{(2)}_p,\ ] ] where @xmath237 , may be obtained by replacing the second spin operator in eq . ( [ a3 ] ) with rotational angular momentum @xmath26 and retracing the steps outlined above . we note , however , that eq . ( [ hsr_bf ] ) is simpler and hence more convenient for the practical purpose of evaluating the matrix elements in the fully uncoupled sf basis ( [ expansion ] ) . since the centrifugal distortion hamiltonian ( [ hcd ] ) is spin - independent , its matrix elements are diagonal in @xmath9 and @xmath62 , so we omit these quantum numbers for the rest of this section . in addition , since the mf angular momentum operators do not couple states with different @xmath189 or @xmath60 ( and independent of @xmath60 ) , we can set @xmath238 and @xmath239 in all the expressions below . the matrix elements of eq . ( [ hcd ] ) in the spherical - top basis @xmath91 can be evaluated by using the standard angular momentum algebra @xcite with the result for the first term @xmath240 the second term @xmath241 \delta_{k_n k_n ' } \biggr{]},\end{gathered}\ ] ] the third term @xmath242 \delta_{k_n , k_n'\pm 2 } \\ + \bigl{\ { } 6[n(n+1)-k_n'^2]^2 - 4[n(n+1)-k_n'^2 ] + 6k_n'^2 \bigr{\ } } \delta_{k_n , k_n ' } \biggr{]}\end{gathered}\ ] ] and , finally , the fourth term @xmath243 \bigl{\ } } \delta_{k_n k_n ' } \biggr{]}.\end{gathered}\ ] ] the matrix element of the centrifugal distortion hamiltonian can be obtained by summing eqs . ( [ b1 ] ) through ( [ b4 ] ) . here , we derive the expressions for the radial expansion coefficients @xmath244 based on the _ ab initio _ pess @xmath66 calculated as described in sec . ii . the expansion of the atom - molecule interaction pes in spherical harmonics ( [ vbf ] ) @xmath245,\ ] ] may be rewritten in terms of real functions of @xmath106 and @xmath107 as @xmath246 where @xmath247 is a normalized associated legendre polynomial @xcite . a disadvantage of eq . ( [ c2 ] ) is that the angular basis functions on the right - hand side are not orthonormal . it is more convenient to use an alternative expansion in terms of orthonormal angular basis functions @xmath248^{1/2 } \cos ( \mu \phi).\ ] ] the expansion coefficients @xmath249 may be obtained by inverting this expression @xmath250^{1/2 } \int_{0}^{2\pi } d\phi \int_0^{\pi } \sin\theta d\theta v(r,\theta,\phi ) \theta_{\lambda\mu}(\theta ) \cos ( \mu \phi).\ ] ] the relationship between the coefficients @xmath251 and @xmath252 can be obtained by comparing eqs . ( [ c2 ] ) and ( [ c3 ] ) . ( [ c4 ] ) , we obtain the final result @xmath253 h. sabbath , l. biennier , i. r. sims , y. georgievskii , s. j. klippenstein , and i. w. m. smith , science * 317 * , 102 ( 2007 ) . s. d. le picard , m. tizniti , a. canosa , i. r. sims , and i. w. m. smith , science * 328 * , 1258 ( 2010 ) . e. herbst and e. f. van dishoeck , annu . . astrophys . * 47 * , 427 ( 2009 ) . m. arndt , o. nairz , j. vos - andreae , c. keller , g. van der zouw , and a. zeilinger , nature ( london ) * 401 * , 680 ( 1999 ) ; k. hornberger , s. uttenthaler , b. brezger , l. hackermller , m. arndt , and a. zeilinger , . ni , s. ospelkaus , m. h. g. de miranda , a. peer , b. neyenhuis , j. j. zirbel , s. kotochigova , p. s. julienne , and j. ye , science * 322 * , 231 ( 2008 ) . j. m. doyle , b. friedrich , j. kim and d. patterson , phys . rev . a * 52 * , r2515 ( 1995 ) . the effective hamiltonian ( [ hmol ] ) is appropriate for a rigid polyatomic molecule in a given vibronic state , and hence can not be used to describe collision - induced transitions between different electronic or vibrational levels . in this work , we consider collisions with the @xmath10he isotope , which is relevant for cryogenic buffer gas cooling experiments @xcite . the isotropic and anisotropic hyperfine interactions arising from the nuclear spin of @xmath10he are extremely weak ( on the order of tens of mhz for @xmath254 atom - he collisions and are neglected here . for more details on these effects , see t. v. tscherbul , p. zhang , h.r . sadeghpour , and a. dalgarno , . note that the spin operators @xmath255 are defined in terms of cartesian components of @xmath37 , while the operators @xmath256 are expressed in a spherical tensor basis @xcite : @xmath257 . for more details , see chap . 6.3 of ref . . isotopic substitution causes a small shift in the origin of the mf coordinate system which should , in principle , be taken into account via a coordinate scaling transformation @xcite . as we are mainly interested in the effect of rotational splitting on collision dynamics , we did not apply the scaling transformation in this work . p. uchowski and j. m. hutson , . |
we use a nearly complete sample of gamma ray bursts ( grbs ) detected by the _ swift _ satellite to study the correlations between the spectral peak energy of the prompt emission , the isotropic energetics and the isotropic luminosity . this grb sample is characterized by a high level of completeness in redshift ( 90% ) . this allows us to probe in an unbiased way the issue related to the physical origin of these correlations against selection effects . we find that one burst , grb 061021 , is an outlier to the correlation . despite this case , we find strong and correlations for the bursts of the complete sample . their slopes , normalizations and dispersions are consistent with those found with the whole sample of bursts with measured redshift and . this means that the biases present in the total sample commonly used to study these correlations do not affect their properties . finally , we also find no evolution with redshift of the and correlations . [ firstpage ] gamma - ray : burst | 1112.4470 | one of the most debated issues in gamma ray bursts ( grbs ) concerns the existence of correlations among the spectral parameters of the prompt emission and its energetic and luminosity . three robust correlations have been identified . each of them involves the rest frame peak energy @xmath0 of the @xmath5 prompt spectrum . this quantity strongly correlates with ( i ) the isotropic energy @xmath1 ( amati relation , amati et al . 2002 ) , ( ii ) the isotropic peak luminosity @xmath2 ( yonetoku relation , yonetoku et al . 2004 ) and ( iii ) the collimation corrected energy @xmath6 ( ghirlanda relation , ghirlanda et al . 2004b ; see also liang & zhang 2005 ) . these correlations are valid for long grbs . there are indications that short bursts ( i.e. bursts with observed duration @xmath72 s ) obey the very same @xmath4 relation defined by long events but they are inconsistent with the @xmath3 correlation ( ghirlanda et al . 2009 ; amati 2010 ) . the spectral energy correlations have relevant implications both for the theoretical understanding of the burst physics and for the application of grbs as cosmological tools ( ghirlanda et al . 2004a ; ghirlanda et al . 2006 ) . while there is no consensus yet on the interpretation of these correlations ( e.g. eichler & levinson 2004 ; levinson & eichler 2005 ; rees & mszros 2005 ; thompson , mszros & rees 2007 ) , an interesting progress has been recently made by studying the comoving properties of grbs ( ghirlanda et al . 2012 ; panaitescu et al . 2009 ) . the @xmath3 and @xmath4 correlations have been discovered with a very small sample of 1216 grbs detected by _ beppo_sax and/or batse ( amati et al . 2002 ; yonetoku et al . 2004 ) . they have been subsequently confirmed by adding bursts detected by other satellites ( e.g. _ hete ii _ , _ konus_/wind , _ integral _ , _ swift _ and _ fermi _ ) . at present the sample of bursts defining the @xmath3 and @xmath4 correlations comprises more than 130 events , with only two well known ( very peculiar ) outliers ( but see ghisellini et al . 2006 ) . despite the considerably large number of bursts consistent with these correlations , their physical origin is debated . some authors claim that they are the result of instrumental selection effects ( nakar & piran 2005 ; band & preece 2005 ; butler et al . 2007 ; shahmoradi & nemiroff 2011 ) . other studies ( ghirlanda et al . 2008 ; nava et al . 2008 ) quantified the possible instrumental selection biases finding that , even if they do affect the sample , they can not be responsible for the existence of the spectral energy correlations . ghirlanda et al . ( 2010 ; 2011a ; 2011b ) added an important piece of information to this debate . they analysed the time resolved spectra of both long and short bursts detected by the _ fermi _ satellite . they found that the time resolved @xmath0 correlates with the luminosity during the temporal evolution of the burst ( and for bursts without @xmath8 they derive a correlation between the time resolved @xmath9 and the flux ) . this is a feature common to short and long grbs . the time resolved correlations found within individual grbs are similar to each other and they are similar to the @xmath4 relation defined by the time integrated spectral properties of different bursts ( i.e. the yonetoku relation ) . this is a strong argument in favour of the physical origin of the correlations because it is very unlikely that within a single burst the instrumental selection effects ( e.g. the detector threshold ) or possible effects related to the dependence of the peak energy , energetic and luminosity on the redshift can play any role . grbs added to the @xmath3 and @xmath4 correlations need to have their redshifts measured . this raised the suspect that these correlations might be biased by the high level of incompleteness in redshift of the samples of grbs so far used . in this paper we study , for the first time , the @xmath3 and @xmath4 correlations with a sample of grbs with a high level of completeness ( 90% ) . the sample is presented by salvaterra et al . ( 2012 ) and it comprises 58 events detected by _ swift _ with a 1s peak photon flux @xmath10 ph s@xmath11@xmath12 ( integrated in the 15150 kev energy range ) . fifty two out of 58 grbs have measured redshift @xmath8 , while in 3 cases an upper limit on @xmath8 can be set from the photometry of the afterglow or of the host galaxy . the completeness level then increases up to @xmath1395% by considering the redshift constraints imposed by the detection of the afterglow or host galaxy in some optical filters . the complete sample used in this work is particularly appropriate to study the possible evolution of the @xmath3 and @xmath4 correlations with redshift . li ( 2007 ) , considered a sample of 48 grbs ( from amati 2006 , 2007 ) and investigated if the @xmath3 correlation evolves with redshift , finding that it does , becoming steeper at higher redshifts . ghirlanda et al . ( 2008 ) repeated the same test with a larger sample of 76 bursts and found no evidence of evolution with redshift . these results , however , were obtained by considering highly incomplete grb samples . our goals are ( i ) to study how @xmath0 is correlated to @xmath1 and @xmath2 within the complete sample of bursts and ( ii ) to study the possible evolution of these correlations with redshift through this unbiased sample . in @xmath142 we present the spectral properties of the grbs composing the complete sample . we study the @xmath3 and @xmath4 correlations defined by this sample in @xmath15 and compare their properties with those defined by the ( incomplete ) larger sample of grbs with measured @xmath8 and @xmath9 . in @xmath16 we discuss and summarise our conclusions . throughout the paper we assume a standard cosmology with @xmath17 and @xmath18 . in order to study a relatively unbiased redshift distribution of grbs , jakobsson et al . ( 2006 ) carefully selected a sample containing bursts which have ` observing conditions ' favourable for redshift determination ( see their web pagepja / grbsample.html ] for an updated compilation of the sample and an explanation of the selection criteria ) . this relatively ` clean ' sample has a redshift recovery rate of roughly 50% . in order to reach a higher level of completeness , salvaterra et al . ( 2012 ) considered this sample and also required that the photon peak flux , measured by _ swift_/bat in the 15150 kev energy band , is @xmath10 ph s@xmath11@xmath12 . up to may 2011 , 58 grbs match this selection criteria and are listed in table [ tab1 ] . fifty - two of them have measured redshift so that our completeness level is 90% . of these 52 , all but two ( namely grb 070521 and grb 080602 ) have spectroscopic confirmed redshift either from absorption lines over - imposed on the grb optical afterglow or from emission lines of the grb host galaxy . moreover , for 3 of the 6 bursts lacking measured @xmath8 the afterglow or the host galaxy have been detected in at least one optical filter , so that @xmath1395% of the bursts in this sample have a constrained redshift . while this sample represents only @xmath1310% of the full swift sample , it contains more than 30% of long grbs with known redshift . the bursts belonging to this sample ( updated to may 2011 ) are listed in table [ tab1 ] . we will refer to this sample as the ` complete ' one . for all bursts in this sample , we collected from the literature the prompt emission spectral parameters and computed the rest frame peak energy @xmath0=@xmath9@xmath19 , the isotropic energy @xmath1 and the isotropic peak luminosity @xmath2 . the two latter quantities have been calculated in the rest frame energy range 1 kev10 mev . the narrow sensitivity range of the bat instrument allows to determine @xmath9 only when it falls approximately between @xmath20 kev and @xmath13150 kev . this condition is necessary but not sufficient : as demonstrated in the spectral simulation studies by sakamoto et al . ( 2009 ) , even when @xmath9 lies in the bat energy range , the value of @xmath9 can be determined only if the spectrum is characterized by a high signal - to - noise . similar conclusions were reached through numerical simulations by ghirlanda et al . ( 2008 ) and nava et al . almost all grbs of the complete sample , however , have been detected also by other instruments with wider sensitivity ranges ( mainly _ konus_/wind and , for bursts after august 2008 , _ fermi _ ) . this allowed to estimate @xmath9 for a large fraction of events . we , therefore , obtained a sample of 46 bursts ( out of 58 ) with @xmath0 , @xmath1 , @xmath2 and redshift @xmath8 , which are the quantities required to investigate the @xmath3 and @xmath4 correlations . for 6 grbs there is a good knowledge of the prompt spectral shape ( i.e. @xmath0 is well constrained ) , but the redshift is unknown ( in 3 cases it is possible to derive an upper limit on @xmath8 from the photometry of the afterglow or of the host galaxy ) . for the remaining 6 bursts the determination of @xmath9 is not possible : in 3 cases the spectrum is well described by a single power law , while in the other 3 cases it is possible to set a lower limit to the observed value of @xmath9 . redshifts and spectral properties of the 58 grbs of the complete sample are reported in table [ tab1 ] . here we also report the spectral parameters of the prompt emission spectrum ( with their uncertainties ) and the fluence and peak flux ( with the energy ranges over which they are computed ) . these are the informations used to compute the rest frame isotropic energy and luminosities of the bursts in our samples . we also report for each burst the instrument used to derive the spectral parameters listed in the table . to summarize , our sample is composed as follows : * 46 grbs have measured @xmath8 and @xmath0 ( and consequently @xmath1 and @xmath2 ) ; * 3 grbs have measured @xmath0 but only an upper limit on @xmath8 ; * 3 grbs have measured @xmath0 but no redshift estimate ; * 3 grbs have measured @xmath8 but only a lower limit on @xmath9 ; * 3 grbs have measured @xmath8 but no @xmath9 estimate ; it is interesting to compare the @xmath3 and @xmath4 correlations defined by the complete sample with the same correlations defined with a larger sample of grbs , comprising all bursts with measured redshifts and @xmath9 detected by different satellites . we updated also this sample ( called in the following the ` total ' sample ) to may 2011 . the total sample contains 136 bursts , for one of which the peak luminosity can not be derived , due to the lack of measured peak flux . for grbs detected by the gamma ray burst monitor ( gbm meegan et al . 2009 ) onboard _ fermi _ after march 2010 we retrieved the public data from the online catalog and we analysed their time integrated spectra following the standard procedure described in nava et al . ( 2011a ) . this sample is based on the one presented in ghirlanda et al . ( 2012 ) to which we added four bursts ( grb 060306 , grb 061021 , grb 081007 , grb 081221 belonging also to the complete sample - see table [ tab1 ] ) . we can thus compare the spectral energy correlations defined by the complete and by the total samples . to point out possible differences we also consider the sample obtained from the total one by excluding the bursts of the complete sample ( we will refer to it as the ` complementary ' sample ) . we first test the @xmath3 and the @xmath4 correlations with the sample of 46 bursts ( included in the complete sample ) with firm estimates of the redshift and of the spectral properties . the results are shown in fig . data points from the total sample are plotted for comparison as grey dots . red squares show the bursts of the complete sample . we estimate the spearman s rank correlation coefficient and the associated chance probability for the complete , total and complementary sample and report these numbers in table [ tab2 ] . we model the distribution of the data points in the logarithmic plane of fig . [ fig1 ] with a linear function . we applied the ordinary least squares bisector method ( isobe et al . 1990 ) for fitting the data . this choice is motivated by the fact that a priori there is no reason to assume either @xmath0 or @xmath1 ( or @xmath2 ) as the independent variable , and also by the large scatter of the data points . the errors on the slope and normalization are determined by fitting in the barycenter of data points , where they are uncorrelated . the results are given in table [ tab2 ] . in fig . [ fig1 ] the best fit lines are shown for each sample as follows : a solid black line for the total sample , a dashed red line for the complete sample and a dot dashed black line for the complementary sample . the shaded region marks the 3@xmath21 of the scatter distribution of the data points around the best fit line for the total sample . the scatter estimated for the other two samples is also listed in table [ tab2 ] . in fig . [ fig1 ] we also report the position of grb 980425 and grb 031203 ( blue filled stars ) , which are well known outliers of both the @xmath3 and the @xmath4 correlation . these two bursts have peculiar properties and they are generally considered as outliers to the confirmed relations . the behaviour of grb 061021 is very peculiar , since it lies at more than the 3@xmath21 from the amati and yonetoku correlations ( considering the scatter defined by the total sample shaded regions in fig . [ fig1 ] ) . twelve events belonging to the complete sample can not be directly used to test the correlations , since they have some unknown property : the redshift or the spectral peak energy ( see sec . 2 ) . however , we can still test their consistency with the correlations ( results are shown in fig . [ fig2 ] ) . to check the consistency different methods are adopted for the following different cases : * lower limit on @xmath9 and secure estimate of @xmath8 ( 3 cases ) : we derive the lower limit on the rest frame peak energy and on @xmath1 and @xmath2 ( right and up arrows in fig . [ fig2 ] ) ; * no estimate of @xmath9 and secure estimate of @xmath8 ( 3 cases ) : we consider the energy range of sensitivity of the instrument and the photon index of the power photon indices greater ( lower ) than 2 [ in the notation @xmath22 suggest peak energies greater ( lower ) than the upper ( bottom ) energy edge of the instrument . an upper or lower limit to @xmath0 ( and consequently on @xmath1 and @xmath2 ) can be set ( empty triangles in fig . [ fig2 ] ) ; * spectral properties well constrained but upper limit on @xmath8 ( 3 cases ) : @xmath0 , @xmath1 and @xmath2 are estimated for a variable redshift ranging from 0.01 to its upper limit ( coloured continuous curves in fig . [ fig2 ] ) ; * spectral properties well constrained but no redshift estimate ( 3 cases ) : @xmath0 , @xmath1 and @xmath2 are estimated for values of the redshift ranging from 0.01 to 10 ( coloured continuous curves in fig . [ fig2 ] ) . the possibility that evolutionary effects play a role in the spectral energy correlations still represents an open issue . to investigate if the slopes of the @xmath3 and @xmath4 correlations have a dependence on the redshift we repeat the same method used by li ( 2007 ) and ghirlanda et al . ( 2008 ) and apply it to our complete sample . we exclude from the fit the burst which lies at more than 3@xmath21 of the amati and yonetoku relations . we define 4 sub samples , with almost the same number of bursts , divided as follows : ( i ) @xmath23 ( 10 events , yellow empty circles in fig . [ fig3 ] ) , ( ii ) @xmath24 ( 11 events , filled purple stars ) , ( iii ) @xmath25 ( 12 events , filled green triangles ) and ( iv ) @xmath26 ( 12 events , empty red squares ) . bursts belonging to the total sample are not used for this test , since we want to take advantage of the completeness of our sample . the best fit slope of each sub sample is reported in the inset of fig . [ fig3 ] , both for the amati ( left panel ) and yonetoku ( right panel ) correlation . the light blue horizontal stripe , shown in the inset , represents the two sided 1@xmath21 interval of the slope derived by considering the whole complete sample and by excluding ( for consistency ) the two outliers . there is no evidence of a systematic evolution of the slope with @xmath8 . the slopes defined by individual bins of redshift are all consistent , within 1@xmath21 , with the slope defined by the entire complete sample . this result does not change by choosing a different number of bins . bursts with the highest redshifts populate only the high energy ( and high luminosity ) region of the amati and yonetoku this is due to the fact that we selected relatively bright bursts , by requiring a minimum peak flux . the effect of this selection can be investigated as follows . we consider a series of spectra characterised by typical values @xmath27 and @xmath28 . we impose that the photon peak flux in the 15150 kev energy range is equal to our limiting flux ( @xmath29 ph s@xmath11@xmath12 ) . for different values of @xmath9 we derive the corresponding limiting luminosity by considering two cases : ( i ) @xmath30 ( dashed curve in fig . [ fig3 ] ) and ( ii ) @xmath31 ( dot dashed curve ) . we consider these redshifts since they are representative of the two sub samples at higher redshifts , which populate only the high energy ( and high luminosity ) part of the planes . from fig . [ fig3 ] it is clear that this effect is due to the use of a flux limited sample . we may wonder if these limiting luminosities derived at different redshifts may affect our study on the redshift evolution . we identify two possible effects : at high redshift the range of available luminosities is smaller , i.e. the reduced leverage in luminosity might harden the estimate of the correlation slope and enlarge the associated uncertainty . moreover , since the introduced cutoff is nearly vertical in the @xmath0@xmath2 plane , we expect a possible steepening of the slopes toward high redshift . this effect has been investigated also by li ( 2007 ) . by performing simulations , he concludes that a spurious evolution of the slope with the redshift can be introduced , with high - redshift bins defining steeper relations . we do not find signs of this effect in our data ( see the inset in fig . [ fig3 ] ) , indicating that , even if at high redshift we are missing low - luminosity events , we are still able to recover the slope of the correlations which are consistent with those derived with low z bins . 4.5pt @lcccccccccccl@ grb & @xmath8 & @xmath32 [ @xmath33 & fluence & range & peak flux & range & mission & @xmath0 & @xmath1 & @xmath2 & ref . + & & & 10@xmath34erg/@xmath35 & kev & erg/@xmath35/s&kev & & kev&10@xmath36erg&10@xmath37erg / s & + 050318 & 1.44 & [email protected] & [email protected] & 15 - 350 & ( [email protected])@xmath39 & 15 - 150 & s & 115@xmath3827 & [email protected] & [email protected] & 1 , 2 + 050401 & 2.90 & -1.0 [ -2.45 ] & [email protected] & 20 - 2000 & ( [email protected])@xmath40 & 20 - 2000 & k & 499@xmath38117 & [email protected] & [email protected] & 3 + 050416a & 0.653 & -1.0 [ -3.4 ] & [email protected] & 15 - 150 & [email protected]@xmath41 & 15 - 150 & s & 26@xmath384 & [email protected] & [email protected] & 4 + 050525a & 0.606 & [email protected] & [email protected] & 15 - 350 & [email protected]@xmath41 & 15 - 350 & s & 127@xmath386 & [email protected] & [email protected] & 5 + 050802 & 1.71 & [email protected] & [email protected] & 15 - 350 & ( [email protected])@xmath42 & 15 - 150 & s & @xmath43192 & @xmath433.19 & @xmath438.90 & 6 , 7 + 050922c & 2.198 & [email protected] & [email protected] & 2 - 400 & ( [email protected])@xmath40 & 20 - 2000 & h / k&416@xmath38118 & [email protected] & [email protected] & 8,9 + 060206 & 4.048 & [email protected] & [email protected] & 15 - 150 & ( [email protected])@xmath39 & 15 - 150&s & 409@xmath38116 & [email protected] & [email protected] & 7 + 060210 & 3.91 & [email protected] & [email protected] & 15 - 150 & [email protected]@xmath41 & 15 - 150 & s&574@xmath38187 & [email protected] & [email protected] & 10,11 + 060306 & 3.5 & -1.2 @xmath380.5 & [email protected] & 15 - 350 & ( [email protected])@xmath42 & 15 - 150&s & 315@xmath44 & [email protected] & [email protected] & 6 , 7 + 060614 & 0.125 & & [email protected] & 15 - 150 & [email protected]@xmath41 & 15 - 150 & s & 55@xmath3845 & [email protected] & [email protected] & 12,13 + 060814 & 1.92 & [email protected] & [email protected] & 20 - 1000 & ( [email protected])@xmath40 & 20 - 1000&k & 750@xmath38245 & [email protected] ) & [email protected] & 14 + 060904a & & [email protected] & [email protected] & 15 - 1000 & ( [email protected])@xmath40 & 20 - 10000 & k / su / s & 235@xmath3825@xmath41 & & & 15 + 060908 & 1.88 & [email protected] & [email protected] & 15 - 150 & ( [email protected])@xmath39 & 15 - 150 & s & 426@xmath38207 & [email protected] & [email protected] & 7 + 060912a@xmath41 & 0.94 & [email protected] & [email protected] & 15 - 1000 & ( [email protected])@xmath40 & 20 - 10000&k / su / s & & & & 15 + 060927 & 5.47 & [email protected] & [email protected] & 15 - 150 & ( [email protected])@xmath39 & 15 - 150 & s & 459@xmath3890 & [email protected] & [email protected] & 7 + 061007 & 1.261 & [email protected] [ [email protected] ] & [email protected] & 15 - 1000 & ( [email protected])@xmath45 & 20 - 10000 & k / su / s & 965@xmath3827&[email protected] & [email protected] & 15 + 061021 & 0.346 & [email protected] & [email protected] & 20 - 2000 & ( [email protected])@xmath40 & 20 - 2000&k & 1046@xmath38485 & [email protected] & [email protected] & 16 + 061121 & 1.314 & [email protected] & [email protected] & 20 - 5000 & ( [email protected])@xmath45 & 20 - 5000&k & 1402@xmath38185 & [email protected] & [email protected] & 17 + 061222a & 2.09 & [email protected] [ [email protected] ] & [email protected] & 15 - 1000 & ( [email protected])@xmath46 & 20 - 10000&k / su / s & 1091@xmath38167 & [email protected] & 140@xmath3838 & 15 + 070306 & 1.50 & [email protected] & [email protected] & 15 - 350 & ( [email protected])@xmath42 & 15 - 150 & s & @xmath43263 & @xmath438.74 & @xmath439.99 & 6,7 + 070328 & @xmath74 & [email protected] [ [email protected] ] & [email protected] & 15 - 1000 & ( [email protected])@xmath46 & 20 - 10000&k / su / s & @xmath47 & @xmath48 & @xmath49&15 + 070521 & 1.35 & [email protected] & [email protected] & 20 - 1000 & ( [email protected])@xmath40 & 20 - 1000&k & 522@xmath3856 & [email protected] & [email protected] & 18 + 071020 & 2.145 & [email protected] & [email protected] & 20 - 2000 & ( [email protected])@xmath40 & 20 - 2000 & k & 1013@xmath38204 & [email protected] & 213@xmath3873 & 19 + 071112c@xmath41 & 0.82 & [email protected] & [email protected] & 15 - 150 & 8@xmath381@xmath41 & 15 - 150 & s & & & & 20 + 071117 & 1.331 & [email protected] & [email protected] & 20 - 1000 & ( [email protected])@xmath40 & 20 - 1000 & k & 648@xmath38317 & [email protected] & [email protected] & 21 + 080319b & 0.937 & [email protected] [ [email protected] ] & 613@xmath3813 & 20 - 7000 & ( [email protected])@xmath45 & 20 - 7000 & k & 1307@xmath3843 & 142@xmath383 & [email protected] & 22 + 080319c & 1.95 & [email protected] & [email protected] & 20 - 4000 & ( [email protected])@xmath40 & 20 - 4000 & k & 1752@xmath38504 & [email protected] & [email protected] & 23 + 080413b & 1.10 & [email protected] & [email protected] & 15 - 150 & ( [email protected])@xmath39 & 15 - 150 & s & 163@xmath3834 & [email protected] & [email protected] & 7 + 080430@xmath41 & 0.77 & [email protected] & [email protected] & 15 - 150 & ( [email protected])@xmath42 & 15 - 150 & s & & & & 7 + 080602 & @xmath131.4 & [email protected] & [email protected] & 20 - 1000 & ( [email protected])@xmath40 & 20 - 1000 & k & @xmath43542 & @xmath434.34 & @xmath4325.3 & 24 + 080603b & 2.69 & [email protected] & [email protected] & 20 - 1000 & ( [email protected])@xmath40 & 20 - 1000 & k & 376@xmath38214 & [email protected] & 116@xmath3830 & 25 + 080605 & 1.64 & [email protected] & [email protected] & 20 - 2000 & ( [email protected])@xmath45 & 20 - 2000 & k & 665@xmath3848 & [email protected] & 308@xmath3864&26 + 080607 & 3.036 & [email protected] & [email protected] & 20 - 4000 & ( [email protected])@xmath50 & 20 - 4000 & k & 1691@xmath38169 & 186@xmath3810 & 2259@xmath38453 & 27 + 080613b & & [email protected] & [email protected] & 20 - 3000 & ( [email protected])@xmath40 & 20 - 3000 & k & 733@xmath38239@xmath41 & & & 28 + 080721 & 2.591 & [email protected] [ [email protected] ] & [email protected] & 20 - 7000 & ( [email protected])@xmath45 & 20 - 7000 & k & 1785@xmath38223 & 121@xmath3810 & 1038@xmath38172 & 29 + 080804 & 2.20 & [email protected] & [email protected] & 8 - 35000 & ( [email protected])@xmath39 & 8 - 35000&f & 810@xmath3845 & [email protected] & [email protected] & 30 + 080916a & 0.689 & [email protected] & [email protected] & 8 - 35000 & ( [email protected])@xmath39 & 8 - 35000 & f & 208@xmath3811 & [email protected] & [email protected] & 30 + 081007 & 0.53 & [email protected] & [email protected] & 25 - 900 & [email protected]@xmath41 & 25 - 900 & f & 61@xmath3815 & [email protected] & [email protected] & 31 + 081121 & 2.512 & [email protected] [ [email protected] ] & [email protected] & 8 - 35000 & ( [email protected])@xmath40 & 8 - 35000&f & 608@xmath3842 & [email protected] & 195@xmath3831 & 30 + 081203a & 2.10 & [email protected] & [email protected] & 20 - 3000 & [email protected]@xmath41 & 15 - 150 & s / k & 1541@xmath38756 & [email protected] & [email protected] & 32,33 + 081221 & 2.26 & [email protected] & [email protected] & 8 - 35000 & ( [email protected])@xmath39 & 8 - 35000&f & 284@xmath382 & [email protected] & 100@xmath382 & 30 + 081222 & 2.77 & [email protected] [ [email protected] ] & [email protected] & 8 - 35000 & ( [email protected])@xmath39 & 8 - 35000&f & 630@xmath3831 & [email protected] & [email protected] & 30 + 090102 & 1.547 & [email protected] & [email protected] & 8 - 35000 & ( [email protected])@xmath39 & 8 - 35000&f & 1174@xmath3838 & [email protected] & [email protected] & 30 + 090201 & @xmath74 & [email protected] [ [email protected] ] & [email protected] & 20 - 2000 & ( [email protected])@xmath40 & 20 - 2000 & k & @xmath7790 & @xmath7229 & @xmath71246 & 34 + 090424 & 0.544 & [email protected] [ [email protected] ] & [email protected] & 8 - 35000 & ( [email protected])@xmath46 & 8 - 35000&f & [email protected] & [email protected] & [email protected] & 30 + [ tab1 ] 4.5pt @lcccccccccccl@ grb & @xmath8 & @xmath32 [ @xmath33 & fluence & range & peak flux & range & mission & @xmath0 & @xmath1 & @xmath2 & ref . + & & & 10@xmath34erg/@xmath35 & kev & erg/@xmath35/s & kev & & kev&10@xmath36erg&10@xmath37erg / s & + 090709a & @xmath73.5 & [email protected] [ [email protected] ] & [email protected] & 20 - 3000 & ( [email protected])@xmath40 & 20 - 3000 & k & @xmath71341 & @xmath7229 & @xmath7442 & 35 + 090715b & 3.00 & [email protected] & [email protected] & 20 - 2000 & ( [email protected])@xmath39 & 20 - 2000&k & 536@xmath38164 & [email protected] & [email protected] & 36 + 090812 & 2.452 & [email protected] & [email protected] & 15 - 1400 & [email protected]@xmath41 & 100 - 1000 & su / k / s & 2023@xmath38663 & [email protected] & [email protected] & 37,38 + 090926b & 1.24 & [email protected] & [email protected] & 8 - 35000 & ( [email protected])@xmath39 & 8 - 35000&f & [email protected] & [email protected] & [email protected] & 30 + 091018 & 0.971 & [email protected] & [email protected] & 20 - 1000 & ( [email protected])@xmath39 & 20 - 1000 & k & 55@xmath3826 & [email protected] & [email protected] & 39 + 091020 & 1.71 & [email protected] [ [email protected] ] & [email protected] & 8 - 35000 & ( [email protected])@xmath40 & 8 - 35000 & f & 507@xmath3868 & [email protected] & [email protected] & 30 + 091127 & 0.49 & [email protected] [ [email protected] ] & [email protected] & 8 - 35000 & ( [email protected])@xmath40 & 8 - 35000 & f & [email protected] & [email protected] & [email protected] & 30 + 091208b & 1.063 & [email protected] & [email protected] & 8 - 35000 & ( [email protected])@xmath39 & 8 - 35000&f & 246@xmath3815 & [email protected] & [email protected] & 30 + 100615a & & [email protected] [ [email protected] ] & [email protected] & 8 - 1000 & [email protected]@xmath41 & 8 - 1000 & f & [email protected]@xmath41 & & & 40 + 100621a & 0.542 & [email protected] [ [email protected] ] & 36@xmath384 & 20 - 2000 & ( [email protected])@xmath40 & 20 - 2000 & k & 146@xmath3823 & [email protected] & [email protected] & 41 + 100728b & 2.106 & [email protected] & [email protected] & 8 - 35000 & ( [email protected])@xmath51 & 8 - 35000&f & 404@xmath3829 & [email protected] & [email protected] & 42 + 110205a & 2.22 & [email protected] & [email protected] & 20 - 1200 & ( [email protected])@xmath51 & 20 - 1200 & k & 715@xmath38238 & [email protected] & [email protected] & 43 + 110503a & 1.613 & [email protected] [ [email protected] ] & 26@xmath382 & 20 - 5000 & ( [email protected])@xmath50 & 20 - 5000&k & 572@xmath3850 & [email protected] & 181@xmath3818 & 44 + [ tab1 ] [ cols="<,<,>,^,<,^,^,^",options="header " , ] [ tab2 ] the complete sample adopted in this work has been well selected and the effects induced by the limiting peak flux have been well understood and quantified in sect . 3.1 . on the other hand , up to now , the spectral energy correlations have been studied by means of a sample which includes all possible grbs with known redshift and @xmath0 , detected by different instruments . this total sample , therefore , is affected by several instrumental effects that can play important roles ( as the different detector sensitivities and the different energy ranges on which the detection is performed ) . to estimate these effects and to understand how they influence the properties of the total sample is not trivial . here we can use a different approach and point out possible biases by comparing the results obtained from the total sample with those from the complete one . from the point of view of correlations , we demonstrated that the total sample does not introduce evident biases on their slopes , normalizations and scatters ( see table [ tab2 ] ) . however , this does not necessarily imply that the complete and total samples share the same properties in terms of distributions of @xmath0 , @xmath1 and @xmath2 . we investigate possible effects present in the total sample by comparing the @xmath0 , @xmath1 and @xmath2 distributions of the complete , total and complementary sample . for each quantity , we also apply the kolmogorov - smirnov ( k - s ) test and estimate the probability that two distributions are drawn from the same parent population . in particular , we apply this test to compare i ) the complete and total sample and ii ) the complete and complementary sample . results are shown in fig . [ fig4 ] . in all cases , the k - s probability between the complete and the total sample is large , and the distributions are consistent within 1@xmath21 . the probability slightly decreases when comparing the complete and the complementary sample . in this case all distributions are consistent within 2@xmath21 . a slight effect is possibly present in the @xmath2 distribution , where the selection cut discussed in the previous section lowers the number of intermediate / low luminosity events . due to the relation between @xmath2 and @xmath0 , this effect also influence the @xmath0 distribution . even if numbers are still small for the complete sample , our results suggest that the total sample is not affected by any relevant bias . 0.3 truecm ( rest frame ) , @xmath1 and @xmath2 for bursts of the total ( filled grey histograms ) , complementary ( dashed dark line ) and complete ( hatched red histograms ) sample ( see legend in the bottom panel ) . in each panel the probability of the k - s test is also reported.,title="fig : " ] the definition of a sample of grbs with a high level of completeness ( 90% ) allows to address several open issues regarding population studies of grbs . in this paper we use this complete sample to study the spectral energy correlations in an unbiased way . out of 58 grbs satisfying the selection criteria 46 have well defined spectral properties and measured redshift . in particular , the rest frame peak energy @xmath0 can be determined and its relation with @xmath1 and @xmath2 can be investigated . we find that this sample defines two strong correlations : ( i ) @xmath3 ( rank s correlation coefficient @xmath52 and chance probability p@xmath53 ) and ( ii ) @xmath4 ( @xmath54 and p@xmath55 ) . the slope of the correlations defined by the complete sample are 0.61 and 0.53 for the amati and the yonetoku correlations , respectively . these slopes are consistent within 1@xmath21 ( 2@xmath21 ) with the slopes of the @xmath4 ( @xmath3 ) correlation defined by bursts not satisfying the completeness criteria ( complementary sample ) . from table [ tab2 ] we also note that the scatter of the correlations defined with the complete , total and complementary sample are similar among themselves and are about @xmath56 and @xmath57 for the amati and yonetoku correlation . we also investigated possible instrumental effects in the correlations found in this complete sample . we defined three sub - samples based on the instrument which derived the burst spectral parameters ( see table 1 ) . in particular , we study the amati and yonetoku correlations within the sample of swift / bat bursts ( 10 events ) , konus bursts ( 18 events ) and fermi / gbm bursts ( 13 events ) . the results derived from different sub - samples are very similar and consistent within the errors . in fact , we found that the best fit slopes ( normalizations ) of the amati correlation are [email protected] ( [email protected] ) , [email protected] ( [email protected] ) and [email protected] ( [email protected] ) for the swift , konus and fermi samples respectively . for the yonetoku correlation we find : [email protected] ( [email protected] ) , [email protected] ( [email protected] ) and [email protected] ( [email protected] ) . our study outlines the presence of one grb ( 061021 ) that lies at 3@xmath21 limit ( or more ) of the tested correlations . in particular , it is an outlier to the amati correlation . grb 061021 has been detected by _ swift_/bat ( moretti et al . 2006 ) and by _ konus_/wind ( golenetskii et al . it shows a single pulse with a duration of @xmath58 s , followed by a weak tail seen up to @xmath59 s. on the basis of the lag analysis ( norris & barthelmy 2006 ) this burst is classified as a long duration event . the time integrated bat spectrum is well described by a power law model with photon index 1.31@xmath60 ( palmer et al . 2006 ) , suggesting @xmath9@xmath61 kev . golenetskii et al . ( 2006 ) report the spectral analysis of _ konus_/wind data separately for the pulse and the tail . the pulse is well described by a power law with a high energy exponential cutoff , with spectral parameters @xmath62 and @xmath9@xmath63 . the soft tail can be modelled as a single power law , with photon index @xmath64 and a fluence which is nearly 1/3 of the fluence of the main pulse . in order to avoid to underestimate the total energetics , we also include the fluence of the soft tail in the computation of @xmath1 . nevertheless , this burst is on average @xmath132 orders of magnitude less energetic as compared to bursts characterized by a similar peak energy . note that the lack of a spectral analysis integrated over the whole burst duration does not affect the estimate of @xmath2 , which is well computed by using the peak flux and the spectral shape of the main peak . also in this case , the peak luminosity is @xmath132 orders of magnitude less than the average @xmath2 of events with @xmath0 similar to grb 061021 . finally , the redshift of this source is @xmath65 ( fynbo et al . 2009 ) . the presence of outliers to the amati relation is somehow expected from the study of the observational plane @xmath9@xmath66 . nava et al . ( 2008 ; 2011b ) studied the distribution of batse and gbm bursts in this plane and derived respectively a 6% and 3% of outliers . we also stress that the definition of the outliers of the @xmath3 correlation refers to the present knowledge of the scatter of this correlation which is assumed to be gaussian . it might well be that the scatter is still not fully known and/or that it is not gaussian . the latter possibility seems to be suggested also by the log normal distribution of the jet opening angles ( ghirlanda et al . 2005 ) if the source of the scatter of the @xmath3 correlation were , indeed , the jet opening angle . there are 12 bursts belonging to the complete sample that can not be used to directly test the correlations , due to the lack of measured redshift and/or @xmath9 . however , we can still check their consistency with the correlations ( fig . [ fig2 ] ) . we find that they are all consistent with both correlations . by taking advantage of our complete sample , for both correlations we investigate the possible evolution with the redshift of the best fit slopes ( fig . [ fig3 ] ) . to this aim we define 4 bins of redshift , chosen in order to have 4 sub samples characterized by a similar number of objects . we find no relation between the slope of each sub sample and the redshift . all slopes are consistent with the one defined by the whole complete sample and this is true both for the @xmath3 and the @xmath4 correlation . we also investigate possible differences ( in terms of distributions of the most fundamental quantities ) between bursts belonging to the complete sample and bursts that do not satisfy the complete sample selection criteria . our aim is to understand if the different instrumental biases present in the total sample affect in some way the distributions of the prompt emission properties . we find that for all the considered quantities ( @xmath0 , @xmath1 and @xmath2 ) the distributions derived from the complete and from the complementary samples are consistent within 2@xmath21 . our results give support to the idea that spectral energy correlations arise because of a robust physical mechanism common to the vast majority of bursts . outliers exist , but they are a few . furthermore , the evidence that grbs do evolve in redshift ( salvaterra & chincarini 2007 ) , while there is no sign of evolution of the slopes of the correlations , also suggests that the spectral peak energy is closely linked to the luminosity / energetics of the burst , another sign that selection effects are not important . we thank the referee for comments that improved the manuscript . this work has been supported by asi grant i/004/11/0 and asi grant i/088/06/0 . this research has made use of the public fermi / gbm data and software obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center . 99 amati l. , frontera f. , tavani m. , et al . , 2002 , a&a , 390 , 81 amati l. , 2006 , mnras , 372 , 233 amati l. , 2007 , ncimb , 121 , 1081 amati l. , della valle m. , frontera f. , malesani d. , guidorzi c. , montanari e. , pian e. , 2007b , a&a , 463 , 913 amati l. , 2010 , arxiv:1002.2232 band d. l. , & preece r. d. , 2005 , apj , 627 , 319 barthelmy s. , barbier l. , cummings j. , et al . , 2006 , gcn 5256 bissaldi e. , mcbreen s. , connaughton v. , 2008 , gcn 8369 blustin a. j. , band d. , barthelmy s. , et al . , 2005 , apj , 637 , 901 butler n. r. , kocevski d. , bloom j. s. , curtis , j. l. , 2007 , apj , 671 , 656 cabrera j. i. , firmani c. , avila - reese v. , ghirlanda g. , ghisellini g. , nava l. , 2007 , mnras , 382 , 342 crew g. , ricker g. , atteia j - l . , et al . , 2005 , gcn 4021 eichler d. , & levinson a. , 2004 , apj , 614 , l13 fynbo j. p. u. , jakobsson p. , prochaska j. x. , et al . , 2009 , apjs , 185 , 526 foley s. , briggs m. , 2010 , gcn 10851 ghirlanda g. , ghisellini g. , lazzati d. , 2004b , apj , 616 , 331 ghirlanda g. , ghisellini g. , lazzati d. , firmani c. , 2004a , apj , 613 , l13 ghirlanda g. , ghisellini g. , frimani c. , 2005 , mnras , 361,10 ghirlanda g. , ghisellini g. , frimani c. , 2006 , njph , 8 , 123 ghirlanda g. , nava l. , ghisellini g. , firmani c. , cabrera j. i. , 2008 , mnras , 387 , 319 ghirlanda g. , nava l. , ghisellini g. , celotti a. , firmani c. , 2009 , a&a , 496 , 585 ghirlanda g. , nava l. , ghisellini g. , 2010 , a&a , 511 , 43 ghirlanda g. , ghisellini g. , nava l. , burlon d. , 2011a , mnras , 410 , 47 ghirlanda g. , ghisellini g. , nava l. , 2011b , mnras , 418 , 109 ghirlanda g. , nava l. , ghisellini g. , celotti a. , burlon d. , covino s. , melandri a. , 2012 , mnras , 420 , 483 ghisellini g. , ghirlanda g. , mereghetti s. , bosnjak z. , tavecchio f. , firmani c. , 2006 , mnras , 372 , 1699 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2005 , gcn 3179 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2005b , gcn 4030 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2006a , gcn 5460 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2006b , gcn 5748 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2006c , gcn 5837 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2007a , gcn 6459 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2007b , gcn 6960 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , oleynik v. , ulanov m. , cline t. , 2007c , gcn 7114 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2008a , gcn 7487 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2008b , gcn 7784 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2008c , gcn 7812 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , cline t. , 2008d , gcn 7854 golenetskii s. , aptekar r. , mazets e. , et al . , 2008e , gcn 7862 golenetskii s. , aptekar r. , mazets e. , et al . , 2008f , gcn 7884 golenetskii s. , aptekar r. , mazets e. , et al . , 2008 g , gcn 8611 golenetskii s. , aptekar r. , mazets e. , et al . , 2009a , gcn 8878 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , oleynik p. , ulanov m. , svinkin d. , 2009b , gcn 9647 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , oleynik p. , ulanov m. , svinkin d. , 2009c , gcn 9679 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , oleynik p. , ulanov m. , svinkin d. , 2009d , gcn 10045 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d. , oleynik p. , ulanov m. , svinkin d. , 2010 , gcn 10882 golenetskii s. , aptekar r. , mazets e. , et al . , 2011a , gcn 11659 golenetskii s. , aptekar r. , mazets e. , et al . , 2011b , gcn 12008 isobe t. , feigelson e. d. , akritas m. g. , babu g. j. , 1990 , apj , 364 , 104 jakobsson p. , levan a. , fynbo j. p.u . , et al . , 2006 , a&a , 447 , 897 krimm h. , barbier l. , barthelmy s. , et al . , 2007 , gcn 7081 lenvinson , a. , & eichler , d. 2005 , apj , 629 , l13 li l .- x . , 2007 , mnras , 379 , 55 liang e. , & zhang , b. 2005 , apj , 633 , 611 markwardt c.b . , chester m. m. , cummings j. r. , et al . , 2007 , gcn 6224 meegan c. , lichti g. , bhat p. n. , et al . 2009 , apj , 702 , 791 moretti a. , barthelmy s. d. , gehrels n. , et al . , 2006 , gcn 5743 nakar e. , & piran t. , 2005 , mnras , 360 , l73 nava l. , ghirlanda g. , ghisellini g. , firmani c. , 2008 , mnras , 391 , 639 nava l. , ghirlanda g. , ghisellini g. , celotti a. , 2011a , a&a , 530 , 21 nava l. , ghirlanda g. , ghisellini g. , celotti a. , 2011b , mnras , 415 , 3453 noda k. , sonoda e. , ohmori n. , et al . , 2009 , gcn 9816 norris j. , & barthelmy s. d. , 2006 , gcn 5749 palmer d. , barbier l. , barthelmy s. d. , et al . , 2006 , gcn 5744 palshin v. , golenetskii s. , aptekar r. , mazets e. , frederiks d. , cline t. , 2009 , gcn 9821 panaitescu a. , 2009 , mnras , 393 , 1010 perri m. , giommi p. , capalbi m. , et al . , 2005 , a&a , 442l , 1 racusin j. l. , karpov s. v. , sokolowski m. , et al . , 2008 , nature , 455 , 183 rees m. , & mszros p. , 2005 , apj , 628 , 847 sakamoto t. , barbier l. , barthelmy s. d. , et al . , 2006a , apj , 636 , l73 sakamoto t. , barbier l. , barthelmy s. d. , et al . , 2006b , gcn 4748 sakamoto t. , barthelmy s. d. , barbier l. , et al . , 2008 , apjs , 175 , 179 sakamoto t. , sato g. , barbier l. et al . , 2009 , apj , 693 , 922 sakamoto t. , palshin v. , yamaoka k. , et al . , 2011a , pasj , 63 , 215 sakamoto t. , barthelmy s. d. , baumgartner w. h. , et al . , 2011b , apjs , 195 , 2 salvaterra r. , & chincarini g. , 2007 , apj , 656 , 49 salvaterra r. et al . , 2012 , apj in press , arxiv:1112.1700 shahmoradi a. , & nemiroff r. j. , 2011 , mnras , 411 , 1843 stamatikos m. , barbier l. , barthelmy s. d. , et al . , 2007 , gcn 6225 starling r. l. c. , rol e. , van der horst a. j. , et al . , 2009 , mnras , 400 , 90 thompson c. , mszros p. , & rees m. , 2007 , apj , 666 , 1012 ukwatta t. n. , barthelmy s. d. , baumgartner w. h. , et al . , 2008 , gcn 8599 yonetoku d. , murakami t. , nakamura t. et al . 2004 , apj , 609 , 935 |
the per ob1 association , which contains the remarkable double cluster and per , is unusual in not having a giant molecular cloud in its vicinity . we show from _ hipparcos _ data that the luminous members of this association exhibits a bulk motion away from the galactic plane , such that their average velocity increases with height above the galactic plane . we find haebe and t tauri stars towards probable remnant molecular clouds associated with per ob1 . these star - forming regions lie well beyond the location of the luminous member stars at heights of 280400 pc above the galactic plane , far higher than that previously found for embedded clusters . we argue that the observed motion of the luminous member stars is most naturally explained if many formed from molecular gas pushed and accelerated outwards by an expanding superbubble driven presumably by stellar winds and perhaps also supernova explosions . a large shell of atomic hydrogen gas and dust that lies just beyond the remnant molecular clouds , believed to be driven by just such a superbubble , may comprise the swept - up remains of the parental giant molecular cloud from which this association formed . in support of this picture , we find a week trend for the younger o star members to lie at higher galactic latitudes than the older supergiant members . the star - forming regions located at even larger heights above the galactic plane presumably correspond to more recent episodes of star formation at or near the periphery of this superbubble . | 0804.4520 | o and b stars are not distributed at random in the sky , but instead are usually concentrated in loose co - moving groups that @xcite first referred to as ob associations . ob associations are generally distributed along spiral arms , and confined in the galactic plane . they have low stellar densities of order 0.1 m@xmath2 pc@xmath3 , and are easily disrupted by galactic tidal forces @xcite so must be relatively young with ages @xmath4 myr ( see , e.g. , reviews by * ? ? ? some of the best evidence that the formation of massive stars can induce the formation of other low- to high - mass stars have been found in ob associations . on scales up to a few tens of parsecs , massive stars can ionize the surfaces of small molecular clouds ( globules ) in their vicinity @xcite ; the pressure from such ionization fronts can then propagate through the globule to help overcome the magnetic , turbulent , and thermal pressure that supports it against collapse , thereby inducing star formation . this process of radiation - driven implosion has been invoked to trigger star formation in many globules within ob associations @xcite , not least by one of us in the orion region @xcite and lac ob1 @xcite . on the other hand , star - forming globules in ob associations may simply be dense cores that have spontaneously collapsed to form stars , and were then uncovered by advancing ionization fronts from nearby massive stars . in the examples described in @xcite and @xcite , classical t tauri stars ( cttss ) are found only between ob stars and bright - rimmed ( i.e. , ionized surface of ) molecular clouds , not within these clouds ; furthermore , those cttss closer to the bright - rimmed clouds are progressively younger , arguing for a genuine sequence in star formation . on larger scales of tens of parsecs or more , expanding superbubbles driven by ionizing radiation and stellar winds from massive stars , together perhaps with supernova explosions , have been invoked to both sweep up and compress gas , and subsequently induce star formation ( see , e.g. , review by * ? ? ? such large - scale triggerings of star formation may be required to explain co - moving ob subgroups with either identical ages or arranged in an apparent age sequence , with at least some subgroups having an age spread that is much smaller than their stellar crossing time . one of the best such examples is the scorpius - centaurus ( scocen ) ob association ( see , e.g. , review by * ? ? ? * ) , which is the closest ob association to the sun . it comprises the three co - moving subgroups upper scorpius ( usco ) with an age of @xmath55 myr , upper centaurus - lupus ( ucl ) with an age of @xmath517 myr , and lower centaurus - crux ( lcc ) with an age of @xmath516 myr distributed from approximately east to west in galactic longitude over a spatial extent of @xmath5130 pc @xcite . the age dispersion in the usco subgroup ( the youngest ob subgroup ) is no more than @xmath512 myr , compared with a stellar crossing time of @xmath520 myr @xcite . @xcite and @xcite proposed that subsequent supernova explosions in lcc induced star formation in the molecular cloud that formed usco . the abovementioned picture of triggered star formation is based on an updated variation of the collect - and - collapse model of @xcite , where a propagating disturbance first collects gas into dense clouds and subsequently induces star formation . here we present our study of the perseus ob1 ( per ob1 ) association , which as shall show has properties that can be most naturally explained by the cloud - shuffling model of triggered star formation as first proposed by @xcite . cloud shuffling takes place on the periphery of an ob association where a superbubble pushes and accelerates a giant molecular cloud outwards , triggering star formation at the dense cloud interface @xcite . until now , cepheus ob3 ( cep ob3 ) has been regarded as one of the best examples where star formation has been triggered by cloud shuffling @xcite . cep ob3 contains two ob subgroups , the younger ( @xmath54 myr ) of which is closer to and has a radial velocity resembling the cep b molecular cloud , and the older ( @xmath5 8 myr ) further away and has a radial velocity resembling the rest of the cloud . @xcite and @xcite proposed that ionizing radiation and stellar winds from the older ob subgroup pushed and accelerated the cep b cloud away , as well as induced star formation in this cloud to form the younger subgroup . our paper is organized as follows . in 2 we collect together previous studies that compile the proposed members of per ob1 , and describe our study of the proper motion of those members included in the _ hipparcos _ catalog ; we also describe our search for low to intermediate - mass pre - main - sequence stars towards per ob1 , and our identification of possible remnant molecular clouds in this association . in 3 we interpret and discuss our results in the context of current star formation models . in 4 we briefly summarize the most important results of our study and their chief implications . the per ob1 association , located in the region of the perseus spiral arm , is commonly ( but not without argument ) thought to be centered about the @xmath0 and @xmath1 per double cluster ( see discussion in * ? ? ? . widely referred to as the double cluster , @xmath0 and @xmath1 per are among the brightest , densest , and closest of open clusters containing moderately massive ( main - sequence b , along with supergiant ) stars . both clusters have been studied extensively over the last century ( see references in * ? ? ? the latest study by @xcite conclude that the two clusters have essentially identical ages of @xmath6 myr , distances of @xmath7 kpc , comparable initial masses ( integrating between 1120 m@xmath2 ) of 3700 m@xmath2 for @xmath0 per and 2800 m@xmath2 for @xmath1 per , and a salpeter imf . although containing more than ten o stars , per ob1 is not associated with any known regions . furthermore , unlike virtually all other ob associations studied in molecular gas , per ob1 is not associated with any known giant molecular cloud . @xcite quote a failed attempt to detect molecular gas in unpublished @xmath8co observations by l. a. hillenbrand & j. m. carpenter using the fcrao . in addition , they find that the reddening to the @xmath0 and @xmath1 per double cluster is indistinguishable and produced entirely by line - of - sight extinction to the perseus spiral arm , with the intracluster reddening essentially zero . @xcite collected from the literature all the then known early and late - type supergiant stars , o stars , and both main - sequence as well as giant b0b1 stars that were proposed to be members of this association . she recomputed their photometric distances using the then latest available luminosity calibration for o and b stars , as well as for late - type supergiants . a total of 105 members were identified , all of which also have comparable radial velocities . all the member stars lie within the range in galactic longitude of @xmath9 and galactic latitude of @xmath10 . the most recent compilation from the literature of all proposed early - type luminous members of per ob1 was made by @xcite . by studying where the early - b ( b0b2 ) main - sequence stars lie on the color - magnitude diagram ( i.e. , cluster fitting method ) , @xcite derived a distance modulus for related members of @xmath511.8 , which corresponds to a distance of @xmath52.3 kpc . next to trigonometric parallaxes , this is the most secure method for identifying cluster members . @xcite also identified possible members from among the early - type giant to supergiant stars based on the then best available color - magnitude relations . in this way , they identified a total of 149 members in per ob1 , including @xmath5@xmath11 of the stars listed in @xcite as members of this association . note that @xcite only considered early - type ( o a ) stars , whereas the cataloged members of @xcite also include late - type ( supergiant ) stars . the per ob1 members identified by @xcite span a significantly larger range in galactic longitude of @xmath12 as well as galactic latitude of @xmath13 . in figure 1 , we plot the proposed members of per ob1 as cataloged by @xcite and @xcite in galactic coordinates . [ we exclude hd 237019 , whose position has been incorrectly entered in table 2a of @xcite . this star is a member of cas ob6 @xcite . ] a clustering of b stars ( green circles ) can be seen towards two closely separated positions , corresponding to the @xmath0 and @xmath1 per double cluster ( indicated by large black circles ) . no clear clustering is seen for the o stars ( blue circles ) , nor the supergiant stars ( red circles ) . the relative deficiency in cluster members at low galactic latitudes ( @xmath14 ) appears to be real ; as can be seen in figure 1 of @xcite for per ob1 , many of the early - type stars at @xmath15 were found to be either field stars or members of other associations . on the other hand , it is not clear to us whether the identification of luminous early - type stars is as complete at low galactic heights ( given the larger extinction and greater confusion ) as at larger heights . from a search of the literature since the work by @xcite , we have found three more luminous stars toward per ob1 that have a comparable distance modulus as inferred by @xcite for this association . these are the wc6 star hd 17638 ( distance 1.91 kpc , * ? ? ? * ) , the o9.5iii(n ) star hd 15137 ( distance 2.65 kpc , * ? ? ? * ) , and the on8v star hd 14633 ( distance 2.15 kpc , * ? ? ? the locations of these three stars are indicated in figure 2 . the radial velocities of hd 15137 and hd 14633 @xcite are closely comparable with the range observed for the per ob1 association . these two stars , especially hd 14633 , lie at a significantly higher galactic latitude than those cataloged by either @xcite or @xcite . the hipparcos astrometry mission @xcite permits a study of the proper motion of stars in ( or towards ) per ob1 . note that the stars in this association are too far away for hipparcos to obtain a measurement of their trigonometric parallax , which is predicted to be @xmath50.4 mas compared with measurement uncertainties of @xmath51 mas @xcite . sixty - one out of a total of 180 stars cataloged by @xcite and @xcite as members of per ob1 , as well as the three new additions mentioned in @xmath16 , are contained in the hipparcos catalog . the positions of these stars , converted to galactic coordinates , are listed in table [ table : members ] together with their hipparcos catalog ( hip ) number , proper motion in galactic longitude and latitude along with their associated standard errors , star name in the henry - draper ( hd ) or bonner durchmusterung ( bd ) catalog , spectral type , and whether these stars are listed as members by @xcite or @xcite or both . most of these stars are either early- or late - type giants and supergiants . in figure 2 , we plot the proper motion of the stars listed in table [ table : members ] in galactic coordinates . the blue , green , and red arrows are o , b , and supergiant stars respectively . the arrows are plotted with lengths proportional to the magnitude of their proper motions : note , however , that the uncertainty in proper motion changes from star to star , and the uncertainty in galactic longitude can be quite different to that in galactic latitude . we also plot the proper motion of the @xmath0 and @xmath1 per clusters ( black arrows ) as determined by @xcite based on space mission and ground - based observations . h per has a proper motion in ( @xmath17 , @xmath18 ) of ( @xmath19 mas yr@xmath20 , @xmath21 mas yr@xmath20 ) , and @xmath1 per a proper motion of ( @xmath22 mas yr@xmath20 , @xmath23 mas yr@xmath20 ) . for comparison , we also plot the proper motions of all the other early - type ( b0b9 ) stars ( yellow arrows ) in the hipparcos catalog within the field encompassed by figure 2 . these stars are referred to hereafter as `` field '' stars , although some may be members of other clusters or associations , and yet others perhaps previously unidentified members of per ob1 as we explain below . the white arrow oriented at a position angle of @xmath24 ( measured anticlockwise from galactic north ) indicates the direction towards the solar apex ; i.e. , the direction ( as projected in the sky ) towards which the sun is moving relative to the local standard of rest . as can be seen in figure 2 , some of the field stars ( yellow arrows ) have very large proper motions that in many cases lie at position angles of roughly @xmath25 . these are relatively nearby stars ( as confirmed from an examination of their trigonometric parallaxes ) that , because of our motion relative to the local standard of rest , appear to be moving in a direction opposite to that of the sun ( white arrow in fig . 2 ) . the proposed members of the per ob1 association , being much more distant , do now show any such appreciable motion . to study the actual proper motion of the proposed per ob1 members relative to the galaxy at their ( presumed ) distance ( 2.3 kpc ) and direction , we subtract our motion relative to the local standard of rest and galactic rotation at a distance of 2.3 kpc in the direction of the individual members . the correction for @xmath0 per is 0.27 mas yr@xmath20 in @xmath17 and @xmath26 mas yr@xmath20 in @xmath18 , giving an actual proper motion of ( @xmath27 mas yr@xmath20 , @xmath28 mas yr@xmath20 ) ; for @xmath1 per , the correction is @xmath29 mas yr@xmath20 in @xmath17 and @xmath26 mas yr@xmath20 in @xmath18 , giving an actual proper motion of ( @xmath30 mas yr@xmath20 , @xmath31 mas yr@xmath20 ) . thus , the double cluster does not exhibit significant motion relative to the galaxy rotation and galactic plane at its location . we tabulate the average measured proper motion of the remaining luminous member stars in galactic latitude strips of width @xmath32 from latitudes of @xmath33 to @xmath34 in table [ table : pm ] . ( note that the measured proper motion of the individual stars is in general of low statistical significance ) . the average proper motion in longitude is not significantly ( i.e. , @xmath35 ) different from zero except for the two individual stars at the two highest latitudes ( hd 15137 at @xmath36 and hd 14633 at @xmath37 ) . the average proper motion in latitude is significantly ( at least 4@xmath38 ) different from zero at latitudes @xmath39 , including the two individual stars at the two highest latitudes . to better visualize the results , in figure 3 we plot the proper motion of the individual members in galactic latitude , as well as their proper motion averaged over the different latitude strips . apart from the abovmentioned motion away from the galactic plane at @xmath39 , there is also a clear trend towards larger motions away from the galactic plane at higher latitudes , with the star at the highest latitude ( hd 14633 ) exhibiting the largest proper motion away from the galactic plane . note that the members of per ob1 cataloged by @xcite have in most cases radial velocities in the range @xmath40 to @xmath41 , the same range in radial velocities as the perseus spiral arm . we therefore conclude that the bulk motion of the stars in per ob1 is predominantly away from the galactic plane , with velocities up to a few tens of km s@xmath20 for most of the stars with measured proper motions . a visual inspection of figure 2 reveals that many of the `` field '' stars at latitudes @xmath42 ( yellow arrows ) spanning approximately the same longitude range as the proposed association members also exhibit a preferential motion away from the galactic plane . the average proper motion of the stars in this latitude strip and in the longitude range @xmath43 is ( @xmath44 mas / yr , @xmath45 mas / yr ) . their average proper motion in galactic latitude is comparable with the values tabulated in table [ table : members ] for the proposed member stars in this latitude strip , and hence these stars are likely genuine members of per ob1 . most of these stars have spectral types in the range b5b9 , and would therefore not have been included in the catalogs of either @xcite or @xcite . to study whether there has been any recent ( within the last @xmath51 myr or so , compared with an age for the @xmath0 and @xmath1 per double cluster for @xmath5 13 myr ) formation of low- to intermediate - mass stars in per ob1 , we have searched the two micron all - sky survey ( 2mass ) point source catalog @xcite for pre - main - sequence ( pms ) star candidates . our selection of such candidate stars was based on the empirical criteria developed by @xcite for low - mass cttss and intermediate - mass herbig ae / be ( haebe ) stars , but with improved criteria here for better selecting the latter stars . with this method , we first select stars showing near - infrared ( near - ir ) excess thereby preferentially picking out pms stars . because haebe stars generally exhibit a stronger near - ir excess than ctts , we then separated these two stellar classes based on their 2mass colors as described below . our initial list comprised all stars in the 2mass catalog at @xmath46 and @xmath47 ( i.e. , spanning a much wider area in the sky than the members of per ob1 as identified by * ? ? ? * ) that exhibit a near - ir excess . in figure 4 , we plot the ( j - h ) versus ( h - k ) color diagram for selecting pms star candidates . ctts candidates are those defined to lie between the two parallel lines @xmath48 and @xmath49 , and above the dereddened ctts locus @xmath50 @xcite . haebe candidates are defined as those lying between the two parallel lines @xmath49 and @xmath51 , and above the line @xmath52 . the line @xmath52 helps discriminates classical be stars from haebe stars . a full explanation for the adopted criteria will be presented elsewhere @xcite . in this way , we found a total of 399 candidate ctts and 264 candidate haebe stars . in figure 5 , we plot the spatial distribution of the candidate pms stars ( red for candidate ctts , green for candidate haebe stars , and blue for early - type members of per ob1 ) . at low galactic latitudes of @xmath53 , apart from three clusters associated with the known hii regions w3 , w4 , and w5 of the cas ob6 association , the candidate pms candidates ( which are presumably confused with cool evolved stars ) are roughly equally distributed in galactic longitude . in figure 5 , we also plot ( in grayscale ) the @xmath54 100-@xmath55 map ( dominated by emission from interstellar dust ) of the per ob1 region . the map in the right - hand panel has been stretched to accentuate features at high galactic latitudes ( @xmath53 ) coincident with the candidate pms stars selected in the manner described above . as can be seen , around the region occupied by the per ob1 association , a number of the candidate pms stars at galactic latitudes as low as about @xmath56 appear to be close to or coincident with dust clouds . an examination of optical images from the 2nd digitized sky survey ( dss2 ) as shown in figure 6 reveals bright - rimmed clouds , comet - shaped clouds , and reflection clouds coincident with these dust clouds . the abovementioned clouds seen in the dss2 images resemble remnant molecular clouds ( rmcs ; ogura et al . rmcs are small molecular clouds that can be commonly found in the vicinity of ob associations . their outlines are shaped by ionization fronts from massive stars , and hence they often have cometary morphologies with heads pointing towards the ionizing source . there is at least one such previously known example in per ob1 , referred to as red nebulous object 6 ( rno6 ; * ? ? ? @xcite found two molecular clouds , each having a mass of @xmath5@xmath57 , around rno6 . they showed that these clouds have cometary morphologies with their heads pointing towards the @xmath0 and @xmath1 per clusters . they also confirmed that rno6 hosts an embedded ir cluster that includes a hbe star responsible for illuminating rno6 , and suggested that the recent star formation in this rmc was triggered by radiatively driven implosion . as can be seen in figure 6 , we find another candidate haebe apparently associated with rno6 . ( the known hbe star in rno6 does not exhibit a near - ir excess in 2mass . ) based on figure 5 , we have identified four other candidate rmcs towards per ob1 that appear to be closely associated with candidate pms stars . ( we emphasize that our selection method does not provide a complete census of rmcs in this region . ) they are henceforth referred to as rmc1rmc4 with locations as indicated in figure 5 and tabulated in table [ table : rmc ] . a search of the catalog of bright nebulae compiled by @xcite reveals that rmc1 , rmc2 , and rmc3 are associated with lbn 638 , lbn 640 , and lbn 666 respectively . all four candidate rmcs have cometary morphologies as shown in figure 5 and 6 with heads pointing towards the galactic plane , rather than to any particular o stars in per ob1 . to determine if there are other sites of recent star formation in per ob1 apart from rno6 , we performed spectroscopic observations of the candidate pms stars close to or coincident with the abovementioned candidate rmcs ( including rno6 ) . the location of these stars are plotted in figure 6 , and tabulated in table [ table : spectra ] . we also observed four stars apparently associated with reflection nebulae located in the vicinity of , and likely associated with , the rmcs . these stars do not exhibit detectable near - ir excess in 2mass , and hence are not included in our list of candidate pms stars . the locations of these stars also are shown in figure 7 , and tabulated in table [ table : spectra ] . these stars will prove useful in obtaining a distant estimate to the rmcs . we obtained spectra of both the candidate pms stars and stars associated with reflection nebulae at the beijing astronomical observatory ( bao ) in 2005 and 2006 . low - dispersion spectra with a dispersion of 200 mm@xmath20 or 4.8 pixel@xmath20 were taken with a 2.16-m optical telescope on 2005 december 68 . we used an omr ( optomechanics research , inc . ) spectrograph together with a tektronix 1024@xmath581024 ccd detector covering the wavelength range 4000 - 9000 . on 2006 november 1113 , we used the same spectrograph but now deployed with a new princeton instrument spec10 400@xmath581300b ccd detector camera covering the wavelength range 3500 - 8500 . we processed the spectroscopic data in the standard manner using the noao / iraf package . after correcting for bias and flat - fielding , we used the iraf program kpnoslit to extract and calibrate the wavelength and intensity of each spectrum . the results for the individual stars are presented in table [ table : spectra ] . ctts and haebe stars are identified by the presence of h@xmath59 emission , and separated by the presence of other balmer lines seen in absorption in haebe stars . we separate intermediate - mass stars into haebe , b , and a stars based on the spectral classification scheme of @xcite . we do not , however , attempt to determine the spectra types of ctts because their spectra exhibit heavy veiling . of the nine candidate pms stars selected by 2mass colors and observed , we confirmed that four are haebe stars and five are possibly t tauri stars . detection of the li line , impossible in our low - dispersion spectra , is required to confirm the nature of the ctts candidates . we find pms stars in the vicinity of or coincident with all four rmcs . as there are no known foreground ( or background ) star - forming regions in their directions , and given their large heights above the galactic plane ( see more detailed discussion in @xmath60 ) , their association with the rmcs is likely genuine . the rmcs are therefore likely to be regions of recent or ongoing star formation . of the four stars ( stars 5 , 6 , 8 , and 10 ) associated with reflection nebulae studied , three were found to be b main - sequence stars ( stars 5 , 6 , and 10 ) and one an a main - sequence star ( star 8) . we now use the b or a stars associated with reflection nebulae to determine a photometric distance to these stars and hence to two of the candidate rmcs . specifically , we will focus on 2mass j01494099@xmath615341341 ( b7v ; star 5 in table [ table : spectra ] ) , 2mass j01500456 + 5354007 ( b3v ; star 6 ) and 2mass j01501546@xmath615341350 ( a2v ; star 8) apparently associated with rmc2 , and 2mass j02220484@xmath615038130 ( b8v ; star 10 ) with rmc3 . to derive the absolute magnitudes in the j - band of stars with the abovementioned spectral classes , we selected eleven b3v , three b7v , forty - one b8v , and 128 a2v stars with good parallax ( @xmath62 ) measurements , [ relative error in parallax of @xmath63 @xcite , goodness - of - fit parameter @xmath64 3 , and percentage of rejected data @xmath65 @xcite ] , and good 2mass photometric qualities . by selecting only stars with small relative errors in parallax , we hope to minimize the bias in determining absolute magnitudes and hence distances using this method . as pointed out by @xcite , the inclusion of stars with large relative errors in parallax biases the final result to a lager parallax than the true parallax . based on measurements of their apparent magnitudes in the 2mass catalog , we then derived median absolute magnitudes of @xmath660.652 , 0.131 , 0.484 , and 1.176 for the b3v , b7v , b8v , and a2v stars respectively in the j - band . based on their apparent magnitudes in the j - band in the 2mass catalog , the corresponding distance modulus of 2mass j01500456 + 5354007 ( b3v ) is 11.324 , 2mass j01494099@xmath615341341 ( b7v ) is 11.831 , and 2mass j01501546@xmath615341350 ( a2v ) is 12.457 ; i.e. , a distance of 1840 pc , 2323 pc , and 3100 pc to rmc2 . similarly , the distance modulus of 2mass j02220484@xmath615038130 ( b8v ) is 12.327 ; i.e. , a distance of 2920 pc to rmc3 . all the estimated distance module agree with the range measured by @xcite to per ob1 of magnitude 10.83 12.97 . the inferred distances to rmc2 and rmc3 are therefore compatible with their location in per ob1 . as mentioned above , all four candidate rmcs appear to be associated with pms stars . if at the distance of the per ob1 association , then these star - forming regions are located at unprecedented heights above the galactic plane for known clusters embedded in molecular clouds . in figure 8 , we plot the height above the galactic plan of seventy - six such embedded clusters compiled by @xcite within @xmath5 2 kpc of the sun , as well as the four rmcs studied here ( assuming a distance to these rmcs of 2.3 kpc ) . as can be seen , @xmath5@xmath67 of these clusters are located within @xmath5160 pc of the galactic plane . only two of the embedded clusters cataloged by @xcite are located beyond a height of 200 pc , ngc 281w and ngc 281e , both of which extend no one higher than 240 pc . by comparison , all four rmcs studied here are situated more than 280 pc away from galactic plane , with rmc3 having the largest height of @xmath5390 pc . as mentioned in @xmath68 , per ob1 is unusual in that it is not associated with a giant molecular cloud . based on the results described in @xmath69 , we now find that per ob1 is unusual in at least two other respects . first , most of its members exhibit a bulk motion away from the galactic plane with those at larger heights moving faster ( on average ) away from the galactic plane . second , star formation has recently occurred ( and may still be occurring ) at unusually large heights of 280400 pc above the galactic plane . we now examine whether all of these features can be explained , and if so how , in the framework of current star formation models . star formation at the present epoch is thought to be relatively inefficient in converting molecular gas to stars , even more so in models that invoke spontaneous as opposed to triggered star formation . to explain the lack of a giant molecular cloud associated with per ob1 without invoking an extraordinarily effective dispersement or destruction mechanism , one would have to propose that the star formation in this association has been more efficient than normal in converting molecular gas to stars . this may , but is not necessarily required to , explain the unusually large stellar densities and masses of the @xmath0 and @xmath1 per double cluster . the high number density of stars and their resultant bipolar outflows , stellar winds , uv ionization , and/or supernova explosions would then be more effective also in sweeping away or destroying much of the remaining molecular gas . in a search for shells and supershells in the galaxy , @xcite identified a large shell of atomic hydrogen ( ) gas in the direction towards per ob1 . @xcite show that this shell occupies the same range in radial velocities as per ob1 , and forms a quasi - circular structure of size @xmath5@xmath70 pc centered at ( @xmath71)=([email protected] , @xmath73.5 ) . the center of this shell therefore lies quite close to the locations of the @xmath0 and @xmath1 per double cluster . @xcite estimate a total mass of atomic gas in the shell of order @xmath74 m@xmath2 . they infer an an upper limit of 10 km s@xmath20 for the expansion velocity of the shell , and a kinetic energy for the shell of @xmath75 ergs . they also compute a minimum mechanical energy to blow the shell of @xmath76 ergs , compared with an energy released through stellar winds by all the known member b stars of @xmath77 ergs . thus , stellar winds from b stars alone in per ob1 is sufficient to blow the observed shell , let alone an additional energy from sources such as supernova explosions . we note that the total kinetic energy of the member stars due to their motion away from the galactic plane ( @xmath78 ergs ) adds insignificantly to the overall energetics . @xcite show from the _ iras _ 100-@xmath79 m image that emission from dust traces the inner border of the shell in the portion at large heights above the galactic plane ( any dust associated with the shell close to the galactic plane is confused with galactic dust ) . this dust feature can be seen as a nearly semi - circular arc in figure 2 of @xcite , extending to a galactic latitude of about @xmath80 at its largest height above the galactic plane . we speculate that this dust shell may have formed from the swept - up giant molecular cloud . we have checked the literature to see if the presumed hot bubble of gas responsible for sweeping up the abovementioned hi and dust shell is visible in x - rays . such x - ray emission is detectable in the rosat all sky survey ( rass ) @xcite from the orion - eridanus superbubble @xcite , which is powered by ori ob1 ( which contains approximately the same number of o stars as per ob 1 ) . this superbubble has approximately the same physical size as that enclosed by the hi shell associated with per ob1 . the kinematic energy of the orion - eridanus superbubble , as derived from its associated expanding shell , is about @xmath81 ergs @xcite , which is between one and two orders of magnitude larger than that of the presumed per ob1 superbubble . if the orion - eridanus superbubble was located at the distance of per ob1 , the count rate range of 3/4 kev band would be about 515 counts s@xmath20 . by comparison , the background count rate towards the presumed per ob1 superbubble is about 110 counts s@xmath20 . perhaps not surprisingly then , we did not find any significant x - ray emission from rass in 1/4 , 3/4 , and 1.5 kev bands towards the presumed per ob1 superbubble . we have searched the literature to find maps in molecular gas towards per ob1 . @xcite has mapped in co(1 - 0 ) the sky in the direction towards per ob1 at an angular resolution of 85 . compact and isolated co emission is detectable towards rmc1 , rmc2 , and rmc3 , but not towards rmc4 . if these sources are genuinely associated with the abovementioned remnant molecular clouds , the corresponding molecular gas mass is in these clouds @xmath5 8000 m@xmath2 . no co emission was detected towards rno6 , which has a molecular gas mass of @xmath5 400 m@xmath2 , consistent with the upper limit in sensitivity of the @xcite map . no co emission was detected towards the dust/ shell either ; assuming a co linewidth of 10 km s@xmath20 , the corresponding 3@xmath38 upper limit in molecular gas along this shell is @xmath82 m@xmath2 . if this estimate is accurate , the bulk of the molecular gas in this shell may therefore have been dissociated into atomic hydrogen gas . much of the star formation in the per ob1 association appears to have occurred in or within the vicinity of the @xmath0 and @xmath1 per double cluster . if so , then the observed increase ( on average ) in velocity with height ( acceleration ) away from the galactic plane may be a natural consequence of the dissolution of these clusters @xcite . over time , stars moving faster away from the double cluster would travel furthest from their initial locations , thus producing the observed linear increase in velocity with height above the galactic plane . although this model could explain the velocity pattern of the b main - sequence and supergiant stars , which can have ages up to @xmath5 10 myr of more , it can not easily explain the same velocity pattern observed for the o stars . the latter stars have ages of only a few millions years , much younger than the b stars in the @xmath0 and @xmath1 per double cluster ( @xmath5 13 myr ) , and could not have reached their present locations if traveling at their observed velocities from these clusters . this model also predicts that the member stars should form an approximately spherical distribution moving away from the @xmath0 and @xmath1 per double cluster , which is not seen ( c.f . fig . 2 ) . given that the molecular gas from which the stars in per ob1 formed appears to have been effectively swept outwards by the luminous stars in this association ( @xmath83 ) , we now consider whether the observed velocity dependence with height could simply be a reflection of stars forming in this swept - up gas . the situation envisaged is akin to that proposed in models of triggered star formation , in particular the scenario of cloud shuffling . cloud shuffling takes place on the periphery of an ob association where an over - pressured superbubble of accumulated stellar winds , supernova explosions , and/or ionized gas pushes and accelerates a giant molecular cloud outwards @xcite . such cloud shuffling is proposed to induce star formation , and also extend the lifetimes of molecular clouds by pushing them away from massive stars and their destructive effects . in the present case , the dust shell at the inner boundary of the shell presumably lies just beyond the expansion front of the superbubble associated with per ob1 . assuming typical physical conditions in ob associations , @xcite suggests that cloud shuffling can accelerate molecular gas by about 1 km s@xmath20 myr@xmath20 ( or equivalently 1 pc myr@xmath84 ) . let us assume that the @xmath0 and @xmath1 per double cluster was the original source of the superbubble ( with luminous stars formed later from the swept - up molecular gas adding to the energy of this superbubble ) , consistent with its location close to the center of the shell . the age of this double cluster is estimated to be @xmath513@xmath851 myr @xcite , allowing sufficient time for the molecular gas at the edge of the superbubble to be accelerated by over @xmath86 . this is the correct order of magnitude to produce , given the uncertainties involved in estimating the magnitude of the acceleration and its possible variation with time , the observed velocity dependence with height of the stars in per ob1 . the proposed scenario provides a natural explanation for the relatively large height of hd 15137 above the galactic plane , and its motion . @xcite suggest that both the massive stars hd 15137 and hd 14633 are runaway stars ejected from the open cluster ngc 654 in the perseus spiral arm . this suggestion , however , has a serious timescale problem , with inferred flight times for both stars that are longer than their ages @xcite . if hd 15137 is a member of per ob1 as proposed in @xmath87 , then in our proposed picture it formed close to its present location and inherited its relatively large velocity away from the galactic plane from its swept - up parent molecular cloud . note that the o star lying at the next highest galactic latitude , hd 15137 ( location as indicated in fig . 2 ) , has a comparably large proper motion away from the galactic plane . this explanation , however , can not be simply applied to hd 14633 , which we also proposed in @xmath87 to be a possible member of per ob1 . this star is located at a galactic latitude of @xmath88 , which is far beyond the presumed outer edge of the superbubble at a galactic latitude of about @xmath80 . the triggered formation of stars in the cloud - shuffling model naturally predicts an age gradient from the center of the superbubble , such that stars at larger radii have younger ages . if this model is applicable , the supergiant stars in per ob1 should therefore be closer to the @xmath0 and @xmath1 per double cluster ( the presumed center of the superbubble ) than the o stars . in figure 9 we plot the spatial distributions of o stars ( dotted lines ) and supergiant stars ( dashed lines ) proposed to be members of per ob1 from @xcite . using the kolmogorov - smirnov test @xcite , we find a @xmath89 probability that these two populations have different distributions in galactic height , such that the younger o tend to be located higher above the galactic plane than the older supergiant stars ( the visual impression given in fig . thus , there does seem to be a tendency , albeit only of marginally significance , for younger stars to be located at larger heights above the galactic plane . in the cloud shuffling model as applied to per ob1 , the relatively recent ( within the last @xmath51 myr ) formation of low- to intermediate - mass stars in remnant molecular clouds at unusually large heights above the galactic plane simply reflects the transport of molecular gas to such heights by the expanding superbubble . the dust and shells described in @xmath83 extend to galactic latitudes of about @xmath80 ( height @xmath5 450 pc above the galactic plane ) , and so the remnant molecular clouds described in @xmath90 lies in projection inside the superbubble . these clouds may actually comprise swept - up molecular gas located at the periphery of the superbubble . alternatively , they may comprise the densest condensations in the giant molecular cloud that were not completely swept up by the expanding superbubble . as mentioned in @xmath91 , all the rmcs have head - tail morphologies with their heads pointing towards the galactic plane rather than in the direction of any particular o star . these morphologies of these rmcs may therefore have been sculpted by the expanding superbubble rather than ionizing radiation from o stars . the per ob1 association comprises the remarkable @xmath0 and @xmath1 per double cluster ( among the brightest and densest open clusters known ) , surrounded by luminous o / b and both early- and late - type supergiant stars . unlike virtually all other known ob associations in the galaxy , there is no known giant molecular cloud in the vicinity of per ob1 . we have gathered together the available information on the stellar population of per ob1 , primarily from the work by @xcite and @xcite . for those luminous early- and late - type stars included in the hipparcos astrometry mission , we have extracted measurements of their proper motions . we found that : we also searched for candidate intermediate - mass ( haebe ) and low - mass ( classical t tauri star ; ctts ) pre - main - sequence stars in per ob1 . we find a number of these candidate stars to be associated with four candidate remnant molecular clouds ( rmcs ) identified as dust clouds in _ iras _ 100-@xmath79 m images and optical nebulae in the 2nd digitized sky survey . all these rmcs are located high ( 280400 pc in projection ) above the galactic plane . from our spectroscopy of both candidate pre - main - sequence as well as main - sequence b a stars apparently associated with these rmcs , we : 1 . show that the distance moduli to the main - sequence b a stars are consistent with the rmcs being located in per ob1 2 . confirm that some of the candidate pre - main - sequence stars are indeed haebe stars and identify possible cttss , implying that there has been recent ( within the last 1 myr ) star formation at galactic heights far above the locations of known luminous members in per ob1 1 . point out that the large shell of atomic hydrogen ( ) gas ( mass @xmath5@xmath92 ) enclosing a slightly smaller shell of dust , studied by @xcite and which they propose to be created by a stellar - wind - blown superbubble from the @xmath0 and @xmath1 per double cluster and ob stars in per ob1 , likely comprises ( in part ) the swept - up remains of the giant molecular cloud from which per ob1 formed 2 . argue that if the observed motion of the luminous member stars is attributed to the dissolution of the @xmath0 and @xmath1 per double cluster ( at their outskirts ) , this can only explain the motion of the main - sequence b and late - type supergiant stars . the main - sequence o stars , which are much younger than the double cluster , would not have had sufficient time to reach their present locations high above the galactic plane if they originated from the close vicinity of the double cluster 3 . show that the observed motion of the luminous member stars is better explained in the cloud - shuffling model @xcite , in which the abovementioned superbubble pushes and accelerates the giant molecular cloud outwards and at the same time induces star formation at the dense cloud interface . this would naturally result in those stars forming locally in the compressed molecular gas to have increasingly large velocities away from galactic plane with height 4 . show that there is a weak trend for the younger o member stars to lie at higher galactic latitudes than the older late - type supergiant member stars , consistent with the simplest prediction of the cloud - shuffling but not cluster dissolution model 5 . suggest that the recent star formation in the remnant molecular clouds that we found at very large heights above the galactic plane , far above the locations of the luminous member stars , likely occurred in the swept - up molecular gas at the periphery of the superbubble , or in the densest condensations that were not entirely swept away by the expanding superbubble h .- lee thanks wen - ping chen for sharing observation time , wen - shan hsiao for helping spectral observations , rue - ron hsu and pin - gao gu for discussing , and the staff at the beijing astronomical observatory for their assistance during observing runs . we are grateful to the anonymous referee for constructive suggestions . this research makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation . this research made use of montage , funded by the national aeronautics and space administration s earth science technology office , computation technologies project , under cooperative agreement number ncc5 - 626 between nasa and the california institute of technology . montage is maintained by the nasa / ipac infrared science archive . this research made use of data products from the digitized sky surveys produced at the space telescope science institute . ambartsumian , v. a. 1947 , in stellar evolution and astrophysics , armenian acad . of sci . ( german transl . , 1951 , sowjetischen astron , 1 , 33 ) bachiller , r. , fuente , a. , & kumar , m. s. n. 2002 , , 381 , 168 bok , b. j. 1934 , harvard college observatory circular , 384 , 1 boyajian , t. s. , et al . 2005 , , 621 , 978 bragg , a. e. , & kenyon , s. j. 2005 , , 130 , 134 brown , a. g. a. , hartmann , d. , & burton , w. b. 1995 , , 300 , 903 brown , a. g. a. , arenou , f. , van leeuwen , f. , lindegren , l. , & luri , x. 1997 , hipparcos - venice 97 , 402 , 63 cappa , c. e. , & herbstmeier , u. 2000 , , 120 , 1963 cohen , m. 1980 , , 85 , 29 cutri , r. m. , et al . 2003 , 2mass all sky catalog of point sources ( pasadena : ipac ) dame , t. m. , hartmann , d. , & thaddeus , p. 2001 , , 547 , 792 de geus , e. j. , de zeeuw , p. t. , & lub , j. 1989 , , 216 , 44 de zeeuw , p. t. , hoogerwerf , r. , de bruijne , j. h. j. , brown , a. g. a. , & blaauw , a. 1999 , , 117 , 354 elmegreen , b. g. & lada , c. j. 1977 , , 214 , 725 elmegreen , b. g. 1979 , , 231 , 372 elmegreen , b. 1992 , in star formation in stellar systems , ed . g. tenorio - tagle , m. prieto , & f. sanchez ( cambridge : cambridge univ . press ) , 381 esa . 1997 , the hipparcos and tycho catalogs , 1997 , esa sp-1200 garmany , c. d. , & stencel , r. e. 1992 , , 94 , 211 geyer , m. p. , & burkert , a. 2001 , , 323 , 988 heiles , c. 1979 , , 229 , 533 humphreys , r. m. 1978 , , 38 , 309 jacoby , g. h. , hunter , d. a. , & christian , c. a. 1984 , , 56 , 257 keller , s. c. , grebel , e. k. , miller , g. j. , & yoss , k. m. 2001 , , 122 , 248 kharchenko , n. v. , piskunov , a. e. , rser , s. , schilbach , e. , & scholz , r .- d . 2005 , , 438 , 1163 kovalevsky , j. 1984 , space science reviews , 39 , 1 kroupa , p. , aarseth , s. , & hurley , j. 2001 , , 321 , 699 lada , c. j. , & lada , e. a. 2003 , , 41 , 57 lee , h .- t . , chen , w. p. , zhang , z. , & hu , j. 2005 , , 624 , 808 lee , h .- , & chen , w. p. 2007a , , 657 , 884 lee , h .- t . et al . 2007b , in preparation lutz , t. e. , & kelker , d. h. 1973 , , 85 , 573 lynds , b. t. 1965 , , 12 , 163 mamajek , e. e. , meyer , m. r. , & liebert , j. 2002 , , 124 , 1670 meyer , m. r. , calvet , n. , & hillenbrand , l. a. 1997 , , 114 , 288 ogura , k. , & sugitani , k. 1998 , publ . astron . australia , 15 , 91 oudmaijer , r. , groenewegen , m. a. t. , & schrijver , h. 1999 , , 341 , l55 perryman , m. a. c. , et al . 1997 , , 323 , l49 preibisch , t. , & zinnecker , h. 1999 , , 117 , 2381 preibisch , t. , brown , a. g. a. , bridges , t. , guenther , e. , & zinnecker , h. 2002 , , 124 , 404 preibisch , t. , & zinnecker , h. 2007 , iau symposium , 237 , 270 press , w. h. , teukolsky , s. a. , vetterling , w. t. , & flannery , b. p. numerical recipes : the art of scientific computing ( 3rd ed . ; cambridge university ) reipurth , b. 1983 , , 117 , 183 sargent , a. i. 1979 , , 233 , 163 slesnick , c. l. , hillenbrand , l. a. , & massey , p. 2002 , , 576 , 880 snowden , s. l. , et al . 1995 , , 454 , 643 snowden , s. l. , burrows , d. n. , sanders , w. t. , aschenbach , b. , & pfeffermann , e. 1995 , , 439 , 399 sugitani , k. , fukui , y. , & ogura , k. 1991 , , 77 , 59 van der hucht , k. a. 2001 , new astronomy review , 45 , 135 van steenberg , m. e. , & shull , j. m. 1988 , , 67 , 225 per ( east ) and @xmath1 per ( west ) double cluster . the locations of remnant molecular clouds ( rmcs ) identified in this work , as well as the remnant molecular cloud rno6 are indicated . the dashed line indicates the approximate geometry of the shell found by @xcite and described in 3.1 . ] and @xmath1 per double cluster . yellow arrows are for other stars in the _ hipparcos _ catalog lying in the region shown . the length of each arrow is proportional to the magnitude of its proper motion . the grayscale is the @xmath54 100 @xmath55 image . ] and @xmath1 per double cluster ( triangles ) by @xcite . also plotted are the tangential velocities for all proposed members in the _ hipparcos _ averaged over latitude strips of @xmath93 ( squares ) . note that all of the proper motions have been corrected for the solar motion . ] 100 @xmath55 image using two different stretches . the left panel emphasizes regions with the strongest extinction , and the right panel regions with lower extinctions to more clearly show the rmcs identified in this work . note that only those candidate ctts and haebe stars lying close to or against the rmcs were selected for follow - up spectroscopy . ] 100 @xmath55 image . locations of the rmcs identified in this work , as well as rno6 , are indicated . also indicated are the locations of main - sequence b or a stars ( blue ) used to estimate distances to the rmcs . the stars are numbered according to their identification numbers in table 4 . ] 130.900 & -3.441 & 8725 & 1.05 & 2.49 & 1.38 & 1.13 & hd 236894 & o8v & a , b + 131.479 & -2.433 & 9211 & 0.65 & -1.88 & 1.03 & 0.87 & hd 236915 & m2iab & b + 131.867 & -5.138 & 9017 & -0.85 & -3.01 & 1.46 & 1.33 & bd+55 441 & b1vpe & a + 132.691 & -2.588 & 9886 & -0.12 & -1.71 & 1.11 & 1.05 & hd 236947 & m2iab - ib & b + 132.742 & -1.700 & 10078 & 1.40 & -0.03 & 1.49 & 1.10 & hd 13036 & b0.5iii & a , b + 132.975 & -5.802 & 9523 & 0.77 & -5.80 & 1.00 & 1.02 & hd 12323 & on9v & a , b + 132.977 & -2.504 & 10060 & -0.13 & -0.26 & 1.00 & 0.98 & hd 13022 & o9.5ia & a , b + 132.981 & -2.880 & 9990 & -0.30 & -1.80 & 0.58 & 0.61 & hd 12953 & a1ia & a , b + 133.105 & -4.282 & 9826 & 0.72 & -3.10 & 1.12 & 0.98 & hd 12727 & b2iii & a + 133.173 & -6.153 & 9582 & 1.25 & -3.21 & 0.71 & 0.78 & hd 12401 & m4ib & b + 133.168 & -1.650 & 10323 & -0.30 & -1.28 & 1.49 & 1.10 & hd 13402 & b0.5ib & a , b + 133.174 & -3.334 & 10024 & -0.87 & -1.39 & 1.16 & 0.99 & hd 12993 & o6.5v & a , b + 133.270 & -4.170 & 9935 & -0.51 & -2.15 & 0.94 & 0.87 & hd 12856 & b0iiipe & a , b + 133.505 & -4.217 & 10055 & 0.72 & -2.47 & 1.09 & 0.98 & hd 13051 & b1ive & a + 133.565 & -2.549 & 10379 & 0.09 & -0.54 & 0.65 & 0.65 & hd 13476 & a3iab & a , b + 133.571 & -3.509 & 10227 & -0.14 & -1.92 & 0.66 & 0.61 & hd 13267 & b5ia & a , b + 133.888 & -2.882 & 10489 & -3.66 & 0.02 & 1.16 & 1.11 & hd 13658 & m1iab & b + 134.013 & -1.219 & 10904 & -0.90 & 0.50 & 1.69 & 1.43 & hd 14242 & m2iab & b + 134.030 & -4.924 & 10228 & 1.90 & -2.83 & 0.92 & 0.74 & hd 13268 & o7 : & a , b + 134.064 & -3.223 & 10527 & 0.90 & -2.15 & 1.03 & 0.90 & hd 13716 & b0.5iii & a , b + 134.130 & -1.931 & 10829 & 1.78 & -2.03 & 1.36 & 1.08 & hd 14142 & m2iab & b + 134.445 & -3.868 & 10624 & 0.23 & -2.39 & 0.77 & 0.69 & hd 13841 & b2ib & a , b + 134.448 & -3.840 & 10633 & -0.60 & -1.02 & 0.54 & 0.52 & hd 13854 & b1iab & a , b + 134.525 & -4.149 & 10615 & -0.51 & -2.68 & 0.86 & 0.81 & hd 13831 & b0iiie & a , b + 134.571 & -4.154 & 10641 & -0.85 & -1.85 & 0.68 & 0.66 & hd 13866 & b2ib & a , b + 134.649 & -4.891 & 10541 & 1.16 & -4.52 & 0.83 & 0.76 & hd 13745 & o9.7ii((n ) ) & a , b + 134.685 & -2.129 & 11093 & 3.21 & -2.09 & 1.98 & 1.71 & hd 14528 & m4ie & b + 134.705 & -5.589 & 10443 & -0.90 & -2.69 & 0.99 & 0.92 & hd 13621 & b0.5iii - iv((n ) ) & a + 134.705 & -3.665 & 10805 & 0.24 & -2.03 & 0.68 & 0.59 & hd 14134 & b3ia & a , b + 134.714 & -3.626 & 10816 & -0.64 & -1.76 & 0.66 & 0.59 & hd 14143 & b2ia & a , b + 134.792 & -2.861 & 10995 & 0.56 & -3.45 & 0.84 & 0.77 & hd 14404 & m2iab & b + 135.033 & -3.433 & 11020 & -0.50 & -2.13 & 0.61 & 0.57 & hd 14433 & a1ia & a , b + 135.058 & -1.678 & 11394 & 0.21 & -0.20 & 1.33 & 0.97 & hd 14947 & o5if+ & a , b + 135.120 & -3.247 & 11115 & -0.84 & -1.52 & 0.72 & 0.63 & hd 14542 & b8ia & a , b + 135.143 & -3.752 & 11018 & 1.46 & -2.86 & 0.99 & 0.84 & hd 14434 & o6.5 : & a , b + 135.155 & -3.385 & 11098 & -1.34 & -1.31 & 0.75 & 0.66 & hd 14535 & a2iab & a , b + 135.334 & -4.743 & 10926 & 0.10 & -1.02 & 0.79 & 0.63 & hd 14322 & b8ib & a , b + 135.381 & -4.815 & 10935 & 0.90 & -1.00 & 1.05 & 0.83 & hd 14331 & b0iii & a , b + 135.400 & -3.088 & 11284 & 2.05 & -1.52 & 1.32 & 1.36 & hd 14826 & m2iab & b + 135.488 & -2.795 & 11391 & 0.94 & -1.21 & 0.62 & 0.59 & hd 14956 & b2ia & a , b + 135.573 & -4.724 & 11060 & 0.76 & -0.99 & 0.66 & 0.54 & hd 14489 & a2iab & a , b + 135.683 & -3.866 & 11279 & -0.13 & -0.75 & 0.51 & 0.50 & hd 14818 & b2ia & a , b + 135.835 & -2.506 & 11625 & 0.07 & 0.58 & 0.68 & 0.63 & hd 15316 & a2ia & a , b + 136.119 & -2.524 & 11769 & 0.55 & 0.37 & 0.73 & 0.67 & hd 15497 & b6ia & a , b + 136.128 & -1.754 & 11953 & 0.73 & -0.30 & 1.21 & 0.98 & hd 15752 & b0iiin & a , b + 136.152 & -2.260 & 11841 & -0.92 & 0.24 & 1.01 & 0.90 & hd 15620 & b8iab & a , b + 136.323 & -0.371 & 12416 & -2.23 & -1.21 & 1.23 & 1.13 & bd+58 501 & m2iab & b + 136.351 & -3.293 & 11722 & 1.10 & -1.38 & 1.22 & 1.14 & hd 15450 & b1iiie & a + 136.385 & -2.587 & 11898 & 0.33 & -3.51 & 1.88 & 1.75 & hd 15690 & b2iab & a , b + 136.535 & -3.468 & 11782 & 0.34 & -1.86 & 1.37 & 1.30 & hd 15548 & b1v & a + 136.695 & -0.046 & 12750 & -0.06 & -0.04 & 0.74 & 0.67 & hd 16778 & a2ia & b + 137.159 & -4.660 & 11850 & 0.05 & -3.11 & 1.11 & 1.05 & hd 15642 & o9.5iii : n & a , b + 137.185 & -2.779 & 12302 & 1.33 & -3.09 & 0.86 & 0.86 & hd 236979 & m1iab & b + 137.351 & -1.300 & 12769 & -0.84 & -1.73 & 1.50 & 1.20 & hd 16808 & b0.5ib & a , b + 137.367 & -1.044 & 12840 & -0.58 & -1.36 & 1.27 & 1.09 & hd 236995 & a0ia & a , b + 137.532 & -7.507 & 11473 & 2.53 & -4.48 & 0.67 & 0.72 & hd 15137 & o9.5iii(n ) & c + 137.796 & -2.660 & 12676 & -0.65 & -4.71 & 1.04 & 1.14 & hd 16691 & o5f & a + 137.932 & -1.681 & 12972 & 1.01 & -1.89 & 0.94 & 0.93 & hd 17088 & b9ia & a , b + 138.022 & -1.701 & 13022 & 0.69 & -1.81 & 1.10 & 1.13 & hd 17145 & b8ia & a , b + 138.383 & -1.326 & 13290 & -0.36 & -1.51 & 1.61 & 1.46 & hd 237010 & m2iab & b + 138.537 & -2.110 & 13178 & -0.03 & -1.72 & 0.63 & 0.69 & hd 17378 & a5ia & a , b + 138.718 & -2.133 & 13262 & 1.21 & -5.23 & 1.01 & 1.03 & hd 237008 & m3iab & b + 138.930 & -2.081 & 13380 & 1.49 & -3.08 & 1.44 & 1.54 & hd 17638 & wc6 & c + 140.863 & -18.126 & 11099 & 3.13 & -6.05 & 0.87 & 0.69 & hd 14633 & on8v & c + [ table : members ] references(a ) @xcite ; ( b ) @xcite ; ( c ) this paper rrrrrr 0 @xmath94 & @xmath95 & @xmath96 & 0.72 & 0.66 & 2 + @xmath94 @xmath97 & 0.44 & @xmath98 & 0.42 & 0.35 & 11 + @xmath97 @xmath99 & 0.46 & @xmath100 & 0.26 & 0.24 & 18 + @xmath99 @xmath101 & 0.41 & @xmath102 & 0.24 & 0.22 & 16 + @xmath101 @xmath103 & 0.67 & @xmath104 & 0.28 & 0.25 & 11 + @xmath103 @xmath105 & 0.06 & @xmath106 & 0.68 & 0.64 & 3 + @xmath105 @xmath107 & 1.64 & @xmath108 & 0.71 & 0.78 & 1 + @xmath107 @xmath109 & 2.58 & @xmath110 & 0.67 & 0.72 & 1 + ......... + @xmath111 @xmath112 & 2.93 & @xmath113 & 0.87 & 0.69 & 1 + [ table : pm ] 1 & j01380790 + 5453463 & h@xmath114 & haebe ( a0 ) & around rmc1 + 2 & j01424061 + 5449135 & h(@xmath6652.5 ) & ctts & in rmc1 + 3 & j01433441 + 5451323 & h(@xmath666.1 ) & ctts & in rmc1 + 4 & j01435138 + 5526098 & h@xmath114 & haebe ( b3 ) & in rmc1 + 5 & j01494099 + 5341341 & & b ( b7v ) & with a reflection nebula in rmc2 + 6 & j01500456 + 5354007 & & b ( b3v ) & with a reflection nebula in rmc2 + 7 & j01501404 + 5341107 & h(@xmath6689.1 ) & ctts & in rmc2 + 8 & j01501546 + 5341350 & & a ( a2v ) & with a reflection nebula in rmc2 + 9 & j02151126 + 5519500 & h@xmath114 & haebe ( b8 ) & around rno 6 + 10 & j02220484 + 5038130 & & b ( b8v ) & with a reflection nebula in rmc3 + 11 & j02220662 + 5040013 & h(@xmath662.2 ) & wtts & in rmc3 + 12 & j02261522 + 5243131 & h(@xmath6616.1 ) & ctts & with a reflection nebula in rmc4 + 13 & j02262049 + 5244288 & h@xmath114 & haebe ( b6v ) & with a reflection nebula in rmc4 + [ table : spectra ] |
in this paper we examine properties of the variable source sgr a * in the near - infrared ( nir ) using a very extensive ks - band data set from naco / vlt observations taken 2004 to 2009 . we investigate the variability of sgr a * with two different photometric methods and analyze its flux distribution . we find sgr a * is continuously emitting and continuously variable in the near - infrared , with some variability occurring on timescales as long as weeks . the flux distribution can be described by a lognormal distribution at low intrinsic fluxes ( mjy , dereddened with ) . the lognormal distribution has a median flux of.1 mjy , but above 5 mjy the flux distribution is significantly flatter ( high flux events are more common ) than expected for the extrapolation of the lognormal distribution to high fluxes . we make a general identification of the low level emission above 5 mjy as _ flaring _ emission and of the low level emission as the _ quiescent state_. we also report here the brightest ks - band flare ever observed ( from august 5th , 2008 ) which reached an intrinsic ks - band flux of 27.5 mjy ( ) . this flare was a factor 27 increase over the median flux of sgr a * , close to double the brightness of the star s2 , and 40% brighter than the next brightest flare ever observed from sgr a*. | 1008.1984 | the _ very _ center of our galaxy houses the variable source named sgr a * , first discovered in the radio as a compact source @xcite . the fact that this source is motionless ( to better than 1 km / s ) at the dynamical center of the galaxy @xcite , and its coincidence with the common focus of elliptical orbits of stars tracked over the last decade and a half , clearly associates it with a supermassive black hole of @xmath4 @xcite . sgr a * has been detected across the electromagnetic spectrum , at radio , submm , nir and x - ray wavelengths . at nir and x - ray wavelengths @xcite the emission is highly variable ( factors up to @xmath2160 and 27 in the x - ray and nir respectively ; @xcite , this work ) compared to the comparatively steady emission at longer wavelengths . nir peaks are detected more often than in the x - ray ( peaks occur @xmath21 and 4 times a day for x - ray and nir variable emission , respectively ; @xcite ) . some nir flares have been detected without any accompanying x - ray flare @xcite . however , when both nir and x - ray exhibit increases in emission , the peaks in emission occur simultaneously ( e.g. , * ? ? ? * ; * ? ? ? the near - infrared lightcurves from sgr a * exhibit @xmath51 hour long increases in emission that are often called ` flares ' in the literature . a number of these have exhibited very suggestive substructural features with timescales of @xmath6 minutes @xcite , possibly quasi - periodic oscillations ( qpos ) . however the existence of qpos and even the use of the term flare to describe the nir variability of sgr a * has been questioned by @xcite ( see also @xcite and @xcite ) who argue that there is no true quasi - periodicity , just a variability process with a featureless red noise power spectrum ( e.g. a power spectrum @xmath7 where @xmath8 is frequency ) . a stochastic source with a red noise power spectrum has higher variability at longer timescales and could potentially be responsible for the structures on longer timescales seen in the real lightcurves . the authors suggest that apparent flare peaks may simply be the highest observed flux excursions in such a purely stochastic source and are not isolated events . a main reason for the two rather contrasting interpretations of the variable emission from sgr a * has been that the nature of the faint emission from sgr a * and its relationship to the high flux emission is uncertain . the nir emission from the galactic center is dominated by the central cluster of bright stars , and adaptive optics at 8-meter class telescopes is required in order to separate the faint source sgr a * from the closest s - stars ( even at this resolution sgr a * is still on occasion confused with a relatively bright star ) . additional , faint stars may be present very close to sgr a * which have not yet been tracked and identified as stars from astrometric monitoring programs ( e.g. * ? ? ? * ) . while the dramatic high flux variability can be unambiguously attributed to the black hole , when a faint source is detected at the position of sgr a * , it is not necessarily clear that the source is sgr a * , faint stars , or a combination of both . accordingly , it is not clear whether sgr a * continues to emit at all at low fluxes , whether it exhibits a ` quiescent state ' ( a non- or weakly active low state ) , or whether the low flux emission continues to vary constantly with the same statistical properties as the high flux emission . cccccc & & & & & + _ brightest states _ & & _ peak fluxes _ & & & + & & & & & + gen+03 & 15 jun 2003 & 13.2 ( 10.5 + 2.7 ) & 1.10 & 0.76 & 11.0 + & & & & & + gen+03 & 16 jun 2003 & 10 ( 7.3 + 2.7 ) & 1.10 & 0.76 & 8.4 + & & & & & + mey+07 & 6 oct 2003 & 22@xmath9 & 1.2 & 0.76 & 20.1 + & & & & & + tri+07/mey+06 & 31 may 2006 & 16/23 ( + s17 ) & - & - & ( 16.7 ) 13.5@xmath10 + & & & & & + hor+07 & 31 jul 2005 & 11.6 & @xmath51.06@xmath11 & 0.52 & 6.4 + & 2 may 2006 & 26.8 & & & 14.8 + & 17 jul 2006 & 6.8 & & & 3.7 + & & & & & + do+08 & 3 may 2006 & 0.8 & @xmath51.06@xmath11 & - & 8.5 + & 20 jun 2006 & 0.65 & & & 6.9 + & 21 jun 2006 & 0.4 & & & 4.2 + & 17 jul 2006 & 0.3 & & & 3.2 + & 18 may 2007 & 0.6 & & & 6.4 + & 12 aug 2007 & 0.2 & & & 2.1 + & & & & & + eck+08 & 15 may 2007 & 24 ( + s17 ) & 0.76 & 0.76 & ( 13.9 ) 10.7 + & & & & & + this work & 5 aug 2008 & 30.7 ( + s17 ) & - & - & ( 30.7 ) 27.5 + & & & & & + & & & & & + _ faintest states _ & & & & & + & & & & & + hor+02 & 9 may 2001 & @xmath12 mjy & @xmath51.06@xmath11 & - & @xmath13 + & & ( upper limit , not dered ) & & & + & & & & & + sch+02 & 5 jun 2006 & @xmath14 mjy & ? & 0.76 & @xmath15 + & & & & & + do+08 & 2006 : may 3 , & @xmath16 mjy & + & jun 20 , 21 , jul 17 & & ( median flux , not dered ) & 1.06@xmath17 & - & 2.0 + & 2007 : may 18 , aug 12 & @xmath18 & 1.06 & - & @xmath19 + & & ( faintest flux , not dered ) & & & + & & & & & + sab+10 & 23 sep 2004 & 2.4 mjy & @xmath20 & 0.76 & 1.4 + & & ( upper limit , & & & + & & no stellar contam . ) & & & + & 23 sep 2004 & 0.9 mjy & & & 0.5 + & & ( upper limit , & & & + & & full stellar contam . ) & & & + in addition to this , an unbiased overview of the properties of the near - infrared emission from sgr a * can be difficult to obtain from the published literature because of publication bias ( bright events have individual interest and are often published alone ) . however , some studies have looked at the statistical properties . @xcite and @xcite presented lightcurves and flux distributions for sgr a * for about @xmath511 hours of 1.6@xmath21 m and @xmath532 hours of 1.7@xmath21 m data observed with nicmos on the hubble space telescope ( hst ) . with the resolution of the hst the close stellar sources are not as well separated from sgr a * as with the vlt or keck telescopes , and the stars s17 and s2 overlap with the sgr a * source in these observations . the flux distributions were fitted with a gaussian at low fluxes , which was attributed to the observational noise on constant sources ( the contribution from s2 , s17 and possible quiescent emission ) and a power - law at high fluxes , which was attributed to transient flares . the best fit models implied that sgr a * was active ( above the noise at low levels ) more than 40% of the time . @xcite presented an analysis of six nights of k-band ( and one l-band ) observations at the keck observatory , using an unbiased set of observations taken between 2005 and 2007 . a source at the position of sgr a * was always detected in this dataset , with an estimated maximum 35% contribution from stellar contamination . these authors reported that the source sgr a * was continuously variable , based on the larger variance of sgr a * compared to stars of similar brightness on five of the six k-band observation nights . this was the data set used to investigate timing properties of sgr a * in which it was claimed the data set was consistent with a featureless red noise power spectrum with no quasi - periodicity . however , with a sum duration for the k - band observations of about 12.1 hours , this data set did not sample well the higher fluxes of sgr a * , i.e. the source was relatively faint compared to publications where variable emission with suggestive quasi - periodic structure have been reported ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * for a comparison of ks - band peak emission from the literature see table [ table2 ] in the appendix ) . although the studies of @xcite , @xcite and @xcite have gone some way towards understanding the statistical properties of sgr a * in the near - infrared , there has not yet been a study on a very large , unbiased dataset of the variability of sgr a * where the rare high fluxes are also well sampled . in this paper , we analyse the ks - band flux distribution of sgr a * for the years 2004 - 2009 from 117 observation nights carried out with the vlt in paranal , chile with the aim of seeking the flux - dependent characteristics of the variability of sgr a * at both high and low fluxes . we do this through investigation of the flux distribution of sgr a*. the dataset of this paper is @xmath512 times larger than the data set of @xcite . in order that our results might be compared with other publications , we give a summary in table [ table2 ] ( in the appendix ) of the brightest and/or notable ks / k-band variable emission reported in the literature , as well as the faintest values or upper limits . all are reported in the literature with different calibrations / extinction values , and we have done our best to determine the corrections to scale them to the calibration and extinction values used in this paper . in section [ section_dataset ] we present our observations and the results of using two different methods of photometry : a six year lightcurve ( aperture photometry ) , and the 2009 data set ( psf photometry ) . in section [ section_results ] we present the flux distribution of sgr a * and our results from various model fits to the flux distribution . in section [ section_discussion ] we discuss our results in the context of two variability states for sgr a * : a low - level lognormally varying _ quiescent state _ and sporadic high flux _ flares_. we summarize in section [ section_conclusions ] . since 2002 we have observed the galactic center with the near - infrared adaptive optics - assisted diffraction limited imager naco at the vlt @xcite . much of the observation time 2002 - 2009 was spent concentrated on the central few arcesconds , measuring the positions of the s - stars and monitoring sgr a*. by now we have amassed a large dataset with which we can investigate statistical properties of sgr a * s variability . while we utilised l ( 3.8@xmath21 m ) , ks ( 2.18@xmath21 m ) and h ( 1.66@xmath21 m ) bands , by far the largest proportion of our data is taken in ks - band , collected in either 13 mas pixel scale imaging , 13 mas pixel scale polarimetry , or 27 mas pixel scale imaging mode . the presence of many stars close to the central black hole complicates the attempt to acquire accurate photometry for the near - infrared sgr a * source . sgr a * is usually fainter than the surrounding s - stars , and the stars _ move_. a star can on occasion get so close to sgr a * in projection that the two sources become confused . for example , the @xmath516 year orbit star s2 @xcite was confused with sgr a * as it passed pericenter in 2002 . more recently , the @xmath22 mag star s17 was confused with sgr a * in 2006 - 2008 . we use two different methods in this paper : 1 . we first attempted to obtain as large a homogeneous , unbiased , dataset as possible , in order to obtain a best overview of the statistical properties of the variability of sgr a * , and to address in particular the variability at ( the rarer ) high fluxes . to do this we used aperture photometry because this allowed us to address the most important complication to the photometry of sgr a * in 2004 to 2009 the star s17 , confused with sgr a * in 2006 - 2008 . to deal with s17 we used a two - aperture photometry method to determine the combined flux of sgr a * and the star s17 in all years 2004 to 2009 with comparable accuracy . this is important because to understand the flux distribution we have to understand the photometric errors at low fluxes . if all years are measured with comparable accuracy the long - term errors can be well approximated with a gaussian ( with width dependent on flux ; see section [ section_obserrors ] ) . we did not include the ks - band data from 2002 - 2003 because the star s2 was so close to sgr a * during this time that it contributed to the flux measured with this method . since s2 is much brighter than s17 the flux of sgr a*+s17 could not be measured with an accuracy comparable to the other years . we secondly looked at a subset of the data in more detail in order to address the nature of the variability at low fluxes . determining the nature / existence of low level variability ( and in particular to distinguish it from observational errors ) requires more precise photometry than could be achieved with aperture photometry , in particular with the inclusion of s17 . we address the question of the nature of the low level variability by analyzing the ( high quality ) 2009 data , in which s17 is not confused with sgr a * , in greater detail with a psf - fitting photometric method starfinder , @xcite ) . the flux errors for this method are much smaller and allow us to distinguish with more certainty between true variability and flux errors at low fluxes ; this data set is not completely unbiased since sgr a * was only reliably detected in 87% of the selected images . by using these two methods in combination , we can overcome the individual difficulties of the methods and piece together a consistent picture of the near - infrared variability of sgr a*. our 2004 - 2009 ks - band data set consists of @xmath23 images . the data were reduced in the standard way by applying a sky subtraction , a flat field correction and a hot / dead pixel correction ( see e.g. , * ? ? ? * ) . we found it necessary to apply a quality cut to eliminate the worst data and ensure we obtained a homogeneous data set on which we could perform accurate photometry . this quality cut was carried out by eye , and the criteria for elimination included : * the two calibration stars were not in the image ( this happened rarely , and only in the polarimetric data ) * image ghosts close to sgr a * or psf artefacts * simply a bad quality image mostly corresponding to low strehl ratio data resulting from poor atmospheric conditions . the quality of images for this criterion was judged by the general visibility of s - stars , not by the visibility of sgr a * to avoid bias towards bright fluxes from sgr a*. we additionally eliminated 854 images from july in 2009 which were taken in a triggered mode , so that our dataset remains unbiased and representative of the overall variability of sgr a*. the remaining data set totaled 6774 images which were used to obtain a lightcurve for sgr a*. to extract fluxes for sgr a * as well as for several control stars , we carried out aperture photometry on each image . however , if we used a standard circular aperture centred on sgr a * , we would have a varying contribution from s17 ( the brightest star to be confused with sgr a * 2004 - 2009 ) to the lightcurve as the star moves through the aperture during 2006 - 2008 . a second star within the aperture , but not centered within the aperture also increases the error on our measurements of the flux of sgr a*. we would obtain more accurate results if s17 could somehow also be well - centered in the aperture at all times , as well as sgr a*. to solve the problem of s17 we thus used a two - aperture method with two circular sub - apertures , one centered on sgr a * and one centered on s17 and measure their combined flux , averaging the results where the two sub - apertures are each 40 , 53 and 66 mas in size . we used this method for all data 2004 to 2009 , so that we always measure the combined flux of s17 + sgr a*. while there may be additional effects due to confusion with other , fainter stars , at first order our method ensures that we measure the fainter fluxes of sgr a * with similar measurement errors from year to year than we would have if we neglected to include a sub - aperture about s17 . to further ensure the self - consistency of our dataset , we used the same set of calibrator stars for calibration of each image . this restricted us to only two suitable calibration stars , s30 and s65 , due to the small field of view of some ( especially polarimetric ) images . to measure the raw counts of control stars and the calibration stars we used the same aperture as for sgr a * and s17 , but centered the star in just one of the two sub - apertures . for each image , the raw counts of sgr a*/s17 and the control stars are then divided by the counts of the two calibrator stars and the calibrated counts are then flux calibrated by scaling relative to the median of s65 for the observation night , using @xmath24 mag for s65 ( e.g. * ? ? ? our photometric calibration is also consistent with @xcite ( s65 : @xmath25 mag ) . to convert magnitudes to fluxes ( mjy ) we used the zeropoint from @xcite ( @xmath26 mag corresponds to 667 jy ) . we computed the expected positions of sgr a * and of the control stars using the orbital ( polynomial ) fits of @xcite . for polarimetric data , we added the raw photon counts recorded in the apertures for all targets in both ordinary and extraordinary images first in order to recover the unpolarized intensities from the ordinary image ) . ] , before the flux calibration was carried out . we dereddened the final flux values using @xmath27 ( * ? ? ? * fritz 2010 in prep . ) . note that the absolute photometric calibration and extinction correction is not important for the analysis of relative fluxes presented in this paper , where we have focused on maximising the long - term relative accuracy of our photometric measurements . however , where we compare with previous publications , we additionally scale the reported fluxes to the photometric calibration ( as best as possible ) and extinction of this paper . the lightcurve we produced in this way for sgr a * is shown in figure [ fig_dataset ] . the data presented here is representative of the overall variability of sgr a * since it is not biased towards the presence of obvious ` flare events ' and the only data selection carried out concerned data quality and photometric consistency as explained above . it is by far the most extensive dataset for sgr a * that has been published : the observations we present here consist effectively a @xmath5184 hour long lightcurve ( if one subtracts from the total time all gaps larger than about 20 min , i.e. gaps not due to sky observations ) . the total amount of exposure time ( i.e. the time spent with the shutter open , excluding overheads ) is smaller , and totals @xmath272 hours , though 184 hours is a better indication of the amount of continuous coverage . . this flare is the brightest ks - band flare that has ever been observed ; the source reached an intrinsic peak flux of 27.5 mjy , a factor @xmath21.7 the flux of the star s2 . while the high flux excursion is preceded by what looks like a flatter , background level of emission , this is however , at@xmath57 mjy , several mjy brighter than the low flux levels from most other observation nights , and indicates there was increased source activity even at this time . the long - term trend for 2009 has not been subtracted from the lightcurve shown in this plot , and the last six datapoints did not make the data quality cut for the 2004 - 2009 flux distribution . , width=302 ] figure [ fig_dataset ] shows the 2004 - 2009 lightcurve of sgr a * and comparison star s7 . on top of the more rapid variability , there is a general trend for the lowest fluxes of the lightcurve to wander from year to year . it is at its lowest in 2006 . in 2004 the extra flux appears to be due to a combination of a faint star , a contribution from the halo of s2 , which was much closer to sgr a * during that year , and s19 , which was confused with s17 . the lightcurve also shows some increased flux in 2008 and 2009 ; and as we will show in section [ sec_stellarcontamination ] , there is at least one additional ( previously unidentified ) faint star which must also contribute to the increased flux in those years . since we can generally identify that stars are responsible there is good justification for subtracting these long ( i.e. year ) timescale trends from the lightcurve . for want of a better method of determining the stellar contribution to the lightcurves we subtracted the difference between the median of the data from each year and the minimum median value ( in 2006 ) this brings the offset in the lightcurve to roughly 3.5 - 4 mjy , much of which can be reasonably attributed to s17 ( between 2.9 and 3.3 mjy in the lightcurve can be attributed to s17 ; see section [ section_psfphot ] ) . it is apparent from figure [ fig_dataset ] that sgr a * is much more variable than the comparison source s7 . some of the flux excursions are much more dramatic than others , with fluxes above 8 mjy ( @xmath55 mjy intrinsic upon subtraction of s17 ) . many of these have been previously published for example , the second brightest peak of the dataset , in may 2006 , was published previously in @xcite and @xcite and has 16.7 mjy ( @xmath513.5 mjy intrinsic ; see table [ table2 ] in the appendix ) . on 5th august 2008 , we saw a particularly extreme event which can be seen as the most dramatic flux excursion in the lightcurve in figure [ fig_dataset ] . this event is also shown in higher time resolution in figure [ fig_superflare ] . on this particular night the source at the position of sgr a * brightened by a factor @xmath28 in a period of @xmath540 minutes , reaching a peak flux of 30.7 mjy ( intrinsic flux of @xmath29 mjy minus s17 , a factor increase over the upper limits on the lowest fluxes from sgr a * of @xmath227 ) . an increase in ks - band emission of this brightness is unparalleled in the literature for sgr a * : the second brightest that has been published is a flare from october 2003 @xcite which reached an intrinsic flux of @xmath520.1 mjy ( see table [ table2 ] in the appendix ) . 2 - 3 mas ; for the purposes of this plot the zero position of declination was fixed to the position for which the brightest flux from sgr a * was recorded ( the @xmath30mjy , @xmath3 mag , event from august 5th , 2008 ) . [ fig_contaminatingstar],title="fig:",width=302 ] + to create the 2004 - 2009 flux distribution of sgr a * we subtracted a trend from the data that we suspected to be due to faint stars . however the trend for higher fluxes in 2008 - 2009 is not explained by the presence of any known ( previously identified ) stars . here , however , we are able to present evidence that there was indeed one to two faint stars confused with sgr a * in the years 2007 to 2009 . in figure [ fig_contaminatingstar ] we show that there are deviations in the source position of sgr a * measured from imaging mosaics taken between 2007 and 2009 . the systematic deviations imply there was at least one , previously unidentified , faint star confused with sgr a*. in fact in new , very high quality observations of early 2010 , it is possible to identify two new faint stars very close to sgr a*. one or both of these stars may have undergone a close pericenter passage . from figure [ fig_contaminatingstar ] it appears that one of the stars was present in 2007 also , while our offset for 2007 does not argue for any extra star . however as we discuss in section [ section_previousmeasurements ] there appears to be a relatively steady ( on the order of @xmath50.5 mjy ) stellar contribution to the flux of sgr a * ( including the years 2006 - 2007 ) , which , together with the subtracted offsets , is probably the result of exactly such faint stars moving slowly in and out of confusion with sgr a * : probably one of the newly discovered stars does indeed already contribute to the flux in 2007 ( with the additional increase in flux in 2009 due to the second star ) . it is somewhat a mystery why these stars have not been identified previously . the stars responsible for the observed deviations in the source position have not yet been identified in data prior to 2007 when they should have been resolved separately . perhaps they have until now always been confused with other stellar sources . the star close to sgr a * in 2004 ( identified as s62 in * ? ? ? * though we disagree with this identification ) may be a candidate . + + + + we additionally carried out psf photometry on the ks - band data from 2009 ( 13 mas pixel scale imaging only ) using starfinder @xcite , which we use especially for investigating the question of whether sgr a * is continuously variable at low fluxes ( section [ sec_2009 ] ) . we use an automated program for running starfinder on many images developed by @xcite . the measured fluxes were again calibrated on the s - star s65 ( @xmath31 ; @xcite ) and the results are shown in figure [ fig_2009 ] . we note that due to the presence of the two previously unidentified faint stars ( section [ sec_stellarcontamination ] ) , we improve our chances of detecting a source in most images , while at the same time the faintness of this additional stellar contribution ensures our flux measurements have very small error , such that this data set is optimum for investigating the variability of sgr a * at low fluxes . to produce the lightcurve we first required that a faint s - star ( s21 , of @xmath51.3 mjy ) was detected in a given image as a basic data quality cut . from these images we then examined the detected positions of sources in the near neighbourhood ( @xmath5100 mas ) of the nominal position of sgr a * ( see figure [ fig_2009 ] ) and selected sources that were within 20 mas of the mean position of this detected source ( by this method a source was detected in 87% of the good quality selected images ) . due to this method of source selection , the dataset we present here is not completely unbiased , due to the 13% of images in which no source was detected , but the fraction in which sgr a * was not detected is small enough that the data set still serves well to assess whether sgr a * is in general more variable than stars of comparable flux . from the figure , sgr a * certainly appears most of the time more variable than the comparison stars of similar flux ( s21 and s60 in figure [ fig_2009 ] ) . s60 , in particular provides a very good comparison since it is also close to s31 , a star of similar brightness to s17 ; the two sources s60 and s31 also have a very similar separation to sgr a * and s17 during 2009 ( see figure [ fig_2009 ] ) . the flux distribution of s60 is fit by a gaussian with @xmath32 and @xmath33 which does not give any indication that the photometric accuracy is signficantly worse for sources of this separation and flux ratio ( the more isolated star s21 is fit by a gaussian with @xmath34 and @xmath35 ) . in general the starfinder and aperture photometry fluxes agree very well , with a constant offset of @xmath36 mjy . however , there is a systematic increase in the offset between the measured fluxes from the aperture photometry method and the starfinder method towards the end of 2009 , when the source(s ) confused with sgr a * has(have ) moved the furthest to the south ( see section [ sec_stellarcontamination ] ) . this is likely explained by starfinder s one - psf fit to the distorted psf of two sources ( which has the effect of decreasing the flux measured with the starfinder method ) ; the sources are even beginning to become separated at the end of the year ( figure [ fig_2009 ] ) . this method also provides us with an estimate for the flux contribution of s17 to the 2004 - 2009 flux distribution . the offset of @xmath36 mjy is slightly fainter than the flux of s17 ( @xmath37 mjy when separated from sgr a * in the 2009 data using starfinder ) . the lower total fluxes obtained with the aperture photometry method is probably a result of the higher stellar background surrounding sgr a * ( i.e. in the background aperture ) compared to the more isolated flux calibration star and/or a side effect of the double aperture method . 2.9 mjy is then a low estimate for the contribution of s17 to the combined sgr a*+s17 2004 - 2009 lightcurve , while 3.3 mjy from starfinder photometry is a high estimate . therefore we estimate the contribution of the flux of s17 to our 2004 - 2009 lightcurve produced via aperture photometry to be in the range @xmath38 mjy . for the flux distribution , we define the detection frequency in bin @xmath39 as @xmath40 where @xmath41 and @xmath42 are the exposure time and flux of the @xmath43th image , respectively , there are @xmath44 images and @xmath45 and @xmath46 denote the bin edges . the detection frequency in each bin is divided by the bin width so that the area under the measured flux distribution is equal to 1 . in the flux distribution each lightcurve data point was weighted by its exposure time so that the flux distribution is not biased towards observations with shorter time samplings . we chose a logarithmic binning with bins spaced at intervals separated by a factor of @xmath47 . the errors for each bin are computed as the square root of the exposure times in the bin added in quadrature ( with the same normalization as in equation [ equation_detfreq ] ) i.e. @xmath48 the flux distribution of sgr a * for the years 2004 - 2009 as derived from aperture photometry is shown in figure [ fig_fluxdist ] with various model fits ( described in the next section ) , as well as for a comparison star . the parameters and fit results for the model fits are given in table [ table1 ] . ( equation [ equation_sfobserror2009 ] ) , which are almost negligible ( compare to the intrinsic flux distribution shown in gray ) . for the same data , but analyzed with aperture photometry , the observational errors are important and have the effect of slightly widening and flattening the peak of the distribution . the red dashed line shown is not a fit to the flux distribution produced from the aperture photometry data , but rather , it is the best fit model for the starfinder photometry with a constant flux offset and convolved with the best fit error law _ determined from stars _ for 2009 aperture photometry data ( see footnote [ footnote_apobserror2009 ] ) . _ lower panel _ : comparison stars with fluxes close to the mode the flux distributions shown above , demonstrating the much greater variance in the flux distribution for sgr a * than that expected from observational noise in the photometric data set obtained with starfinder ( see section [ sec_2009 ] ) . [ fig_comparefluxdists],title="fig:",width=302 ] ( equation [ equation_sfobserror2009 ] ) , which are almost negligible ( compare to the intrinsic flux distribution shown in gray ) . for the same data , but analyzed with aperture photometry , the observational errors are important and have the effect of slightly widening and flattening the peak of the distribution . the red dashed line shown is not a fit to the flux distribution produced from the aperture photometry data , but rather , it is the best fit model for the starfinder photometry with a constant flux offset and convolved with the best fit error law _ determined from stars _ for 2009 aperture photometry data ( see footnote [ footnote_apobserror2009 ] ) . _ lower panel _ : comparison stars with fluxes close to the mode the flux distributions shown above , demonstrating the much greater variance in the flux distribution for sgr a * than that expected from observational noise in the photometric data set obtained with starfinder ( see section [ sec_2009 ] ) . [ fig_comparefluxdists],title="fig:",width=302 ] the main difficulty with fitting the distribution of fluxes is the influence of observational errors , which are important enough at low fluxes to widen the peak of the distribution from its intrinsic shape . we characterized the observational errors by fitting gaussian functions to the flux distributions of 22 nearby stars of fluxes between @xmath20.6 and 25 mjy . fitting a power law we obtain @xmath49 similar to that of @xcite who found a typical dependence of @xmath50 for their data . the @xmath20.5 power - law dependence of the errors is consistent with a photon noise origin @xcite . we do not find any flattening of the power law dependence of the noise towards low fluxes with statistical significance . the 13 mas pixel data from 2009 is of much higher quality and for the starfinder photometry has noise properties @xmath51 the noise dependence was determined in the same way as for the 2004 - 2009 dataset through comparison of 22 stars of fluxes 0.5 to 25 mjy . ] which is a significant improvement on the overall 2004 - 2009 aperture photometry accuracy , but still not as good as the accuracy obtained with starfinder for the same dataset . [ footnote_apobserror2009 ] ] we include the effect of observational errors in our model fits by convolving the intrinsic flux distribution with a gaussian of width @xmath52 , which has the dependence on flux of equation [ equation_apobserror ] : @xmath53 where @xmath54 is the intrinsic flux distribution ( without observational errors ) . the effect is only important at low fluxes @xmath55 mjy ( see figure [ fig_fluxdist ] ) and is responsible for broadening the low flux peak in the observed flux distribution for sgr a*. + + cccc|cccccc & + & + & & & & & & & & + @xmath56 & @xmath57 & @xmath58 & @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 & @xmath64 & @xmath65 + & & & & & & & & + & + & + & & & & & & & & + @xmath66 & @xmath67 & @xmath68 & @xmath69 & @xmath70 & @xmath71 & @xmath72 & @xmath73 & @xmath74 & @xmath75 + & & & & & & & & a lognormal fit to the 2004 - 2009 flux distribution from aperture photometry is shown in figure [ fig_fluxdist ] , with the best fit parameters given in table [ table1 ] . this model has intrinsic probability distribution @xmath76 where @xmath77 where @xmath78 represents the observed flux ( measured in mjy here and elsewhere in this paper ) , and @xmath79 with @xmath80 a flux offset due to some constant contribution ( e.g. s17 + contaminating stars ) . the parameters @xmath81 and @xmath82 of the lognormal distribution have a natural analogy to the normal distribution when exponentiated : the source can be thought of as having a median flux of @xmath83 with a multiplicative standard deviation of @xmath84 ( i.e. the interval @xmath85 to @xmath86 contains @xmath268% of the probability ) . the residuals at low fluxes of the lognormal model are probably just due to the low flux wing of the error distribution which is not perfectly gaussian ( see star s7 in figure [ fig_fluxdist ] ) . the residuals for the lognormal model at high fluxes , on the other hand , can not arise from measurement error , and we obtain a much better fit to the flux distribution with a lognormal model which makes a transition to a flatter tail at high fluxes . the intrinsic probability distribution in this case is @xmath87 where @xmath88 is the lognormal distribution of equation [ equation_p_logn ] , @xmath89 is the flux at which the distribution makes the transition to the flatter tail , @xmath90 is the power - law slope of the tail , and @xmath39 is a renormalizing factor ] such that the total probability is 100% . the flattening of the distribution towards high fluxes in the 2004 - 2009 aperture photometry flux distribution is significant at @xmath91 . however , the fits are not formally ` good fits ' . this is to the most part due to residuals in the lowest flux bins , most likely caused by the deviation of the observational errors from the plain gaussian description used in our model . if we instead only fit only to fluxes @xmath92mjy ( ignoring the lowest three bins ) , we obtain a more reasonable @xmath93 for the lognormal plus tail model of 17.4/13 ( while the best fit lognormal model for the same has 37.6/15 ) . in this case the flatter tail of the distribution is still significant at @xmath94 . interestingly , in the 2009 starfinder photometric dataset the tail is also present with similar best fit parameters , although in this smaller dataset ( which only reaches fluxes of @xmath510 mjy ) the tail has a significance of only 1.4@xmath95 . what is not clear in the 2004 - 2009 data set is whether it is appropriate to fit only continuously emitting , continuously variable models to the flux distribution . a plausible alternative is a model in which the source of near - infrared emission emits only sporadically ( i.e. exhibits ` on ' and ` off ' periods ) . if this is the case observational errors alone ( on the measurement of a constant component to the flux ) should be responsible for observed variations in the flux level during ` off ' periods . the goal of this section is to determine whether there is too much variation at low levels for the data to be explained by observational errors in ` off periods ' of a sporadic model . we use the following prescription for a sporadic model for the flux distribution : @xmath96 where @xmath97 is the fraction of time the source spends in an ` on ' state . using @xmath98 ( equation [ equation_p_logn ] ) and fitting to both the 2004 - 2009 aperture photometry flux distribution and to the 2009 starfinder distribution with the measured noise laws of equations [ equation_apobserror ] and [ equation_sfobserror2009 ] , we find in both cases the best fits to the distributions require @xmath99 and @xmath100 ( at @xmath101 ) respectively , i.e. the measured noise laws can not account for the observed amount of variability at low fluxes . to look at the question from a different angle , we investigate what level of observational noise would be required to produce the observed variation at low fluxes . to do this we allow the normalization of the noise law to be a free parameter in the model fits , i.e. @xmath102 we find @xmath103 and @xmath104 for the 2004 - 2009 aperture photometric datset and the 2009 starfinder dataset respectively . we also refit the noise laws to the stars ( see equations [ equation_apobserror ] and [ equation_sfobserror2009 ] ) fixing the slope of the error law to 0.5 and obtain @xmath105 and @xmath106 for the 2004 - 2009 aperture photometric datset and the 2009 starfinder dataset respectively . the best fit noise law to the 2004 - 2009 flux distribution from aperture photometry allows @xmath107 but the normalization of the noise law is inconsistent with that of the stars at a level of 3.3@xmath95 . for the 2009 flux distribution from starfinder photometry the required noise law is even more unlikely , and is ruled out at a level of 4.5@xmath95 . we conclude that the observed behavior of sgr a * at low fluxes is best consistent with a continuously emitting , continuously varying source . we also detect longer timescale trends in the low flux level of sgr a * in the 2009 data ( figure [ fig_2009 ] ) , such as an @xmath50.8 mjy increase in the minimum level from the fourth to the fifth data sections , with a gap in time of only 12 days , and a similar size decrease from the second to the fourth data sections with a gap in time of only 11 days . it might be that a passing star ( such as the two new stars discovered in 2010 , see section [ sec_stellarcontamination ] ) is responsible for these trends , either passing in and out of confusion very quickly , or even just moving within the sgr a * psf and distorting the flux : the further apart the two sources the lower the fitted flux . however these stars can not move quickly enough for this : the change in flux is so large that in either case the star must move by at least @xmath550 mas within a period of 12 days to distort the flux as much as observed , which at the 8kpc distance of the gc requires an average speed of at least 0.1@xmath108 over a two week timescale . even assuming the most extreme case a star moving only in the plane of the sky and falling towards sgr a * in a parabolic orbit at the escape speed a star should take at least @xmath5110 days ( 0.02@xmath108 in average ) , to cover such a distance . a larger centroid shift between the datasets would also be expected if it were the case that a star were distorting the source psf ( @xmath520 mas instead of the observed @xmath109 mas ) . furthermore , even if this could explain the trend between the fourth and fifth datasets , the same explanation would have to be invoked to again explain the variation between the second and fourth datasets , also on a timescale of only @xmath511 days . we therefore conclude that the observed trends on timescales of a few weeks in the lightcurve are most likely intrinsic . though they can not be due to direct flux contamination from passing stars , the trends are possibly still related to the feeding of the accretion flow from stellar winds of the passing stars , which may trigger increased activity as the star passes orbital pericenter ( e.g. @xcite ) . we have shown in section [ section_results ] that the flux distribution of sgr a * is best described by a continuously emitting , continuously variable component with a lognormal flux distribution at low fluxes , but additionally that there is a significantly flatter tail above about 5 mjy , implying that _ there is something different about the high flux emission from sgr a*_. put together , it seems justified to identify * emission below 5 mjy as _ quiescent state _ emission * emission above 5 mjy as _ flaring _ emission . this is the ( phenomenological ) definition by which we hereafter refer to either ` flares ' or ` quiescent ' emission in sgr a*. we present how the two states may fit into a possible physical picture below . that the low fluxes of sgr a * are fit by a continuously present , variable component suggests that there is truly a _ _ quiescent state _ _ in sgr a * , i.e. a low - level , lognormally varying component ( factor @xmath52 in flux of the lognormal component corresponds to 1@xmath95 variability ) . our finding of continuously emitting , continuously variable component confirms the findings of @xcite who first reported continuous variability of sgr a * from the larger variance of the flux of sgr a * , as compared to stellar sources , in 5 out of 6 k-band observation nights . lognormal distributions of fluxes are seen in x - ray binaries , where they are interpreted as a sign that multiplicative processes are behind the variability ( e.g. @xcite ) . a good example of a multiplicative process is the model of @xcite where variability is produced through the inwards propagation of accretion rate fluctuations at different radii . these variations , produced in regions of the flow which themselves do not emit near - infrared emission , all combine multiplicatively as they propagate inwards . the end result of combined fluctuations is then imprinted on the variability produced by the ( say , near - infrared ) emitting region of the flow . a very similar idea may work with magnetic turbulence , which has been shown in grmhd / mhd simulations to be capable of producing the factor 40 - 50% millimeter variability @xcite . from where exactly the quiescent emission arises is however a difficult question . nonthermal electrons are required to emit in the near - infrared , and it s not clear how these should be related to the thermal electrons emitting at submm wavelengths , or how they should necessarily be arranged within the accretion inflow / outflow . * see also @xcite ) require a hybrid thermal / non - thermal electron distribution to explain the spectrum of sgr a * ( in particular the excess radio emission at low frequencies ) . the models predict that the same non - thermal electrons ( with @xmath51.5% of the thermal energy ) could produce emission in the near - infrared . thus it might be interesting to compare the flux distribution of the nir quiescent state emission ( @xmath110 mjy ) with that of the low - frequency radio emission , as well as correlations in timing properties , such as whether the medium - timescale trends ( @xmath52 weeks ) observed in the 2009 data ( figure [ fig_2009 ] ) correlate with any similar long - term trends in the radio regime . more detailed comparison of the low - level nir emission from sgr a * with other wavelengths ( e.g. submm ) could also shed light on how non - thermal electrons are related to the thermal population . it seems unlikely that the same variability process is responsible for both high and low flux emission from sgr a*. there is no obvious reason for the underlying process , if multiplicative at low fluxes , to deviate at high fluxes in order to explain the flatter tail above @xmath63 mjy . the flatter tail of the flux distribution in sgr a * requires at the very least some transition within the source ( to a _ flaring state _ ) , triggered around an intrinsic flux of 5 mjy , such that either large fluctuations or their emission undergo some kind of runaway amplification . on the other hand the physical mechanism behind the flares could be unrelated to ( though possibly still triggered by ) the mechanism producing the low level variability . the flare tail could arise from additional sporadic _ flare events _ which dominate the distribution of fluxes above 5 mjy . in this case 5 mjy is not a physically interesting flux outside the fact that it is the flux at which one is equally likely to observe either a flare event or a low level lognormal variation . spontaneous magnetic reconnection is a good candidate for a physical process that could give rise to sporadic flares in sgr a * ( e.g. * ? ? ? * ; * ? ? ? the decrease in magnetic field strength that would accompany the conversion of magnetic energy to high energy electron acceleration in a magnetic reconnection event can explain the non - one - to - one rise in the lightcurves of the so far brightest nir / x - ray flare from sgr a * @xcite . a number of other properties also indicate that there are two types of nir emission from sgr a * , with differences between high and low flux emission , which include : * _ spectral index_. there are indications that the spectral index at low fluxes is redder than at higher fluxes @xcite , although this has been disputed by other authors @xcite . * _ polarization_. changes in polarization in the rising / decaying phases of flares are seen @xcite , and low level emission appears to be generally more highly polarized ( @xmath111 ) than high flux emission ( @xmath112 ) @xcite . * _ association with x - ray flares_. the best fit model ( _ lognormal+tail+errors _ ) to the 2004 - 2009 flux distribution has 4.5% of the probability above the best fit transition flux of @xmath113 mjy , which , interestingly , is close to the x - ray flaring rate of @xmath114 ( for flares occurring once a day with a duration of 60 - 100 min ; @xcite ) . estimates of the nir flaring rate including emission below 5 mjy , on the other hand , typically put the nir flaring rate much higher at @xmath115 @xcite . perhaps the additional low - level component is even responsible for differences in the nir / x - ray peak flux ratios . there are a number of high flux ks / k-band events ( @xmath116 mjy ) , not part of the 2004 - 2009 dataset presented here but published elsewhere in the literature , which are also of @xmath1175 mjy intrinsic flux and can be identified as belonging to the flare tail of the flux distribution . to list some of the brightest ( see table [ table2 ] in the appendix for details ) : @xmath211 and 8.4 mjy ( 15 , 16 june 2003 ; * ? * ) ; 20.1 mjy ( 6 october 2003 ; * ? ? ? * ) ; 13.5 mjy ( 31 may 2006 ; @xcite , @xcite , and part of our 2004 - 2009 dataset ) ; 10.7 mjy ( 15 may 2007 ; * ? ? ? * ) ; and 14.8 mjy @xcite . our results are reasonably consistent with the median flux of 2.0 mjy over six separate observation nights from 2006 - 2007 found by @xcite ( see table [ table2 ] in the appendix for the scaling factors applied to scale the reported values to our calibration / extinction ) . we find a median flux of @xmath21.6 mjy for our 2004 - 2009 lightcurve with the long - timescale trend and s17 ( assuming 3.1 mjy ) removed ( note however that this still includes some constant contribution : the median flux of the lognormal component alone is 1.1 mjy ) . a number of previous measurements have put upper limits on any long - term steady quiescent state at near - infrared wavelengths which have been on the @xmath51 - 1.5 mjy level ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * see table [ table2 ] in the appendix ) . these measurements are consistent with the picture we have put forward here for sgr a * at low levels , of a continuously emitting , continuously varying source . for a lognormally varying quiescent state with the properties of our best fit model to the 2004 - 2009 flux distribution , an upper limit of 1 - 1.4 mjy in selected observations is not surprising : the source is expected to be at @xmath1181 mjy roughly @xmath119 of the time . however , more likely the upper limits of 1 - 1.4 mjy include a stellar contribution to the flux . @xcite estimated an upper limit on the _ intrinsic _ flux from sgr a * of @xmath1200.5 mjy in the presence of a likely @xmath50.8 - 1 mjy stellar contribution . in the context of our best fit lognormal model , this low flux measurement is a rarer , but not unlikely occurrence , with a probability of @xmath121 . the 0.8 - 1.0 mjy diffuse component reported by @xcite might be associated with the @xmath50.3 - 0.7 mjy offset we find in our fits to the flux distribution of sgr a * , between the fitted constant component ( 3.6 mjy ) and the expected contribution of s17 ( 2.9 - 3.3 mjy ) . @xcite also estimated from the color of sgr a * at low fluxes , that they most likely had @xmath535% contribution from stellar contamination ( 0.7 mjy ) for their observations in 2006 . these results all seem to point towards an additional stellar component to the flux of sgr a * with a constant flux of @xmath50.5 mjy ( our data ) , though it does not necessarily always consist of the same stars from year to year . lognormal distributions of fluxes are also seen in other astrophysical objects , for example x - ray binaries ( prototype cyg x-1 in the low hard state , * ? ? ? * ) , implying a connection between sgr a * other low - accretion rate black holes @xcite . the physical picture of the low - hard state of x - ray binaries is that compared to higher luminosity flows the inner optically thick accretion disc disappears and is replaced by a hot inner flow @xcite . similar pictures , with hot inner flows , exist for sgr a * ( e.g. * ? ? ? sgr a * is , however , at the extreme low end of the accretion rate range of observable sources ( the ` quiescent ' state ) . connections between sgr a * and low - hard state x - ray binaries of higher accretion rate such as cygnus x-1 can shed light on whether or not the quiescent state is indeed a distinct state or whether there is simply a smooth continuation of the hard state down to such low accretion rates ( @xcite ) . if timescale scales with mass , then the flares in sgr a * of @xmath510 - 100 minute duration , correspond to @xmath2 millisecond timescales in a stellar mass black hole binary . it is also true that millisecond - timescale flares are seen in cyg x-1 , though it was shown by @xcite that the majority of these ( and all those in the low / hard state ) were not more frequent than expected in the context of the extension of the low - level lognormal distribution in this source to higher fluxes . there was one flare however , a 12.4@xmath95 event that occurred ( in fact when cyg x-1 was in the high / soft state ) that could not be accounted for in the context of the underlying variability and was posited by the authors to have a different physical origin . in this vein , it could be that the flares in sgr a * are due to the source s `` flickering attempts at outburst activity from out of quiescence '' , as suggested by @xcite . a connection between the near - infrared variability of sgr a * and the variability of x - ray binaries has also been made previously from a long timescale break in the psd of sgr a * which can be associated with a break timescale in the low - hard state of x - ray binaries ( meyer et al . 2009 ) , though the single dataset used to constrain the high frequency power spectrum in this case may not be representative . the lognormal distribution for low - level fluxes makes however an independent case for an analogy between the variability of sgr a * and low - hard state x - ray binaries . the flux distributions of @xcite and @xcite at 1.6@xmath21 m and 1.7@xmath21 m display a qualitatively similar shape to our ks - band distribution , although these datasets are still much smaller than ours , and the low - level activity appears to be dominated by the observational noise . nevertheless , with more data at 1.6 and 1.7 @xmath21 m and modeling of the observational errors , it might be possible to obtain constraints on the colours of the low level and high flux components . in future work we intend to use the same techniques as we have used in this paper at ks - band , in other bands ( l and h ) to investigate in detail the dependence of the flux distribution with wavelength . our work is not the first to point out that the flux distribution of sgr a * is not well - described by a gaussian : @xcite , @xcite and @xcite have all noted that the flux distribution has a high flux tail . however , the timing analyses which have been carried out to date @xcite in which sgr a * was compared to simulated red noise lightcurves have nevertheless used simulated lightcurves which have had , by construction , a gaussian flux distribution . sgr a * s intrinsic mean and variance are such that if one were to plot the simulated lightcurves scaled to sgr a * s mean and variance ( for the 2004 - 2009 dataset @xmath122 mjy subtracting 3.8 mjy for s17 plus other faint stars , and @xmath123 mjy ) , the model red noise lightcurves would have negative intrinsic fluxes much of the time which is rather unphysical . in @xcite the high flux tail was excluded when fitting the flux distribution , and only fluxes @xmath1183.2 mjy were fitted with a gaussian ( see table [ table2 ] in the appendix for scaling factors ; the reported value in @xcite was a reddened flux of 0.3 mjy ) . thus their model lightcurves best resemble the low level emission of sgr a*. indeed the entire dataset used by @xcite does not include much bright emission @xmath116 mjy , and thus their results really apply to the low level emission of sgr a*. that the low level emission from sgr a * is consistent with red noise fits together with our finding of the lognormal distribution of fluxes at low fluxes . both of these aspects of the variability are seen in x - ray binaries such as cyg x-1 @xcite . however , given that the flux distribution is not gaussian , it remains questionable that the power spectrum has yet been reliably constrained from the monte carlo simulations of previous timing studies . in particular , high and low flux states of sgr a * may require separate timing analysis . in this respect it is interesting that @xcite finds a quasi - periodic signal with false alarm probability of only @xmath124 ( 4.2@xmath95 significance in gaussian equivalent terms ) , if one considers just the window 385 - 445 minute subset from 30 - 31 july 2005 , though the signal was deemed insignificant when the entire lightcurve from the night was analyzed . we have determined the distribution of fluxes for the variable source sgr a * in the near - infrared and thereby investigated flux - dependent properties of the emission . we summarize our main results as follows : * sgr a * is continuously emitting and variable in the near - infrared . * at low ks - band fluxes the variability follows a lognormal distribution with a median flux of 1.1 mjy and a multiplicative standard deviation factor of 2.1 . we identify this continuously variable low - level state as the quiescent state of sgr a * in the near - infrared . * at least 0.5 mjy of the flux of sgr a * is due to a faint stellar component at all times during 2004 - 2009 . * there is variability on a @xmath2two - week timescale in the low level component , which can not be attributed to stars . * at high fluxes ( above @xmath55 mjy ) the flux distribution flattens which seems to be most likely explained by the presence of sporadic flare events additional to the quiescent emission . * on august 5th 2008 we observed a very bright ks - band flare of 27.5 mjy , the brightest ks - band flare yet observed from sgr a*. differences in spectral index , polarization properties and association with x - ray flares may already be further indications of the different nature of the high flux near - infrared emission from sgr a*. in future work we aim to search for and quantify further possible differences between high and low flux emission of sgr a*. balick , b. , & brown , r. l. 1974 , , 194 , 265 baganoff , f. k. , et al . 2001 , , 413 , 45 baganoff , f. k. 2003 , bulletin of the american astronomical society , 35 , 606 chan , c .- k . , liu , s. , fryer , c. l. , psaltis , d. , zel , f. , rockefeller , g. , & melia , f. 2009 , , 701 , 521 diolaiti , e. , bendinelli , o. , bonaccini , d. , close , l. , currie , d. , & parmeggiani , g. 2000 , astronomical data analysis software and systems ix , 216 , 623 dexter , j. , agol , e. , & fragile , p. c. 2009 , , 703 , l142 ding , j. , yuan , f. , & liang , e. 2010 , , 708 , 1545 do , t. , ghez , a. m. , morris , m. r. , yelda , s. , meyer , l. , lu , j. r. , hornstein , s. d. , & matthews , k. 2009 , , 691 , 1021 do , t. , ghez , a. m. , morris , m. r. , lu , j. r. , matthews , k. , yelda , s. , & larkin , j. 2009 , , 703 , 1323 dodds - eden , k. , et al . 2009 , , 698 , 676 dodds - eden , k. , sharma , p. , quataert , e. , genzel , r. , gillessen , s. , eisenhauer , f. , & porquet , d. 2010 , arxiv:1005.0389 done , c. , gierliski , m. , & kubota , a. 2007 , , 15 , 1 eckart , a. , et al . 2004 , , 427 , 1 eckart , a. , et al . 2006 , , 450 , 535 eckart , a. et al . 2006 , , 455 , 1 eckart , a. , et al . 2008 , , 479 , 625 eckart , a. , et al . 2008 , , 492 , 337 eisenhauer , f. , et al . 2005 , , 628 , 246 falcke , h. , krding , e. , & markoff , s. 2004 , , 414 , 895 fritz , t. , et al . 2010 , , 401 , 1177 genzel , r. , schdel , r. , ott , t. , eckart , a. , alexander , t. , lacombe , f. , rouan , d. , & aschenbach , b. 2003 , , 425 , 934 ghez , a. m. , et al . 2004 , , 601 , l159 ghez , a. m. , et al . 2008 , , 689 , 1044 gillessen , s. , et al . 2006 , , 640 , l163 gillessen , s. , eisenhauer , f. , trippe , s. , alexander , t. , genzel , r. , martins , f. , & ott , t. 2009 , , 692 , 1075 goldston , j. e. , quataert , e. , & igumenshchev , i. v. 2005 , , 621 , 785 hornstein , s. d. , ghez , a. m. , tanner , a. , morris , m. , becklin , e. e. , & wizinowich , p. 2002 , , 577 , l9 hornstein , s. d. , matthews , k. , ghez , a. m. , lu , j. r. , morris , m. , becklin , e. e. , rafelski , m. , & baganoff , f. k. 2007 , , 667 , 900 lenzen , r. , et al . 2003 , , 4841 , 944 loeb , a. 2004 , , 350 , 725 lyubarskii , y. e. 1997 , , 292 , 679 markoff , s. 2010 , lecutre notes in physics , berlin springer verlag , 794 , 143 mchardy , i. m. , koerding , e. , knigge , c. , uttley , p. , & fender , r. p. 2006 , , 444 , 730 meyer , l. , schdel , r. , eckart , a. , karas , v. , doviak , m. , & duschl , w. j. 2006 , , 458 , l25 meyer , l. , schdel , r. , eckart , a. , duschl , w. j. , karas , v. , & doviak , m. 2007 , , 473 , 707 meyer , l. , do , t. , ghez , a. , morris , m. r. , witzel , g. , eckart , a. , blanger , g. , & schdel , r. 2008 , , 688 , l17 meyer , l. , do , t. , ghez , a. , morris , m. r. , yelda , s. , schdel , r. , & eckart , a. 2009 , , 694 , l87 zel , f. , psaltis , d. , & narayan , r. 2000 , , 541 , 234 porquet , d. et al . 2008 , , 488 , 549 quataert , e. 2003 , astronomische nachrichten supplement , 324 , 435 rank , s. 2007 , masters thesis , technische universitt mnchen . reid , m. j. , & brunthaler , a. 2004 , , 616 , 872 rousset , g. , et al . 2003 , , 4839 , 140 sabha , n. , et al . 2010 , , 512 , a2 schdel , r. , et al . 2002 , , 419 , 694 schdel , r. , eckart , a. , mui , k. , meyer , l. , viehmann , t. , & bower , g. c. 2007 , , 462 , l1 schdel , r. , najarro , f. , muzic , k. , & eckart , a. 2010 , , 511 , a18 tokunaga , a. t. 2000 , allen s astrophysical quantities , 143 trippe , s. , paumard , t. , ott , t. , gillessen , s. , eisenhauer , f. , martins , f. , & genzel , r. 2007 , , 375 , 764 uttley , p. , mchardy , i. m. , & vaughan , s. 2005 , , 359 , 345 yusef - zadeh , f. , et al . 2006 , , 644 , 198 yusef - zadeh , f. , et al . 2009 , , 706 , 348 yuan , f. , quataert , e. , & narayan , r. 2003 , , 598 , 301 |
according to recent models , the accretion disk and black hole in active galactic nuclei are surrounded by a clumpy torus . we investigate the nir flux variation of the torus in response to a uv flash for various geometries . anisotropic illumination by the disk and the torus self - occultation contrast our study with earlier works . both the waning effect of each clump and the torus self - occultation selectively reduce the emission from the region with a short delay . therefore , the nir delay depends on the viewing angle ( where a more inclined angle leads to a longer delay ) and the time response shows an asymmetric profile with a negative skewness , opposing to the results for optically thin tori . the range of the computed delay coincides with the observed one , suggesting that the viewing angle is primarily responsible for the scatter of the observed delay . we also propose that the red nir - to - optical color of type-1.8/1.9 objects is caused by not only the dust extinction but also the intrinsically red color . compared with the modest torus thickness , both a thick and a thin tori display the weaker nir emission . a selection bias is thus expected such that nir - selected agns tend to possess moderately thick tori . a thicker torus shows a narrower and more heavily skewed time profile , while a thin torus produces a rapid response . a super - eddington accretion rate leads to a much weaker nir emission due to the disk self - occultation and the disk truncation by the self - gravity . a long delay is expected from an optically thin and/or a largely misaligned torus . a very weak nir emission , such as in hot - dust - poor active nuclei , can arise from a geometrically thin torus , a super - eddington accretion rate or a slightly misaligned torus . | 1107.0678 | active galactic nuclei ( agns ) are powered by gas accretion onto supermassive black holes ( bhs ) at the center of each galaxy . a variety of observations suggest that the accretion disk and the bh are surrounded by an optically and geometrically thick torus ( telesco et al . 1984 ; antonucci & miller 1985 ; miller & goodrich 1990 ; radovich et al . since the torus potentially plays a role of a gas reservoir for the accretion disk , its nature , such as the structure , the size and the mass , has long been investigated ( pier & krolik 1992 , 1993 ; fukue & sanbuichi 1993 ; granato & danese 1994 ; efstathiou & rowan - robinson 1995 beckert & duschl 2004 ; mor et al . 2009 ) . a large geometrical thickness of the torus revealed by various observations ( antonucci 1993 ; pogge 1989 ; wilson & tsvetanov 1994 ; schmitt & kinney 1996 ) indicates that numerous dusty clumps , rather than a smooth mixture of gas and dust , constitute the torus with a large clump - to - clump velocity dispersion @xmath0 ( krolik & begelman 1988 ; wada & norman 2002 ; hnig & beckert 2007 ) . temperature of clumps is less than a critical temperature @xmath1k above which dust grains are sublimated ( barvainis 1987 ) . infrared ( ir ) emission and absorption features provide unique opportunities to probe the clumpy torus ( nenkova et al . 2002 , 2008 ; dullemond & van bemmel 2005 ; hnig et al . 2006 ; geballe et al . 2006 ; shirahata et al . 2007 ; ibar & lira 2007 ; schartmann et al . 2008 ; deo et al . 2011 ) . clumps are heated by illumination from the central accretion disk , and closer clumps have higher temperature . inner edge of the dusty torus is determined by the sublimation process so that clumps temperature equals @xmath2 there , and radiates at near - ir ( nir ) as `` 3@xmath3 m bump '' ( rees et al . 1969 ; neugebauer et al . 1979 ; edelson & malkan 1986 ; kobayashi et al . based on the energy balance of the clump closest to the bh , barvainis ( 1987 ) derived the innermost radius of the torus ( dust sublimation radius , denoted as @xmath4 in this study ) : @xmath5 where @xmath6 and @xmath7 are uv luminosity and the size of dust grains , respectively . indeed , nir emission from type-1 agns lags behind optical variation by an order of a month ( clavel et al . 1989 ; glass 1992 , 2004 ; nelson 1996 ; oknyanskij et al . 1999 ; minezaki et al . 2004 ; suganuma et al . moreover , the luminosity dependency of the time lag also coincides with the theoretical prediction as @xmath8 ( suganuma et al . 2006 ; gaskell et al . however , the nir - to - optical time lag is systematically smaller than the lag predicted from equation ( [ eq : bar87 ] ) by a factor of @xmath9 ( oknyanskij & horne 2001 ; kishimoto et al . 2007 ; nenkova et al . 2008 ) . to tackle with this conflict , kawaguchi & mori ( 2010 , hereafter paper i ) pointed out that the illumination by an optically thick disk is inevitably anisotropic , which is a fact missing in deriving equation ( [ eq : bar87 ] ) . there is a systematic difference between the inclination angle at which we observe the disk in type-1 agns and the angle at which an aligned torus observes the disk . the effects of the anisotropic illumination naturally resolve the puzzle of the systematic deviation of a factor of @xmath9 ( paper i ) . in paper i , we assumed the configuration appropriate for a typical type-1 agn . in this study , we investigate the expected characteristics of the nir emission for various geometries of the disk , the torus and the observer . anisotropic illumination by the disk and the effect of the torus self - occultation contrast our study with earlier works . the next section describes the calculational methods of our model . then , properties of nir emission from an aligned ( section [ sec : aligned ] ) and a misaligned ( section [ sec : misaligned ] ) tori are presented . finally , we make a summary of this study in section [ sec : summary ] . we calculate nir reverberation / echo from the inner part of the torus in response to a flash of disk illumination . the calculational method is essentially the same as paper i , except for the incorporation of the torus self - occultation in this study . by considering the anisotropy of the disk illumination , we solved the inner structure of the torus , and explained why the observed time delay of nir emission is systematically shorter than equation ( [ eq : bar87 ] ) . a large grain size and/or extinction between the torus and the disk are possibly alternative concepts to reduce the inner radius of the torus ( maiolino et al . 2001 ; gaskell & benker 2007 ; gaskell et al . 2007 ; kishimoto et al . 2007 ) . while the geometry of the torus , the disk and the observer appropriate for a typical type-1 agn was assumed in paper i , we here investigate how the nir response differs with various possible geometries . a variety of type-1 objects with a common @xmath4 are compared . in other words , we compute for objects with the same isotropic - equivalent luminosity ( luminosity presuming isotropic emission " ) in optical / uv . barvainis ( 1992 ) examined the nir response due to the dust reverberation of an agn torus . since the inner radius of the torus is an input parameter there , the mean time delay of nir emission behind the optical / uv flux variations is simply coupled with the assumed inner radius . on the other hand , in paper i and this study , both the inner radius and the time delay are output of calculations . moreover , he supposed that the whole torus , an ensemble of cube - shaped , optically thick clumps , is optically thin . since the inner part of an agn clumpy torus is likely optically thick ( appendix ) , we take into account the torus self - occultation as well as the anisotropic emission from spherical , optically thick clumps . similarities and differences in the results for the time response between our and his calculations are discussed in sections [ sec : phi_contri ] and [ sec : qobs ] . in radiative transfer calculations of dusty clumpy tori ( e.g. , hnig et al . 2006 ; nenkova et al . 2008 ) , the isotropic illumination , mainly in optical and uv , is assumed for simplicity . we take into account the fact that the disk emission is inevitably anisotropic . then , the broadband color between the optical / uv radiation from the disk and the nir emission from the torus can be computed appropriately . in sections [ sec : qobs ] and [ sec : qmin ] , we compare our results with theirs . below , our calculational method is summarised . radiation flux ( @xmath10 ) from a unit surface area of an optically thick disk toward a unit solid angle at the polar angle of @xmath11 decreases with an increasing @xmath12 as follows ( netzer 1987 ) : @xmath13 here , the first term represents the change in the projected surface area , while the latter represents the limb darkening effect for plasma , whose opacity is dominated by electron scattering over absorption ( sunyaev & titarchuk 1985 ; phillips & meszaros 1986 ) . figure [ fig : flux_theta_dep ] shows the @xmath11-dependency of @xmath14 , where the former effect alone is also drawn for comparison . an accretion disk emits lesser radiation in the direction closer to its equatorial plane ( i.e. , larger @xmath12 ; laor & netzer 1989 ; sun & malkan 1989 ; hubeny et al . if the torus and the disk are aligned , the assumption of isotropic emission from accretion disks ( e.g. , equation ( [ eq : bar87 ] ) ) obviously overestimates the radiation flux toward the torus , leading the overestimation of the inner radius of the torus . this effect works even if the disk is infinitesimally thin . as shown in section [ sec : qmax ] , a nonzero thickness of the disk brings about another anisotropy of illumination flux , such that the torus is not illuminated below the disk height at @xmath12 larger than a critical angle @xmath15 . except in section [ sec : qmax ] , we throughout adopt a thin disk with an aspect ratio of @xmath16 , like the standard accretion disk model ( shakura & sunyaev 1973 ) . the effects of anisotropic emission and orientation have been discussed in the context of the baldwin effect in the line fluxes of photo - ionised emission ( netzer 1985 ; francis 1993 ; bottorff et al . these effects upon the torus were examined for the first time in paper i. the inner edge of the torus is determined so that the temperature of a clump ( at the irradiated surface ) equals @xmath2 there . since the radiation flux from the disk @xmath10 varies with the polar angle @xmath12 , the sublimation radius of the torus @xmath17 is also a function of @xmath12 . namely , @xmath17 is the distance between the torus edge and the central bh for various @xmath12 . in contrast , we express the sublimation radius estimated under the isotropic emission assumption ( equation ( [ eq : bar87 ] ) ) as @xmath4 . the anisotropic illumination given in equation ( [ eq : flux ] ) results in @xmath18^{0.5}. \label{eq : rsub}\ ] ] here , @xmath19 is the polar angle toward the observer seen from the central accretion disk . outside this radius , there are numerous clumps with their temperature below @xmath2 . in the case of an isotropic emission from the disk , the torus edge is supposed to stand at a distance of @xmath4 . although various grain sizes result in the sublimation process occurring over a transition zone rather than a single distance ( nenkova et al . 2008 ) , we employ a sharp boundary for simplicity . in paper i , we found that ( 1 ) the torus inner edge is located closer to the central bh than suggested by previous estimations ( equation ( [ eq : bar87 ] ) ) and that ( 2 ) the structure of the edge is concave / hollow . moreover , ( 3 ) our result indicated that the innermost edge of the torus may connect with the outermost edge of the accretion disk continuously ( e.g. , emmering et al . 1992 ; elitzur & shlosman 2006 ) . figure [ fig : schematic ] shows a schematic view of the torus inner region . if the torus is indeed a reservoir of gas for the disk , angular momenta of the infalling gas will align these axes . the misalignment between the torus and the disk is investigated in section [ sec : misaligned ] . current interferometric nir observations ( e.g. , swain et al . 2003 ; kishimoto et al . 2009 ) and future adaptive optics imaging with @xmath20 m telescopes can not spatially resolve the innermost region ( radius and shape ) of the torus in nearby seyfert galaxies . thus , observations of time variability will continue to be powerful tools to probe the innermost structure of the torus even in the coming decade . we calculate the time variation of nir emission in response to a @xmath21-function like variation of the optical / uv illumination [ transfer function @xmath22 . transfer functions contain various information of the re - emitting region , such as the shape and the emissivity profile etc ( blandford & mckee 1982 ; netzer 1990 ; gaskell et al . time variation of the reprocessed radiation ( nir in this study ) is a convolution of the illumination flux variation with @xmath23 . we calculate @xmath23 and its centroid @xmath24 , which corresponds to the observed time lag . since the time profile is also one of the characteristics of this study compared to the earlier work ( sections [ sec : phi_contri ] and [ sec : theta_contri ] ) , we also calculate the width @xmath25 and the skewness @xmath26 to describe the shape of @xmath23 : @xmath27^{0.5}\\ s & = & \int ( t - t_{\rm delay})^3 \ , \psi ( t ) \ , dt \ , / \int \psi ( t ) \ , dt \ , /\ , rms^3 \end{aligned}\ ] ] a negative ( or positive ) @xmath26 means that the distribution has a left ( or right ) tail . the computed @xmath25 and @xmath26 would be useful to interpret and predict the cross correlation functions between nir and optical / uv light curves . a larger @xmath25 will correspond to a larger uncertainty in the measurements of the time delay . as to the detectability of nir variations , a small @xmath25 and a high @xmath23 indicate the relative ( i.e. in mag ) and absolute ( e.g. , in ergs@xmath28 ) variability , respectively . when the illumination flux varies , the inner edge of the torus shifts in principle ( laor 2004 ) . depending on whether dust grains in the clumps are sublimated or not , clumps belong to either the broad emission line region or the dusty torus ( netzer & laor 1993 ; suganuma et al . however , it takes @xmath29year for the inner region of the torus to adjust to the varying illumination flux ( koshida et al . 2009 ; pott et al . thus , we regard that the inner structure of the dusty torus is steady in the timescale of nir - to - optical time lag ( @xmath30months ) . to calculate @xmath23 for the clumpy torus , we consider the following items ; ( 1 ) the optical path , ( 2 ) nir emissivity of the torus inner region as a function of @xmath12 and ( 3 ) anisotropic emission of each clump . in this work , ( 4 ) we include the effect of torus self - occultation ( i.e. , absorption of nir emission from a clump by other clumps on the line of sight ) . while the torus self - occultation is a minor effect for a typical type-1 agn , it plays a significant role for inclined viewing angles , thick tori and misaligned tori . considering the self - occultation , we ignore the response from the aligned torus at @xmath31 ( appendix ) . first , ( 1 ) the optical path difference is written as @xmath32 , \label{eq : opd}\ ] ] where @xmath33 is the azimuthal angle and defined so that @xmath34 for the observer ( figure [ fig : schematic ] ) . the concave shape of the inner region of the torus reduces the optical path difference . clumps at slightly farther and those at slightly closer than @xmath35 will also emit nir radiation , smearing out the resultant nir response . since this effect unlikely changes the @xmath23 drastically , we consider only optical paths that hit the inner edge of the torus . next , ( 2 ) for the emissivity as a function of @xmath12 , we assume that ( 2 - 1 ) the clump size increases and ( 2 - 2 ) the clump number density decreases when the clump - to - bh distance increases ( e.g. , hnig et al . 2006 ; schartmann et al . following paper i , the emissivity of nir flux per @xmath36 is assumed to be proportional to @xmath37 . third , ( 3 ) the anisotropy of the nir emission from each clump is considered , since clumps are optically thick to nir ( and optical / uv ) radiation ( see appendix for details ) . namely , the question is how extent the illuminated surface of a clump is seen by the observer ( nenkova et al . let us suppose that an observer looks at a clump with an angle @xmath38 , where @xmath39 means a face - on view of the illuminated surface . we adopt the following anisotropic coefficient for the waning effect , @xmath40 . \label{eq : waning}\ ] ] this coefficient is chosen so as to reproduce the monte carlo calculations by hnig et al . ( 2006 ) for a single clump observed from three different @xmath38 . finally , ( 4 ) if the line of sight from a region to the observer passes through the torus , we omit the nir flux from such a region . radiation energy absorbed by the clumps on the way will be re - radiated at mid - ir bands . the dotted line in figure [ fig : phi_contribution ] presents the resultant transfer function of the aligned torus for @xmath19 = 25 , calculated by summing up the nir responses from different portions of the torus at ( @xmath11 , @xmath33 ) . the integration is done from @xmath41 to @xmath42 in the @xmath11-direction and 0 to @xmath43 in the @xmath33-direction . here , the opening angle of the torus @xmath41 is assumed to be 45 , which is roughly consistent with various observational results ( e.g. , the ratio between types-1 and -2 agns and/or opening angles of light cones in the narrow line region ) , and the maximum @xmath11 of the torus @xmath42 is set to 89 ( i.e. , thin disk approximation ) . this parameter set is identical to the one adopted in paper i , and is regarded as fiducial in this study . the mean delay @xmath24 , the width @xmath25 and the skewness @xmath26 of @xmath23 are @xmath44 , @xmath45 and @xmath46 , respectively . for typical seyfert 1 galaxies and quasars ( e.g. , with uv luminosity of @xmath47ergs@xmath28 and @xmath48ergs@xmath28 ) , the expected inner radii @xmath4 are @xmath300.1 and 1pc ( equation [ eq : bar87 ] ) , while our resultant delays @xmath24 for the fiducial parameter set are about a month and a year , respectively ( see figure [ fig : tl ] ) . the long - dashed line is computed by switching off both the waning effect and the torus self - occultation , making the left horn higher than the right one similar to the time response of optically thin tori ( barvainis 1992 ) . then , the short - dashed line takes into account the waning effect , and is identical to the @xmath23 presented in paper i , with @xmath24 , @xmath25 and @xmath26 of @xmath49 , @xmath50 and @xmath51 , respectively . the torus self - occultation results in a slight increase of @xmath24 and a more skewed profile . we will see the reason for these changes in the next subsection . in order to clarify contributions from different @xmath52 , we divide the torus inner edge into three regions equally . the three solid lines in figure [ fig : phi_contribution ] show the @xmath23 from each @xmath52 range . among them , the left one is produced by the low @xmath52 region ( @xmath53 ) , showing the rapid response at @xmath54 ( due to a short optical path difference ) and a strong reduction of flux by the waning effect ( large @xmath38 ) and the torus self - occultation . on the contrary , the right one with the longer delay at @xmath55 comes from the large @xmath52 region ( @xmath56 ) , where clumps tend to direct their illuminated surface toward the observer , suffering from less waning effect . we see that the torus self - occultation affects no influence at @xmath57 with the fiducial parameter set . in other words , both the waning effect and the self - occultation selectively reduce the emission from the region with a short delay , making the right horn higher than the left one , opposing to the results for optically thin tori ( barvainis 1992 ) . the fluences , @xmath58 , from each @xmath52 region are 1.3 , 6.4 and 10.4 from low to high @xmath52 , respectively . later , we change the torus thickness ( section [ sec : qmin ] ) and the disk thickness ( section [ sec : qmax ] ) . to understand how they will affect the time response of nir emission , we draw @xmath23 from different @xmath12 ranges ( at @xmath59 ) separately in figure [ fig : theta_contribution]a . it turns out that most nir flux arises from small @xmath12 regions at @xmath60 ( i.e. , the region with a high latitude from the equatorial plane ) , which show similar time delay . because the emissivity is assumed in proportion to @xmath37 , the fluence from the largest @xmath12 range ( 7589 ) near the equatorial plane , where the torus inner edge is closest to the central bh , is small . the nir fluences from each @xmath12 range , from small to large @xmath12 , is 8.4 , 6.8 and 2.9 , respectively . therefore , little difference in @xmath24 for various @xmath42 and a shorter @xmath24 for an extremely thin ( @xmath61 ) torus are expected . consequences of a thick torus are not drawn straightforwardly , because of the torus self - occultation . in case the torus is optically thin ( cf . appendix ) , we shortly mention the contribution from the far side ( @xmath62 ) of the torus . depending on the origin of optical / uv time variability of agns ( x - ray reprocessing or change in the accretion rate etc . ; e.g. , kawaguchi et al . 1998 ; sakata et al . 2011 ) , the link between the disk fluxes to @xmath62 and to @xmath59 will vary . the nir time variations from @xmath62 is controlled by the variations of the disk flux to @xmath62 , which we can not observe . thus , if the disk illumination to the two sides are random , the nir flux variations from @xmath62 influence the measurements of the optical / uv - to - nir lag as noise . on the other hand , if the time variations of disk illumination toward the two sides are similar , the time delay of nir emission will become longer , as follows . to calculate the nir response from @xmath62 , we replace @xmath63 in equation ( [ eq : rsub ] ) by @xmath64 , and switched off the self - occultation effect . figure [ fig : theta_contribution]b presents nir responses from @xmath62 , which have long time delay due to their long light path . if no extinction affects the nir emission from @xmath62 ( as barvainis 1992 assumed ) , we will obtain the net response from @xmath65 with @xmath24 of @xmath66 . in other words , a long @xmath24 may be a signature of an extremely low volume filling factor of clumps in the torus and/or a very thin torus ( appendix ) . in this section , we present various dependencies of the nir emission from the torus whose rotation axis is aligned to the disk axis . at the end of each subsection , we shortly summarise the obtained dependency by contrasting with the result for the fiducial parameter set . figure [ fig : thetaobs_tf ] shows the transfer functions for various @xmath19 , from an exactly pole - on geometry ( @xmath67 ) to inclined viewing angles . with a large @xmath19 ( @xmath68@xmath69 ) , the line of sight grazes the upper boundary of the torus , which would corresponds to the situation in type-1.8/1.9 agns . here , @xmath41 and @xmath42 are fixed at 45 and 89 , respectively . as @xmath19 increases , we see that ( 1 ) the centroid of the response @xmath24 increases , ( 2 ) the nir fluence @xmath58 also increases , and ( 3 ) the profile becomes broad . in addition , ( 4 ) @xmath23 is quite peaky for a nearly face - on geometry . the first and second results are our new findings . although the third and forth trends are already reported for optically thin tori by barvainis ( 1992 ) , we find here that both trends are also true for optically thick tori . first , ( 1 ) @xmath24 is drawn as a function of @xmath19 in figure [ fig : thetaobs_tdelay ] . for comparison , the result computed without the self - occultation is also shown . the torus veils selectively the region with a short delay ( figure [ fig : phi_contribution ] ) , hence enlarges @xmath24 . the self - occultation shows larger influences for more inclined angles . on the right - hand side , we also draw a histogram of the observed delay in the unit of @xmath70 ( based on figure [ fig : tl ] ) . the computed @xmath24 ranges from 0.27 to 0.74@xmath70 ( 0.27 to 0.60@xmath70 without the self - occultation ) , which covers the range of the observed time delay . in contrast , since barvainis ( 1992 ) assumed an optically thin torus , @xmath24 was expected to have no ( or quite weak ) @xmath19-dependency . since such a broad range is not achieved by the changes of @xmath41 and @xmath42 as we will see later , we propose that the viewing angle is the key parameter responsible for the observed scatter about the regression line in the @xmath24@xmath6 diagram ( oknyanskij & horne 2001 ; suganuma et al . 2006 ) . conversely , the measurements of @xmath24 are potentially useful to estimate the inclination angles . in order to compare our results with the observed data more directly , the @xmath24 v.s . @xmath6 diagram is drawn in figure [ fig : tl ] ( suganuma et al . 2006 and references therein ) . following kishimoto et al . ( 2007 ) , we estimate @xmath6 by @xmath71 . the uncertainty of the @xmath24 is @xmath72 on average . our loci for various @xmath19 ( with @xmath41 and @xmath42 fixed at the fiducial values ) well cover the observed scatter . a type-1.5 seyfert galaxy ngc3227 ( triangle ) is located at an area with a small @xmath19 ( @xmath73@xmath74 ) , which would require a hysteresis effect ( koshida et al . 2009 ) or a thin torus for its relatively short @xmath24 . among the three points for ngc4151 ( green diamonds ) , the lower two data ( collected in 1969 - 1980 and 1990 - 1998 ) are consistent with a pole - on view , while the upper one in 2001 indicates an inclined angle . a change of the viewing angle on the timescale of tens years , due to e.g. , a precession of the disk , is indicated . second , ( 2 ) the top panel of figure [ fig : thetaobs_lssp ] shows that the nir fluence toward the observer , @xmath58 , is insensitive to the viewing angle at @xmath75 and increases with @xmath19 at more inclined angles ( solid line ) . comparing @xmath76 with @xmath77 , the fluence increases by a factor of 1.5 . due to the self - occultation , the nir flux decreases by 19% at @xmath78 and by 2628% at @xmath79@xmath69 . regarding a continuous illumination from the disk as a series of flash , we can draw conclusions other than the time response of nir emission . namely , the obtained nir fluence indicates the nir flux toward the observer under a given steady optical / uv illumination . we recall that we compute for various objects with a common optical flux toward the observer . therefore , the @xmath80-axis ( nir fluence ) directly indicates the nir - to - optical color . in other words , the nir - to - optical color becomes red as @xmath19 increases . we thus argue that the red ir - optical color observed for type-1.8/1.9 agns compared with type-1 objects ( alonso - herrero et al . 2003 ) arises not only from dust extinction but also from their intrinsically red color . the red color for large @xmath19 objects can originate in the reduction of optical flux toward large @xmath11 ( equation ( [ eq : flux ] ) ) and/or in the change of nir flux . here , we try to distinguish its origin . the dotted line shows the disk flux toward @xmath81 , in proportion to @xmath82^{-1}$ ] , normalised so that it coincides with the nir fluence at @xmath76 . if the nir emission is isotropic , the nir fluence in this diagram should change as the dotted line . instead , if the computed nir fluence is smaller ( or larger ) than the dotted line , it means the nir flux decreases ( or increases ) with @xmath11 . our result indicates that the nir flux decreases with @xmath11 due to the torus self - occultation , but its @xmath11-dependency is weaker than that of the disk flux , thereby presenting the red nir - optical color for inclined angles . the reduction of the nir flux toward a larger @xmath19 is consistent with the radiative transfer calculations under an isotropic optical / uv illumination ( hnig et al . 2006 ; nenkova et al . incorporation of the anisotropy of the optical / uv flux from the disk ( equation ( [ eq : flux ] ) ) enables us to predict the nir - optical / uv color . third , ( 3 ) the middle panel of figure [ fig : thetaobs_lssp ] describes the profile of @xmath23 . as @xmath19 increases , the width of @xmath23 @xmath25 also increases , meaning that @xmath23 becomes broader . both @xmath24 and @xmath25 increase with @xmath19 , and their ratio @xmath83 slightly rises toward a smaller viewing angle : a @xmath83 ratio increases twice between @xmath84@xmath69 and @xmath73 . with a small @xmath19 , the echo from various parts of the inner edge arrives at the observer at a similar delay , making the @xmath23 quite narrow . for larger @xmath19 , a variety of light pass difference arise between @xmath85 and @xmath86 , which causes the broader @xmath23 . we expect that the cross correlation function between nir and optical / uv flux variations becomes broader for type-1.51.9 agns compared with a typical type-1 object . the skewness @xmath26 is always negative ( i.e. , with a tail toward a shorter time delay ) , and the degree of asymmetry gets larger for a pole - on view . finally , ( 4 ) we comment on the nir variability amplitude ( to be precise , the ratio of the nir amplitude to the optical / uv one ) . a peaky @xmath23 will result in a large nir amplitude , whereas a less peaky @xmath23 smears out the variability of reprocessed emission , producing a less nir amplitude . in order to see this quantitatively , we draw the nir fluence - to-@xmath25 ratio and the peak value of @xmath23 as a function of @xmath19 in the bottom panel of figure [ fig : thetaobs_lssp ] . for nearly pole - on view with @xmath87 , both quantities rise . for a given optical / uv variability , such pole - on objects will show large nir variability amplitudes . at @xmath88 , the peakiness of @xmath23 is insensitive to the viewing angle . by contrast with the fiducial @xmath19 of 25 , objects with a small viewing angle will exhibit a short time delay with a narrow and peaky response . on the other hand , a more inclined viewing angle leads to a longer delay with a broader profile and to an intrinsically redder nir - to - optical color . the @xmath19-dependent delay contrasts our work with the earlier study for optically thin tori ( barvainis 1992 ) . the computed range of @xmath24 coincides with the observed one . the nir response always shows an asymmetry with a tail toward a shorter delay . in paper i , we assumed that the semi - thickness of the torus ( from the equatorial plane to the upper surface ) is 45 . however , luminous agns ( quasars ) seem to have thinner tori than seyfert galaxies ( lawrence 1991 ; ueda et al . 2003 ; la franca et al . 2005 ; arshakian 2005 ; simpson 2005 ; maiolino et al . 2007 ; hasinger 2008 ; treister et al . moreover , recent hard x - ray observations discovered type-2 agns with very thick tori ( levenson et al . 2002 ; ueda et al . 2007 ; eguchi et al . 2009 ; noguchi et al . 2010 ) . in this subsection , we show the expected characteristics of the nir emission from type-1 agns with thick and thin tori . figure [ fig : thetamin_tf ] shows transfer functions for various @xmath41 , with @xmath19 and @xmath42 fixed at 25 and 89 , respectively . when the torus thickens from @xmath89 to @xmath90 , the self - occultation selectively veils the region with a short delay at @xmath85 , hence enlarges the delay and reduces the fluence , the width and @xmath26 . in other words , the self - occultation by the region with @xmath85 and @xmath91 hides the rapid response ( left horn ) , which originates in the large @xmath11 ( @xmath92 ) region and becomes visible when the torus gets thin ( @xmath93 ) . on the other hand , the self - occultation unlikely influences the result for a thin torus . as shown in section [ sec : theta_contri ] , a rapid response is obtained when the torus is extremely thin ( @xmath61 ) , since @xmath94 is small at large @xmath11 . to see the @xmath41-dependency more quantitatively , we draw @xmath24 as a function of @xmath41 in the top panel of figure [ fig : thetamin_tlss ] . clearly , @xmath24 becomes short for thin tori . therefore , we expect that luminous quasars will show a relatively short time delay in the unit of @xmath70 . for instance , if the torus thickness is reduced for luminous objects , the loci in figure [ fig : tl ] shown by the solid lines become bent ( convex ) . the computed @xmath24 ranges from 0.16 to 0.47@xmath70 for @xmath41 in @xmath95@xmath90 . in contrast to the @xmath19-dependency , it seems difficult to explain the observed range ( shown as the histogram in figure [ fig : thetaobs_tdelay ] ) by a change of @xmath41 alone . as the torus becomes thicker , @xmath23 shows a longer delay with a narrower and more heavily skewed profile at @xmath96 due to the torus self - occultation ( middle panel ) . this trend opposes to the results for various viewing angles , where @xmath24 and @xmath25 are positively correlated each other . although a low @xmath23 is expected for a thin torus ( bottom panel ) , the small @xmath25 implies a large relative variability in the nir emission . next , figure [ fig : thetamin_tf ] also shows that the nir fluence is a strong function of @xmath41 . qualitatively , it is trivial , since various @xmath41 mean various solid angles of the torus subtended at the central bh , @xmath97 . here , @xmath98 however , as shown in numerical results by nenkova et al . ( 2008 ) , the nir flux is not exactly in proportion to @xmath97 . [ therefore , we need a caution when relating the observed nir - to - uv luminosity ratio with @xmath97 ( e.g. , mor & trakhtenbrot 2011 ; see also section [ sec : qmax ] ) . ] moreover , we let the inner radius of the torus @xmath35 vary with @xmath11 . thus , it is not obvious how the nir fluence of our torus model varies with @xmath97 . figure [ fig : thetamin_fluence ] shows that the nir fluence decreases drastically as @xmath41 increases . a thin torus locates its inner radius at a short distance from the bh ( equation ( [ eq : rsub ] ) and figure [ fig : schematic ] ) , where the size of clumps is small and the nir emissivity is low ( section [ sec : tf ] ) . for thin tori with @xmath99 , we find that the nir fluence is roughly in proportion to @xmath100 . thus , luminous quasars are expected to show weak nir emission ( blue nir - to - optical color ) , which is consistent with the observed trend of the decreasing nir - to - optical flux ratio with an increasing optical luminosity ( maiolino et al . 2007 ; treister et al . 2008 ; jiang et al . 2010 ; mor et al . 2011 ) . agns with very weak nir emission ( such as `` hot - dust - poor '' agns named by hao et al . 2010 ) may indicate that their tori are very thin . for comparison , we also draw the flux densities at 2@xmath3 m from clumpy tori seen from @xmath78 , based on a radiative transfer computation ( figure 8 of nenkova et al . 2008 , corrected following its erratum by nenkova et al . 2010 , and scaled to match with our result at @xmath101 ) . they present the results for four different torus semi - thickness @xmath102 , with a gaussian clump distribution . for the smooth boundaries of the torus in the @xmath11 direction , the solid angle is computed as ( see apendix ) : @xmath103 \right ) \sin \theta \ , d\theta.\ ] ] which is larger than @xmath104 . the steeper decline of the nir flux in our result is likely due to the @xmath11 dependency of @xmath105 , in contrast to the constant sublimation radius ( resulted from the presumed istropic illumination ) in their computations . next , we move on to the result for a thick torus . the self - occultation reduces the nir flux by 5567% for @xmath106@xmath90 , whereas it results in a reduction by @xmath107% for the fiducial @xmath41 of @xmath108 . due to the self - occultation , too thick tori with @xmath109 exhibit lesser nir flux ( i.e. , intrinsically bluer nir - optical color ) than a torus with a moderate thickness with @xmath110@xmath108 . the flux reduction of a thick torus compared with a modestly thick torus is again consistent with the result by nenkova et al . ( 2008 ) . bringing together the two behaviours , both a thick and a thin tori show the weak nir emission . namely , a modest thickness of the torus leads to the strongest nir emission . therefore , a selection bias will arise such that nir - selected agns tend to possess moderately thick tori . compared with the modest torus thickness ( @xmath89 ) , both a thick and a thin tori display the weaker nir emission ( i.e. , a bluer nir - optical color ) . a thick torus shows a slightly delayed , narrow and largely skewed nir response . on the other hand , as the torus gets thinner , the nir response becomes more rapid , narrower , closer to time - symmetric and low . when the accretion rate exceeds the eddington rate ( @xmath111 ) , an optically thick advection - dominated accretion flow ( a slim disk ) appears ( abramowicz et al . 1988 ) . since super - eddington disks are geometrically thick ( abramowicz et al . 1988 ; madau 1988 ) , they can not illuminate the directions near their equatorial plane by the disk self - occultation ( fukue 2000 ) . as discussed in paper i , some agns with presumably super - eddington accretion rates show the weak nir emission ( ark564 , tons180 , j0005 and j0303 ; rodrguez - ardila & mazzalay 2006 ; kawaguchi et al . 2004 ; jiang et al . 2010 ; see , however , hao et al . 2010 ) . small @xmath42 due to the self - occultation by a geometrically thick disk can be a reason for the weakness . moreover , the observed data do not support the concept of eddington - limited accretion ( collin & kawaguchi 2004 ) . thus , a strong anisotropy of the disk emission , such as equation ( [ eq : flux ] ) and the disk self - occultation , is required to allow gas infall to super - eddington accreting sources . in this subsection , we examine the influences of the disk thickness and the accretion rate upon the nir emission in more detail . in principle , the disk thickness is a function of the distance from the central bh . both the illumination spectrum from an agn disk and the absorption efficiency of dust have their peaks at far - uv ( e.g. , laor & draine 1993 ) . therefore , we deduce the disk thickness at the far - uv emitting region . based on the work by kawaguchi ( 2003 ; his figure 5 for a @xmath112 bh ) , the semi - thickness of the disk ( = 90@xmath113 ) at the region with the temperature of ( 45)@xmath114k are 1 , 4 , 17 and 39 , for the accretion rates of 1 , 10 , 100 and 1000@xmath115 , respectively . comparing the first and the last cases , the solid angle of the torus illuminated by the disk ( equation [ eq : omega ] ) differs by a factor of @xmath309 ( a smaller @xmath97 with a higher accretion rate ) . transfer functions for four accretion rates are presented in figure [ fig : thetamax_tf ] . as we see in section [ sec : theta_contri ] , @xmath24 is insensitive to @xmath42 ( top panel of figure [ fig : thetamax_tlss ] ) . therefore , the observed broad range of @xmath24 is not reproduced by a change of the accretion rate . larger accretion rates make the disk thicker and the shade of the disk itself larger ( i.e. , smaller @xmath97 ; equation ( [ eq : omega ] ) ) . thus , the nir fluence becomes small as the accretion rate increases , as shown in the second panel . by changing the accretion rate from 1 to 1000@xmath115 , the fluence becomes @xmath116 . this is consistent with the weakness of the x - ray emission line from neutral iron , which potentially originates in the illuminated torus , in a narrow - line quasar ( takahashi et al . 2010 ; cf . page et al . similar to @xmath24 , @xmath25 and @xmath26 are insensitive to @xmath42 ( third panel ) . on the contrary , the height of @xmath23 ( bottom panel ) is affected by the accretion rates in the sense that super - eddington sources will show a low absolute variation . other than the reduction of @xmath42 and @xmath97 , super - eddington accretion rates cause another influence upon the nir flux . sub - eddington accretion disks suffer little from the disk self - gravity , hence extend far away from the central bh , radiating across uv , optical and nir bands ( tomita et al . 2006 ; kishimoto et al . when the accretion rate becomes super - eddington , the disk self - gravity starts to govern the disk and truncate the outer part of the disk . due to the truncation , super - eddington disks do not radiate at nir ( kawaguchi et al . 2004 ) . to sum up , both small @xmath42 of the torus and the small outer radius of the disk , caused by the high accretion rate , provide less nir emission . in contrast to the result for a thin disk with a sub - eddington accretion rate of @xmath115 , a super - eddington accretion rate leads to a much weaker nir emission ( with a help of the disk self - gravity ) and to a low time response . here , we summarise the three dependencies examined above . only the variation of the viewing angles reproduces the observed range of the time delay . therefore , we propose that the viewing angle is the key parameter responsible for the observed scatter in the @xmath24@xmath6 diagram . a weak nir emission ( such as in hot - dust - poor agns in the weakest cases ) indicates either a thin torus , a thick torus or the super - eddington accretion . on the other hand , a stronger nir flux ( a redder nir - optical color ) than the fiducial one is obtained only by a large viewing angle for a modestly thick torus . so far , we have assumed that the rotation axis of the torus is aligned to that of the disk . since a specific angular momentum is likely larger in the torus than in the disk , the assumption sounds plausible . indeed , we have examples , where the rotation axes of maser disks are aligned to the jet directions ( e.g. , miyoshi et al . 1995 ; meisenheimer et al . 2007 ; mamyoda et al . 2009 ) . however , there are also a number of examples showing the misalignment between the maser disks and the jets ( e.g. , yamauchi et al . 2004 ; raban et al . random orientation between jets and galactic axes is also reported ( clarke et al . 1998 ; nagar & wilson 1999 ; schmitt et al . 2001 ) . in this section , we investigate the consequences of the misalignment between the torus and the disk axes . now , we introduce @xmath117 and @xmath118 to specify the rotation axis of the torus relative to that of the disk and the observer ( @xmath34 ) . figure [ fig : schematic ] shows the geometry with @xmath119 and @xmath120 . three @xmath117 are examined ; 30 , 60 and 90 . in the first two cases , we adopt @xmath19 of @xmath121 . since the direction to @xmath19 of @xmath122 is obscured by the torus if @xmath123 , @xmath19 of 65 is chosen in the last case . with @xmath124 and 90 , the inner edge of the torus at @xmath62 becomes visible . thus , for these two @xmath117 , we integrate @xmath11 not only from 0 to @xmath42 but also from @xmath125 to @xmath43 . here , the semi - thickness of the torus and the disk are fixed at 45 and 1 ( i.e. , @xmath126 and @xmath127 ) , respectively . figure [ fig : misalign ] shows the transfer functions ( integrating @xmath52 from 0 to @xmath128 ) , and table [ tbl : misalign ] summarises the results of a series of calculations . for reference , we also show the response of the aligned torus from @xmath129 , which is identical to the twice enlarged version of the dotted line in figure [ fig : phi_contribution ] ( obtained by @xmath130 integration ) . for @xmath117 of 90@xmath131 , the disk illumination flux toward the torus is larger than the flux toward the observer , meaning that the inner radius of the torus is large . blank fields ( and @xmath132 ) in the table mean that the direction to the observer from the central bh is obscured by the torus with such parameter sets . in other words , a large @xmath118 excludes type-1 agns , and is biased to type-2 agns . ccccc + @xmath117 [ deg ] & + & 0 & 30 & 60 & 90 + 0 & + & + & + & + 30 & 0.25 & 0.28 & 0.35 & 0.41 + & 13 & 13 & 13 & 12 + & 0.06 & 0.11 & 0.15 & 0.15 + & @xmath133&@xmath134&@xmath135&@xmath136 + 60 & 0.43 & 0.47 & & + & 16 & 17 & & + & 0.19 & 0.20 & & + & @xmath137 & @xmath138 & & + + 90 & 0.62 & 0.98 & & + & 38 & 44 & & + & 0.28 & 0.50 & & + & @xmath139 & @xmath140 & & + a variety of @xmath24 is achieved from 0.25 to 0.98@xmath70 , which is wide enough to match up with the observed range ( figure [ fig : thetaobs_tdelay ] ) . a short @xmath24 is obtained for @xmath117 of 30@xmath131 , and is associated with a small nir fluence . the geometry with @xmath141 and @xmath142 is similar to a pole - on view of aligned tori ( section [ sec : qobs ] ) , producing a very narrow nir response at a short time delay . on the other hand , a large @xmath24 is realized for @xmath117 of 90@xmath131 , accompanied with a fluence similar to the one for the fiducial parameter . as the torus becomes inclined relative to the disk , the torus hides not only the region with a short delay but also various area of the inner edge of the torus , making the nir fluence small and the time response complicated . contributions from @xmath62 , which are observable for @xmath117 of @xmath143 and @xmath144 , appear at a longer time delay . thus , they tend to increase the time delay , the nir fluence and the skewness ( @xmath145 ) . with @xmath123 , the nir emission from @xmath62 is large enough to compensate the flux reduction due to the torus self - occultation , exhibiting a comparable fluence to the reference result . a long @xmath24 ( @xmath146 ) , if observed in the future , may be a signature of a large @xmath117 ( @xmath147 ; i.e. , heavily misaligned ) . in case the torus self - occultation does not work due to an extremely small volume filling factor of clumps etc ( appendix ) , an aligned torus shows the nir response as described in section [ sec : theta_contri ] with @xmath148 . summing up all the four lines in figure [ fig : theta_contribution]b and multiply the fluence by two ( to convert the @xmath130 integration to the 02@xmath43 integration ) , we obtain the fluence of 140 . in contrast , we find that misaligned tori without the self - occultation effect show a yet longer delay with a huge nir fluence . for instance , for @xmath149 with @xmath150 and for @xmath151 with @xmath152 , @xmath153 and @xmath154 with the fluence of 250 and 960 , respectively . therefore , an extremely long ( @xmath155 ) delay with a big nir flux would mean a misaligned optically thin torus . according to recent models , the accretion disk and the bh in agns are surrounded by a clumpy torus , with its inner radius governed by the dust sublimation process . regarding the inner radius of the torus , there was a systematic deviation between the observational results and the theory . in paper i , we showed that the anisotropy of the disk emission resolves this conflict for a typical type-1 agn . we found that the anisotropy makes the torus inner region closer to the central bh and concave . furthermore , the innermost edge of the torus may connect with the outermost edge of the accretion disk continuously . in this study , we have calculated the nir flux variation of the torus in response to a uv flash for various geometries of the disk , the torus and the observer . anisotropic illumination by the disk and the effect of the torus self - occultation contrast our study with earlier works . we have found that both the waning effect of each clump and the torus self - occultation selectively reduce the emission from the region with a short delay . thus , the resultant nir time response shows a @xmath19-dependent delay and an asymmetric profile with a negative skewness , opposing to the results for optically thin tori ( barvainis 1992 ) . by contrast with the fiducial viewing angle of 25 , a small viewing angle results in a short time delay with a narrow and peaky response . on the other hand , a more inclined viewing angle leads to a longer delay with a broader profile and to a redder nir - to - optical color . we propose that the red nir - optical color of type-1.8/1.9 objects is caused by not only the dust extinction but also intrinsically red color . the computed range of @xmath24 coincides with the observed one . compared with the modest torus thickness of @xmath108 , both a thick and a thin tori display the weaker nir emission , consistent with the work by nenkova et al . in other words , a modest thickness of the torus leads to the strongest nir emission . a selection bias is thus expected such that nir - selected agns tend to possess moderately thick tori . for thin tori , we have found that the nir fluence is in proportion to @xmath100 . as the torus becomes thicker , the nir response shows a slightly longer delay with a narrower and more heavily skewed profile due to the torus self - occultation . this trend opposes to the viewing angle dependency , where the delay and the width are positively correlated each other . on the contrary , as the torus gets thinner , the nir response becomes more rapid , narrower , closer to time - symmetric and low . in contrast to a thin disk with a sub - eddington accretion rate , a super - eddington accretion rate leads to a much weaker ( @xmath156 ) nir emission due to the disk self - occultation and the disk truncation by the self - gravity , and to a low time response . among the three dependencies examined for aligned tori , only the variation of the viewing angles reproduces the observed range of the delay . therefore , we propose that the viewing angle is the key parameter responsible for the observed scatter about the regression line in the @xmath24 v.s . @xmath6 conversely , the measurements of @xmath24 are potentially useful to estimate the inclination angles . we have also investigated the consequences of the misalignment between the torus and the disk axes . a variety of @xmath24 is achieved , which is wide enough to cover the observed range . a short delay is obtained for a small misalignment and is associated with a small nir fluence , while a a long delay is obtained for a largely misaligned torus with an usual fluence . this trend contrasts with the viewing angle dependency for aligned tori , where a short delay is associated with a normal nir fluence while a long delay means a large nir fluence . for highly misaligned cases , contributions from @xmath62 increase the time delay ( up to @xmath146 ) , the nir fluence and the skewness ( @xmath145 ) . in case the torus is optically thin ( with an inefficient self - occultation ) , the time delay of the nir emission from an aligned torus becomes longer . moreover , misaligned optically thin tori show a yet longer delay ( @xmath155 ) with a huge nir flux . from an observational point of view , these numerical results are summarised as follows . if the observed time delay of the nir emission is short , it will mean either a small viewing angle , a geometrically thin torus , or a slightly misaligned torus . if the delay is long , on the other hand , it indicates either an inclined viewing angle , an aligned optically thin torus , or a largely misaligned torus . an extremely long delay ( @xmath155 ) would mean a misaligned optically thin torus . as to the nir flux , a blue nir - to - optical color ( i.e. , a weak nir emission such as in hot - dust - poor agns in the weakest cases ) indicates either a geometrically thin torus , a geometrically thick torus , a super - eddington accretion , or a slight misalignment between the torus and the disk . on the other hand , a red nir - optical color ( a large nir emission ) means a large viewing angle with a modest geometrical thickness of the torus or a largely misaligned optically thin torus . we are grateful to t. minezaki , m. umemura , m. shirahata , t. nakagawa , m. gaskell , j. fukue , h. takahashi , y. miki , m. kishimoto and s. koshida for useful discussions , and the anonymous referee for helpful comments . this work was partly supported by the grants - in - aid of the ministry of education , science , culture , and sport ( 19740105 , 21244013 ) . we briefly summarise the basic properties of clumps in the torus . the main aim here is to deduce the optical thickness of the clumpy torus . if it is optically thin at nir ( as barvainis 1992 assumed ) , we need to consider the nir emission from all the clumps including those at the far side of the torus which is illuminated by the back side ( relative to the observer ) of the accretion disk . in contrast , if it is optically thick , the torus self - occultation should be incorporated appropriately . each clump located at a distance of @xmath157 from the central bh is characterised by the radius of @xmath158 and the mass of @xmath159 . clumps that survive against its own thermal pressure and tidal shearing by the bh must be heavier than the jeans mass and/or more compact than the tidal ( hill or roche ) radius ( e.g. , vollmer et al . 2004 ) . marginally stable clumps , at the boundary of these criteria , have the following various quantities ( hnig & beckert 2007 ) , @xmath160 here , @xmath161 , @xmath162 , @xmath163 and @xmath164 are the mean density in the clump , the column density of the clump , the sound speed in the clump and the proton mass , respectively . the normalization in @xmath161 ( corresponding to @xmath165@xmath166 ) is consistent with the observed lower limit for the mean number density ( @xmath167@xmath166 ; geballe et al . 2006 ; shirahata et al . incidentally , clumps in the broad line region ( at @xmath168 cm ) have their sizes around 10@xmath169 cm ( risaliti et al . 2009 ; maiolino et al . 2010 ) , consistent with the extrapolation of equation ( [ eq : clumpsize ] ) to smaller @xmath157 . however , their density ( @xmath170 ) and @xmath171 [ @xmath172 are much less than those expected from equations ( [ eq : clumpdensity ] ) and ( [ eq : clumnh ] ) . the column density above means compton - thick ( @xmath173 ) , and corresponds to optical depths at @xmath174- and @xmath175-bands of 1400 and 160 , respectively , using the conventional extinction law ( savage & mathis 1979 ; cardelli et al . 1989 ) . in radiative transfer calculations of agn tori , clumps with optical depth at @xmath174-band of @xmath176@xmath177 are often adopted ( nenkova et al . 2008 ; hnig et al . 2008 ; deo et al . krolik & begelman ( 1988 ) estimates @xmath162 to be @xmath178 ( @xmath30one tenth of equation ( [ eq : clumnh ] ) ) . even with such a reduction of the clump opacity by @xmath179@xmath180 , each clump is still opaque to nir photons . therefore , optical thickness of the torus @xmath181 is simply related to the probability that the incoming rays hit clumps . we treat the two directions separately : one in the vertical direction ( parallel to the rotation axis ) with a suffix of @xmath182 , and another along the equatorial plane with a suffix of @xmath183 . the key parameter here is the volume filling factor of clumps in the torus @xmath184 , for which we assume 0.03 ( vollmer et al . 2004 ) . by writing the number density of clumps in the torus by @xmath185 , @xmath186 where we assume the thickness of the torus @xmath187 to be @xmath188 now , @xmath181 means the mean number of clumps along the ray , and a fraction @xmath189 of the incoming rays pass through the torus without encountering any clumps ( natta & panagia 1984 ; nenkova et al . 2002 ) . if we adopt the clump size in equation ( [ eq : clumpsize ] ) , then @xmath190 assuming that clumps are mainly heated by the direct illumination from the central accretion disk ( i.e. ignoring the irradiation from nearby clumps ) , the clump temperature will be proportional to @xmath191 , with @xmath192 being the illumination luminosity ( cf . nenkova et al . therefore , the size of torus emission is larger for larger @xmath192 at longer @xmath193 , in the form @xmath194 ( tristram et al . 2009 ) . then , we assume for @xmath163 as follows , @xmath195 with @xmath196 being the eddington luminosity . in summary , @xmath197 figure [ fig : tautorus ] shows the optical thickness of the clumpy torus as a function of @xmath157 . throughout , we fix @xmath198 , @xmath199 and @xmath184 so that these in parentheses in equation ( [ eq : tautorus ] ) equal unity . for instance , @xmath200 is about five at @xmath157 of 0.1pc . the radius where the torus is opaque @xmath201 is achieved by setting @xmath202 : @xmath203 the inner part within this radius of the torus is likely optically thick . in this study , we therefore consider basically emission from the near side of the torus , as we did in paper i. this assumption can be tested in principle via the profile of broad emission lines and its time variation ( e.g. , peterson 2001 ) . if some indications of the emission from the far side of the torus is observed , the volume filling factor seems extremely low ( @xmath204 ) or the torus is very thin ( @xmath205 ) . similarly , we deduce the optical thickness perpendicular to the rotation axis @xmath206 , which is equivalent to @xmath207 ( the average number of clumps along radial equatorial rays ) denoted by nenkova et al . they showed that @xmath207 is likely between 5 and 15 , consistent with our estimation below . adopting the inner radius of @xmath4 and the clump size in equation ( [ eq : clumpsize ] ) , @xmath208 here , we assume @xmath209 . since this optical thickness is also larger than unity , we restrict ourselves to non - obscured objects ( i.e. , the geometry where the line of sight to the central bh is not blocked by the torus ) . |
a 1984 problem of s.z . ditor asks whether there exists a lattice of cardinality , with zero , in which every principal ideal is finite and every element has at most three lower covers . we prove that the existence of such a lattice follows from either one of two axioms that are known to be independent of , namely ( 1 ) martin s axiom restricted to collections of dense subsets in posets of precaliber , ( 2 ) the existence of a gap- morass . in particular , the existence of such a lattice is consistent with , while the non - existence of such a lattice implies that is inaccessible in the constructible universe . we also prove that for each regular uncountable cardinal and each positive integer , there exists a-semilattice of cardinality and breadth in which every principal ideal has less than elements . | 0804.1781 | various representation theorems , stating that every object of ` size ' @xmath2 belongs to the range of a given functor , rely on the existence of lattices called _ @xmath11-ladders_. by definition , a @xmath11-ladder is a lattice with zero , in which every principal ideal is finite , and in which every element has at most two lower covers . every @xmath11-ladder has cardinality at most @xmath2 , and the existence of @xmath11-ladders of cardinality exactly @xmath2 was proved in ditor @xcite ( cf . proposition [ p:2ladd ] ) . these @xmath11-ladders have been used in various contexts such as abstract measure theory ( dobbertin @xcite ) , lattice theory ( grtzer , lakser , and wehrung @xcite ) , ring theory ( wehrung @xcite ) , or general algebra ( rika , tma , and wehrung @xcite ) . a sample result , established in @xcite , states that _ every distributive algebraic lattice with at most @xmath2 compact elements is isomorphic to the lattice of all normal subgroups of some locally finite group_. ( here and in many related results , the @xmath2 bound turns out to be optimal . ) the basic idea of `` ladder proofs '' is always the same : we are given categories @xmath12 and @xmath13 together with a functor @xmath14 and a ` large ' object @xmath15 of @xmath13 ( of ` size ' @xmath2 ) , that we wish to represent as @xmath16 , for some object @xmath17 in the domain of @xmath18 . we represent @xmath15 as a direct limit @xmath19 of ( say ) ` finite ' objects @xmath20 , where @xmath21 is an upward directed poset of cardinality @xmath2 ( often the lattice of all finite subsets of @xmath15 in case we are dealing with a concrete category ) . then , using the existence of a @xmath11-ladder of cardinality @xmath2 , we can replace the original poset @xmath21 by that @xmath11-ladder . under `` amalgamation - type '' conditions , this makes it possible to represent each @xmath20 as some @xmath22 , with transition morphisms between the @xmath23s being constructed in such a way that @xmath24 can be defined and @xmath25 . we define _ @xmath26-ladders _ the same way as @xmath11-ladders , except that `` two lower covers '' is replaced by `` three lower covers '' . the problem of existence of @xmath26-ladders of cardinality @xmath0 was posed in ditor @xcite . such @xmath26-ladders would presumably be used in trying to represent objects of size @xmath0 . nevertheless i must reluctantly admit that no potential use of the existence of @xmath26-ladders of cardinality @xmath0 had been found so far , due to the failure of a certain three - dimensional amalgamation property , of set - theoretical nature , stated in section 10 in wehrung @xcite , thus making ditor s problem quite ` romantic ' ( and thus , somewhat paradoxically therefore arguably , attractive ) . however , this situation has been evolving recently . for classes @xmath12 and @xmath13 of algebras , the _ critical point _ @xmath27 is defined in gillibert @xcite as the least possible cardinality of a semilattice in the compact congruence class of @xmath12 but not of @xmath13 if it exists ( and , say , @xmath28 otherwise ) . in case both @xmath12 and @xmath13 are finitely generated lattice varieties , it is proved in gillibert @xcite that @xmath27 is either finite , or @xmath29 for some natural number @xmath6 , or @xmath28 . in the second case only examples with @xmath30 have been found so far ( ploica @xcite , gillibert @xcite ) . investigating the possibility of @xmath31 ( i.e. , @xmath32 ) would quite likely require @xmath26-ladders of cardinality @xmath0 . the present paper is intended as an encouragement in that direction . we partially solve one of ditor s problems by giving a ( rather easy ) proof that for each regular uncountable cardinal @xmath5 and each positive integer @xmath6 , there exists a @xmath7-semilattice of cardinality @xmath9 and breadth @xmath10 in which every principal ideal has less than @xmath5 elements ( cf . theorem [ t : largebdedbr ] ) . furthermore , although we are still not able to settle whether the existence of a @xmath26-ladder of cardinality @xmath0 is provable in the usual axiom system @xmath1 of set theory with the axiom of choice , we prove that it follows from either one of two quite distinct , and in some sense ` orthogonal ' , set - theoretical axioms , namely a weak form of martin s axiom plus @xmath33 denoted by @xmath34 ( cf . theorem [ t : map23ladd ] ) and the existence of a gap-@xmath3 morass ( cf . theorem [ t : mor2ladd ] ) . in particular , the _ existence _ of a @xmath26-ladder of cardinality @xmath0 is consistent with @xmath1 , while the _ non - existence _ of a @xmath26-ladder of cardinality @xmath0 implies that @xmath4 is inaccessible in the constructible universe . our proofs are organized in such a way that no prerequisites in lattice theory and set theory other than the basic ones are necessary to read them . hence we hope to achieve intelligibility for both lattice - theoretical and set - theoretical communities . we shall use standard set - theoretical notation and terminology . throughout the paper , ` countable ' will mean ` at most countable ' . a cardinal is an initial ordinal , and we denote by @xmath35 the successor cardinal of a cardinal @xmath5 . more generally , we denote by @xmath9 the @xmath36 successor cardinal of @xmath5 , for each natural number @xmath6 . we denote by @xmath37 ( resp . , @xmath38 ) the domain ( resp . , range ) of a function @xmath39 , and we put @xmath40:={\{{f(x)}\mid{x\in x}\}}$ ] for each @xmath41 . we denote by @xmath42 the powerset of a set @xmath17 , and by @xmath43 the partial operation of disjoint union . a function @xmath39 is _ finite - to - one _ if the inverse image of any singleton under @xmath39 is finite . for a set @xmath44 and a cardinal @xmath45 , we set @xmath46^{\lambda}&={\{{x\in{\mathfrak{p}}(\omega)}\mid{|x|=\lambda}\}}\,,\\ [ \omega]^{{\leqslant}\lambda}&={\{{x\in{\mathfrak{p}}(\omega)}\mid{|x|\leq\lambda}\}}\,,\\ [ \omega]^{<\lambda}&={\{{x\in{\mathfrak{p}}(\omega)}\mid{|x|<\lambda}\}}\ , . \end{aligned}\ ] ] two elements @xmath47 and @xmath48 in a poset @xmath49 are _ comparable _ if either @xmath50 or @xmath51 . we say that @xmath47 is a _ lower cover _ of @xmath48 , if @xmath52 and there is no element @xmath53 such that @xmath54 ; in addition , if @xmath47 is the least element of @xmath49 ( denoted by @xmath55 if it exists ) , we say that @xmath48 is an _ atom _ of @xmath49 . we say that @xmath49 is _ atomistic _ if every element of @xmath49 is a join of atoms of @xmath49 . for a subset @xmath17 and an element @xmath56 in @xmath49 , we set @xmath57 we say that @xmath17 is a _ lower subset _ of @xmath49 , if @xmath58 for each @xmath59 . ( in forcing terminology , this means that @xmath17 is _ open_. ) an _ ideal _ of @xmath49 is a nonempty , upward directed , lower subset of @xmath49 ; it is a _ principal ideal _ if it is equal to @xmath60 for some @xmath61 . a _ filter _ of @xmath49 is an ideal of the dual poset of @xmath49 . we say that @xmath49 is _ lower finite _ if @xmath60 is finite for each @xmath61 . an _ order - embedding _ from a poset @xmath49 into a poset @xmath62 is a map @xmath63 such that @xmath64 iff @xmath50 , for all @xmath65 . an order - embedding @xmath39 is a _ lower embedding _ if the range of @xmath39 is a lower subset of @xmath62 . observe that a lower embedding preserves all meets of nonempty subsets in @xmath49 , and all joins of nonempty finite subsets of @xmath49 in case @xmath62 is a join - semilattice . for an element @xmath47 and a subset @xmath66 in @xmath49 , we denote by @xmath67 the least element of @xmath66 above @xmath47 if it exists . anin s classical _ @xmath68-lemma _ ( cf . jech ( * ? ? ? * theorem 9.18 ) ) is the following . @xmath68-lemma let @xmath69 be an uncountable collection of finite sets . then there are an uncountable subset @xmath70 of @xmath69 and a finite set @xmath71 such that @xmath72 for all distinct @xmath73 . we recall some basic terminology in the theory of forcing , which we shall use systematically in section [ s : ppk ] . a subset @xmath17 in a poset @xmath49 is _ dense _ if it meets every principal ideal of @xmath49 . we say that @xmath17 is _ centred _ if every finite subset of @xmath17 has a lower bound in @xmath49 . we say that @xmath49 has _ precaliber @xmath2 _ if every uncountable subset of @xmath49 has an uncountable centred subset ; in particular , this implies the countable chain condition . for a collection @xmath74 of subsets of @xmath49 , a subset @xmath75 of @xmath49 is _ @xmath74-generic _ if @xmath76 for each _ dense _ @xmath77 . the following classical lemma ( cf . jech ( * ? ? ? * lemma 14.4 ) ) is , formally , a poset analogue of baire s category theorem . [ l : ctblegen ] let @xmath74 be a countable collection of subsets of a poset @xmath49 . then each element of @xmath49 is contained in a @xmath74-generic filter on @xmath49 . for @xmath78 there may be no @xmath74-generic filters of @xmath49 ( cf . jech ( * ? ? ? * exercise 16.11 ) ) , nevertheless for restricted classes of partial orderings @xmath49 we obtain set - theoretical axioms that are independent of @xmath1 . we shall need the following proper weakening of the @xmath2 instance of martin s axiom @xmath79 usually denoted by either @xmath80 or @xmath81 ( cf . jech ( * ? ? ? * section 16 ) and weiss ( * ? ? ? * section 3 ) ) ; the first parameter @xmath2 refers to the cardinality of @xmath74 . @xmath34 for every poset @xmath49 of precaliber @xmath2 and every collection @xmath74 of subsets of @xmath49 , if @xmath82 , then there exists a @xmath74-generic filter on @xmath49 . for a positive integer @xmath6 , trying to build certain structures of size @xmath83 as direct limits of countable structures may impose very demanding constraints on the direct systems used for the construction . the pattern of the repetitions of the countable building blocks and their transition morphisms in the direct system is then coded by a complex combinatorial object called a _ gap-@xmath6 morass_. gap-@xmath6 morasses were introduced by ronald jensen in the seventies , enabling him to solve positively the _ finite gap cardinal transfer conjecture _ in the constructible universe @xmath84 . the existence of morasses is independent of the usual axiom system of set theory @xmath1 . for example , there are gap-@xmath3 morasses in @xmath84 ( cf . devlin ( * ? ? ? * section viii.2 ) ) , and even in the universe @xmath85 $ ] of sets constructible with oracle @xmath86 , for any @xmath87 ; hence if @xmath4 is not inaccessible in @xmath84 , then there is a gap-@xmath3 morass in the ambient set - theoretical universe @xmath88 ( cf . devlin ( * ? ? ? * exercise viii.6 ) ) . conversely , the existence of a gap-@xmath3 morass implies the existence of a kurepa tree while , in the generic extension obtained by levy collapsing an inaccessible cardinal on @xmath4 while preserving @xmath89 , there is no kurepa tree ( cf . silver @xcite ) . in particular , the non - existence of a gap-@xmath3 morass is equiconsistent , relatively to @xmath1 , to the existence of an inaccessible cardinal . however , even for @xmath90 the combinatorial theorems involving morasses are hard to come by , due to the extreme complexity of the definition of gap-@xmath6 morasses . fortunately , the definition of a gap-@xmath3 morass has been greatly simplified by dan velleman @xcite , where it is proved that the existence of a gap-@xmath3 morass is equivalent to the existence of a ` simplified @xmath91-morass ' . we denote the ( noncommutative ) ordinal addition by @xmath92 . for ordinals @xmath93 , we denote by @xmath94 the unique ordinal @xmath95 such that @xmath96 . furthermore , we denote by @xmath97 the order - embedding from ordinals to ordinals defined by @xmath98 we shall use the following definition , obtained by slightly amending the one in devlin ( * ? ? ? * section viii.4 ) by requiring @xmath99 instead of @xmath100 , but all @xmath101s nonzero ( which could not hold for @xmath100 ) . it is easily obtained from a simplified morass as defined in devlin ( * ? ? ? * section viii.4 ) by adding a new zero element to each @xmath102 ( that is , by replacing @xmath102 by @xmath103 ) and replacing any @xmath104 by the unique zero - preserving map sending @xmath105 to @xmath106 , for each @xmath107 . [ d : gap1mor ] let @xmath5 be an infinite cardinal . a _ simplified @xmath108-morass _ is a structure @xmath109 satisfying the following conditions : * * * @xmath99 , @xmath110 for each @xmath111 , and @xmath112 . * * @xmath113 is a set of order - embeddings from @xmath102 into @xmath114 , for all @xmath115 . * @xmath116 , for all @xmath117 . * if @xmath118 , then @xmath119 . * for each @xmath111 , there exists a _ nonzero _ ordinal @xmath120 such that @xmath121 and @xmath122 , where @xmath123 denotes the restriction of @xmath124 from @xmath102 into @xmath125 . * for every limit ordinal @xmath126 , all @xmath127 and @xmath128 , for @xmath129 , there exists @xmath130 with @xmath131 together with @xmath132 , for @xmath129 , and @xmath133 such that @xmath134 for each @xmath129 . * the equality @xmath135}\mid{\xi<\alpha\text { and } f\in{\mathcal{f}}_{\xi,\alpha}}\right)}$ ] holds for each @xmath136 . it is proved in velleman @xcite that for @xmath5 regular uncountable , there exists a @xmath108-morass iff there exists a simplified @xmath108-morass . for the countable case , the existence of a @xmath137-morass is provable in @xmath1 , see velleman @xcite . simplified morasses as above satisfy the following simple but very useful lemma , which is the basis of the construction of the kurepa tree obtained from a @xmath108-morass ( cf . velleman ( * ? ? ? * lemma 3.2 ) ) . [ l : mor2tree ] let @xmath138 , let @xmath139 , and let @xmath140 . if @xmath141 , then @xmath142 and @xmath143 . the classical definition of breadth , see ditor ( * ? ? ? * section 4 ) , runs as follows . let @xmath6 be a positive integer . a join - semilattice @xmath15 has _ breadth at most @xmath6 _ if for every nonempty finite subset @xmath17 of @xmath8 , there exists a nonempty @xmath144 with at most @xmath6 elements such that @xmath145 . this is a particular case of the following definition of breadth , valid for every poset , and , in addition , self - dual : we say that a poset @xmath49 has breadth at most @xmath6 , if for all @xmath146 , @xmath147 in @xmath8 , if @xmath148 for all @xmath149 in @xmath150 , then there exists @xmath151 such that @xmath152 . [ s : ladder ] let @xmath153 be a positive integer . a _ @xmath153-ladder _ is a lower finite lattice in which every element has at most @xmath153 lower covers . every @xmath153-ladder has breadth at most @xmath153 . the diamond @xmath154 has breadth @xmath11 but it is not a @xmath11-ladder . every finite chain is a @xmath3-ladder . the chain @xmath155 of all non - negative integers is also a @xmath3-ladder . note that @xmath153-ladders are called _ @xmath153-frames _ in dobbertin @xcite ; the latter terminology being already used for a completely different lattice - theoretical concept ( von neumann frames ) , we will not use it . the following is proved in ditor @xcite . [ p : uppbdbr ] let @xmath153 be a positive integer . then every lower finite lattice of breadth at most @xmath153 has at most @xmath156 elements . proposition [ p : uppbdbr ] is especially easy to prove by using kuratowski s free set theorem , see kuratowski @xcite . the converse is obviously true for @xmath157that is , there exists a @xmath3-ladder of cardinality @xmath158 ( namely , the chain @xmath159 of all natural numbers ) ; also for @xmath160 , by the following result of ditor @xcite , also proved by dobbertin @xcite . we include a proof for convenience . [ p:2ladd ] there exists an atomistic @xmath11-ladder of cardinality @xmath2 . we construct inductively a @xmath89-sequence @xmath161 of countable atomistic @xmath11-ladders , such that @xmath162 implies that @xmath163 is a proper ideal of @xmath164 . once this is done , the @xmath11-ladder @xmath165 will clearly solve our problem . we take @xmath166 . if @xmath167 is a limit ordinal and all @xmath163 , for @xmath130 , are constructed , take @xmath168 . suppose that @xmath163 is constructed . if @xmath163 is finite , pick outside objects @xmath56 , @xmath3 and put @xmath169 , with the additional relations @xmath170 and @xmath171 for each @xmath172 . if @xmath163 is infinite , then , as it is a countable lattice , it has a strictly increasing cofinal sequence @xmath173 . consider a one - to - one sequence @xmath174 of objects outside @xmath163 and put @xmath175 with the additional relations @xmath176 for @xmath177 and @xmath178 for @xmath179 . the @xmath89-sequence @xmath180 thus constructed is as required . the most natural attempt at proving the existence of a @xmath26-ladder of cardinality @xmath0 , by imitating the proof of proposition [ p:2ladd ] , would require that every @xmath26-ladder of cardinality @xmath2 has a cofinal meet - subsemilattice which is also a @xmath11-ladder ( cf . the proof of theorem [ t : map23ladd ] ) . we do not know whether this statement is a theorem of @xmath1 , although , by theorem [ t : exgenaleph1 ] , it is consistent with @xmath1 . the following example shows that the most straightforward attempt at proving that statement , by expressing structures of cardinality @xmath2 as directed unions of countable structures , fails . [ ex : llfnof ] there exists a countable @xmath26-ladder @xmath181 with an ideal @xmath182 and a cofinal meet - subsemilattice @xmath66 of @xmath182 which is also a @xmath11-ladder , although there is no cofinal meet - subsemilattice of breadth at most two of @xmath181 containing @xmath66 . we denote by @xmath181 the lattice represented in figure [ fig : no2ladd ] , and we put @xmath183 ] suppose that there is a cofinal meet - subsemilattice @xmath184 of @xmath181 , with breadth at most two , containing @xmath66 . as @xmath184 is cofinal in @xmath181 , there exists @xmath185 such that @xmath186 . as @xmath187 belongs to @xmath66 , it also belongs to @xmath184 , thus @xmath188 also belongs to @xmath184 , so @xmath189 is contained in @xmath184 , and so , as @xmath184 has breadth at most two , @xmath190 belongs to @xmath191 , a contradiction . a particular case of the problem above is stated on top of page 58 in ditor @xcite : let @xmath66 be a @xmath11-ladder of cardinality @xmath2 . does the @xmath26-ladder @xmath192 , endowed with the product order , have a cofinal meet - subsemilattice which is also a @xmath11-ladder ? the answer to that question is affirmative , due to the following easy result . [ p : ktimesom ] let @xmath182 be an infinite , lower finite lattice . then @xmath193 has a cofinal meet - subsemilattice isomorphic to @xmath182 . as @xmath182 is an infinite , lower finite lattice , it has a strictly increasing sequence @xmath194 . as @xmath182 is lower finite , for each @xmath195 , there exists a largest natural number @xmath6 such that @xmath196 ; denote this integer by @xmath197 . we define @xmath181 as the graph of @xmath39 , that is , @xmath198 . as @xmath39 is a meet - homomorphism , @xmath181 is a meet - subsemilattice of @xmath193 . furthermore , @xmath199 defines a lattice isomorphism of @xmath182 onto @xmath181 , and @xmath181 is obviously cofinal in @xmath193 . in this section we shall see that weak analogues of @xmath153-ladders , obtained by replacing @xmath158 by a regular uncountable cardinal and the condition that every element has at most @xmath153 lower covers by the breadth being below @xmath153 , can be easily constructed ( cf . theorem [ t : largebdedbr ] ) . the following lemma [ l : breadthgen ] is a slightly more general version of ( * ? ? ? * proposition 4.3 ) . it says that breadth can be verified on the generators of a join - semilattice , the difference with ( * ? ? ? * proposition 4.3 ) lying in the definition of a generating subset . a subset @xmath75 in a join - semilattice @xmath8 _ generates _ @xmath8 , if every element of @xmath8 is a ( not necessarily finite ) join of elements of @xmath75 . equivalently , for all @xmath200 such that @xmath201 , there exists @xmath202 such that @xmath203 and @xmath204 . [ l : breadthgen ] let @xmath8 be a join - semilattice , let @xmath75 be a generating subset of @xmath8 , and let @xmath6 be a positive integer . then @xmath8 has breadth at most @xmath6 iff for every subset @xmath205^{n+1}$ ] , there exists @xmath206 such that @xmath207 . we prove the nontrivial direction . let @xmath208^{n+1}$ ] , and suppose that @xmath209 for each @xmath210 . as @xmath75 generates @xmath8 , there exists @xmath211 such that @xmath212 and @xmath213 , for all @xmath210 . hence @xmath214 , for each @xmath210 , a contradiction as all elements @xmath215 belong to @xmath75 and by assumption . an _ algebraic closure operator _ on a poset @xmath49 is a map @xmath216 such that @xmath217 ( we say that @xmath39 is _ idempotent _ ) , @xmath218 for each @xmath219 , @xmath50 implies that @xmath64 for all @xmath65 , and for each nonempty upward directed subset @xmath17 of @xmath49 admitting a join , the join of @xmath40 $ ] exists and is equal to @xmath220 . ( we shall refer to the latter property as the _ join - continuity _ of @xmath39 . ) [ l : largealcl ] for each regular uncountable cardinal @xmath5 and each positive integer @xmath6 , there exists an algebraic closure operator @xmath39 on @xmath221^{<\kappa}$ ] such that @xmath222^{n+2})(\exists\xi\in u)\bigl(\xi\in f(u\setminus{\{{\xi}\}})\bigr)\ , , \label{eq : nofree}\\ f(x)=x\quad\text{for each } x\in[\kappa^{+n}]^{{\leqslant}n}\,.\label{eq : f(x)=x } \end{gathered}\ ] ] it follows from kuratowski @xcite ( see also theorem 45.7 in erds _ et al . _ @xcite ) that there exists a map @xmath223^{n+1}\to[\kappa^{+n}]^{<\kappa}$ ] such that @xmath224^{n+2})(\exists\xi\in u ) \bigl(\xi\in f_0(u\setminus{\{{\xi}\}})\bigr)\,.\ ] ] we set @xmath225^{n+1}}\right)}\,,\quad \text{for each } x\in[\kappa^{+n}]^{<\kappa}\,.\ ] ] as @xmath5 is regular , @xmath226 is a self - map of @xmath221^{<\kappa}$ ] . it obviously satisfies both and , together with all properties defining an algebraic closure operator except idempotence . now we set @xmath227^{<\kappa}\,.\ ] ] as @xmath5 is regular uncountable , @xmath39 is a self - map of @xmath221^{<\kappa}$ ] . it obviously satisfies both and , together with all properties defining an algebraic closure operator except idempotence . furthermore , for every @xmath228^{<\kappa}$ ] , @xmath229 so @xmath217 . therefore , we obtain a positive answer to the specialization of question a , page 57 in ditor @xcite to regular uncountable cardinals . [ t : largebdedbr ] for each regular uncountable cardinal @xmath5 and each positive integer @xmath6 , there exists an atomistic @xmath7-semilattice @xmath8 of breadth @xmath10 and cardinality @xmath9 such that @xmath230 for each @xmath231 . let @xmath39 be an algebraic closure operator as in lemma [ l : largealcl ] . we endow @xmath232^{<\omega}}\}}$ ] with containment . obviously , @xmath8 is a @xmath7-semilattice and @xmath230 for each @xmath231 . as @xmath233 , it follows from ( * ? ? ? * theorem 5.2 ) that @xmath8 has breadth at least @xmath10 . as @xmath6 is nonzero and by , every singleton @xmath234 , for @xmath235 , belongs to @xmath8 , so @xmath8 is atomistic , and so @xmath236 generates @xmath8 in the sense required by the statement of lemma [ l : breadthgen ] . therefore , it follows from together with lemma [ l : breadthgen ] that @xmath8 has breadth at most @xmath10 . for @xmath237 , it it not known whether theorem [ t : largebdedbr ] extends to singular cardinals , for instance @xmath238 ( cf . problem 2 in ditor @xcite ) . for @xmath239 , it is not known whether theorem [ t : largebdedbr ] extends to @xmath240 ( cf . problem 1 in ditor @xcite ) , although we shall prove in two different ways that a positive answer is consistent with @xmath1 ( cf . theorems [ t : map23ladd ] and [ t : mor2ladd ] ) . [ d : nmdlatt ] a _ normed lattice _ is a pair @xmath241 , where @xmath182 is a lattice and @xmath242 is a join - homomorphism from @xmath182 to the ordinals ( the ` norm ' ) . in such a case we put @xmath243 observe that each @xmath244 is either empty or an ideal of @xmath182 ( we will say that it is an _ extended ideal _ of @xmath182 ) . we call the subsets @xmath245 the _ levels _ of @xmath182 . we say that @xmath241 is _ transitive _ if its _ range _ , that is , the range of @xmath242 , is an ordinal . of course , in such a case , @xmath246 . observe that conversely , every increasing well - ordered sequence @xmath247 of extended ideals of @xmath182 with union @xmath182 defines a norm @xmath242 , _ via _ the rule @xmath248 in case @xmath182 is _ lower finite _ , the set @xmath249 is a finite ideal of @xmath182 , for every @xmath195 and every ordinal @xmath95 , hence it has a largest element , that we shall denote by @xmath250 . the assignment @xmath251 is _ isotone _ , that is , @xmath50 and @xmath252 implies that @xmath253 . we shall put @xmath254 observe that @xmath255 is a _ chain_. furthermore , as @xmath255 is a subset of @xmath256 , it is _ finite_. the binary relation @xmath257 defined by the rule @xmath258 is a partial ordering on @xmath182 in which all principal ideals are finite chains . we shall always denote by @xmath242 the norm function on a normed lattice . we shall repeatedly use the following easy observation . [ l : proj2meet ] the following implications hold , for any elements @xmath47 and @xmath48 in a lower finite normed lattice @xmath182 : 1 . @xmath50 implies that @xmath259 . 2 . @xmath260 iff @xmath261 . ( i ) . from @xmath50 it follows that @xmath262 , hence @xmath263 . we need to prove only the direct implication . it is trivial that @xmath264 . for the converse , observe that @xmath265 , thus , as @xmath266 , we obtain that @xmath267 , and so the equality holds . [ d : norm2ladd ] a _ preskeleton _ of a lower finite normed lattice @xmath182 is a subset @xmath66 of @xmath182 satisfying the following conditions : 1 . @xmath268 is a ( possibly empty ) chain , for every ordinal @xmath95 ; 2 . @xmath66 is _ projectable _ , that is , @xmath250 belongs to @xmath66 , for any @xmath269 and any ordinal @xmath95 . if , in addition , @xmath268 is cofinal in @xmath245 for each @xmath95 , we say that @xmath66 is a _ skeleton of @xmath182_. [ l : norm2ladd ] every preskeleton @xmath66 of a lower finite normed lattice @xmath182 satisfies the following properties : 1 . @xmath66 is a meet - subsemilattice of @xmath182 ; 2 . every element of @xmath66 has at most two lower covers in @xmath66 . in particular , if @xmath66 is upward directed , then it is a @xmath11-ladder . let @xmath270 and put @xmath271 . hence @xmath272 . from @xmath273 it follows that @xmath274 , thus @xmath275 . similarly , @xmath276 . as both @xmath277 and @xmath278 belong to @xmath66 , it follows that they are comparable , and therefore @xmath279 . let @xmath269 . we denote by @xmath280 the largest element of @xmath281 smaller than @xmath47 if it exists ( i.e. , as @xmath281 is a chain , if @xmath281 has an element smaller than @xmath47 ) , and by @xmath282 the largest element of @xmath283 if it exists ( i.e. , if there exists an element in @xmath255 smaller than @xmath47 ) . it suffices to prove that every element @xmath284 smaller than @xmath47 lies either below @xmath280 or below @xmath282 . from @xmath51 it follows that @xmath285 and @xmath286 . if @xmath287 then , as @xmath288 , @xmath280 exists and @xmath289 . if @xmath290 , then , as @xmath51 and by lemma [ l : proj2meet](i ) , @xmath291 , thus @xmath282 exists and @xmath292 . now let @xmath66 be upward directed . together with ( i ) and the lower finiteness of @xmath66 , this implies that @xmath66 is a lattice . therefore , by ( ii ) , @xmath66 is a @xmath11-ladder . in particular , every skeleton of @xmath182 is both a @xmath11-ladder and a meet - subsemilattice of @xmath182 . easy examples show that a skeleton of @xmath182 may not be a _ join_-subsemilattice of @xmath182 . as the present section involves martin s axiom in an essential way , we shall use forcing terminology ( _ open _ , _ dense _ ) rather than poset terminology ( _ lower subset _ , _ coinitial _ ) throughout ( cf . section [ s : basic ] ) . [ d : ppk ] let @xmath182 be a lower finite normed lattice ( cf . definition [ d : nmdlatt ] ) . we denote by @xmath293 the set of all _ finite _ preskeletons of @xmath182 ( cf . definition [ d : norm2ladd ] ) , ordered under _ reverse _ containment . furthermore , we put @xmath294 and @xmath295 . it is clear that @xmath296 is an open subset of @xmath293 . [ l : ppkadense ] the subset @xmath296 is dense in @xmath293 , for each @xmath297 . let @xmath298 , we must find @xmath299 containing @xmath300 . put@xmath301 . as the finite chain @xmath302 is projectable , the subset @xmath303 is projectable . furthermore , for each @xmath304 , from @xmath305 it follows that @xmath306 , and so @xmath66 is a preskeleton of @xmath66 . put @xmath307 . from @xmath308 it follows that @xmath309 and @xmath310 ( cf . lemma [ l : proj2meet](i ) ) . therefore , the element @xmath311 witnesses that @xmath66 belongs to @xmath296 . [ d : locctble ] a normed lower finite lattice is _ locally countable _ if all its levels are countable . [ l : locctble ] every locally countable normed lower finite lattice @xmath182 has cardinality at most @xmath2 and range of order - type at most @xmath89 . without loss of generality , @xmath312 is an ordinal . for each @xmath313 and each @xmath314 , the map ( @xmath315 , @xmath316 ) is finite - to - one , thus , as @xmath317 is countable , @xmath318 is countable as well . as @xmath319 , it follows that @xmath320 . therefore , @xmath321 has cardinality at most @xmath2 . in the following lemma , we shall use the @xmath68-lemma together with the notions `` centred '' and `` precaliber '' ( cf . section [ s : basic ] ) . [ l : precalppk ] suppose that @xmath182 is locally countable . then the poset @xmath293 has precaliber @xmath2 . let @xmath322 be an @xmath89-sequence of elements of @xmath293 , we must find an uncountable @xmath323 such that @xmath324 is centred . put @xmath325 ( an element of @xmath182 ) , @xmath326 ( a finite subset of @xmath182 ) , and @xmath327 $ ] ( a finite subset of @xmath328 ) , for each @xmath329 . two successive applications of the @xmath68-lemma yield an uncountable subset @xmath330 of @xmath89 and finite sets @xmath17 , @xmath15 such that @xmath331 as @xmath15 is a finite subset of @xmath328 , it is contained in @xmath332 for some @xmath333 . for each @xmath334 , @xmath335 is a finite subset of @xmath336 , hence , as @xmath336 is countable , there are a finite subset @xmath66 of @xmath336 and an uncountable subset @xmath337 of @xmath330 such that @xmath338 we claim that @xmath339 is centred . it is sufficient to prove that for each nonempty finite subset @xmath86 of @xmath337 , the union @xmath340 belongs to @xmath293 . as @xmath341 is obviously projectable , it suffices to prove that any elements @xmath342 , for @xmath343 , are comparable . let @xmath344 such that @xmath345 . if @xmath346 then , as @xmath163 is a preskeleton of @xmath182 , we are done . suppose that @xmath347 . as @xmath348 belongs to @xmath349 ( cf . ) , we get @xmath350 , thus , by , @xmath351 , so @xmath352 , and so , as @xmath353 is a chain , @xmath47 and @xmath48 are comparable . the following lemma describes @xmath354-generic filters over @xmath293 . [ l : genonppk ] let @xmath182 be a lower finite normed lattice and let @xmath355 be a @xmath354-generic filter of @xmath293 . then @xmath356 is a skeleton of @xmath182 . in particular , @xmath75 is a cofinal meet - semilattice in @xmath182 and it is a @xmath11-ladder . as @xmath355 is an upward directed ( for containment ) set of preskeletons of @xmath182 , @xmath75 is also a preskeleton of @xmath182 . as @xmath355 meets @xmath296 for each @xmath297 , @xmath75 is cofinal on each level of @xmath182 . in particular , @xmath75 is cofinal in @xmath182 , thus it is upward directed , and thus ( cf . lemma [ l : norm2ladd ] ) it is a @xmath11-ladder . an immediate application of lemmas [ l : ctblegen ] , [ l : ppkadense ] , and [ l : precalppk ] yields the following theorem . however , as there is an easy direct proof , we provide it as well . [ t : large2ladd ] every countable lower finite normed lattice @xmath182 has a skeleton . as @xmath182 is countable , it has a cofinal chain @xmath357 . it is obvious that the subset @xmath358 is projectable . for any @xmath343 , any two elements of @xmath268 have the form @xmath359 and @xmath360 , for some @xmath361 . as @xmath362 and @xmath363 are comparable , so are @xmath359 and @xmath360 . this proves that @xmath66 is a preskeleton of @xmath182 . as @xmath357 is cofinal in @xmath182 , every @xmath195 lies below some @xmath364 . hence @xmath365 and , by lemma [ l : proj2meet](i ) , @xmath366 . therefore , @xmath281 is cofinal in @xmath367 . for lattices of cardinality @xmath2 we get the following result : [ t : exgenaleph1 ] assume that the axiom @xmath34 holds . let @xmath182 be a lower finite lattice of cardinality at most @xmath2 . then @xmath182 has a skeleton . write @xmath368 , and denote by @xmath369 the ideal of @xmath182 generated by @xmath370 , for each @xmath329 . finally , for each @xmath195 , denote by @xmath371 the least ordinal @xmath372 such that @xmath373 . then @xmath241 is a locally countable , lower finite normed lattice . now apply lemmas [ l : ppkadense ] and [ l : precalppk ] . we obtain the following result . [ t : map23ladd ] suppose that @xmath34 holds . then there exists an atomistic @xmath26-ladder of cardinality @xmath0 . we argue as in the proof of proposition [ p:2ladd ] . we construct inductively a @xmath4-sequence @xmath374 of atomistic @xmath26-ladders of cardinality at most @xmath2 , such that @xmath162 implies that @xmath317 is a proper ideal of @xmath375 . once this is done , the @xmath26-ladder @xmath376 will clearly solve our problem . we take @xmath377 . if @xmath378 is a limit ordinal and all @xmath317 , for @xmath130 , are constructed , take @xmath379 . let @xmath380 and suppose that @xmath317 is constructed , of cardinality at most @xmath2 . by theorem [ t : exgenaleph1 ] and lemma [ l : norm2ladd ] , @xmath317 has a cofinal meet - subsemilattice @xmath66 which is also a @xmath11-ladder . we consider an isomorphic copy @xmath381 of @xmath66 disjoint ( set - theoretically ) from @xmath317 , _ via _ an isomorphism @xmath382 , and we consider the partial ordering on @xmath383 obtained as the union of the respective partial orderings of @xmath317 and @xmath381 together with the additional pairs @xmath384 as @xmath66 is a meet - subsemilattice of @xmath317 , @xmath385 is a lattice . it is easily seen to be an atomistic @xmath26-ladder ( the only atom in @xmath385 not in @xmath317 is @xmath386 ) . in particular , the lower covers of @xmath197 , for @xmath269 , are @xmath47 and ( in case @xmath387 ) @xmath388 , @xmath389 , where @xmath390 and @xmath391 are the lower covers of @xmath47 in @xmath66 . the @xmath4-sequence @xmath392 thus constructed is as required . [ d : stramalg ] a poset @xmath49 is the _ strong amalgam _ of two subsets @xmath86 and @xmath393 over a subset @xmath21 , if @xmath394 , @xmath395 , and for all @xmath396 , @xmath308 ( resp . , @xmath397 ) iff there exists @xmath398 such that @xmath399 ( resp . , @xmath400 ) . the proof of the existence of a @xmath26-ladder of cardinality @xmath0 from a @xmath91-morass is based on the following lemma ( the maps @xmath401 are defined in ) . [ l : amalg ] let @xmath182 be a countable , transitive , normed @xmath26-ladder with range @xmath402 , let @xmath403 be an ordinal with @xmath404 , and put @xmath405 . then there are a countable , transitive , normed @xmath26-ladder @xmath406 with range @xmath407 , containing @xmath182 as an ideal , and a lower embedding @xmath408 such that the following conditions hold : 1 . @xmath409 $ ] is the strong amalgam of @xmath182 and @xmath410 $ ] over @xmath21 ; 2 . every element of @xmath406 is the join of two elements of @xmath409 $ ] ; 3 . @xmath411 is the identity map on @xmath21 ; 4 . the equality @xmath412 holds for each @xmath195 . it is straightforward to construct a ( set - theoretical ) copy @xmath413 of @xmath182 such that @xmath414 , isomorphic to @xmath182 _ via _ a bijection @xmath415 such that @xmath416 . defining the ordering on @xmath417 according to definition [ d : stramalg ] , we obtain that @xmath417 is a strong amalgam of @xmath182 and @xmath418 $ ] over @xmath21 . ( as @xmath419 , @xmath420 remains the least element of @xmath417 . ) we extend the norm @xmath242 to @xmath417 by setting @xmath421 observe that the extension of @xmath242 thus defined preserves all finite joins within each block @xmath182 or @xmath410 $ ] . pick @xmath422 and fix a cofinal chain @xmath357 in @xmath182 with @xmath423 as least element . we set @xmath424 ( cf . ) . observe that the equality @xmath425 holds for each @xmath426 . we refer the reader to section [ s : basic ] for the notation @xmath67 . [ cl : fmsk2 ] the subset @xmath66 is a cofinal meet - subsemilattice of @xmath427 . furthermore , the equality @xmath428 holds for each @xmath429 . as @xmath430 belongs to @xmath66 for each @xmath364 , @xmath66 is cofinal in @xmath427 . let @xmath431 , we must prove that @xmath432 . we may assume that @xmath433 . it is straightforward to verify that @xmath434 , and so @xmath435 belongs to @xmath66 . now let @xmath429 , put @xmath436 and @xmath437 . as @xmath438 and @xmath439 , it follows from lemma [ l : proj2meet](i ) that @xmath440 , hence @xmath441 belongs to @xmath66 ; it obviously lies above @xmath442 . for each @xmath426 above @xmath442 , it follows from @xmath443 , @xmath444 , and @xmath364 that @xmath445 . from @xmath446 it follows that @xmath447 , thus , as @xmath448 , we get @xmath449 , and thus @xmath450 ( the latter equality following from the relation @xmath451 ) . therefore , @xmath452 . as @xmath66 is a cofinal meet - subsemilattice of @xmath427 , it is a lower finite lattice ; its least element is @xmath453 . [ cl : f2ladd ] the lattice @xmath66 is a @xmath11-ladder . we prove that every element @xmath426 has at most two lower covers in @xmath66 . we denote by @xmath454 the unique lower cover of @xmath362 in @xmath357 if it exists ( i.e. , @xmath455 ) , and by @xmath456 the largest element of @xmath457 such that @xmath458 if it exists ( i.e. , @xmath459 ) . if @xmath454 exists , then @xmath460 is an element of @xmath66 smaller than @xmath461 . if @xmath456 exists , then @xmath462 is an element of @xmath66 smaller than @xmath461 . now let @xmath463 such that @xmath464 . if @xmath465 , then @xmath454 exists and @xmath466 , but @xmath467 , thus @xmath468 and so @xmath469 . now suppose that @xmath470 . hence @xmath471 , but @xmath472 , thus @xmath473 , so @xmath456 exists and @xmath474 , and so @xmath475 . therefore , each lower cover of @xmath461 in @xmath66 is equal to @xmath476 for some @xmath129 . now we endow the disjoint union @xmath477)\sqcup f$ ] with the partial ordering obtained as the union of the respective partial orderings of the strong amalgam @xmath409 $ ] and the @xmath11-ladder @xmath66 together with the pairs @xmath478 for each @xmath195 and each @xmath426 . it is important to observe that there is no ambiguity between and in case @xmath398 ( i.e. , @xmath479 ) , because of the equality @xmath425 . it is straightforward to verify that @xmath406 is a lower finite poset in which both @xmath182 and @xmath410 $ ] are lower subsets . furthermore , by using claim [ cl : fmsk2 ] , we obtain that @xmath406 is a @xmath7-semilattice , in which the join operation on pairs of elements not in the same block @xmath182 or @xmath410 $ ] is given by @xmath480 for all @xmath195 , all @xmath481 , and all @xmath482 . for example , @xmath483 is the least element of @xmath66 above @xmath484 , thus , by claim [ cl : fmsk2 ] and as @xmath485 , it is equal to @xmath486 . observe that implies immediately that every element of @xmath406 is the join of two elements of @xmath409$]indeed , @xmath487 for each @xmath426 . we set @xmath488 in order to verify that the extension of @xmath242 defined in and is a join - homomorphism , we consider the expressions and we use repeatedly the fact that @xmath489 holds for each @xmath426 , hence @xmath490 absorbs @xmath371 for each @xmath195 . the verifications of these facts are straightforward . as @xmath491 has range @xmath402 , @xmath492}$ ] has range @xmath493 , and @xmath494 has range contained in ( and in fact equal to ) @xmath495 , we obtain that the range of @xmath242 is @xmath496 . as both @xmath182 and @xmath410 $ ] are lower subsets of @xmath406 and @xmath26-ladders , in order to verify that @xmath406 is a @xmath26-ladder , it suffices to verify that each @xmath426 has at most three lower covers in @xmath406 . we use the notations @xmath454 and @xmath456 introduced in the proof of claim [ cl : f2ladd ] . by using claim [ cl : f2ladd ] , it is easy , although a bit tedious , to verify that every lower cover of @xmath461 in @xmath406 has one of the following forms : * @xmath497 , @xmath498 , and @xmath499 , if @xmath455 and @xmath459 ; * @xmath362 , @xmath500 , and @xmath499 , if @xmath455 and @xmath501 ; * @xmath423 and @xmath502 , if @xmath503 . in any case , @xmath461 has at most three lower covers in @xmath406 . [ t : mor2ladd ] suppose that there exists an @xmath91-morass . then there exists an atomistic @xmath26-ladder of cardinality @xmath0 . as discussed in section [ s : morass ] , there exists a simplified @xmath91-morass . we shall use the same notation as in definition [ d : gap1mor ] . we shall construct inductively a system @xmath504 satisfying the following conditions : * each @xmath245 is an atomistic normed @xmath26-ladder with range @xmath505 . + _ accordingly , we shall denote by @xmath242 , or @xmath506 in case @xmath95 needs to be specified , the norm function on @xmath245_. * @xmath245 is countable for each @xmath507 . * if @xmath508 and @xmath509 , then @xmath510 is a lower embedding . * if @xmath511 , then the equality @xmath512 holds for all @xmath513 . * if @xmath508 , then the equality @xmath514 holds for each @xmath509 . provided we can carry out this construction we shall be done , since then @xmath515 is a normed @xmath26-ladder with range @xmath4 , thus of cardinality at least @xmath0 . as every @xmath26-ladder has cardinality at most @xmath0 ( cf . proposition [ p : uppbdbr ] ) , @xmath515 has cardinality exactly @xmath0 . we start with @xmath516 ( endowed with containment ) with the norm operation defined by @xmath517 , @xmath518 , and @xmath519 . suppose the construction carried out to the level @xmath372 ( that is , replace @xmath520 by @xmath521 in ( k0)(k4 ) above ) , we shall show how to extend it to the level @xmath522 . we put @xmath523 by lemma [ l : amalg ] , there exists a countable , transitive , normed @xmath26-ladder @xmath385 with range @xmath524 , containing @xmath317 as an ideal , together with a lower embedding @xmath525 such that the following conditions hold : 1 . @xmath526 $ ] is the strong amalgam of @xmath317 and @xmath527 $ ] over @xmath369 ; 2 . every element of @xmath385 is the join of two elements of @xmath526 $ ] ; 3 . @xmath528 is the identity map on @xmath369 ; 4 . the equality @xmath529 holds for each @xmath530 ( _ recall that @xmath123 is the restriction of @xmath124 from @xmath102 into @xmath125 _ ) . in particular , it follows from the induction hypothesis and from ( ii ) that @xmath385 is also atomistic . we put @xmath531 . hence @xmath532 is a lower embedding from @xmath317 into @xmath385 , for each @xmath533 . it follows immediately from the definition that @xmath534 we must define @xmath532 , for each @xmath535 and each @xmath536 . it follows from ( p2 ) ( cf . definition [ d : gap1mor ] ) that there are @xmath537 and @xmath538 such that @xmath539 . we have no other choice than to define @xmath540 . it follows easily from lemma [ l : mor2tree ] that the map @xmath541 is uniquely determined , however the map @xmath226 is not necessarily uniquely determined . this uniqueness problem arises iff @xmath542 , in which case the range of @xmath541 is contained in the intersection of @xmath102 and the range of @xmath123 , which is @xmath101 . in this case , we need to prove that @xmath543 , that is , @xmath544 . this is obvious by , as @xmath545 , so , by using the induction hypothesis , the relations @xmath546 hold for each @xmath547 . at this stage we are able to define @xmath548 as above . once this is done , the verification of ( k3 ) is straightforward . this concludes the successor case . now let @xmath549 be a limit ordinal and suppose that the construction of @xmath392 has been carried out to the level @xmath550 ( i.e. , replace @xmath520 by @xmath550 in the formulations of ( k0)(k4 ) ) . the proof of this case runs as the proof of the limit case of the gap-@xmath11 transfer theorem using a simplified morass presented in devlin ( * ? ? ? * section viii.4 ) , so we shall merely outline it . we put @xmath551 and we define @xmath552 , for @xmath553 , as the unique @xmath130 such that @xmath554 . for @xmath555 , let @xmath556 hold , if @xmath557 and there exists a ( necessarily unique ) @xmath558 such that @xmath559 . it follows from the induction hypothesis that the map @xmath560 is a lower embedding and @xmath561 it follows from ( p2 ) and ( p4 ) that @xmath562 is an upward directed strict ordering on @xmath563 , so we can form the direct limit @xmath564 in the category of lattices and lattice homomorphisms . as all @xmath565s are lower embeddings , so are all @xmath532s . as @xmath566 is the directed union of all the ranges of the lower embeddings @xmath567 and all the @xmath568s are atomistic @xmath26-ladders , @xmath566 is an atomistic @xmath26-ladder . furthermore , it follows from that there exists a unique map @xmath569 such that @xmath570 holds for each @xmath553 , and it is straightforward to verify that @xmath571 is a join - homomorphism . as @xmath572 has range @xmath573 for each @xmath553 and by ( p5 ) , the map @xmath571 has range @xmath574 . this completes the construction . as reminded in section [ s : morass ] , the nonexistence of a @xmath91-morass implies that @xmath4 is inaccessible in the constructible universe @xmath84 . hence , [ c : mor2ladd ] if there is no @xmath26-ladder of cardinality @xmath0 , then @xmath4 is inaccessible in @xmath84 . we still do not know whether the nonexistence of a @xmath26-ladder of cardinality @xmath0 is consistent with @xmath1 . we thus reach the intriguing conclusion that the existence of a @xmath26-ladder of cardinality @xmath0 follows from either one of two axioms of set theory , namely @xmath34 and the existence of a gap-@xmath3 morass , which do not imply each other . indeed , as there are morasses in @xmath84 and @xmath84 satisfies the generalized continuum hypothesis , while on the other hand @xmath34 implies that @xmath575 ( for the martin - solovay forcing , constructed from @xmath2 almost disjoint subsets of @xmath159 , used in the proof of jech ( * ? ? ? * theorem 16.20 ) , has precaliber @xmath2 ) , the existence of a @xmath91-morass does not imply @xmath34 . a simpler argument , pointed by the referee , runs as follows . if @xmath576 ( i.e. , @xmath81 restricted to _ countable _ notions of forcing ) holds , then , following the argument of jech ( * ? ? ? * exercise 16.10 ) ( using cohen forcing ) , it is easy to prove that @xmath33 . in particular , the existence of a gap-@xmath3 morass does not even imply @xmath576 . conversely , devlin @xcite finds a model of @xmath577 without any kurepa tree , thus _ a fortiori _ without any @xmath91-morass . the question whether there exists a @xmath153-ladder of cardinality @xmath156 , for @xmath578 , is also raised in @xcite . this question seems to get harder as @xmath153 grows larger ( we have no proof of this , although the similar implication with lower finite lattices of breadth @xmath153 is easy to establish ) . the proof of theorem [ t : mor2ladd ] suggests ( without proving it ) that the existence of a gap-@xmath6 morass would entail the existence of a @xmath579-ladder of cardinality @xmath83 , for every positive integer @xmath6 . however , such a program would run into a number of technical difficulties , starting with the lack of a formally established equivalence between gap-@xmath6 morasses and their simplified versions for @xmath580 ( cf . morgan @xcite for the case @xmath581 ) , or even the lack of formally established proofs that the latter exist in @xmath85 $ ] for any @xmath87 . furthermore , our other result that @xmath34 implies the existence of a @xmath26-ladder of cardinality @xmath0 ( theorem [ t : map23ladd ] ) suggests that morasses may not even be the optimal tool required to solve that problem . another objection is then that unlike morasses , higher and higher versions of martin s axiom would apparently not be able to construct ladders of cardinality beyond the continuum still , without the question whether @xmath1 implies the existence of a @xmath26-ladder of cardinality @xmath0 being settled , the morass track and the martin s axiom track seem to be all we have . i am grateful to rich laver for inspiring early e - mail discussions about the consistency problem of @xmath26-ladders of cardinality @xmath0 . i also thank michael pinsker for his many remarks and corrections , that lead in particular to substantial simplifications of several proofs . p. erds , a. hajnal , a. mt , and r. rado , combinatorial set theory : partition relations for cardinals " . studies in logic and the foundations of mathematics * 106*. north - holland publishing co. , amsterdam , 1984 . 347 p. isbn : 0 - 444 - 86157 - 2 j. silver , _ the independence of kurepa s conjecture and two - cardinal conjectures in model theory _ , 1971 axiomatic set theory ( proc . sympos . pure math . , vol . * 13 * , part i , univ . california , los angeles , calif . , 1967 ) p. 383390 . soc . , providence , r.i . |
we have searched for optical identifications for 79 _ chandra _ x - ray sources that lie within the half - mass radius of the nearby , core - collapsed globular cluster ngc 6397 , using deep hubble space telescope advanced camera for surveys wide field channel imaging in , , and . photometry of these images allows us to classify candidate counterparts based on color - magnitude diagram location . in addition to recovering nine previously detected cataclysmic variables ( cvs ) , we have identified six additional faint cv candidates , a total of 42 active binaries ( abs ) , two millisecond pulsars ( msps ) , one candidate active galactic nucleus , and one candidate interacting galaxy pair . of the 79 sources , 69 have a plausible optical counterpart . the 15 likely and possible cvs in ngc 6397 mostly fall into two groups : a brighter group of six for which the optical emission is dominated by contributions from the secondary and accretion disk , and a fainter group of seven for which the white dwarf dominates the optical emission . there are two possible transitional objects that lie between these groups . the faintest cvs likely lie near the minimum of the cv period distribution , where an accumulation is expected . the spatial distribution of the brighter cvs is much more centrally concentrated than those of the fainter cvs and the active binaries . this may represent the result of an evolutionary process in which cvs are produced by dynamical interactions , such as exchange reactions , near the cluster center and are scattered to larger orbital radii , over their lifetimes , as they age and become fainter . | 1106.1927 | gravitational interactions involving `` hard '' binary stars ( i.e.those with orbital velocities that exceed the local velocity dispersion ) provide the energy source that drives globular cluster evolution from the time that the cluster core approaches collapse @xcite . these interactions have a profound effect on the long - term fates of both the cluster and the binaries . among the hardest binaries in clusters are compact , accretional x - ray sources . bright low - mass x - ray binaries ( lmxbs ) , which contain a neutron star primary , are strongly overabundant in dense clusters and thus almost certainly have a dynamical origin there ( see * ? ? ? * ; * ? ? ? * ) . there is also growing evidence that a majority of the more numerous cataclysmic variables ( cvs ) in clusters may also have a dynamical origin , rather than representing a primordial population @xcite . this is a complex issue , given the many possible channels for the formation of cvs in clusters , which include : the evolution of primordial binaries , mediated by hardening interactions ; exchange interactions between singles and primordial binaries , possibly involving multiple exchanges ; tidal captures which do not result in mergers ; and three - body interactions ( see * ? ? ? deep surveys of compact binary populations in globular clusters will help to clarify the relative importance of these cv formation mechanisms . ngc 6397 is the second closest globular cluster ( @xmath2 ; * ? ? ? * ) and is by far the nearest cluster to have undergone core collapse . it thus provides a special laboratory for studying the formation and evolution of compact binary systems in the extreme environment of a collapsed core . ngc 6397 has been a key target for _ rosat _ , hubble space telescope ( _ hst _ ) , and chandra x - ray observatory ( _ chandra _ ) studies of compact binaries in clusters . @xcite first detected cataclysmic variables in ngc 6397 , as counterparts to faint _ rosat _ sources , using _ hst _ wide - field / planetary camera 1 and @xmath0-band imaging to select -excess objects . @xcite confirmed the power of the photometric technique , providing _ _ faint object spectrograph verification of the first three cv identifications in ngc 6397 . @xcite carried out the first _ chandra _ observations of ngc 6397 , which detected a rich population of 25 sources within the cluster half - mass radius ( @xmath3 = 2.33 ; * ? ? ? wfpc2 imaging in , @xmath0 , @xmath4 , and @xmath5 has led to the identification of many of these sources , including nine cvs and a number of active binaries @xcite . since the study by @xcite was carried out , we have obtained considerably deeper _ chandra _ and _ hst _ imaging of ngc 6397 . the results of the _ chandra _ imaging , which detected 79 sources within the half - mass radius , are reported by @xcite . these x - ray sources , designated by a prefix u , are listed in table [ t : counterparts ] . in this paper , we report a search for optical counterparts to these sources using a deep _ hst _ acs / wfc dataset . we describe the data , the analysis method , and the results in the following sections . the optical imaging used in this study is the _ hst _ go-10257 dataset ( pi : anderson ) , which provides deep , highly dithered acs / wfc imaging of the central region of ngc 6397 in f435w ( @xmath1 ) , f625w ( @xmath0 ) , and f658n ( ) . one of the main motivations of this study was to search for wobbles in the positions of stars , indicative of massive unseen companions , such as white dwarfs , neutron stars , or black holes . to provide good time sampling , the center of ngc 6397 was imaged over 10 single - orbit epochs , spaced at approximately one - month intervals between 2004 july and 2005 june . a by product of this observing strategy is that we imaged the cluster at many different orientations , allowing for an exquisite mitigation of detector - dependent photometric and psf - related errors . in addition , the stacked images show round psfs with no sign of diffraction spikes . the dataset consists of 5 short @xmath1 ( 13s ) , 5 long @xmath1 ( 340s ) , 5 short @xmath0 ( 10s ) , 5 long r ( 340s ) , and 40 ( 390s ) exposures . the short exposures were designed to provide accurate photometry for stars above the main - sequence turnoff ( msto ) . saturation only becomes an issue near the tip of the giant branch . with the large number of frames , the psf sampling is particularly good for this band the _ hst _ data were analyzed using software based on the program developed for the acs globular cluster treasury project , described in @xcite . the routine first finds the astrometric and photometric mapping from each exposure into a master reference frame . it then searches through the master reference frame one small patch at a time , identifying stars where a significant number of exposures indicate coincident detections . these stars are then measured two ways : ( 1 ) they are measured in each individual exposure where they could be found , and ( 2 ) they are also measured simultaneously in all the exposures ( which provides a better measurement for the fainter stars ) . stars were measured with spatially variable , library psfs ( see anderson & king acs isr 2006 - 01 ) constructed from the go-9444 data set ( pi - cool ) of @xmath6cen , using an aperture of whole pixels that was optimized for the brightness of the star and its particular surroundings . since our best coverage was in , we did our star - finding on those images , and then used the positions found to identify and measure the stars in the b and r images . the final photometry we report here comes from averaging the photometry from the individual stars , with sigma clipping applied to remove outlying values due to cosmic rays , defective pixels , etc . a total of 25004 stars were detected . for the @xmath1 and @xmath0 bands , the photometry was performed independently for the short and long frames . photometric calibration to the vegamag system was adopted from @xcite . in addition to the star finding and measuring , we also produced stacked images of the scene , in a procedure that is akin to the iraf / stsdas _ drizzle _ algorithm with pixfrac set to zero . the stacked images were oversampled by a factor of two , in order to take advantage of the heavy dithering to increase the effective resolution . the resulting supersampled stacked images have a 12,000@xmath712,000 format and cover an approximately circular field of diameter 5 with a pixel scale of 0.025 . figures [ f : mosaic ] and [ f : mosaic_zoom ] show a combination of the 40 images , together with the error circles for the 79 _ chandra _ sources . the error circles are plotted at their actual sizes in both figures . as can be seen in these figures , the combination of images at various orientations results in a very smooth symmetric psf , with a distinct lack of diffraction spikes . color - magnitude diagrams ( cmds ) were constructed from the photometry using the @xmath0 magnitudes , and the and color indices . the short - exposure @xmath1 and @xmath0 magnitudes were used for stars brighter than about 2.5 mag below the msto , since the long - exposure frames saturate above this point . figures [ f : cmd_cv ] and [ f : cmd_ab ] show the resulting cmds . the efficacy of the photometric procedure is indicated by the tightness of the fiducial sequences . no proper - motion cleaning has been performed at this point in the analysis . proper - motion determination is described in [ astrometry ] . in [ source_id ] we will use proper motions in our source identification . unfortunately the quality of the cluster - field separation depends too sensitively on brightness to provide a definitive membership determination for all stars . the ( , @xmath0 ) cmd reaches deepest for the bluest stars , since the faintest red main - sequence ( ms ) stars are below the detection limit in @xmath1 . the upper part of the white dwarf ( wd ) cooling sequence is clearly detected in the ( , @xmath0 ) cmd , extending to nearly 10 mag below the msto in @xmath0 . there is also an indication of a second wd sequence above the primary sequence , which @xcite interpret as a he wd sequence . note that in the ( , @xmath0 ) cmd , the wds lie to the -deficit side of the ms , reflecting the strong absorption lines in wd spectra relative to those of faint ms stars . the blue stragglers and the blue horizontal branch stars similarly lie to the -deficit sides of the subgiant branch and the giant branch , respectively . the stacked _ hst _ acs / wfc images for each filter were rectified to a common astrometric system . we calculated an absolute astrometric zeropoint for the stacked images relative to the icrs using a two - step process . we first computed a plate solution for an eso wfi frame ( wfi.2002 - 06 - 21t01:45:08.153 ) using 357 primary astrometric standards from the usno ucac2 catalog . we then selected 350 secondary astrometric standards from the wfi frame and used these to compute a plate solution for our stacked acs / wfc images . we determined a boresight correction for the _ chandra _ source coordinates from @xcite by computing the mean offsets between the _ hst _ and _ chandra _ coordinates for the six brightest cvs and the millisecond pulsar ( msp ) psr j1740 - 5340 . this resulted in a shift of the _ chandra _ coordinates to the icrs of @xmath8 in ra and @xmath9 in dec , where the quoted uncertainties are the standard deviations of the offsets for the seven objects . as an external test of our boresight correction , we note that our optical position for psr j1740 - 5340 agrees with that of @xcite to @xmath10 in ra and @xmath11 in dec . they followed a similar astrometric procedure to what we used , with wfpc2 rather than acs / wfc imaging . we searched for optical counterparts to the _ chandra _ sources by overlaying the _ chandra _ error circles on the _ hst _ stacked images , with the boresight correction applied to the _ chandra _ source positions from @xcite . the 95% confidence error circle sizes were computed following @xcite . since the uncertainty in the optical positions @xmath12 ) was small compared with the size of the x - ray error circle radii ( @xmath13 ) , we neglected the contribution of the former to the total positional uncertainty . a preliminary proper - motion analysis was carried out in order to do a first - pass cluster - field separation . the present dataset was used for the reference epoch , with the stellar positions averaged over the 10 orbits . two different datasets were considered for the second epoch , go-7335 ( obtained with the wfpc2 in 1999 ) and go-10775 ( obtained with the acs / wfc in 2006 ) . the former provides a longer time baseline , while the later provides more depth and spatial coverage . to compute proper motions , we measured stars in each early - epoch exposure image , corrected these positions for distortion , then transformed these positions into the reference frame using linear transformations based on only cluster members . by using only members to define the transformations , we automatically get displacements ( and hence motions ) with respect to the cluster s systemic motion . the cluster - field separation that can be achieved for moderately bright stars is illustrated in fig . [ f : pm ] , which shows the proper - motion distribution for stars with @xmath14 . cluster and field regions are defined by a magnitude - dependent radius about the centroids of the apparent cluster and field clumps . stars that do not reside in either of these regions are considered to have an unknown proper - motion membership status . since the proper - motion error increases with magnitude , this proper - motion separation becomes less effective with increasing magnitude . eventually the cluster and field regions of the proper - motion plane show a strong overlap . it was generally possible to obtain an approximate discrimination between cluster and field stars to a limiting magnitude of @xmath15 . in a few cases where the 2006-second - epoch results were resulted in an unknown proper - motion membership status , the 1999-second - epoch results were definitive . optical cmds provide a powerful means of selecting likely _ chandra _ source counterparts and investigating their properties . for each of the 79 _ chandra _ sources within the half - mass radius , we checked the locations of all objects within the x - ray error circle in the cmds shown in figures [ f : cmd_cv ] and [ f : cmd_ab ] . objects that fell on the main sequence were considered to be unlikely counterparts , given the relatively low x - ray to optical flux ratio , @xmath16 , of most ms stars , in contrast to the ranges for chromospherically active binaries and cataclysmic variables . table [ t : counterparts ] summarizes the result of this counterpart search . plausible identifications were obtained for nearly all sources and are listed in this table , along with previous source designations , inferred object types , proper - motion information , and photometry . the only sources for which the error circle was empty are the quiescent low - mass x - ray binary ( qlmxb ) u24 @xcite , u104 , u106 , u111 , u113 , u119 , u121 , u122 , and u123 . in the case of u24 , there is a small `` blip '' near the center of the error circle in the stacked @xmath0 image , but it likely represents the combination of airy ring artifacts from two bright nearby stars . several other sources , u5 , u16 , u77 , and u84 , lie outside of the @xmath0 field and thus are difficult to classify , from these data , due to the lack of and color measurements . however , the u5 and u77 counterparts , v30 and v36 respectively , have been shown to be variable using ground - based photometry @xcite . @xcite classified v30 as an eclipsing binary that lies to the far red of the main sequence ( suggestive of it being a non - member ) . @xcite have recently classified v36 as a single - line spectropsopic binary . we have classified both of these stars as uncertain active binaries ( abs ) . u5 and u84 have very red @xmath17 colors , suggesting that their colors are likely very red as well , which supports an ab identification . table [ t : counterparts ] also gives a summary of the results of the proper - motion analysis , which indicates whether the object is consistent with being a cluster member , a field member , or neither , based on a comparison of its proper motion to the proper - motion distribution for stars of similar magnitude . based on the location of the proposed counterparts in the cmds ( or , in a few cases , based on ground - based detection of variability ) , we primarily assigned types of cataclysmic variable ( cv ) and chromospherically active binary ( ab ) . cvs were defined as being significantly to the blue of the ms and having significant excesses ( either relative to the ms or to the white dwarf sequence ) . abs were defined as lying within @xmath18 mag above the ms ( and thus within @xmath19 mag to the red of the ms ) and having small excesses ( @xmath20 mag ) , based on the previous analysis of abs in ngc 6397 by @xcite . in three cases where the only object in the error circle was an apparent ms star , u41 , u91 , and u112 , we note its presence in table [ t : counterparts ] . similarly , an apparent msto star is the only object present in the error circle for source u118 . we note that an ab with a low mass ratio and weak lines could look like a ms star in both cmds . two objects , u28 and u108 , were classified as background galaxies based on the extended appearance of their images . u28 resembles an edge - on spiral , while u108 has a more complex structure suggestive of interacting galaxies . the high x - ray to optical flux ratio of u28 ( see [ flux_ratio ] ) suggests that it is an active galactic nucleus ( agn ) . while the flux ratio of u108 was not determined , due to the complex nature of the optical image , it is clearly much lower than that of u28 . one possibility is that we are detecting x - ray emission from u108 that is produced by a galaxy collision . figure [ f : cmd_cv ] shows the location of the cv candidates in the cmds . there is a suggestion of an evolutionary sequence from young , bright cvs to old , faint ones . we return to this point in [ summary ] . the six brightest cvs mostly lie about 0.2 0.8 mag to the blue of the ms . the optical emission of the systems appears to be dominated by the secondary in the @xmath0 band , with a larger contribution from the disk in the @xmath1 band . the fairly high @xmath0-band flux indicates that the secondaries are relatively massive , @xmath21 , as inferred from the isochrones of @xcite . all of the bright cvs have substantial excesses relative to the ms ; these excesses generally increase with magnitude . the cv candidates 18 are numbered according to the scheme used in previous papers , as listed by @xcite . a prefix cv is used to designate each cv candidate . @xcite identify u28 as a cv ; however we find that this chandra source appears to be a galaxy . here we use the identification by @xcite of u60 as cv9 . cvs 1015 are newly identified in the present study ; @xcite assigned numbers for cvs 10 and 11 . we have numbered cvs 1215 based on optical luminosity within this group . the translation between _ chandra _ source numbers and cv numbers is given in table [ t : counterparts ] . there is a two magnitude gap between the bright cvs and the fainter cvs . just below this gap are two possible transitional objects , u7 and u61 , which lie between the white dwarf sequence and the main sequence . we note that the proper - motion data produce an unknown membership status for u61 , while u7 registers as consistent with membership . this leaves open the possibility that u61 is a background agn with an emission line that is shifted into the window . however , the image of the u61 counterpart does not show obvious evidence of extension . the fainter cv candidates , below u7 and u61 , lie in the vicinity of the wd cooling sequence . the optical fluxes for the faint cv candidates are clearly dominated by the contribution of the wd . the indices for these faint cvs nearly all lie at least 0.5 mag to the excess side of the wd clump , which is centered at about @xmath22 . this suggests that the faint cvs have a strong -emission core ( due to an accretion disk ) within the broad absorption lines of the wd continuum . this inferred spectroscopic property of the faint cvs appears to be generally consistent with those of the wz sge class of evolved cataclysmic variable @xcite . another possibility is that some of these objects are am cvn type double - degenerate systems , in which a low - mass he wd donor feeds a much more massive carbon - oxygen wd . such systems show no evidence of hydrogen lines in their spectra and instead have he i and sometimes he ii lines , which are typically in emission @xcite . am cvn stars could have evolved into contact from the known he wds in ngc 6397 , which have been shown to have ( detached ) heavy co wd companions @xcite . such an object should lie close to the main sequence in the @xmath23 cmd . possible candidate am cvn stars among the cv candidates reported here are u25 , u31 , u60 , u80 , and u83 . we note that most of the wd - like cv candidates cluster near a magnitude of @xmath24 , which corresponds to a @xmath25 , using the distance modulus determined by @xcite . we note that @xcite have reported a sharp peak ( `` spike '' ) near the period minimum ( @xmath26 ) in the period distribution of cvs selected from the sloan digital sky survey ( sdss ) . the average absolute magnitude for the cvs in the spike is @xmath27 . allowing for the difference in photometric band , this is reasonably close to the characteristic magnitude of the faint cvs in our study . this suggests that these faint cvs may well belong to the spike population , i.e. have periods in the vicinity of the period minimum . this could be tested by determining orbital periods for the faint cvs from suitable time - resolved photometry . @xcite find that of 33 cvs with orbital periods less than 86 min , 20 have spectra that are wd - dominated . in our sample , it appears that all seven faint cvs are wd - dominated . given the small sample size , it is not clear whether this represents a significant difference between faint field and faint cluster cvs . figure [ f : cmd_ab ] shows the location of the objects identified as ab stars in the cmds . we note that these objects sometimes lie near the edge of the ms in either the right or left panel , but deviate by a larger amount in the other panel . of the 42 ab counterparts listed in table [ t : counterparts ] , 26 are newly identified in our study . overall , the abs appear to form a relatively homogeneous binary sequence alongside the ms , presumably mostly differing in mass . @xcite have argued that these stars are likely by draconis stars and represent a hard binary population in ngc 6397 . in a large study of photometric variability in 47 tuc , @xcite found a substantial population of by draconis binaries , as well as w uma and other contact binaries , and eclipsing binaries . in addition to the likely by dra population , there are a number of counterparts that show a significantly different distribution in the cmd . these include the `` red straggler '' counterpart to u12 ( the millisecond pulsar ) and the similar counterpart to u18 , the blue - straggler counterpart to u70 , a group of five stars with very red colors , and a group of five stars with strong excesses . there are two stars in common between these latter two groups . the counterparts to u12 and u18 both have proper motions consistent with cluster membership . given their similar optical and x - ray properties , it appears likely that u18 is a msp @xcite . the counterpart to u70 , which registers as a proper - motion cluster member , has been identified as variable star v20 by @xcite . @xcite find that it is a double - line spectroscopic binary with a period of @xmath28 and a mass ratio of @xmath29 . the he wd candidate pc-5 @xcite , which has been previously suggested as a possible counterpart to u70 , lies well outside of the current error circle . the counterparts to u42 , u63 , u65 , u86 , and u92 differ strikingly from the typical abs , lying up to 1 mag to the red of the ms , and thus well above it . of these , the counterparts to u42 and u92 have been previously identified as optically variable stars @xcite , as confirmed here ( see [ variability ] and fig . [ f : variability ] ) ; the counterparts to u63 , u65 , and u86 are new identifications . the u92 counterpart ( v7 ) is classified as a w uma system by @xcite , while the u42 counterpart ( v26 ) is classified as an irregular variable by @xcite . u92 is consistent with the cluster proper - motion distribution . the proper - motion information available for u42 , u65 , and u86 produce an unknown membership status . these stars have proper motions that are well outside of the cluster distribution but they are also inconsistent with the field distribution . there is no proper - motion information for u63 . one reasonable possibility is that these stars are foreground active binaries superposed on the cluster , as discussed by @xcite in the case of u5 ( v30 ) . in this interpretation , the foreground binaries contain k / m type dwarfs with distance moduli that put them well above the cluster main sequence in the @xmath30 cmd . figure [ f : cmd_ab ] also shows several ab candidates that have larger than typical excesses ; these include the counterparts to u62 , u65 , u66 , u86 , and u120 . there is no proper - motion information available for u62 . the proper - motion information for u65 , u86 , and u120 produces an unknown membership status . u66 is consistent with the field proper - motion distribution . thus , the proper - motion information does not offer much assistance in determining whether these stars are likely cluster members . @xcite have drawn attention to two of the variable stars that we have identified as ab counterparts , v17 and v36 , which correspond to the sources u76 and u77 . these similar objects both lie close to the main - sequence turnoff in the cmd presented by @xcite . both objects are single - line spectroscopic binaries , in which the unseen primary is inferred to have a mass in excess of 1 . they suggest that these two objects are members of a class of dormant degenerate binaries which may have much larger peak x - ray luminosities . as an additional aid to the classification of sources , we have examined the x - ray to optical flux ratio , @xmath16 , where we take @xmath31 from @xcite and set @xmath32 . the latter conversion factor is computed from the _ hst _ flux calibration constants . the x - ray count to flux transformation assumes a power - law spectrum with a photon index of @xmath33 @xcite . the resulting flux ratio is plotted versus @xmath34 in fig . [ f : fx_fopt ] . the ratio is observed to be higher for accretional sources , such as cvs , lmxbs , and agns , than for chromospherically active binaries ( see e.g. * ? ? ? * who studied the _ chandra _ source distribution in m4 ) . consistent with this , fig . [ f : fx_fopt ] shows that the cv and ab candidates form two distinct groups with no overlap in flux ratio . the median flux ratio is about 500 times larger for the cvs than for the abs . within each group there is a broad range of the flux ratio , about a factor of 40 for the cvs and a factor of 100 for the abs . there is no apparent dependence of the flux ratio on @xmath34 over a range of more than three decades in @xmath34 for the cvs and nearly two decades for the abs . the apparent agn , u28 , has the highest flux ratio of all objects . the known millisecond pulsar , u12 , has a higher flux ratio than most of the abs and is about 40@xmath7 brighter in the x - ray than a typical ab . the object u18 , which has a similar optical counterpart to that of u12 , has even higher values of @xmath34 and @xmath16 . this supports the interpretation of u18 as a msp not yet detected at radio wavelengths @xcite . interestingly , the x - ray flux of cv5 ( u22 ) is considerably lower in the 2007 observations than in the previous 2002 observations ; the flux differs more than an order of magnitude . this gives cv5 the lowest @xmath16 value of all the cvs when the 2007 x - ray flux is used together with the 2004 - 05 optical flux . since the x - ray and optical fluxes were not measured contemporaneously , it is possible that cv5 typically has a flux ratio that is more consistent with the other bright cvs . we note that the x - ray flux of a bright cv typically drops during outburst ( e.g. * ? ? ? * ) , suggesting that cv5 may have been in outburst in 2007 . the objects u62 , u66 , and u86 , which lie among the abs with the four highest flux ratios , all have significantly larger excesses than the typical ab . it is not surprising that two measures of chromospheric activity , x - ray flux and flux , are correlated for these stars . since our dataset provides a 40-exposure time sequence of exposures , with four exposures per orbit , it was possible to investigate optical variability . the time sequence samples time scales shorter than about one hour ( the visibility period per _ hst _ orbit ) and also time scales from one month to one year . the analysis of variability is complicated by the presence of outliers in the time sequences , some of which represent photometric anomalies . we investigated several measures of variability and adopted the rms deviation about the mean of the magnitude measurements , computed using an iterative 3@xmath35 clip to reduce the impact of outliers . we plot @xmath36 versus mean magnitude in figure [ f : variability ] . this procedure results in a measure of variability that is most sensitive to orbital variability of binary systems , rather than large - amplitude fluctuations of cvs , given the outlier filtering . thus , we also investigated the total range of the magnitudes for each object and constructed light curves for each of the cv candidates in order to test for outburst behavior . as can be seen in fig . [ f : variability ] , most stars fall on a well - defined sequence of increasing @xmath36 with increasing magnitude . stars that lie significantly above this sequence generally show evidence for variability , although some of this scatter is likely due to photometric issues . we note that the mean trend of the variability index @xmath36 with mean @xmath37 magnitude differs for stars brighter than @xmath38 , for which @xmath36 increases from a minimum of about 0.003 mag to about 0.04 mag for the brightest stars plotted . this behavior for bright stars is due to saturation issues that increase in size with decreasing magnitude . we have plotted the positions of the previous and new chandra source counterparts in fig . [ f : variability ] . many of these counterparts show significant variability as measured by the @xmath35 of the time series . this group includes all of the bright cvs 16 : sources u10 , u17 , u19 , u21 , u22 , and u23 . of the newly identified faint cv counterparts , only u61 registers as clearly variable by this measure and u25 shows a possible slight variability . while the other faint cv candidates have @xmath36 values that are comparable in size to those of the bright cvs , these values are not significantly larger than those of other stars of similar magnitude . thus , the failure to detect statistically significant variability for the other faint cv counterparts may well be due to the decreasing sensitivity of the variability test with increasing magnitude . the total magnitude range for each of the six bright cvs was about 0.61.3 mag , which is less than the 25 mag amplitude for a dwarf nova outburst @xcite . source u21 ( cv4 ) had the largest amplitude ( 1.3 mag ) of these bright cvs , with a light curve that a high phase of about 50 days duration . source u61 ( cv12 ) displayed a total range of 2.3 mag . the faintest magnitude measurements , which contribute significantly to this large range , have a high degree of uncertainty . the light curve shows a clearly differentiated low state in the first half of the year of observation and a high state during the second half . the mean offset between these two states is about 1 mag and thus also falls short of a dwarf nova outburst . a number of the candidate ab counterparts also show evidence of variability , including six newly identified abs : sources u79 , u97 , and u114 , which show strong variability at the level of the six bright cvs ; sources u88 and u101 , which show moderate variability ; and source u103 , which shows slight evidence of variability . five previously identified ab counterparts , u14 , u18 , u42 , u70 , and u92 , show clear variability , and u76 and u81 show weaker evidence of variability . we note that u18 has a similar level of variability to the msp counterpart u12 , again suggesting that these are similar objects . we carried out a period - folding light curve analysis for the bright cvs 16 and the msp u12 . given the very unequal spacing of the data , it is difficult to determine a period ab initio , due to the high degree of aliasing . however , we were able to confirm the previously determined orbital periods obtained by @xcite and @xcite for cvs 1 and 6 , as well as that of u12 @xcite . we also see evidence for additional long - term variability for cvs 1 and 6 , at the level of a few tenths of a magnitude . we determined the cluster center by iterative centroiding in a 20 radius aperture using a sample of ms stars with magnitudes in the range @xmath39 . the resulting center of @xmath40 lies within about 1 of cv2 ( u19 ) . experimentation with the centroiding aperture size and the stellar sample definition indicates that the center position is uncertain by about 1 . we then determined the cumulative radial distributions of a number of stellar groups out to a radius of 100 , which is the approximate size of the central surface density cusp of ngc 6397 @xcite . figure [ f : radial ] shows the distribution functions for the stellar groups listed in the figure legend . as has been previously noted , the bright cvs in ngc 6397 show a strong central concentration relative to other stellar groups , including the fainter cvs . with the exception of cv6 , the five other bright cvs lie within 11 ( @xmath41 ) of the cluster center . in comparison , with the exception of u25 , the faint cvs lie at least 20 ( @xmath42 ) from the cluster center . we carried out kolmogorov - smirnov intercomparisons of the stellar samples defined in fig . [ f : radial ] . the comparison of each group to a msto sample , defined by @xmath43 , is given in table [ t : cored_pl_fits ] , where the probability , @xmath44 , of the two samples being drawn from the same parent distribution is listed . the bright cv and blue straggler ( bs ) samples both differ very significantly from the msto sample ( @xmath45 ) , while the overall cv and ab samples differ from the msto sample at the 4% level and 1% level , respectively . the bright and faint cv samples differ from each other at a significance level of 3% and the bright cv and ab samples differ at a level of 2% . lrcccccr msto & 1111 & 1.0 & @xmath46 & @xmath47 & @xmath48 & & + cv & 14 & @xmath49 & @xmath50 & @xmath51 & @xmath52 & 2.3 & 4.0% + bright cv & 6 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & 2.9 & 0.07% + faint cv & 8 & @xmath57 & @xmath58 & @xmath59 & @xmath60 & 0.2 & 89% + ab & 36 & @xmath61 & @xmath62 & @xmath63 & @xmath64 & 2.6 & 1.1% + bs & 23 & @xmath65 & @xmath66 & @xmath67 & @xmath68 & 3.7 & 0.03% + in order to further investigate the spatial distribution of the _ chandra _ sources in ngc 6397 and the implications for object masses , we carried out maximum likelihood fits of what has been termed a `` generalized king model '' to the surface density distributions . this model can be described more generally as a `` cored power law , '' in that it has a smooth transition from an inner core region to an outer power law . it takes the form , @xmath69^{\alpha/2},\ ] ] with the core radius @xmath70 related to the scale parameter @xmath71 by , @xmath72 an `` analytic king model , '' which provides a good fit to the surface density profile of the turnoff - mass stellar group for a normal - core cluster ( e.g. 47 tuc ; * ? ? ? * ) , corresponds to @xmath73 and @xmath74 . since ngc 6397 has a central - cusp structure , due to its core - collapsed status , we expect that the turnoff - mass group will have a flatter power - law slope than @xmath75 @xcite . we performed a maximum - likelihood fit of eqn . [ eqn : cored_pl ] to a turnoff - mass group , defined by @xmath43 and radial offset within @xmath76 of the cluster center , to determine the best - fit values of @xmath77 and @xmath70 . we used nonlinear optimization to maximize the likelihood and bootstrap resampling to estimate the parameter uncertainties , as discussed by @xcite . the resulting best - fit parameter values are @xmath78 and @xmath79 . by fitting eqn . [ eqn : cored_pl ] to the surface density distributions of individual groups such as the cvs and abs , the characteristic masses of these objects can be estimated , as described by @xcite and @xcite . in order to do this , we adopt the approximation here that stellar groups with masses exceeding the turnoff mass are in approximate thermal equilibrium with the turnoff mass group and that the velocity dispersion of each mass group is approximately constant with radius within the central cusp region . this approach is motivated by fokker - planck simulations of post - collapse core oscillations which indicate that a cluster core spends a majority of the time near a maximally expanded state in which the more massive stellar groups are close to thermal equilibrium @xcite . more recent monte - carlo and n - body simulations suggest that the presence of primordial binaries may delay the gravothermal oscillation phase to well beyond a hubble time and that even clusters with an observationally core - collapsed designation such as ngc 6397 may still be in the binary burning phase @xcite . in any case , the assumption of thermal equilibrium among the most massive components in the central region of an evolved cluster is a useful starting point . in the thermal equilibrium approximation , the surface density profile for a mass group with mass @xmath80 is given by eqn . [ eqn : cored_pl ] with a slope parameter @xmath77 related to the turnoff - mass slope @xmath81 by @xmath82 where @xmath83 . thus , fitting this model involves just one free parameter , the mass ratio @xmath84 , as @xmath85 is determined by normalization and @xmath71 and @xmath81 are determined by the fit to the turnoff - mass group . table [ t : cored_pl_fits ] gives the results of maximum - likelihood fits of eqn . [ eqn : cored_pl ] to the turnoff mass stars , cvs , abs , and a comparison group of blue stragglers ( bss ) . the cumulative radial distributions for these groups and the model fits are shown in fig . [ f : radial ] . as can be seen from the table , the @xmath84 values for all of the groups exceed unity , indicating that the characteristic masses exceed the turnoff mass . for all cases other than the faint cvs , the excesses are significant at the 2@xmath35 or higher level . the groups with the strongest evidence for a mass significantly above the turnoff mass are the bright cvs ( @xmath86 ) and the bss ( @xmath87 ) . we note that the results of this analysis agree with the k - s comparison results given in the last column of the table . the inferred mass range for the bright cvs ( @xmath88 ) is consistent with a system made up of a heavy white dwarf ( e.g. @xmath89 ) with a main sequence secondary of mass @xmath90 . the inferred mass range for the faint cvs ( @xmath91 ) is consistent with a somewhat less massive white dwarf ( e.g @xmath92 ) and a secondary mass that has been whittled down to @xmath93 . in such a system , the optical flux would be dominated by the white dwarf , as observed here . we presume that the white dwarf in this case has been heated by the accretion process , since noninteracting heavy white dwarfs are too faint to detect due to the degree of cooling that has occurred over their long lifetimes . as discussed in [ source_types ] , an alternative interpretation is that some of the cvs are am cvn systems , in which the secondary is a he wd . in this case , the mass of the secondary would likely be about 0.20.3 @xcite . it is interesting to compare the white dwarf masses inferred here with those for field cvs . @xcite have recently determined a mass of @xmath94 for the wd in a field cv that lies above the period gap . @xcite have determined the masses of the wds in seven cvs that lie below period gap , with a resulting mean mass and standard deviation of @xmath95 . thus , given the uncertainty ranges in the cv masses inferred here , our results for likely wd masses are generally consistent with these measured wd masses . we have shown that with moderately deep @xmath1 , @xmath0 , and _ hst_-acs / wfc imaging , it is possible to identify and classify nearly all of the 79 currently detected _ chandra _ sources within the half - mass radius of ngc 6397 . our classifications are based on cmd location and are tested for consistency with expected ranges of the x - ray to optical flux ratio . in only a few cases was it not possible to classify the source , generally due to a lack of coverage in one filter . the vast majority of the sources in the cluster can be classified as either cataclysmic variables ( a total of 15 ) or active binaries ( a total of 42 ) . the x - ray to optical flux ratio for the cvs ( 0.2 6 ) substantially exceeds that for most of the abs ( @xmath96 ) . there is one previously known msp and one potential msp ; the latter classification is based on the similar photometric properties of the two objects . the proper - motion test for cluster membership produces an unknown status for many of the _ chandra _ source counterparts . this is largely a consequence of the short time baseline of 1.5 years between the two epochs of acs / wfc data used for determining proper motions . while the 1999 wfpc2 observations give a longer baseline , fewer stars are detected by the wfpc2 . thus , the proper - motion part of this study would benefit substantially from new _ hst _ observations of ngc 6397 with either the acs / wfc or the wfc3 . while it appears likely that the vast majority of the proposed counterparts are cluster members , proper - motion confirmation would be valuable . it would be particularly interesting to know how many of the atypical abs are cluster members . while it is likely that most are foreground stars , one of the very red ones has appears to be a member on the basis of our preliminary proper - motion measurements . the cv distribution in ngc 6397 has a bimodal character . we detect distinct bright and faint populations with only two possible transition objects . the optical emission of the six brightest cvs appears to be dominated by a relatively massive secondary , while that of the faint cvs appears to be dominated by the wd , with very little contribution from a very low - mass secondary . alternatively , some of the faint cvs may be am cvn systems , with a he wd secondary . the distribution of cvs in optical luminosity and color is consistent with expected cv evolution due to the effects of mass transfer and angular momentum loss . as cvs age and the secondary loses mass to the primary , the orbit tightens due to magnetic braking and ultimately gravitational wave radiation . in this picture , the bright cvs 16 represent young , recently formed systems , while the faint cvs represent old , highly evolved systems . while cvs can be born at any orbital period , depending on the nature of the secondary star , there is a bias for cvs formed in exchange interactions in collapsed cluster cores to have more massive secondaries and thus longer orbital periods than newly formed field cvs . this is both the result of mass segregation in the cluster , which may invert the mass function in the cluster core , and of the tendency for more massive stars to displace less massive ones in exchange interactions . thus , while it is possible for a faint cv to be a young system , we expect that the bulk of the faint cvs evolved from an earlier bright state . given the likely ongoing formation of cvs by dynamical processes in the central region of the cluster , it is plausible that we are observing a roughly equilibrium cv population in ngc 6397 . however , there remains a possibility that the current population of six bright cvs represents the result of a recent core collapse event that resulted in an enhanced production of cvs . in this case , the bright cvs would be overpopulous relative to the faint cvs . the six brightest cvs in ngc 6397 have a much more centrally concentrated spatial distribution than do either the fainter cvs or the abs . indeed , five of six bright cvs are located within about 2@xmath70 of the cluster center , while all but one of the fainter cvs lie outside of 4@xmath70 . this suggests that cvs are preferentially formed by dynamical processes within the core and surrounding region and migrate to larger radii as they age and undergo repeated scattering interactions that act to increase the typical size of their orbits in the cluster potential . since repeated strong interactions between compact binaries and other stars ultimately result in the ejection of the binaries from clusters , the lifetime of a compact binary in a cluster is extended if the amount of time it spends in the central region of the cluster is reduced . thus , it is reasonable to expect that long - lived cvs in globular clusters should generally be found outside of the densest regions . this is consistent with the spatial distribution of the cvs in ngc 6397 . , s. f. , becker , a. c. , haggard , d. , prieto , j. l. , knapp , g. r. , sako , m. , halford , k. e. , jha , s. , martin , b. , holtzman , j. , frieman , j. a. , garnavich , p. m. , hayward , s. , ivezi , . , mukadam , a. s. , sesar , b. , szkody , p. , malanushenko , v. , richmond , m. w. , schneider , d. p. , & york , d. g. 2008 , , 135 , 2108 , s. f. , haggard , d. , homer , l. , joshi , n. r. , margon , b. , silvestri , n. m. , szkody , p. , wolfe , m. a. , agol , e. , becker , a. c. , henden , a. , hall , p. b. , knapp , g. r. , richmond , m. w. , schneider , d. p. , stinson , g. , barentine , j. c. , brewington , h. j. , brinkmann , j. , harvanek , m. , kleinman , s. j. , krzesinski , j. , long , d. , neilsen , jr . , e. h. , nitta , a. , & snedden , s. a. 2005 , , 130 , 2230 , b. t. , dillon , m. , southworth , j. , thorstensen , j. r. , rodrguez - gil , p. , aungwerojwit , a. , marsh , t. r. , szkody , p. , barros , s. c. c. , casares , j. , de martino , d. , groot , p. j. , hakala , p. , kolb , u. , littlefair , s. p. , martnez - pais , i. g. , nelemans , g. , & schreiber , m. r. 2009 , , 397 , 2170 , d. , & hut , p. 2003 , the gravitational million - body problem : a multidisciplinary approach to star cluster dynamics ( the gravitational million - body problem : a multidisciplinary approach to star cluster dynamics , by douglas heggie and piet hut . cambridge university press , 2003 , 372 pp . ) , d. , lewin , w. h. g. , anderson , s. f. , baumgardt , h. , filippenko , a. v. , gaensler , b. m. , homer , l. , hut , p. , kaspi , v. m. , makino , j. , margon , b. , mcmillan , s. , portegies zwart , s. , van der klis , m. , & verbunt , f. 2003 , , 591 , l131 llrlllllcccl u5 & 17:40:54.531 @xmath9753:40:44.57 & 1.85 & v30 & 1 & ab ? & ? & & & 18.73 & 19.23 & outside of @xmath0 field + u7 & 17:40:52.832 @xmath9753:41:21.77 & 1.81 & cv10@xmath98 & 1 & cv & c & & 22.58 & 23.44 & 24.73 & + u10 & 17:40:48.978 @xmath9753:39:48.62 & 1.21 & cv6 , v13 & 1 & cv & c & c & 19.14 & 20.34 & 21.53 & + u11 & 17:40:45.781 @xmath9753:40:41.52 & 0.58 & cv7 & 1 & cv & ? & c & 23.65 & 24.26 & 25.88 & + u12 & 17:40:44.621 @xmath9753:40:41.60 & 0.42 & by - wf4 - 1 , v16 & 1 & msp & c & c & 16.25 & 17.43 & 18.99 & + u13 & 17:40:44.084 @xmath9753:40:39.17 & 0.33 & cv8 & 1 & cv & ? & c & 24.00 & 24.33 & 25.95 & + u14 & 17:40:43.328 @xmath9753:41:55.46 & 1.46 & v35 & 1 & ab & c & & 18.31 & 19.72 & 20.96 & + u15 & 17:40:42.910 @xmath9753:40:33.81 & 0.14 & by - pc-2 & 1 & ab & c & & 17.25 & 18.53 & 19.93 & + u16 & 17:40:42.636 @xmath9753:42:15.24 & 1.78 & & 1 & ? & ? & & & & 25.83 & outside of @xmath0 and @xmath1 fields + u17 & 17:40:42.651 @xmath9753:40:19.30 & 0.17 & cv3 , v33 & 1 & cv & c & c & 18.52 & 19.10 & 20.67 & + u18 & 17:40:42.606 @xmath9753:40:27.62 & 0.07 & by - pc-8 , v31 & 1 & msp & c & c & 16.10 & 17.51 & 18.73 & + u19 & 17:40:42.306 @xmath9753:40:28.70 & 0.02 & cv2 , v34 & 1 & cv & c & c & 18.87 & 20.02 & 21.03 & + u21 & 17:40:41.830 @xmath9753:40:21.37 & 0.13 & cv4 & 1 & cv & c & c & 19.82 & 20.94 & 21.82 & + u22 & 17:40:41.701 @xmath9753:40:29.00 & 0.07 & cv5 & 1 & cv & c & c & 20.15 & 21.50 & 22.04 & + u23 & 17:40:41.597 @xmath9753:40:19.30 & 0.18 & cv1 , v12 & 1 & cv & c & c & 17.88 & 18.93 & 20.25 & + u24 & 17:40:41.468 @xmath9753:40:04.43 & 0.42 & & 0 ? & qlmxb & & & & & & hint of detection in @xmath0 only + u25 & 17:40:41.237 @xmath9753:40:25.79 & 0.15 & cv13@xmath98 & 1 & cv ? & & & 23.48 & 23.44 & 26.01 & uncertain photometry + u28 & 17:40:38.904 @xmath9753:39:51.09 & 0.79 & & 1 & agn & ? & & 22.72 & 24.73 & 25.44 & + u31 & 17:40:34.202 @xmath9753:41:15.28 & 1.41 & cv11@xmath98 & 1 & cv & f & & 23.32 & 23.93 & 26.07 & + u41 & 17:40:45.008 @xmath9753:39:55.21 & 0.70 & & 1 & ms & c & c & 16.74 & 17.69 & 19.50 & only object in error circle + u42 & 17:40:43.059 @xmath9753:38:31.29 & 1.96 & v26 & 1 & ab & ? & & 15.52 & 17.47 & 18.20 & very red + u43 & 17:40:40.543 @xmath9753:40:22.79 & 0.26 & by - pc-4 & 1 & ab & c & c & 17.68 & 19.03 & 20.38 & + u60 & 17:40:47.807 @xmath9753:41:28.40 & 1.30 & cv9 & 1 & cv & c & c & 23.38 & 23.63 & 26.04 & + u61 & 17:40:45.223 @xmath9753:40:28.60 & 0.45 & cv12@xmath98 & 1 & cv ? & ? & & 22.08 & 23.19 & 24.19 & uncertain photometry + u62 & 17:40:30.422 @xmath9753:39:17.47 & 2.11 & & 1 & ab ? & & & 20.90 & 23.33 & 23.17 & uncertain photometry + u63 & 17:40:31.663 @xmath9753:38:46.36 & 2.31 & & 1 & ab ? & & & 18.15 & 20.39 & 20.72 & very red , moderate excess + u65 & 17:40:37.558 @xmath9753:39:17.85 & 1.36 & & 1 & ab ? & ? & & 18.72 & 21.27 & 21.11 & very red , large excess + u66 & 17:40:38.918 @xmath9753:38:49.80 & 1.71 & & 1 & ab ? & f & & 21.93 & 24.56 & 24.29 & red , large excess + u67 & 17:40:40.065 @xmath9753:40:16.59 & 0.37 & & 1 & ab & c & c & 19.34 & 21.17 & 22.00 & + u68 & 17:40:40.730 @xmath9753:38:32.63 & 1.95 & & 1 ? & ? & & & & & & marginal detection in @xmath0 only + u69 & 17:40:40.867 @xmath9753:40:17.17 & 0.27 & by - pc-5 & 1 & ab & c & c & 17.84 & 19.03 & 20.52 & + u70 & 17:40:41.693 @xmath9753:40:33.33 & 0.11 & v20 & 1 & ab , bs & c & c & 15.50 & 16.21 & 18.30 & + u73 & 17:40:42.681 @xmath9753:39:28.73 & 1.00 & & 1 & ab & c & c & 18.19 & 19.47 & 20.88 & + u75 & 17:40:43.666 @xmath9753:40:30.61 & 0.22 & & 1 & ab & c & & 19.71 & 21.78 & 22.32 & + u76 & 17:40:43.818 @xmath9753:41:16.32 & 0.83 & v17 & 1 & ab & c & c & 15.81 & 16.81 & 18.57 & + u77 & 17:40:44.125 @xmath9753:42:11.48 & 1.74 & v36 & 1 & ab ? & & & & & 24.21 & outside of @xmath0 field + u79 & 17:40:46.409 @xmath9753:40:04.05 & 0.75 & & 1 & ab & ? & & 19.82 & 21.98 & 22.41 & + u80 & 17:40:46.455 @xmath9753:41:56.50 & 1.60 & cv14@xmath98 & 1 & cv ? & & & 25.40 & 25.84 & 27.91 & weak detection + u81 & 17:40:46.481 @xmath9753:41:15.44 & 1.01 & v14 & 1 & ab & c & c & 18.62 & 20.27 & 21.29 & + u82 & 17:40:48.537 @xmath9753:39:39.53 & 1.25 & by - wf2 - 1 & 1 & ab & c & & 18.72 & 20.29 & 21.39 & + u83 & 17:40:49.615 @xmath9753:40:43.02 & 1.13 & cv15@xmath98 & 1 & cv ? & ? & ? & 26.22 & 27.18 & 28.56 & very faint + u84 & 17:40:54.807 @xmath9753:40:19.79 & 1.88 & & 1 & ab ? & ? & ? & & 17.27 & 18.23 & very red , outside of @xmath0 field + u86 & 17:40:37.473 @xmath9753:41:47.24 & 1.48 & & 1 & ab ? & ? & & 21.16 & 24.06 & 23.39 & very red , large excess + u87 & 17:40:42.877 @xmath9753:40:26.45 & 0.11 & by - pc-3 & 1 & ab & c & c & 17.46 & 18.69 & 20.14 & + u88 & 17:40:42.863 @xmath9753:40:23.43 & 0.13 & by - pc-1 , v19 & 1 & ab & c & c & 16.73 & 17.77 & 19.43 & + u89 & 17:40:43.613 @xmath9753:40:24.60 & 0.23 & & 1 & ab & c & c & 19.76 & 21.82 & 22.38 & + u90 & 17:40:41.779 @xmath9753:40:14.42 & 0.24 & by - pc-6 & 1 & ab & c & c & 18.45 & 19.88 & 21.11 & + u91 & 17:40:42.430 @xmath9753:40:41.65 & 0.22 & & 1 & ms & c & & 18.94 & 20.44 & 21.63 & only object in error circle + u92 & 17:40:43.916 @xmath9753:40:35.39 & 0.28 & by - wf4 - 2 , v7 & 1 & ab ? & c & c & 16.72 & 18.26 & 19.53 & very red , slight deficit + u93 & 17:40:42.393 @xmath9753:40:46.62 & 0.30 & & 1 & ? & & & & & 24.95 & uncertain photometry + u94 & 17:40:42.868 @xmath9753:40:49.07 & 0.36 & & 1 & ab & c & c & 18.88 & 20.58 & 21.53 & + u95 & 17:40:40.320 @xmath9753:40:44.58 & 0.38 & & 1 & ab ? & c & & 20.81 & 23.02 & 23.41 & ms color , excess + u96 & 17:40:39.097 @xmath9753:40:23.09 & 0.47 & v24 & 1 & ab & c & & 18.40 & 20.15 & 21.42 & + u97 & 17:40:43.918 @xmath9753:40:05.90 & 0.46 & & 1 & ? & c & & 20.90 & 22.88 & 23.56 & blue , normal + u98 & 17:40:40.994 @xmath9753:40:58.40 & 0.53 & & 1 & ab & c & & 19.18 & 21.09 & 21.82 & + u99 & 17:40:46.431 @xmath9753:40:30.40 & 0.63 & & 1 & ab & c & & 18.49 & 19.96 & 21.16 & + u100 & 17:40:38.201 @xmath9753:40:46.55 & 0.66 & & 1 & ab & c & & 20.81 & 23.14 & 23.40 & + u101 & 17:40:45.399 @xmath9753:41:01.40 & 0.73 & & 1 & ab ? & c & c & 18.48 & 19.94 & 21.19 & red , normal + u102 & 17:40:38.845 @xmath9753:39:43.12 & 0.90 & & 1 & ab & c & & 21.27 & 23.80 & 23.87 & + u103 & 17:40:35.698 @xmath9753:40:12.56 & 1.00 & & 1 & ab & c & & 19.26 & 21.05 & 21.93 & + u104 & 17:40:43.124 @xmath9753:39:29.04 & 1.01 & & 0 & & & & & & & empty error circle + u105 & 17:40:36.521 @xmath9753:41:07.85 & 1.06 & & 1 & ab & ? & & 19.78 & 21.84 & 22.37 & + u106 & 17:40:43.737 @xmath9753:39:17.52 & 1.21 & & 0 & & & & & & & empty error circle + u107 & 17:40:34.115 @xmath9753:40:17.01 & 1.21 & & 1 & ab & c & & 19.61 & 21.63 & 22.23 & + u108 & 17:40:52.099 @xmath9753:39:48.25 & 1.62 & & 1 & glx ? & & & & & & extended object + u109 & 17:40:52.728 @xmath9753:40:52.88 & 1.61 & & 1 & ab ? & c & c & 19.77 & 21.83 & 22.44 & red , normal + u110 & 17:40:33.455 @xmath9753:39:16.83 & 1.76 & & 1 & ab & c & & 19.92 & 22.04 & 22.53 & + u111 & 17:40:29.845 @xmath9753:40:26.99 & 1.82 & & 0 & & & & & & & empty error circle + u112 & 17:40:50.374 @xmath9753:39:06.00 & 1.84 & & 1 & ms & f & & 19.60 & 21.25 & 22.29 & only object in error circle + u113 & 17:40:42.764 @xmath9753:40:20.76 & 0.16 & & 0 & & & & & & & empty error circle + u114 & 17:40:43.469 @xmath9753:40:34.34 & 0.21 & & 1 & ? & c & & 19.49 & & 22.10 & uncertain photometry + u116 & 17:40:42.236 @xmath9753:40:19.97 & 0.18 & & 1 & ab & c & c & 18.18 & 19.49 & 20.88 & + u117 & 17:40:42.153 @xmath9753:40:25.56 & 0.14 & & 1 & ab & c & c & 16.82 & 17.84 & 19.55 & + u118 & 17:40:41.576 @xmath9753:40:15.88 & 0.05 & & 1 & msto & c & c & 15.60 & 16.60 & 18.34 & only object in error circle + u119 & 17:40:41.261 @xmath9753:40:19.35 & 0.22 & & 0 & & & & & & & empty error circle + u120 & 17:40:46.517 @xmath9753:40:15.64 & 0.21 & & 1 & ab ? & ? & ? & 21.09 & 23.70 & 23.39 & moderately red , large ha excess + u121 & 17:40:33.631 @xmath9753:39:34.96 & 0.68 & & 0 & & & & & & & empty error circle + u122 & 17:40:47.903 @xmath9753:39:24.83 & 1.55 & & 0 & & & & & & & empty error circle + u123 & 17:40:49.621 @xmath9753:38:45.93 & 2.04 & & 0 & & & & & & & empty error circle + |
we present kinematic and photometric evidence for the presence of seven candidate tidal dwarf galaxies in stephan s quintet . the central regions of the two most probable parent galaxies , n7319 and n7318b , contain little or no gas whereas the intragroup medium , and particularly the optical tails that seem to be associated with n7318b are rich in cold and ionized gas . two tidal - dwarf candidates may be located at the edge of a tidal tail , one within a tail and for four others there is no obvious stellar / gaseous bridge between them and the parent galaxy . two of the candidates are associated with hi clouds , one of which is , in addition , associated with a co cloud . all seven regions have low continuum fluxes and high h luminosity densities ( f(h ) = 1 60 10 erg s ) . their magnitudes ( m 16.1 to 12.6 ) , sizes ( 3.5 h kpc ) , colors ( typically ) and gas velocity gradients ( 8 26 h km s kpc ) are typical for tidal dwarf galaxies . in addition the ratios between their star formation rates determined from h and from the b band luminosity are typical of other tidal dwarf galaxies . the masses of the tidal dwarf galaxies in stephan s quintet range from 2 10 to 10 m and the median value for their inferred mass - to - light ratios is 7 m/l . at least two of the systems may survive possible `` fall - backs '' or disruption by the parent galaxies and may already be or turn into self - gravitating dwarf galaxies , new members of the group . epsf.sty | astro-ph0101226 | optical and hi studies of interacting galaxies ( e.g. hibbard & van gorkom 1996 , duc & mirabel 1998 ) have shown that dwarf galaxies may be produced during galactic collisions . duc & mirabel ( 1998 ) have presented an especially convincing case of ngc 5291 , where more than ten star - forming high - metallicity dwarf galaxies may have been formed in a recent merger . these newly formed tidal - dwarf galaxies may be good sites to study galaxy formation in the nearby universe . although the concept of the formation of self - gravitating objects in tidal tails is an old one , considered by zwicky as early as in 1956 , it is only recently that these objects have received more attention and have been searched for systematically . hi observations have so far been the principal means used in searches of tidal dwarf galaxies since these are usually gas - rich systems . hi maps can reach large radii but have the disadvantage of being ( mostly ) of too poor spatial resolution for a direct investigation of the velocity field of the spatially - small and low - velocity - dispersion objects ( typically 5 - 20 arcsec in extent , and velocity gradients of 2 20 km s@xmath3 arcsec@xmath3 ; duc , 1995 ) . the study of the kinematics of the _ ionized _ gas of a galaxy , when it is present , with high spatial and spectral resolution velocity maps , can alternatively provide a useful tool to probe the nature of the candidate dwarf galaxies at small scales , complementing the lower - resolution radio studies . this paper presents new data on the kinematics of the ionized gas for seven star - forming tidal dwarf galaxy candidates and many other giant hii regions of stephan s quintet . combining fabry perot h@xmath0 maps with deep b and r images , we study the photometric and kinematic properties of 23 emission - line regions of the group . stephan s quintet ( arp 319 , vv288 , hcg 92 ) has been the subject of studies in x - rays , radio continuum , hi and h@xmath0 . most of its gaseous material is not concentrated around the bright galaxies but in the intragroup medium , suggesting that collisions among the group members may have taken place . arp ( 1973 ) was the first to obtain an h@xmath0 image of the group and to identify hii regions around galaxies n7318a , b , n7319 and n7320 . as part of an imaging survey of compact groups , in the r band , hunsberger et al . ( 1996 ) identified 27 tidal dwarf galaxy candidates possibly associated with n7318a , b and n7319 . one of these regions was observed by xu et al . ( 1999 ) to be a bright infrared source , when observed with the iso satellite ( their source a ) . xu et al . ( 1999 ) also presented h@xmath0 + [ nii ] images of the group highlighting the spatial overlap of the two emission clouds of ionized gas in the intragroup medium centered around 5700 km s@xmath3 and 6700 km s@xmath3 . a similar result was found by plana et al . ( 1999 ) based on fabry - perot h@xmath0 maps . an interesting possible scenario for the interaction history of stephan s quintet has been put forward by shostak et al . ( 1984 ) and refined by moles et al . ( 1997 , 1998 ) . the scenario is based on the analysis of optical images of the quintet , spectroscopy of a number of systems ( moles et al . 1998 ) and hi maps which show three velocity components associated with the group , at @xmath6 5700 km s@xmath3 , 6000 km s@xmath3 and 6700 km s@xmath3 ( balkowski et al . 1973 , allen & sullivan 1980 , shostak et al . 1984 ) . in their scenario , one or more major collisions occurred between n7320c and n7319 ( at velocities @xmath6 6000 km s@xmath3 and @xmath6 6700 km s@xmath3 respectively ) , which may have resulted in the removal of the hi gas from the inner parts of n7319 . more recently , stephan s quintet would have received a new member , n7318b ( it has a radial velocity of 5774 km s@xmath3 while the other core - group members , n7317 , n7318a and n7319 , have a mean velocity of 6664 km s@xmath3 , hickson et al . n7318b would be entering the group for the first time and it would be colliding with the intragroup medium ( moles et al . we identify a number of possible tidal dwarf galaxy candidates most probably formed as a result of the ongoing collision that involves n7318b . only one of the candidates studied here may be associated with the collision n7320c n7319 and hence the hi cloud removed from the center of n7319 . an alternative possibility is that the latter originated from n7318a ( see sections 4.7 and 4.8 ) . all the regions we measure in this work have been previously identified as hii regions by arp ( 1973 ) . a number of them have also been listed by hunsberger et al . ( 1996 ) as tidal dwarf galaxy candidates and gallagher et al . ( 2000 , using hst images ) as massive cluster candidates . however , these studies did not have velocity information . radial velocity information for some of the regions is summarized in table 1 of plana et al . determination of the velocity gradients within the regions is the main contribution of the present work . the organization of this paper is as follows . section 2 describes the observations and reduction procedure . in section 3 , we present the new photometric and kinematic results . in section 4 , we discuss the results and section 5 summarizes the paper . throughout this paper h@xmath9 is the hubble constant in units of h@xmath13 = 75 km s@xmath3 mpc@xmath3 . we assume that all the regions are at the same distance , and the differences in cz are just kinematical . for the figures and the determination of the star - formation rates we assumed a distance to the group of 80 mpc , for an average velocity of 6000 km s@xmath3 and a hubble constant of 75 km s@xmath3 mpc@xmath14 . the observations were carried out with a scanning fabry - perot instrument mounted on the canada - france - hawaii 3.6 m telescope ( cfht ) and the russian sao 6 m telescopes . the observations and characteristics of the set - ups are summarized in table 1 . during the cfht run we mostly detected sources with radial velocities v@xmath15 @xmath6 5700 km s@xmath3 , since we used a narrow - band filter that cut the higher velocities . we , therefore , did not detect the 6700 km s@xmath3 velocity component of the group . the filter used during the observations at the russian sao 6 m telescope allowed the detection of a much broader range of wavelengths and both velocity components were detected . however , the data are of much lower signal - to - noise and spectral and spatial resolution than the cfht data . llll observations & telescope & cfht 3.6 m & sao 6 m + & equipment & mos / fp @ cassegrain & cigale @ primary + & date & august , 23th 1996 & august , 5th 1991 + & seeing & @xmath16 1 `` & @xmath6 1.2 '' + interference filter & central wavelength & 6697 & 6726 + & fwhm & 17 & 50 + & transmission ( maximum ) & 0.7 & 0.55 + calibration & neon comparison light & @xmath17 6598.95 & @xmath17 6598.95 + perot fabry & interference order & 1162 @ 6562.78 & 796 @ 6562.78 + & free spectral range at h@xmath0 & 265 km s@xmath3 & 380 km s@xmath3 + & finesse at h@xmath0 & 12 & 12 + & spectral resolution at h@xmath0 & 27344 at the sample step & 18750 + sampling & number of scanning steps & 24 & 24 + & sampling step & 0.24 ( 11 km s@xmath3 ) & 0.35 ( 16 km s@xmath3 ) + & total field & 440@xmath18440 ( 512@xmath18512 px@xmath19 ) & 170@xmath18170 ( 256@xmath18256 px@xmath19 ) + & pixel size & 0.86 & 0.67 + detector & & stis 2 ccd & ipcs + exposures times & total exposure time & 1.2 hours & 0.8 hour + & each scanning exposure time & 180 s & 20 s per channel + & total exposure time per channel & 180 s & 120 s + reduction of the data was performed using the cigale / adhoc software ( http://www-obs.cnrs-mrs.fr/adhoc/adhoc.html ) . the data reduction procedure has been extensively described in amram et al . ( 1995 , 1996 ) , plana et al . ( 1998 ) and references therein . wavelength calibrations were obtained by scanning the narrow ne 6599 line during the observations . the relative velocities with respect to the systemic velocities are very accurate , with an error of a fraction of a channel width ( @xmath20 ) over the whole field . the signal measured along the scanning sequence was separated into two parts : ( 1 ) a constant level produced by the continuum light in a narrow passband around h@xmath21 ( continuum map ) , and ( 2 ) a varying part produced by the h@xmath21 line ( monochromatic map ) . the continuum level was taken to be the mean of the three faintest channels . the monochromatic map was obtained by integrating the monochromatic profile in each pixel . the cfht data were roughly calibrated in flux using a common galaxy also observed in a previous run when a flux calibrator was observed ( amram et al . we estimate an error of 30% in the flux calibration . the h@xmath0 profiles were measured to a minimum flux density of 4.9 @xmath1 10@xmath22 erg s@xmath3@xmath4 arcsec@xmath14 . we were not able to flux calibrate the data obtained at the sao 6 m telescope for lack of calibrators . images of stephan s quintet in b ( 5 x 900s ) and r ( 6 x 350s ) were obtained with the cfht and the subarcsec imaging instrument . the values of seeing on the final images were 0.8 and 0.6 arcsec respectively . preprocessing of all frames through flat - fielding was carried out using iraf tasks . the individual frames were registered and combined with a clipping scheme that eliminated one lowest and one highest pixel values in each registered stack to exclude cosmic rays , hot pixels and low pixels . the program sextractor ( bertin & arnouts 1996 ) was used to obtain the photometry of all objects in the b and r images with a sky `` meshsize '' of 128 pix . `` automatic aperture magnitudes '' were obtained , which give total magnitudes for the objects ( based on the kron s , 1980 , first - moment algorithm ) . calibration to the standard system for the b and r band observations was done via observations of standard stars from the list of landolt ( 1992 ) and updated magnitudes for the kpno consortium fields in m92 and ngc 7006 . the transformations from instrumental to standard systems were made using the following : @xmath23 = @xmath24 0.015(@xmath25 ) @xmath26 @xmath27 and @xmath28 = @xmath29 0.047(@xmath25 ) @xmath26 @xmath30 . the final combined images of h92 can be obtained upon request . r map was obtained after the images were calibrated and their seeing profiles were matched ( shown in fig . the h@xmath0 contours superimposed on the b r image of stephan s quintet is shown in fig . 1a . the 23 emitting regions detected by plana et al . ( 1999 ) are marked . dark is blue and white is red . the letters `` l '' and `` h '' are used to identify the systems for which we have an unambiguous determination of their radial velocities . the ones marked with `` l '' belong to the low - velocity system associated with n7318b , with velocities of 57006000 km s@xmath3 and the only system marked with `` h '' , region 6 , is the high - velocity system probably associated with n7319 , at a velocity of @xmath6 6700 km s@xmath3 . in fig . 1b the corresponding r image of the group ( after subtraction of a heavily smoothed sky ) is shown . the optical blobs corresponding to the h@xmath0 emitting regions are marked in this figure ( e.g. regions 1a and 1b correspond to the b and r counterparts of region 1 in h@xmath0 , see fig . all regions detected in h@xmath0 have counterparts on the b and r image . region 17 is faint , it is affected by dust , and it is too close to the galaxy n7318a . although it is not evident given the contrast of fig . 1b , it does have a counterpart on the b and r images . the b r image of the group shown in fig . 1a reveals some interesting features . the galaxy n7319 may contain several dust spots ( white regions in the figure , which are those with the reddest values of b r ) , particularly the one to the north of the galaxy corresponds to strong co emission ( gao & xu 2000 ) . we also note that the center of galaxy n7318a has a dust lane with a very peculiar shape , suggesting that this galaxy may have taken part in the interaction history of the group ( not contemplated in the scenario described by moles et al . the complicated dust lane structure of n7318a is much more impressive in the hst image of the group obtained by charlton et al . further evidence that galaxy n7318a may be in interaction is given by fig . 1c , which is discussed in section 4.7 . lcccccccc 1 & 20.7/21.9 & 14.2 & 0.76 & 3.1 & 1.5 & 0.23 & 2.2 & 9.0 x 8.4 2 & 20.6 & 14.0 & 0.53 & 2.3 & 28.7 & 4.26 & 12.2 & 34 @xmath1 17 3 & 20.3/20.9 & 14.7 & 0.55 & 3.3 & & & & 4 & 21.2 & 13.4 & 0.67 & 3.2 & & & & 5 & 19.6/20.7 & 15.3 & 0.47 & 3.9 & & & & 6 & 19.5 & 15.0 & 1.32 & 3.6 & & & 4.5 & 16.0 x 13.0 7 & 21.5 & 13.0 & 0.75 & 3.2 & 0.3 & 0.04 & 0.7 & 2.6 x 3.4 8 * & 21.4/22.3/21.2 & 14.2 & 0.79 & 3.0 & 2.5 & 0.38 & 2.2 & 13.7 x 7.5 9 & 23.2 & 11.4 & 1.00 & & 2.5 & 0.38 & 0.2 & 7.4 x 4.6 10 & 20.1 & 14.5 & 0.92 & 6.3 & 0.7 & 0.09 & 2.9 & 3.1 x 3.1 11 & 20.8 & 13.7 & 0.58 & 4.8 & 0.7 & 0.09 & 1.4 & 4.7 x 3.1 12 & 19.6 & 4.9 & 0.43 & 4.6 & 0.7 & 0.10 & 4.1 & 8.3 x 4.8 13 & 21.5 & 13.0 & 0.77 & & 2.2 & 0.32 & 0.7 & 8.6 x 5.6 14 & 19.3/21.2 & 15.4 & 0.94 & 6.3 & 8.1 & 1.20 & 6.5 & 9.2 x 6.7 15 & 20.8/20.4 & 14.7 & 1.00 & 4.7 & 9.6 & 1.42 & 3.4 & 10.8 x 9.3 16 & 21.1/21.6/22.0/21.5 & 14.3 & 0.86 & 5.4 & 3.2 & 0.46 & 2.4 & 3.3 x 3.1 17 & & & & & 2.7 & 0.40 & & 5.4 x 3.5 18 & 19.7/21.0 & 15.1 & 0.56 & 4.0 & 13.4 & 2.00 & 4.9 & 11.9 x 9.0 19 & 20.3/21.5 & 14.6 & 0.71 & 2.9 & 9.0 & 1.35 & 3.1 & 15.0 x 7.2 20 & 21.9 & 12.6 & 0.76 & 0.3 & 3.0 & 0.45 & 0.5 & 10.5 x 8.0 21 & 20.1 & 14.4 & 0.34 & 2.0 & 4.9 & 0.73 & 2.6 & 11.3 x 10.5 22 & 21.0/21.7/22.2 & 14.2 & 0.42 & 1.0 & 4.2 & 0.63 & 2.2 & 11.8 x 9.7 23 & 21.5 & 13.0 & 0.63 & 0.2 & 1.3 & 0.20 & 0.7 & 8.2 x 8.7 table 2 shows the main photometric parameters for all the 23 h@xmath0-emitting regions shown in figs . 1a and 1b ( with exception of a few entries for region 17 , for the reason explained previously ) . total b magnitudes ( `` automatic - aperture magnitudes '' given by sextractor ) and b r colors are listed in columns 2 and 3 . when one h@xmath0 region corresponds to more than one identification on the b and r images we list the apparent b magnitudes of the different blobs in a single entry , separated by slashes ( magnitudes and colors are corrected for galactic extinction only ) . r color was measured inside an aperture of 1.0 arcsec radius centered on each blob . the total absolute magnitudes given in column 3 correspond to the sums of the individual luminosities ( e.g. the absolute magnitude for region 8 , see fig . 1a , corresponds to the sum of regions 8a , 8b and 8c , see fig . the i d s marked with a `` c '' on table 2 correspond to the candidate tidal dwarf galaxies ( see section 3.3 ) . the magnitude and color of region 14a is contaminated by light from a red background galaxy that overlaps with this region . there is an object to the northwest of region 2 which also seems to be a background object , given its red color . we measure a very low continuum level ( on the fabry - perot map ) for several of the h@xmath0-emitting regions . since we were not able to perform calibration in flux of the continuum map ( for lack of calibrators ) , we determined the ratio of the total flux contained in each blob to the flux within an arbitrary aperture of radius 1.8 " centered on galaxy n7318b . this fraction ( in percentage ) is shown in column 5 of table 2 . this procedure is intended to show , in a relative way , which regions are the ones with brightest and faintest continuum emission . the regions with the lowest fluxes are regions 20 and 23 . they have @xmath6 10 times less flux in the continuum than the median for the remaining regions . regions 9 , 13 and 17 could not be measured due to contamination from the bright galaxies n7318a , b . we show color gradients for each emission - line blob in fig . the ( b r)@xmath13 ( corrected for galactic extinction ) profiles in general become bluer toward each hii region , despite the fact that the r - band flux is contaminated by the h@xmath0-line emission . we can see from column 4 of table 2 that all but one emitting region detected in h@xmath0 are blue ( the median value of the color is 0.7 ) . the exception is region 6 , with a b r color of 1.3 . plana et al . ( 1999 ) detected 23 h@xmath0 emitting regions around the n7318a / b system . for seven of them ( six systems and one complex with four regions ) a velocity field is derived and shown in figs . 3a to 3 g . these correspond to the objects which show rotation . the velocity fields were used to derive the rotation curves shown in fig . 4 . in all cases the velocities were measured within @xmath31 of the major axis in the plane of the sky , except for regions 25 and 6 , where a cone of @xmath32 was used instead . the rotation curves were obtained using the position angle and inclination listed in columns 2 and 3 of table 3 . the position angle of the major axis ( given in column 2 ) is taken to be the axis of symmetry of the main kinematic body of each galaxy . the center of the velocity field is whenever possible chosen to be a point along the major axis which makes the rotation curve symmetric , with similar amplitudes for the receding and approaching sides ( this is , however , not always possible , e.g. region 6 ) . the values of inclination with respect to the plane of the sky ( given in column 3 ) are determined from the velocity fields of the candidates . specifically we used a least square s fit program which gives the best value for the inclination fixing the position angle , the central systemic velocity and the kinematic center and by fitting a classical rotation velocity model for data points inside crowns at increasingly growing distances from the center . in the top - left panels of figs . g we show the h@xmath0 contours determined from the cfht fabry - perot maps , superimposed on the deep b - band image of the galaxy ( see also plana et al . 1999 ) . for region 6 the h@xmath0 map is derived from the sao 6 m data ( since its velocity was outside the filter range for the cfht observations ) . the lower - left panels show the continuum - subtracted emission - line profiles . these have been smoothed spectrally by a gaussian function having a fwhm of three channels ( or 35 km s@xmath3 ) and spatially ( 3x3 pixels ) , except for fig . 3a ( see below ) . each pixel represents 0.86 `` on the sky . although the pixel size for the image of region 6 was originally 0.67 '' , it was degraded to 0.86 " to match the cfht data . in the right panels we show the velocity fields of the galaxies superimposed on the corresponding monochromatic h@xmath0 images . we mark the absolute value for each isovelocity for regions 25 , 6 and 21 , but for regions 8 , 20 , 22 and 23 , only relative velocities are marked , since we do not know the systemic velocities of these regions . we classify systems 25 , 6 , 8 , 20 , 21 , 22 and 23 as tidal dwarf galaxy candidates on the basis of the clear signature of rotation , above a cut - off value of 8.0 h@xmath9 km s@xmath3 kpc@xmath3 ) . examples of non - rotating structures , possible associations of giant hii regions , are regions 9 , 10 , 11 , 14 , 15 , 18 and 19 . we see continuous profiles between regions 9 to 11 ( the data have s / n @xmath33 20 in between the emission blobs ) showing that these are connected . these regions are most probably attached to the southern tail of n7318b . no velocity gradients are detected within them . regions 14 and 15 are attached to one or more tidal tails ( these regions correspond to double velocity components , plana et al . even if these complexes have h@xmath0 fluxes , sizes and colors consistent with their being tidal dwarf galaxies ( like the ones studied by duc 1995 ) , we detect no velocity gradient within the objects . in the same way , regions 18 and 19 also show bright h@xmath0 emission with no velocity gradients . these are very close to n7318a but , as indicated by their radial velocities ( @xmath6 5700 km s@xmath3 , plana et al . 1999 ) , they may belong to n7318b . regions 18 and 19 do not have any obvious contact to a tidal arm . region 1 is isolated . we detected no velocity gradient within the region but the s / n of the data may be too low to allow a gradient measurement . we summarize below the properties of each object that we classify as candidate tidal dwarf galaxies . * regions 25 : * the complex formed by regions 25 has a radial velocity of 6020 km s@xmath3 ( velocity obtained by moles et al . 1998 for region 2 ) and it shows an h@xmath0 diameter ( full width at zero intensity ) of @xmath6 34 " ( corresponding to a radius of r=6.5 h@xmath7 kpc ) . in order to increase the s / n ratio we performed a spatial smoothing of 3x3 pixels in the center and 5x5 pixels outside the inner region . since this complex is large , smoothing does not affect the results . we can see on the lower - left panel of fig . 3a that the complex formed by regions 25 is connected by continuous velocity profiles . although with a much lower s / n , the effect is also measurable in the unbinned data . this suggested to us that these regions may form one system , perhaps within a tidal tail . 4a shows the corresponding rotation curve . the approaching and receding sides of the rotation curve differ in shape and amplitude . the curve is increasing regularly , reaching the average maximum velocity of 50 km s@xmath3 . as the monochromatic map shows , we have at least four emitting regions and therefore the velocity field and rotation curve may be averages of internal motions of these various regions . this could explain , at least partially , the disturbances of the velocity field . the r image in fig . 1b shows at least six bright condensations . the system as a whole has no defined center . the conditions needed for the complex of regions 25 to become an independent galaxy are discussed in sections 4.1 and 4.6 . we conclude that there is little chance that this region will survive as a single object . it could , however , form several condensations that could then be stable . we point out that there is an hi cloud encompassing the region with velocity similar to the optical velocity measured from the h@xmath0 emission ( @xmath6 6000 km s@xmath3 , williams et al . 1999 ) . * region 6 : * region 6 has a radial velocity of 6680 km s@xmath3 ( moles et al . it shows a disturbed and fairly extended velocity curve out to a radius of r=3.5 h@xmath7 kpc . the profiles shown in the lower - left panel of fig . 3b indicate a larger velocity dispersion ( i.e. they have larger widths ) than for the other regions ( @xmath6 115 km s@xmath3 compared to 80 km s@xmath3 for the others ) . a larger dispersion is expected when there is , for example , a second component blended with the main velocity component . however , in the case of region 6 , the components are not resolved and any attempt to decompose them would be arbitrary . as can be seen in fig . 3b , some of the profiles do show double peaks but no consistent solution for a moving second component could be found , perhaps due to low s / n . the rotation curve presented in fig . 4b is peculiar . there is complete disagreement between the amplitudes of the receding and the approaching sides . the approaching side has approximately constant velocity at @xmath6 10 km s@xmath3 while for the receding side the velocity reaches 90 km s@xmath3 . the maximum rotational velocity listed in table 4 is taken to be the average between both sides . the mass estimate obtained from this number is therefore very uncertain . williams et al . ( 1999 ) have detected an hi cloud at the position of region 6 with similar velocity to that measured from the h@xmath0 emission ( @xmath6 6700 km s@xmath3 ) . it was also detected in co by gao & xu ( 2000 ) . in addition , this region was detected with iso by xu et al . ( 1999 ) and it was found to be a very bright source in the infrared . this could therefore be a peculiar object , given its infrared luminosity and its disturbed rotation curve . * region 8 : * it shows an extent of r @xmath6 4.5 h@xmath7 kpc . the velocity field is quite regular . 4c shows the rotation curve rising up to an average maximum velocity of 115 km s@xmath3 in 12 arcsec ( @xmath6 4.5 h@xmath7 kpc ) . this is the largest maximum rotational velocity we detect for a candidate tidal dwarf galaxy . interestingly enough , this is the only tidal dwarf galaxy that seems to be embedded in a tidal tail and it is the one closest to the parent galaxy . nonetheless the very high velocity gradient measured along a very short path and the optical condensations coincident with the emission - line region motivated us to list this as a possible candidate . * region 20 : * it has an extent of r @xmath6 3.2 h@xmath7 kpc . as the profiles of fig . 3d show , the emission is centrally concentrated . the rotation curve presented in fig . 4d is regular except outside a radius of 2 h@xmath7 kpc , where we can see a discrepancy between the approaching and the receding sides . the average maximum velocity is 30 km s@xmath3 . the r image in fig . 1b shows a single center and no bridge to the parent galaxy ( possibly n7318b ) . * region 21 : * it has an extent of r @xmath6 3.5 h@xmath7 kpc . the velocity field ( fig . 3e ) does not show perturbed isovelocities , but the position angle changes along the radius . the rotation curve , shown in fig . 4e , is derived along a mean position angle of the major axis of @xmath6 80 . the approaching side of the rotation curve has a normal behavior , with velocities increasing up to 65 km s@xmath3 . for the receding side , due to the position angle change , the velocities are scattered beyond 2 h@xmath7 kpc . the r image in fig . 1b shows two condensations , the one most to the west being the brightest . * region 22 : * it shows an extent of r @xmath6 3 h@xmath7 kpc . the velocity field is disturbed . a second component may also be present in this case , as suggested by the profiles of fig . we were not able , however , to do a satisfactory decomposition of the two components . the rotation curve , shown in fig . 4f , presents a strong rise in the first 2 `` , a decrease in velocities after that and then a new increase around r @xmath6 6 '' . the error bars are very large for radii beyond 6 " indicating a higher scatter in the velocities . three condensations are seen in the optical image , the ones most to the east being the brightest . it is interesting to note that the blueshifted isovelocities of region 22 point towards the blueshifted isovelocities of region 21 . this suggests the velocity gradients measured are probably intrinsic to the regions and are not larger scale motions within a possible faint tidal tail that joins the regions . * region 23 : * it has an extent of r @xmath6 2 h@xmath7 kpc . the velocity field ( fig . 3 g ) does not show disturbed isovelocities , but the rotation curve ( fig . 4 g ) shows that the velocities for the approaching side increase in the first 2 " and then decline while the velocities for the receding side increase with radii . the receding and the approaching sides are therefore not in agreement . the rotation curves for four of the galaxies rise linearly with radius , without reaching a significant plateau . for two dwarf galaxy candidates , regions 22 and 23 , the shape of the curve is disturbed . for region 6 , the shapes of the approaching and receding sides are very different . the rotation curves presented here can be compared with those for other tidal dwarf galaxies and blue compact galaxies . this is done in section 4.2 . table 2 gives h@xmath0 fluxes for each emitting region , derived from the calibrated monochromatic map , measured above a threshold of 4.9 @xmath1 10@xmath22 erg s@xmath3 @xmath4 arcsec@xmath14 , within apertures of diameters given in column 9 ( defined as the diameter of the aperture of zero intensity ) . for region 6 no flux is listed given that the data for this region is uncalibrated ( see section 2.1 ) . the current star formation rate can be derived from the formula sfr(h@xmath34 ) . this formula is valid for 0.1 100 m@xmath35 ( hunter & gallagher 1986 ) . the h@xmath0 fluxes given in column 6 of table 2 were transformed to luminosities for use in the formula above integrating over the sphere of space , assuming spherical geometry and a distance to the group of 80 mpc . in contrast to the h@xmath0 luminosity , the b luminosity is more sensitive to the older stellar population of a system ( larson & tinsley 1978 ) . using the formula proposed by gallagher & hunter ( 1984 ) it is possible to derive the star formation rate averaged for a period of 10@xmath36 years : sfr(l@xmath37 ) = 0.29 @xmath1 10@xmath38 ( @xmath39 ) . the ratio sfr ( l@xmath37 ) / sfr @xmath40 gives a rough measure of the evolution of the star - formation rate inside the emission - line region during the last billion years . the star formation rates are listed in columns 7 and 8 of table 2 and are discussed in section 4.3 . lccccccccc 25 & 125 & 55 @xmath41 10 & 6.2 & 6.5 & 23 & 50 & 38 & 6 & 123 6 & 130 & 60 @xmath41 15 & 2.7 & 3.5 & 24 & 50 & 20 & 7 & 12 8 & 5 & 60 @xmath41 5 & 1.9 & 4.5 & 13 & 115 & 138 & 73 & 225 20 & 70 & 65 @xmath41 10 & 0.4 & 3.2 & 23 & 30 & 6.7 & 17 & 15 21 & 80 & 60 @xmath41 5 & 1.4 & 3.5 & 27 & 55 & 24 & 17 & 12 22 & 75 & 55 @xmath41 10 & 0.9 & 3.0 & 24 & 25 & 4.3 & 5 & 11 23 & 40 & 50 @xmath41 10 & 0.4 & 2.0 & 23 & 23 & 2.4 & 6 & 3.6 the fate of a tidal dwarf galaxy is basically driven by the ratio of its mass to the virial mass and the ratio of its mass to the so called tidal mass ( binney and tremaine 1987 , duc 1995 ) . the virial mass condition tells us if the tidal dwarf candidate is massive enough to be gravitationally stable against internal motions . the tidal mass condition will tell us if it is massive enough to survive the tidal forces exerted by the parent galaxy . in this work we only estimate the tidal mass since for the virial mass we would need a measurement of the internal velocity dispersion of the galaxy which we do not have . in table 3 we give , for each dwarf - galaxy candidate , in ( 1 ) the region identification number , ( 2 ) the position angle ( 3 ) , the inclination ( 4 ) the total luminosity of the region , in the b band , measured inside an aperture of size similar to that within which the ionized gas was measured , ( 5 ) the radius @xmath23 that corresponds to the maximum rotational velocity ( except for region 22 , where @xmath23 is the radius within which emission was detected ) , ( 6 ) the distance to the parent galaxy , ( 7 ) the maximum rotational velocity ( derived from the rotation curves , shown in fig . 4 ) , ( 8) the total galaxy mass , ( 9 ) the mass - to - light ratio and ( 10 ) the tidal mass . the values given in columns ( 6 ) , ( 8) and ( 10 ) are based on the following assumptions : 1 ) the tidal dwarf galaxy candidate was drawn from material from galaxy n7318b ( or n7319 only for region 6 ) , 2 ) the dwarf candidate presents a velocity gradient due to rotation , it is virialized , and therefore a value for the mass can be derived from the rotation curve , and 3 ) galaxies n7319 and n7318b have masses of 1.3 and 1.8 @xmath1 10@xmath42 m@xmath35 respectively ( see below ) . * total masses of the tidal dwarf galaxy candidates : * we estimated the total mass of the candidate tidal dwarf galaxies from their rotation curves , derived from the velocity fields ( figs . 3a3f ) using the simple virial estimator ( lequeux 1983 ) in which the mass within a radius r of a rotating disk system is given by m(r)=f @xmath1 r @xmath1 v(r)@xmath19 @xmath1 g@xmath3 . here v(r ) is the rotational velocity at radius r , g the gravitational constant and f is a constant value between 0.5 and 1 . we chose a value of 1 for simplicity . m(r ) is a lower limit on the estimate of the mass , since most likely these regions are not pure rotators . random motions of the gas could provide significant dynamical support for the tidal dwarf galaxies as indicated by their line widths . as an approximation , if we assume that the gas is in equilibrium , and that the gas velocity ellipsoid is isotropic , a very rough estimate of the contribution from the pressure support could be made using an `` asymmetric drift correction '' following oort ( 1965 ) . for gaseous objects in formation , such as those present in the stephan s quintet , and under the above assumptions , a general rule may be : 1 ) for rotation curves with amplitudes v@xmath43 @xmath6 25 km s@xmath3 , i.e. for the most internal regions of the self gravitating structures or for the least massive objects , the contributions from the dispersion and the rotation components may be comparable ; 2 ) for v@xmath43 @xmath6 50 km s@xmath3 , the contribution of the dispersion velocity component may be approximately half of that for the rotational velocity component and 3 ) for v@xmath43 @xmath33 50 km s@xmath3 , i.e. at larger distances from the center of the objects and/or for more massive structures , the dispersion velocity may be negligible at first approximation . * m / l of the tidal dwarf galaxy candidates : * the mass - to - light ratios of the candidate dwarf galaxies were obtained by dividing the values in column 8 by those in column 4 of table 3 . the median value for the mass - to - light ratios of the tidal dwarf galaxy candidates is 7 @xmath44 , with a large scatter , with values as high as 74 @xmath44 for region 8 . the mass - to - light ratio of region 8 reflects its steep velocity gradient of @xmath6 26 h@xmath9 km s@xmath3 kpc@xmath3 , the highest observed among the tidal dwarf candidates observed ( see a discussion of this point in section 4.4 ) . * total masses of the progenitor galaxies : * in order to determine the tidal masses , an estimate of the total mass of the progenitor galaxies is necessary . as we were not able to measure the rotation curves of the parent galaxies n7319 and n7318b due to the lack of ionized gas in their disks , we estimated their masses using the tully - fisher relation ( m@xmath37 = 5.85 log ( 2 v@xmath43 ) 5.61 , pierini ( 1999 ) , the factor 2 within the log was included to match our definition of v@xmath43 to theirs ) . for absolute magnitudes of 21.4 and 21.3 and r@xmath45 of 10.4 and 13.2 h@xmath9 kpc for n7319 and n7318b respectively we obtain maximum velocities of 249 and 240 km s@xmath3 . using the virial estimator we then obtain masses of 1.3 and 1.8 @xmath1 10@xmath42 m@xmath46 . * tidal masses : * the tidal masses listed in column 10 of table 3 were derived from binney and tremaine ( 1987 ) , m@xmath47 = 3m(r / d)@xmath48 , where @xmath49 is the mass of the parent galaxy , @xmath23 the radius of the tidal dwarf and @xmath50 the distance to the parent galaxy . we assumed that the masses of the parent galaxies ngc 7318b ( for regions 2 - 3 - 4 - 5 , 8 , 20 , 21 , 22 and 23 ) and ngc 7319 ( for region 6 ) are 1.3 and 1.8 @xmath1 10@xmath42 m@xmath35 respectively , as calculated above . the results in table 3 show that the dwarf masses determined are in two cases larger than the tidal masses ( for regions 6 and 21 ) , suggesting that the regions considered may survive tidal forces . this is not the case , however , for the complex of regions 25 and for regions 8 , 20 and 22 , for which the tidal mass is larger than the total mass estimated . * escape velocities of the candidates : * we can also estimate the escape velocity of the tidal dwarf candidates and compare this velocity to the systemic velocity difference between the dwarf candidate and the parent galaxy . we have found that the velocity difference between region 21 and n7318b is of the order of the escape velocity calculated at the distance of the region ( 27 h@xmath3 kpc ) , indicating that it may be able to escape from the parent galaxy potential . this is not the case for regions 25 and 6 ( for the latter n7319 is the parent galaxy ) . for regions 8 , 20 , 22 and 23 we do not have an unique systemic velocity determination nevertheless we can give some estimate based on the possible systemic velocities ( see plana et al . for these four regions ( 8 , 20 , 22 and 23 ) , the minimum possible velocities are @xmath6 5935 , 5855 , 5975 and 5935 km s@xmath3 , respectively ( plana et al . 1999 , their table 1 ) . if we assume that ngc 7318b ( with velocity 5774 km s@xmath3 ) is the parent galaxy , then the differences between the velocity of the parent galaxy and that for each of the tidal dwarf candidates are 161 , 81 , 201 and 161 km s@xmath3 ( or larger ) for regions 8 , 20 , 22 and 23 respectively . on the other hand , the escape velocities are 344 , 259 , 253 and 259 km s@xmath3 respectively . therefore , if the minimum possible velocities are the true radial velocities , only region 22 of the four may be able to escape the potential of the parent galaxy , since for this region the escape velocity is similar to the velocity difference . we then conclude that only two out of the seven candidates may have velocities such that they can escape the gravitational field of the parent galaxy . just a few kinematic studies are available on tidal dwarf galaxies and dwarf galaxies in general . duc and mirabel ( 1998 ) showed velocity curves of dwarf galaxies around ngc 5291 measured from hi kinematics . the curves show in general stronger gradients than those measured for the galaxies in stephan s quintet . however , the diameter of the objects , as computed from the optical images , are similar . another set of velocity curves and rotation curves for dwarf galaxies comes from stlin et al . ( 1999 ) and stlin et al . their sample is composed of luminous blue compact galaxies ( hereafter bcg s ) . the radial extent of the rotation curves are comparable to those for the candidate tidal dwarf galaxies in stephan s quintet and the velocity amplitudes also seem to be similar . the shapes of the rotation curves , however , are very different . while in our sample the rotation curves show in general a typical shape for solid body rotation , in stlin et al . ( 1999 ) the rotation curves of the bcg s have flat plateaus . in stlin et al . ( 1999 , 2000 ) , the velocity fields are irregular and often contain secondary dynamical components although they in most cases display overall rotation . by comparison of the stellar masses ( by means of multicolor images and spectral evolutionary synthesis analysis ) to the total masses of the galaxies ( derived from the rotation curves ) , they showed that at least half of the galaxies can not be supported by rotation . they found that the morphologies and dynamics of the bcg s suggest that the starburst activity in these galaxies are most likely triggered by mergers of dwarf galaxies and/or massive gas clouds . in the case of the candidate tidal dwarf galaxies , it is possible that they are also partially supported by random motions as it is the case for the bcg s . moreover , the starburst activity in the tidal dwarf galaxies could also be nurtured by merging of gas clouds during the ongoing tidal process . hunter et al . ( 2000 ) have recently made a compilation of possible dwarf galaxies in many different environments which could have tidal origin . a comparison of the maximum rotation velocities as a function of the b absolute magnitudes of the tidal dwarf galaxy candidates in stephan s quintet and galaxies in hunter s sample is shown in fig . 5 . as can be seen , the velocities and magnitudes of the h92 s tidal dwarf candidates are within the ranges of parameters of other possible candidates . there are several classes of emission - line objects which have similarities in some respects with the objects we observed in stephan s quintet . a comparison of the star formation rates ( sfr ) for five samples of dwarf galaxies , one sample of hii regions in nearby spiral galaxies and our sample is shown in fig . we identified each of the galaxies in our sample with its corresponding i d number . included in the comparison are ( 1 ) a sample of irregular galaxies from the local group ( mateo 1998 ) , ( 2 ) a sample of isolated irregular galaxies of low luminosity and central surface densities , ( 3 ) a sample of blue compact galaxies studied by sage et al . ( 1992 ) with blue luminosities similar to the systems observed in stephan s quintet , ( 4 ) another sample of blue compact galaxies studied by stlin et al . ( 1999 ) , ( 5 ) a sample of tidal dwarf galaxy candidates around n5291 studied by duc and mirabel ( 1998 ) , and ( 6 ) a sample of hii regions in the spiral galaxies ngc 1365 , ngc 1566 , ngc 2366 , ngc 2903 , ngc 2997 , ngc 3351 , ngc 4303 , ngc 4449 , ngc 5253 studied by mayya ( 1994 , their table 1 , only objects with data quality 1 or 2 were included ) . 6 shows that the local group and isolated irregular galaxies have sfr(@xmath51 ) s that are on average lower than those for all the other samples . hii regions in nearby galaxies and blue compact galaxies have values for the sfr(@xmath51 ) in between those observed for isolated galaxies and those for tidal dwarf galaxies and giant hii regions of stephan s quintet . while in the sample of sage et al . ( 1992 ) the blue compact galaxies have a comparable range of luminosities to the galaxies in the other samples , stlin et al.s sample is dominated by blue compact galaxies with high b - luminosities . still , the sfr(@xmath51 ) for these two samples agree in the region of overlap . the faintest blue compact galaxies seem to overlap with the location of the figure occupied by the hii regions in nearby spiral galaxies , although the latter extend to much lower luminosities . the tidal dwarf galaxies studied by duc & mirabel ( 1998 ) have average sfr(@xmath51 ) s that are comparable to the objects studied in this paper . however , duc & mirabel s measurements were made through long slit observations and the values may be a lower limit on the h@xmath0 fluxes and therefore to the sfr(@xmath51 ) s derived for these galaxies . the sfr(@xmath51 ) s for the bcg s may be affected by dust , which would artificially lower the rates derived for these objects . therefore , in reality , the sfr(@xmath51 ) s for tidal dwarf galaxies and bcg s may be similar although these galaxies are fundamentally two distinct classes ( for instance they have completely different metallicities , with the bcg s being metal poor objects and the tidal dwarf galaxies having the high metallicities of their parent galaxies ) . one should also be aware that there exists a very large range of sfr(@xmath51 ) s for both the bcg s and the tidal dwarf galaxies . only a crude comparison between the sfr(@xmath51 ) s of the two populations is then possible . region 8 as well as the complex of regions 25 seem to be a part of a tidal tail . in addition , region 6 may be at the edge of a tidal tail , if we assume that it is associated with n7318a and not with n7319 ( see section 4.7 ) . one obvious concern is then that the observed velocity gradients for these regions could be a result of streaming motion in the tail within which the region is embedded . in the case of region 8 , if we assume that it is attached to n7318b ( at a velocity of @xmath6 5700 km s@xmath3 ) and it has a velocity of @xmath6 5900 km s@xmath3 ( j. sulentic , private communication ) , the gradient of the motion in the tidal tail ( lower velocities in the south and higher velocities in the north ) has opposite sense to what is observed in the region itself ( the red side of the region is to the south , see fig . this represents clear evidence that the strong velocity gradient observed for region 8 can not be due to streaming motions within the tidal tail . in addition , we note that there is a large misalignment of at least 30 degrees between the pa of the major axis of region 8 and that of the tidal tail within which it seems to be embedded ( see fig . this again indicates that region 8 can not be just a part of a tidal tail and it is , instead , an object with its own rotation pattern . in the case of the complex of regions 25 , we find that the velocity gradient we observed for it ( these regions are located at the edge of a tidal tail that emanates from the northern arm of n7318b ) could not be strongly affected by streaming motions for the following reason . for this complex we have information on the hi velocity field kindly provided by l. verdes - montenegro ( private communication ) . the motion observed in hi encompasses an area that goes from the base of the northern tail of n7318b to our region 1 and it has a very similar mean radial velocity ( @xmath6 6000 km s@xmath3 ) to that of the h@xmath0 emission in complex 25 . we compared the h@xmath0 and the hi velocity fields for regions 25 and we found that while the h@xmath0 gradient reaches 50 km s@xmath3 over a radius of 13 arcsec in extent , the hi gradient reaches only about 15 km s@xmath3 over the same area . assuming that the hi velocity gradient gives us an upper limit on the large - scale motion of the tidal tail ( upper limit because some internal motion is also expected to be present in the hi component ) the difference between the two gradients ( 35 km s@xmath3 ) is due to internal motions within the complex 25 . with this correction the new mass - to - light ratio of the complex of regions 25 would be 3 @xmath44 . in a scenario where region 6 were at the edge of the tidal tail that emanates from n7318a , there would be no concern about its velocity gradient being due to streaming motion since the center of the parent galaxy and region 6 have almost exactly the same velocity . we conclude that streaming motions within the tails may not significantly change the observed velocity gradients measured in this study . dubinski et al . ( 1999 ) find that a good criterion for making tidal tails during a collision appears to be that the ratio of the escape velocity ( v@xmath52 ) over the circular velocity ( v@xmath43 , in our notation ) of the galaxy , measured at a radius of 2r@xmath53 , must be lower than 2.5 ( r@xmath53 is the scale length of the galaxy ) . according to dubinski et al . ( 1999 ) this condition is valid for a wide range of models including those with disk - dominated and halo - dominated rotation curves . they also find that galaxies with declining rotation curves are the best candidates to produce tidal tails during collisions . we do not have information on the rotation curves of the galaxies ( due to their lack of ionized gas ) but we can still check the velocity ratio condition v@xmath54/v@xmath43 @xmath16 2.5 assuming that the galaxies follow the tully - fisher relation ( see section 4.1 ) . the masses inside 2r@xmath53 are 1.1 and 0.8 @xmath1 10@xmath42 for n7318b and n7319 respectively ( if we assume that the scale length of the galaxies r@xmath53 @xmath6 r@xmath45/3.2 ) . the ratio v@xmath54/v@xmath43 is , then , well below 2.5 for both galaxies , indicating that they may develop tidal tails during collisions . this is a comfortable confirmation since both galaxies obviously have developed one or more tidal tails . it is interesting to note that the inclination of the seven tidal dwarf candidates lie in between the values of 50and 65 , which are very similar to the inclinations of the progenitor galaxies of 59and 66(from the lyon - meudon extragalactic database , leda ) , for n7319 and n7318b respectively . if the candidate tidal dwarf galaxies are formed in the plane of the disks of the parent galaxies , they probably could keep the angular momentum of their progenitor disks . this inclination concordance would then be expected by the conservation of angular momentum of the system . however , the inclination found for the candidates could be biased by the fact that standard methods for deriving inclinations favor high values of inclination ( low inclination values are harder to be determined ) . in addition , since we do not have information on which side of the galaxy is pointing to us , the two inclined disks with 60inclinations could actually be separated by 120 ! lccccccc 25 & + & + & + & & & end of tail & + 6 & & + & + & + & & end of tail & + 8 & + & + & + & & & inside a tail & 20 & + & + & + & & & isolated & 21 & + & + & + & + & + & isolated & 22 & + & + & + & & + & isolated & 23 & + & + & + & & & isolated & table 4 summarizes the characteristics of each of the seven tidal dwarf candidates of stephan s quintet . a `` plus '' in an entry means ( 1 ) in column two if the color is bluer than b r = 0.8 , ( 2 ) in column 3 if the sfr(h@xmath0)/sfr(l@xmath37 ) ( see section 3.4 ) is typical of a tidal dwarf galaxy , ( 3 ) in column 4 , if the galaxy presents a velocity gradient larger than 8.0 h@xmath9 km s@xmath3 kpc@xmath3 ( this will be true for all of the objects since they were classified as tidal dwarf candidates based on the presence of the velocity gradients ) , ( 4 ) in column 5 if the mass of the galaxy as measured from the internal kinematics of the ionized gas is larger than the tidal mass obtained in section 4.1 , ( 5 ) in column 6 if the escape velocity of the parent galaxy is lower than the velocity difference between the systemic velocity of the tidal dwarf candidate and that of the parent galaxy ( assuming that the parent galaxy is n7318b for all regions except for region 6 , for which the parent galaxy is assumed to be n7319 ) . the last two columns of table 4 mark if the candidate belongs to a tidal tail and if it is associated with an hi cloud . regions 6 and 21 seem to be the ones with the highest chance of surviving as single entities . this is specially true for region 6 where co emission at the redshift and location of the galaxy has been recently detected ( gao & xu 2000 ) indicating a molecular gas mass of a few times 10@xmath10 m@xmath35 . other less massive candidates like regions 20 , 22 and 23 may not be massive enough to become independent entities . however , we should keep in mind that the values for the mass were determined under the assumption the regions are pure planar rotators and hence these are lower limits . in fact , random motions of the gas and stars may provide significant dynamical support for the tidal dwarf galaxies as indicated by their line widths . larger values for the mass of the candidates ( obtained if the internal dispersion velocity is also taken into account ) would result in an increase of the number of objects that could survive . although the complex of regions 25 may not survive as a single entity ( the tidal mass is very large mainly because of the large radius of the complex ) , it may be split into several smaller candidates which could then survive the tidal forces exerted by n7318b . the hi map of stephan s quintet presented by williams et al . ( 1999 , their fig . 5 ) shows that the hi component at velocity 66006700 km s@xmath3 is distributed along a fragmented ring around the galaxy n7319 , possibly formed after a tidal interaction with n7320c . such hi rings have previously been observed in other interacting systems like e.g. n5291 ( malphrus et al . 1997 ) . the northwestern side of the hi ring overlaps with the dwarf galaxy candidate identified as region 6 . it is possible that there are other tidal dwarf galaxy candidates along what would be the continuation of the ring , to the north , beyond region 6 , north of n7319 , where hi is not measured by williams et al . ( 1999 ) . comparing the eastern side of the hi ring with an optical image of the quintet we notice that the end of the tidal tail of n7319 that points to n7320c ( in the southeastern side of the group ) has a number of optical condensations ( hunsberger et al . 1996 , gallagher et al . 2000 ) that overlaps with maxima in the hi distribution . these regions probably have similar velocities to that of the hi gas ( 6700 km s@xmath3 ) and are most likely tidal dwarf candidates formed during a past encounter between n7319 and n7320c . other faint condensations can be seen in this region , between n7319 and n7320c . the velocities of these objects have to be measured in order to confirm their nature . what is very intriguing at first is the exact superposition of the hi component at 6700 km s@xmath3 with the galaxy ngc 7320 , at 800 km s@xmath3 . this superposition may be understood if the following is considered . moles et al . ( 1998 ) noted that while in a narrow - band image ( e.g. h@xmath0 ) the galaxy ngc 7320 appeared symmetric , in a broad - band image it appeared very asymmetric , showing an extension to its northwest . in fact , what may appear to be the northwestern extension of the nearby galaxy n7320 in broad - band images may be in reality a chain of smaller background objects ( perhaps other tidal dwarf galaxy candidates ? ) superimposed on the line of sight , but with velocities that coincide with that of the hi component at 6700 km s@xmath3 . other regions may have velocities of @xmath6 5700 km s@xmath55 ( possibly associated with n7318b ) . support for this scenario comes from the mix of velocities measured in this region by moles et al . ( 1998 ) . again , more velocity measurements of the condensations present in this region are needed . other tidal dwarf galaxies may be associated with n7318a . although moles et al . ( 1998 ) did not consider this galaxy in the interacting history of the quintet , it must have been part of collisions given the peculiar pattern of dust seen in the hst image ( charlton et al . 2000 ) and due to the several linear features that seem to emanate from the galaxy . these were marked with arrows in fig . a few of these could be projected material from n7318a into the intragroup medium , in the same way it has been observed for arp 105 by duc ( 1995 ) . alternatively it could be material in the process of re - accretion onto the parent galaxy after an interaction . there seems to be two different types of linear features , a few that are long and filamentary , with very low surface brightnesses and others that are small and have high surface brightnesses . perhaps also region 6 may be born out of gas from n7318a and not of n7319 as it has been assumed throughout this paper . this would be a specially appealing scenario if we could confirm that there is material in between galaxy n7318a and region 6 at a common velocity ( of @xmath6 6700 km s@xmath3 ) . however , it is more probable ( see fig . 2 of xu et al . 1999 ) that the apparent optical connection between regions 8 and 4 ( seen as a filament in the hst image ) is at a lower velocity of 59006000 km s@xmath3 . also important would be to confirm if the features marked in fig . 1c coincide with the velocity of n7318a . we would also like to point out that the emission detected in regions 7 and 9 and in the features in between these two regions might be associated with the higher velocity component of the group at 6700 km s@xmath3 instead of being related to the lower velocity component ( as it has been assumed in this paper ) . this is supported by the h@xmath0[nii ] map presented by xu et al . ( 1999 , their fig . 2 ) , which shows clearly that these regions are dominated by emission from the high - velocity component . further indirect evidence is their low h@xmath0 content despite their high optical luminosities ( see fig . 1a and 1c ) , which suggests that the emission from this area may perhaps fall at the very edge of the cfht filter , where the transmission is minimum . these regions would not have been detected in the lower s / n data taken with the sao 6 m , where only very bright sources like region 6 were detected . as can be seen from fig . 1c the _ optical _ luminosities of the sources in the area in between regions 7 and 9 are comparable to those of other regions identified by us . this only illustrates the incredible mixture of matter at two different velocities in this group . we assumed in the previous sections that region 6 is formed of remnant gas from n7319 ( moles et al . we favored this scenario based on the evidence from the hi observations ( fig . 5 from williams et al . 1999 ) which showed a structure similar to a ring ( an hi incomplete ring ) around galaxy n7319 . the hi gas associated with region 6 seems to be part of this large ring , in which case region 6 would be formed from gas originally in n7319 . on the other hand , there is also the possibility that region 6 was formed from gas spelled from n7318a . support for this is given by fig . 1c , where linear features apparently associated with n7318a are seen . in this scenario region 6 might be located at the tip of a tidal tail that may originate at the northern arm of n7318a . we need new velocity measurements of the region to decide between the two scenarios . in the following we give the parameters for region 6 in the case where it is physically associated with n7318a : ( 1 ) it would be at a distance of 25 h@xmath7 kpc to the parent galaxy ; ( 2 ) it would have a tidal mass of 6 @xmath1 10@xmath56 m@xmath35 , smaller than its inferred mass ; ( 3 ) the difference between its radial velocity and that of the parent galaxy is still lower than the escape velocity of the parent galaxy ; ( 4 ) its inclination ( 60@xmath41 15 ) would still be similar to ( within the errors ) to that measured for the parent galaxy ( 49 , value taken from the lyon - meudon extragalactic database , leda ) . we presented b and r photometry and a fabry - perot h@xmath0 map of 23 emission - line regions in stephan s quintet . all but one region may be associated with n7318b ( region 6 may be attached either to n7319 or to n7318a ) . * we find that seven of the regions have velocity gradients greater than 8 h@xmath9 km s@xmath3 kpc@xmath3 . we classify those as tidal dwarf galaxy candidates . * two tidal dwarf candidates may be located at the edge of a tidal tail , one within a tail and for four others there is no obvious stellar / gaseous bridge between them and the parent galaxy . * two of the candidates are associated with hi clouds , one of which is , in addition , associated with a co cloud . * the tidal dwarf candidates have low continuum fluxes and high h@xmath0 luminosity densities of f(h@xmath0 ) = 1 60 @xmath1 10@xmath2 erg s@xmath3 @xmath4 , magnitudes of m@xmath5 16.1 to 12.6 , sizes of typically @xmath6 3.5 h@xmath7 kpc , colors @xmath25 between 0.3 and 1.3 , gas velocity gradients of @xmath6 8 26 h@xmath9 km s@xmath3 kpc@xmath3 , sfr(h@xmath57)/sfr(@xmath58 ) between @xmath6 1100 and 6000 , masses of @xmath6 2 @xmath1 10@xmath10 to 10@xmath11 m@xmath12 and a median mass - to - light ratio of 7 m@xmath12/l@xmath12 . * the lower limits of the masses of the candidate tidal dwarf galaxies determined from their rotation curves ( assuming that they are pure rotators ) are in two cases larger than their tidal masses . * two out of the seven candidates may have velocities such that they can escape the gravitational field of the parent galaxy . * possible streaming motions within the tails may not have significantly affected the observed velocity gradients of those tidal dwarf candidates that are located within tails . * the dynamical criterion for formation of tidal tails during a collision v@xmath54/v@xmath43 @xmath16 2.5 is followed by galaxies ngc 7319 and ngc 7318b . we would like to thank dr . jacques boulesteix and the sao observatory staff ( drs v. afanasiev , a. burenkov , s. dodonov , v. vlasiuk and s. drabek ) for their help during the observations . we would also like to thank drs . s. hunsberger , m. marcelin and j. sulentic for insightful discussions and lourdes verdes - montenegro for sending us the hi map of the quintet . cmdo thanks the marseille observatory for their hospitality and funding of a trip in july 1999 . pa thanks the brazilian funding agency fapesp for the financial support for two visits to the instituto astronomico e geofisico , project numbers 99/05514 - 3 and 99/03744 - 1 . hp acknowledges the financial support of fapesp through a pos - doc position , project number 96/06722 - 0 , and the mexican conacyt under project 32303-e . williams , b.a . , van gorkom , j.h . , yun , m. , & verdes - montenegro , l. 1999 , _ in galaxy interactions at low and high redshift _ , proceedings of the iau symposium 186 , held at kyoto , japan , 26 - 30 august , 1997 . edited by j. e. barnes , and d. b. sanders , p.375 |
cr invariant differential operators on densities with leading part a power of the sub - laplacian are derived . one family of such operators is constructed from the `` conformally invariant powers of the laplacian '' via the fefferman metric ; the powers which arise for these operators are bounded in terms of the dimension . a second family is derived from a cr tractor calculus which is developed here ; this family includes operators for every positive power of the sub - laplacian . this result together with work of cap , slovk and soucek imply in three dimensions the existence of a curved analogue of each such operator in flat space . | math0301092 | invariant differential operators have a long history of importance and this is particularly the case for operators of laplace type . the conformally invariant laplacian is the basic example in conformal geometry . a family of higher order generalizations of the conformal laplacian with principal part a power of the laplacian was constructed in @xcite and has been the subject of recent interest . in cr geometry , the cr invariant sub - laplacian of jerison - lee ( @xcite ) plays a role analogous to that of the conformal laplacian . in this paper we construct and study generalizations of the jerison - lee sub - laplacian which are cr analogues of the `` conformally invariant powers of the laplacian '' . one can deduce the existence of some such operators in the cr case from the conformal operators via the fefferman metric . fefferman @xcite showed that to a cr manifold @xmath0 of dimension @xmath1 one can associate a conformal structure on a circle bundle @xmath2 of dimension @xmath3 . on conformal manifolds of dimension @xmath4 , the construction in @xcite produces for each @xmath5 , such that @xmath6 and @xmath7 if @xmath4 is even , a conformally invariant natural differential operator , with principal part @xmath8 , where @xmath9 denotes the space of conformal densities of weight @xmath5 . as we explain in the next section , on a cr manifold one can consider cr densities @xmath10 parametrized by @xmath11 . it is necessary that @xmath12 in order that @xmath10 be well - defined and this is always implicitly assumed . the space @xmath10 of cr densities on @xmath0 can be naturally regarded as a subspace of the space of conformal densities @xmath13 on @xmath2 . the operators @xmath14 for the fefferman metric can be shown to preserve the subspaces of cr densities and it is straightforward to calculate the principal part of the induced operators . one thereby obtains : [ ambient ] let @xmath15 with @xmath16 . there is a cr invariant natural differential operator @xmath17 whose principal part agrees with that of @xmath18 . in the case that @xmath0 is a flat cr manifold , the hypothesis @xmath16 is unnecessary in this construction and in the flat case the operators @xmath19 exist for all @xmath20 such that @xmath21 . in the conformal case , it is conjectured that if @xmath4 is even and @xmath22 , there is no natural operator with principal part @xmath8 mapping @xmath23 . this has been established for @xmath24 , @xmath25 ( @xcite ) and @xmath26 , @xmath27 ( @xcite ) . our main result is the existence in the cr case of the following family of operators , which includes operators of higher orders . set @xmath28 . [ intromain ] for each @xmath20 such that @xmath15 and @xmath29 , there is a cr invariant natural differential operator @xmath30 whose principal part agrees with that of @xmath18 . in order to construct the operators @xmath31 , we derive a tractor calculus for cr geometry which we anticipate will be of independent interest . the main ingredients are to define a tractor bundle , a complex vector bundle of rank @xmath32 over @xmath0 , together with a cr invariant connection , and an extension of this connection to a so - called tractor @xmath33 operator between weighted tractor bundles . we give a direct derivation of these in terms of the pseudohermitian tanaka - webster connection induced by a choice of contact form on @xmath0 , parallel to the derivation in @xcite in the projective and conformal cases . the tractor bundle and connection can also be derived from the cr cartan connection of cartan , tanaka , chern - moser , but we found it preferable to proceed directly , especially for the tractor @xmath33 calculus needed for the construction of the invariant operators . see @xcite , @xcite , @xcite , @xcite for further discussion and background about tractors . given this machinery , the construction of the operators @xmath31 follows a similar construction of eastwood and gover in the conformal case described in @xcite , involving iterating tractor @xmath33 s . the operators so constructed are _ strongly invariant _ in the sense of @xcite , i.e. they act invariantly not just on scalar densities but also on density - valued tractors . note that the tractor construction does not yield an operator in the case @xmath34 even though the construction via the fefferman metric does ; this is the only weight for which this occurs . however , when @xmath35 , a refinement of the tractor construction does produce a cr invariant ( but not necessarily strongly invariant ) operator @xmath36 , which by direct calculation can be seen to agree with @xmath37 . based on the situation in the conformal case ( @xcite ) , we do not anticipate that the operators @xmath19 and @xmath31 agree in general when both are defined . in fact , one must make choices of orderings and conjugations in defining @xmath31 , and one expects that different choices will in general lead to different operators . in the flat case , the tractor construction ( suitably interpreted ) also produces operators for all @xmath20 for which @xmath21 ; these do agree with the corresponding operators constructed via the fefferman metric . a construction of invariant natural operators is given in @xcite for general parabolic geometries , of which cr geometry is a special case . this construction produces curved analogues of many invariant operators on the flat homogeneous model , in the case when the flat operators are standard and act between irreducible bundles corresponding to integral non - singular highest weight vectors . for cr geometry with @xmath38 , any flat operator @xmath19 satisfying these criteria has @xmath39 . so in these dimensions theorem [ intromain ] substantially extends the results of @xcite . however , the three dimensional case is special : the relevant parabolic is a borel subalgebra , so all flat operators are standard . there do exist bundles @xmath10 with @xmath40 corresponding to non - integral or singular highest weight vectors , for which @xcite does not apply . however , one can check that if @xmath41 , then the highest weight vector for the bundle @xmath10 is integral and non - singular . we are not aware that it has been established in general that a given specific operator constructed in @xcite has the desired principal part , or even is nonzero . however , it is possible to show that this is the case for all the operators which arise in three dimensional cr geometry . therefore , in three dimensions theorem [ intromain ] and @xcite together yield the existence of a curved version of each of the flat operators : for each @xmath20 satisfying @xmath42 , there is a natural differential operator on three dimensional cr manifolds mapping @xmath43 , whose principal part agrees with that of @xmath18 . kengo hirachi has informed us that he has established the existence of such an operator in three dimensions mapping @xmath44 via the cr ambient metric with ambiguity derived in @xcite . we are grateful to andi cap for pointing out to us the special applicability of @xcite in the three dimensional case . in section [ pseu ] we review the basic facts about cr structures and pseudohermitian geometry . we derive the tractor bundle , its connection , curvature , and the tractor @xmath33 operators in section [ tractors ] . these in turn are used in section [ maint ] to construct the operators @xmath31 . the main point in proving theorem [ intromain ] is that there are natural candidate operators which can be written down in terms of iterated tractor @xmath33 operators whose principal part is a multiple of @xmath18 and it must be determined when the multiple is nonzero . this leads to some numerology in calculating the multiple for flat space . finally in section [ amb ] we review lee s formulation of the fefferman metric and the gjms construction of the conformal operators and present the details of the proof of theorem [ ambient ] . we show how this derivation via the fefferman metric can be reformulated in terms of fefferman s ambient metric and also in terms of the khler - einstein metric of cheng - yau when @xmath0 is a hypersurface in @xmath45 , and close with a brief discussion of @xmath46-curvature in the cr case . our presentation in this section in terms of the fefferman metric is influenced by the point of view explicated in @xcite . in sections [ maint ] and [ amb ] we also prove that the operators @xmath31 and @xmath19 are self - adjoint ( for @xmath31 the orderings of the @xmath33 operators in the iteration must be chosen appropriately ) . it is a pleasure to thank andi cap , mike eastwood and kengo hirachi for helpful discussions . a _ cr - structure _ on a @xmath47-dimensional smooth manifold @xmath0 is an @xmath48-dimensional complex subbundle @xmath49 such that @xmath50 and satisfying the integrability condition @xmath51\subseteq t^{1,0}$ ] , where we have used the same notation @xmath52 for the bundle and the space of its smooth sections . set @xmath53 and define @xmath54 by @xmath55 . the _ canonical bundle _ @xmath56 is a complex line bundle on @xmath0 . we will assume that @xmath57 admits an @xmath58 root and we fix a bundle denoted @xmath59 which is a @xmath60 power of @xmath57 . the bundle @xmath61 of @xmath20-densities is defined for @xmath62 satisfying @xmath63 . ( if @xmath64 is viewed as a @xmath65-principal bundle , then @xmath10 is the bundle induced by the representation @xmath66 of @xmath65 . ) without further comment , we will assume implicitly whenever we write @xmath10 that @xmath67 and @xmath68 . we will also usually write @xmath10 for the space of sections of this bundle ; the interpretation should be clear from context . the bundle @xmath69 is a real @xmath70-dimensional subbundle of @xmath71 . @xmath72 carries a natural almost complex structure map given by @xmath73 for @xmath74 , which induces an orientation on @xmath72 . we will assume that @xmath0 is orientable , which implies that the bundle @xmath75 admits a nonvanishing global section . pseudohermitian structure _ on @xmath0 is the choice of such a contact 1-form @xmath76 . we fix an orientation on the bundle @xmath77 and restrict consideration to @xmath76 s which are positive relative to this orientation . the _ levi - form _ of @xmath76 is the hermitian form @xmath78 on @xmath52 defined by @xmath79 we assume that the cr structure is nondegenerate , i.e. @xmath78 is a nondegenerate form , whose signature we denote by @xmath80 , @xmath81 . given a pseudohermitian form @xmath76 , define @xmath82 to be the unique vector field on @xmath0 satisfying @xmath83 an _ admissible coframe _ is a set of @xmath84 forms @xmath85 , @xmath86 , which satisfy @xmath87 and whose restrictions to @xmath52 form a basis for @xmath88 . we will use lower case greek indices to refer to frames for @xmath52 or its dual . we may also interpret these indices abstractly , so will denote by @xmath89 the bundle @xmath52 ( or its space of sections ) and by @xmath90 its dual , and similarly for the conjugate bundles or for tensor products thereof . by integrability and , we have @xmath91 for a smoothly varying hermitian matrix @xmath92 , which we may interpret as the matrix of the levi form in the frame @xmath93 , or as the levi form itself in abstract index notation . it is shown in lemma 3.2 of @xcite ( see also ( 4.3 ) of @xcite ) that if @xmath94 is a ( locally defined ) nonvanishing section of @xmath57 , then there is a unique pseudohermitian form @xmath76 with respect to which @xmath94 is volume normalized in the sense that @xmath95 this is equivalent to the requirement that @xmath94 have unit length with respect to the natural norm induced by @xmath76 and @xmath78 . upon replacing @xmath94 by @xmath96 , where @xmath97 is a smooth @xmath65-valued function , @xmath76 is replaced by @xmath98 . now @xmath99 is a section of @xmath100 , so @xmath101 defines a section of @xmath102 which depends only on the cr structure . the levi form of @xmath76 scales the same way , so defines a canonical section @xmath103 of @xmath104 , where in general we denote the tensor product with a density bundle by omitting the second @xmath105 : @xmath106 . we will use exclusively @xmath103 and its inverse @xmath107 to raise and lower indices , so that raising and lowering indices changes weight . the choice of pseudohermitian form @xmath76 on @xmath0 is equivalent to the choice of a _ cr - scale _ @xmath108 related by @xmath109 , and we also have @xmath110 . if @xmath111 , the change of pseudohermitian structure @xmath112 results in @xmath113 and @xmath114 . a pseudohermitian structure is said to be _ flat _ if it is locally equivalent to the heisenberg group with its standard pseudohermitian structure . it is said to be _ cr flat _ if there is a rescaling of @xmath76 which is pseudohermitian flat . a choice of pseudohermitian structure determines a connection on @xmath71 , the tanaka - webster , or pseudohermitian , connection ( @xcite , @xcite ) . it is given in terms of an admissible coframe by @xmath115 where webster s connection 1-forms @xmath116 satisfy @xmath117 and the _ pseudohermitian torsion tensor _ @xmath118 is symmetric . ( according to our conventions for raising indices , @xmath119 is a section of @xmath120 . ) in particular , this connection preserves @xmath52 , so induces connections on @xmath89 and @xmath90 . one has @xmath121 . there is an induced connection on the canonical bundle , and therefore also on all the density bundles @xmath10 . [ parallel ] @xmath122 the definition is @xmath123 , where @xmath94 is volume normalized with respect to @xmath76 . since @xmath124 , it suffices to show that @xmath125 for such @xmath94 . choose an admissible coframe @xmath126 such that @xmath127 . the condition that @xmath94 is volume normalized is equivalent to @xmath128 . by we have @xmath129 it follows that @xmath130 as desired . from proposition [ parallel ] we conclude that @xmath131 , where @xmath132 is the scale associated to @xmath76 , and that @xmath133 . the pseudohermitian connection preserves the splitting @xmath134 therefore , if we decompose a tensor field relative to this splitting ( and/or its dual ) , we may calculate the covariant derivative componentwise . each of the components may be regarded as a section of a tensor product of @xmath89 or its dual or conjugates thereof . therefore we will often restrict consideration to the action of the connection on @xmath89 or @xmath90 . we will use indices @xmath135 for components with respect to the frame @xmath136 and its dual , so that the 0-components incorporate weights . if @xmath137 is a ( possibly density - valued ) tensor field , we will denote components of the ( tensorial ) iterated covariant derivatives of @xmath137 in such a frame by preceding @xmath138 s , e.g. @xmath139 . as usual , such indices may alternately be interpreted abstractly . so , for example , if @xmath140 , we will consider @xmath141 as the triple @xmath142 , @xmath143 , @xmath144 . the torsion and curvature of @xmath138 can be described by structure equations for the forms @xmath145 ( see above and ( 1.3 ) of @xcite ) , or by commuting second derivatives on functions and 1-forms ( see lemma 2.3 of @xcite . we generally follow the same conventions as @xcite , @xcite , except that we substitute @xmath146 for @xmath147 so that all quantities are naturally weighted . ) these can be expressed in terms of the _ pseudohermitian curvature tensor _ @xmath148 , @xmath149 , @xmath150 , and conjugates of the latter two . flat pseudohermitian structures are characterized by the vanishing of @xmath151 and @xmath149 . the webster - ricci tensor is defined by @xmath152 and the webster scalar curvature by @xmath153 from these we define @xmath154 we will also need to know how to commute derivatives of densities . [ commprop ] if @xmath155 , then @xmath156 we first note that holds when @xmath34 , since in that case it agrees with lemma 2.3 of @xcite . now it is a straightforward consequence of the definitions that if @xmath157 is a vector bundle with connection over a manifold @xmath0 , which itself has a linear connection , if @xmath158 is a section of @xmath157 , then @xmath159v^a = \omega_{ijb}{}^a v^b - t_{ij}^k{\nabla}_kv^a,\ ] ] where here @xmath160 label @xmath71 , @xmath161 denotes the curvature of the connection on @xmath157 , @xmath162 denotes the torsion of the connection on @xmath71 , and the second covariant derivative is with respect to the coupled connection on @xmath163 . we first apply taking @xmath157 to be the trivial bundle with the flat connection over our pseudohermitian manifold with its connection : the case @xmath34 . the curvature term vanishes , and we conclude that the torsion term in is exactly the right hand side of for @xmath34 . next let @xmath164 , and take @xmath165 with its induced connection . the form of the torsion term in is independent of the choice of @xmath157 , so takes exactly the same form as in the previous case . to identify the curvature term , observe that since @xmath57 is a line bundle , its curvature is a scalar 2-form on @xmath0 . choose an admissible coframe and set @xmath166 . then we have , so the connection form for @xmath57 relative to the frame @xmath167 is @xmath168 . the curvature of @xmath57 is therefore @xmath169 . however , ( 2.3 ) , ( 2.4 ) of @xcite state that @xmath170 thus reduces to for @xmath164 . the general case now follows upon conjugating , taking powers , and adding . proposition [ commprop ] together with the formulae in lemma 2.3 of @xcite enable one to calculate the effect of commuting covariant derivatives on any weighted tensor field . if @xmath112 is another pseudohermitian form , we can express the connection , curvature , and torsion for @xmath171 in terms of that for @xmath76 . we consider first the relation between the induced connections on tensor products of @xmath89 or its dual or conjugates thereof . for efficiency in expressing the transformation laws , we will denote covariant derivatives of @xmath172 with indices : @xmath173 . one observes first that if @xmath126 is an admissible coframe for @xmath76 , then @xmath174 defines an admissible coframe for @xmath171 . since the restrictions of @xmath175 and @xmath93 to @xmath52 agree , the components of a section @xmath137 of @xmath89 or its dual ( or conjugates or tensor products ) are the same in the two frames . we will denote by @xmath176 , @xmath177 , and @xmath178 the components of @xmath179 relative to @xmath180 . for example , if @xmath137 is an unweighted function , one obtains upon changing frames that @xmath181 the connection forms @xmath182 for @xmath183 can be expressed in terms of @xmath145 and @xmath172 ; this is lemma 3.4 of @xcite . it is straightforward to use this to calculate the analogue of for a section @xmath184 of @xmath185 . one obtains : @xmath186 we also need to know how the connection transforms on densities . [ denhat ] if @xmath187 , then @xmath188f \end{split}\ ] ] as in the proof of proposition [ commprop ] , if we establish the result for @xmath189 , the general case follows by conjugating , taking powers , and adding . let @xmath127 be a section of @xmath190 . then @xmath191 is given by . however , also @xmath192 , so applying again gives @xmath193 . therefore @xmath194 now lemma 3.4 of @xcite gives @xmath195 in terms of @xmath172 . contracting , adding @xmath196 , and reformulating the result in terms of components yields the desired formulae . the curvature and torsion also transform under the pseudohermitian change . from lemma 2.4 of @xcite one obtains @xmath197 and @xmath198 we also will need the transformation laws for two other objects . set @xmath199 and define @xmath200 @xmath201 straightforward calculation using the formulae discussed above and the bianchi identities of lemma 2.2 of @xcite gives : @xmath202 we derived the expressions for @xmath203 and @xmath204 from the condition that they satisfy transformation laws of this form , which we needed to construct the tractor connection in the next section . only later did we observe that they occur already in @xcite as components of the connection forms and ricci tensor of the fefferman metric . also , @xmath205 and a variant of @xmath204 arise in @xcite , where they were combined with @xmath206 and @xmath149 to form a 2-tensor on @xmath0 which was used to construct a version of pseudohermitian normal coordinates . in that work , only the part of the transformation laws involving highest derivatives ( counted non - isotropically ) of @xmath172 was relevant . the version of @xmath204 in @xcite therefore does not include the @xmath207 term , which however is important for us . the pseudohermitian connection depends on the choice of @xmath76 . in this section we construct a vector bundle of rank @xmath32 over @xmath0 , the tractor bundle , together with a cr invariant tractor connection . we also show how to extend this connection to a cr invariant tractor @xmath33 operator on weighted tractors which can be iterated . for a given choice of @xmath76 , the ( co-)tractor bundle @xmath208 is realized as a direct sum @xmath209 the realization corresponding to @xmath210 is identified with that for @xmath76 via @xmath211 where @xmath212 , @xmath213 , @xmath214 , and @xmath215 recalling , it is easily checked that these identifications are consistent upon changing to yet a third @xmath76 . we may therefore mod them out to obtain a bundle @xmath216 determined solely by the cr structure . ( more formally , the total space of the bundle @xmath216 can be defined to be the disjoint union of one copy of the total space of @xmath217 for each global section @xmath76 , modulo the equivalence relation . the smooth structure on this total space and the linear structure on the fibers are inheirited from that of any of the realizations . ) the subspaces @xmath218 and @xmath219 are preserved by the identifications , so determine canonical subbundles of @xmath216 . this composition series structure is what remains of the direct sum decomposition after making the identifications . the vector given by @xmath220 is preserved by @xmath221 , so defines an invariant section of @xmath222 which we denote by @xmath223 . the matrix @xmath221 is in @xmath224 , where @xmath225 it follows that @xmath226 defines an invariant hermitian metric on @xmath216 , and that @xmath216 has an invariant volume form . we denote by @xmath227 the conjugate bundle , by @xmath228 the dual bundle ( the tractor bundle ) , and we use @xmath226 and its inverse to raise and lower tractor indices . we define a connection on @xmath216 as follows . if @xmath229 is a section of @xmath216 in the realization determined by @xmath76 , set @xmath230 @xmath231 @xmath232 where @xmath203 and @xmath204 were defined in 2 , and where the @xmath138 s on the right hand side refer to the pseudohermitian connection on the appropriate weighted bundles . this operation satisfies a leibnitz formula so defines a connection on @xmath216 . it is a direct calculation using the transformation formulae in the previous section to show that the definition is independent of the choice of @xmath76 , and therefore cr invariant . for example , consider the second component of @xmath233 . using , proposition [ denhat ] , and , we obtain @xmath234 as desired . ( here we verified the full transformation law , but it suffices to check invariance to first order in @xmath172 . ) this connection induces connections on the dual and conjugate and tensor product bundles , and a calculation shows that @xmath235 ( reflecting a hermitian symmetry inherent in ) . therefore differentiation commutes with raising and lower tractor indices . one can calculate the curvature of the tractor connection from , using the commutation formulae and bianchi identities of @xcite . the result is the following : @xmath236 @xmath237 where the component tensors are given by @xmath238 @xmath239 @xmath240 @xmath241 @xmath242 @xmath243 and have the properties : @xmath244 @xmath245 @xmath246 @xmath247 this should be compared with the curvature of chern s cartan connection ( see @xcite , @xcite ) . we observe for future reference that the vanishing trace conditions above imply @xmath248 next we use the tractor connection to construct a cr invariant second order differential operator @xmath33 between tractor bundles . let us write @xmath249 to indicate any weighted tractor bundle ( with arbitrary lists of conjugated and/or unconjugated upper and/or lower indices ) . the _ tractor d _ operator @xmath250 is defined by @xmath251 in the realization determined by a choice of @xmath76 . here @xmath252 refers to the tractor connection defined above coupled with the pseudohermitian connection on the relevant density bundle . the @xmath253 in the bottom component additionally couples the pseudohermitian connection on @xmath254 . that this definition is cr invariant is again a direct calculation of its transformation under change of @xmath76 . the tractor indices on @xmath137 play no role whatsoever in this calculation since the tractor connection is invariant . the first component does not depend on @xmath76 , as required . for the second component , by proposition [ denhat ] we have @xmath255 , which is precisely what is required by the identification . the third follows upon transforming the connection using and proposition [ denhat ] and substituting the contraction of . conjugation produces the operator @xmath256 both @xmath257 and @xmath258 clearly commute with raising and lowering tractor indices . a straightforward calculation from the definition shows that one has @xmath259 for @xmath260 , where here and in what follows we view @xmath261 as a multiplication operator and we omit the parentheses when composing operators . one motivation for the @xmath257 operator is its interpretation in flat space . the flat model can be taken to be a real hyperquadric in @xmath262 . a section of @xmath10 can be interpreted as a function on the associated affine null cone @xmath263 which is homogeneous of degree @xmath20 . if @xmath264 , such a function @xmath137 has a homogeneous extension @xmath265 off @xmath263 which is harmonic to first order with respect to the corresponding laplacian , and one has @xmath266 a similar interpretation can be made in the curved case when @xmath0 is a hypersurface in @xmath45 ; in this situation the extension is required to be harmonic to first order with respect to fefferman s ambient metric . details of this and further connections between tractors and the ambient construction will be given in @xcite . it is often useful to decompose @xmath257 into two pieces . given a choice of @xmath76 , let us write @xmath267 for the operator appearing in the bottom slot of the above formula for the tractor @xmath268 operator ; that is @xmath269 and by @xmath270 the corresponding conjugate operator given by @xmath271 we also define @xmath272 by @xmath273 with conjugate @xmath274 . then clearly we have @xmath275 a direct calculation shows that as operators on any weighted tractor bundle , @xmath276 = { \tilde{d}}_a~.\ ] ] another calculation ( using ) gives @xmath277 on @xmath249 . the operators @xmath278 and @xmath279 depend on the choice of @xmath76 and are not cr invariant . a first attempt to use tractors to construct an invariant operator from densities to densities would be to form @xmath280 . however , direct calculation shows that this vanishes identically for @xmath137 a tractor field of any weight . note , though , that if @xmath137 has weight @xmath20 satisfying @xmath281 , then from ( [ crexp ] ) @xmath282 therefore , @xmath278 is a cr invariant differential operator @xmath283 this is a generalization of the sub - laplacian of jerison - lee ( @xcite ) , which is @xmath278 on @xmath284 . the operator @xmath285 is also cr invariant when @xmath281 , but from we conclude that @xmath286 for such @xmath20 . it follows immediately from these observations and the invariance of @xmath257 that for @xmath287 , the operator @xmath288 is cr invariant if @xmath289 . variants of this operator can also be formed which are invariant between the same spaces by replacing some of the indices by barred indices and by reordering the @xmath33 factors independently on either side of the @xmath278 . it is not hard to see that in any choice of cr scale , the principal part of any such operator agrees with a multiple of that of @xmath18 , where @xmath290 . however , the multiple can vanish . by calculating these operators explicitly in flat space , we will determine when the multiple is nonzero and thereby prove the following strengthening of theorem [ intromain ] . [ main ] for each @xmath20 such that @xmath291 and @xmath29 , there is a cr invariant natural differential operator @xmath292 whose principal part agrees with that of @xmath18 . for fixed @xmath20 as in theorem [ main ] we shall in general obtain several operators @xmath31 corresponding to different choices of barred and unbarred indices and different orderings of the @xmath33 factors . we shall see that these operators all agree for a flat cr structure . it is straightforward but tedious to check directly from the definition that for a flat pseudohermitian structure , @xmath293=0 , \quad [ d_b , d_{\bar{c}}]=0 , ~\mbox { and } ~ [ d_{\bar{b}},d_{\bar{c}}]=0\ ] ] as operators on any weighted tractor bundle . since @xmath257 is cr invariant , this remains true if the structure is only cr flat . so in the flat case , it is clear that the order of the @xmath33 factors is irrelevant . the following proposition calculates the operators in flat space . the proof of theorem [ main ] only uses the cases @xmath294 or @xmath295 , but we include the general case for completeness . [ flatgoody ] let @xmath20 satisfy @xmath296 . let @xmath297 be such that @xmath298 . for a flat pseudohermitian structure , we have as operators on @xmath249 : @xmath299 @xmath300 @xmath301 we have written @xmath302 for the @xmath303-fold composition of @xmath278 with itself . note that each factor is acting on a different density space , so each @xmath278 in the composition is really a different operator . in case @xmath304 or @xmath305 equals @xmath306 , the empty product in is to be interpreted as @xmath307 . _ proof of theorem [ main ] using proposition [ flatgoody ] . _ conjugating if necessary , we may assume without loss of generality that @xmath308 . set @xmath309 , @xmath295 . the numerical factor on the right hand side of is then nonzero . since the principal part of @xmath310 agrees with that of @xmath311 , we may define @xmath31 by multiplying the left hand side of by the appropriate constant . it is clear that @xmath31 is a cr invariant operator , and expanding shows that it is natural in the sense that it is given by a universal formula in terms of the pseudohermitian metric , curvature , torsion , and connection . in flat space it has the correct principal part , and it is easily seen that the correction terms in curved space are of lower order . we remark that the proof shows that in theorem [ main ] a stronger statement can be made about the principal part of @xmath31 : its nonisotropic principal part , in which derivatives with respect to @xmath82 are weighted by a factor of 2 , agrees with that of @xmath312 . the proof of proposition [ flatgoody ] is preceeded by three lemmas . [ one ] let @xmath297 and let @xmath20 satisfy @xmath313 . for a cr flat structure , we have as operators on @xmath249 : @xmath314 for a differential operator @xmath315 . it follows from that if @xmath316 , then @xmath317 skewing on @xmath318 and @xmath319 and recalling gives @xmath320}d_c \cdots d_i d_{\bar{j}}\cdots d_{\bar{q } } f = 0.\ ] ] again using , we can commute the @xmath33 s to conclude that the commutator can be taken on any of the unbarred indices : @xmath321}d_g\cdots d_i d_{\bar{j}}\cdots d_{\bar{q } } f=0\ ] ] for @xmath322 . recalling , we obtain similarly @xmath323}d_{\bar{p}}\cdots d_{\bar{q } } f=0\ ] ] for @xmath324 . therefore @xmath325 must be proportional to each of @xmath326 , and the result follows . [ two ] in lemma [ one ] , we have @xmath327 if @xmath328 . it suffices to show that for @xmath329 we have @xmath330 as operators on @xmath249 with @xmath20 as in lemma [ one ] . this is a consequence of , and as follows : @xmath331 [ three ] for flat pseudohermitian structures , we have @xmath332=0 , \qquad [ \box^k , z_a ] = k\box^{k-1}{\tilde{d}}_a\ ] ] on any weighted tractor bundle . the first equation follows by direct calculation from the definitions . the second is a consequence of the first together with . _ proof of proposition [ flatgoody ] . _ according to lemmas [ one ] and [ two ] , in order to establish it suffices to show that for @xmath287 , we have for flat pseudohermitian structures @xmath333 as operators on @xmath249 for @xmath334 . we prove this by induction on @xmath303 . the case @xmath335 is clear . suppose the statement is true for @xmath303 and let @xmath316 with @xmath336 . note that it follows from lemma [ one ] that @xmath337 combining this with the induction hypothesis and then using and lemma [ three ] gives @xmath338 as desired . now follows upon repeatedly applying and its conjugate to , and follows upon comparing for @xmath295 with the conjugate of for @xmath294 , recalling from that @xmath339 on @xmath249 for @xmath281 . for cr flat structures , lemmas [ one ] and [ two ] show that the operator @xmath340 is cr invariant and depends only on @xmath341 , and its principal part is easily determined . it follows that for cr flat structures we can extend the definition of @xmath31 to the case @xmath291 without the restriction @xmath16 by setting @xmath342 . then for cr flat structures @xmath31 is independent of the choice of barred and unbarred indices ( and of their ordering , as already observed ) . in the curved case , one expects by analogy with the conformal case ( see the discussion in @xcite , @xcite ) that it is no longer generally true that @xmath343 is proportional to @xmath344 , necessitating the application of @xmath345 as in and leading to the condition @xmath29 in theorem [ main ] . however , for @xmath346 this is not necessary even in the curved case : if @xmath187 with @xmath347 , one can see by direct calculation from the definitions that the first two slots of @xmath348 vanish , so we have @xmath349 for a fourth order operator @xmath350 whose principal part agrees with that of @xmath351 . applying @xmath352 shows that @xmath353 if @xmath354 , so in this case nothing new is obtained . however , if @xmath35 we conclude the existence of an invariant operator @xmath355 not covered by theorem [ main ] . for @xmath35 , we may therefore extend the definition of the family @xmath31 by setting @xmath356 . we shall recover @xmath357 and show the existence of higher dimensional analogues mapping @xmath358 in the next section using the ambient metric . next we show that as long as the @xmath33 s are ordered consistently , the operators @xmath31 constructed in theorem [ main ] are self - adjoint . observe first that under the change of scale @xmath112 , the volume form @xmath359 multiplies by @xmath360 , so the bundle of volume densities can be canonically identified with @xmath361 and @xmath362 is invariantly defined for @xmath363 . ( throughout this discussion we assume that all sections are compactly supported . ) in general , if @xmath364 is a differential operator between complex vector bundles @xmath365 and @xmath366 , then the formal adjoint @xmath367 is defined by @xmath368 for @xmath369 , @xmath370 , where @xmath371 denotes the conjugate of the dual bundle to @xmath365 . we have @xmath372 . if @xmath291 , then by our implicit assumption that @xmath373 it follows that @xmath374 , so the operators @xmath375 constructed in theorem [ main ] have the property that also @xmath376 . recall that the operators @xmath31 depend on a choice of ordering of barred and/or unbarred indices before and after the middle @xmath278 . we shall say that the indices are ordered consistently if the order of the indices before the @xmath278 is the opposite of those after the @xmath278 . [ duality ] if @xmath20 is as in theorem [ main ] and if the indices are ordered consistently , then @xmath375 is self - adjoint . the proposition clearly follows if we show that @xmath278 is self - adjoint on @xmath249 for @xmath281 and that @xmath377 for @xmath316 , @xmath378 , where the supressed tractor indices on @xmath137 and @xmath379 are the same and we have denoted by ` @xmath380 ' the full contraction of these tractor indices . an easy calculation shows that the divergence operator @xmath381 is cr invariant @xmath382 and integration by parts ( see ( 2.18 ) of @xcite ) shows that @xmath383 for @xmath384 . also it is easily seen from , and that @xmath385 + w(n+w - w')p\ ] ] on @xmath249 . therefore for @xmath260 , @xmath386 we have @xmath387f + w(n+w - w')pf)\cdot g\\ & = \int_m f\cdot(n{\nabla}_{{\alpha}}{\nabla}^{{\alpha}}g + w[{\nabla}_{{\alpha}},{\nabla}^{{\alpha}}]g + w(n+w - w')pg)\\ & = \int_m f\cdot(n{\nabla}^{{\alpha}}{\nabla}_{{\alpha}}g -(n+w)[{\nabla}^{{\alpha}},{\nabla}_{{\alpha}}]g + w(n+w - w')pg)\\ & = n\int_m f\cdot ( \box g + ( n+w+w')pg ) . \end{split}\ ] ] self - adjointness of @xmath278 when @xmath281 follows immediately from this together with the fact that @xmath339 when @xmath281 . finally , if @xmath316 and @xmath388 then the definition of @xmath257 gives @xmath389 while a computation shows that @xmath390 therefore follows from and . in this section we review lee s formulation ( @xcite ) of the fefferman conformal structure and the construction in @xcite of the invariant powers of the laplacian on a conformal manifold via the ambient metric . combining these produces cr invariant powers of the sub - laplacian . for a hypersurface in @xmath45 , this construction can be expressed in terms of fefferman s original formulation , and reexpressed in terms of the associated cheng - yau metric . in lee s formulation ( see also @xcite , @xcite ) , the fefferman metric of a cr manifold @xmath0 lives on the circle bundle @xmath391 , where @xmath57 denotes the canonical bundle of @xmath0 , and @xmath392 . a section @xmath94 of @xmath393 determines a fiber variable @xmath394 on @xmath391 by the requirement that @xmath395 be in the given @xmath396-equivalence class . suppose that @xmath76 is a choice of pseudohermitian form , that @xmath94 is volume normalized with respect to @xmath76 , and choose an admissible coframe @xmath397 such that @xmath127 . the 1-form @xmath398 on @xmath391 defined by @xmath399 is independent of the choice of @xmath94 and @xmath93 , so depends only on @xmath76 and is globally defined . the fefferman metric associated to @xmath76 is the metric of signature @xmath400 given by @xmath401 in @xcite it is shown that if @xmath112 , then @xmath402 , so that the conformal class of @xmath403 is cr invariant . lee also explicitly calculated the connection forms and ricci curvature of @xmath403 . from his expression for the connection , it follows that all components of the curvature tensor of @xmath403 and its iterated covariant derivatives are given by universal expressions in the pseudohermitian metric , curvature and torsion and their pseudohermitian covariant derivatives . as in section 2 , we assume that @xmath57 admits an @xmath58 root , which we fix . for our purposes it will be useful to work on the circle bundle @xmath404 . the fefferman metric pulls back via the @xmath58 power map to a metric on @xmath2 which we shall also denote by @xmath403 . a fiber variable @xmath405 on @xmath2 satisfies @xmath406 . the metric bundle of a manifold @xmath2 of dimension @xmath4 with a conformal class of metrics @xmath407 $ ] of signature @xmath408 is the ray subbundle @xmath409 of multiples of the metric : if @xmath403 is a representative metric , then the fiber of @xmath410 over @xmath411 is @xmath412 . the bundle of conformal densities of weight @xmath413 is @xmath414 , where by abuse of notation we have denoted by @xmath410 also the line bundle associated to the ray bundle defined above . the main result of @xcite is the existence , for @xmath415 satisfying @xmath7 if @xmath4 is even , of a conformally invariant natural differential operator @xmath416 with principal part equal to that of @xmath8 . these operators are constructed in @xcite using the ambient metric of @xcite . denote by @xmath417 the natural projection of the metric bundle , and by @xmath418 the tautological symmetric 2-tensor on @xmath419 defined for @xmath420 and @xmath421 by @xmath422 . there are dilations @xmath423 for @xmath424 given by @xmath425 , and we have @xmath426 . denote by @xmath204 the infinitesimal dilation vector field @xmath427 . define the ambient space @xmath428 . identify @xmath410 with its image under the inclusion @xmath429 given by @xmath430 for @xmath431 . the dilations @xmath432 and infinitesimal generator @xmath204 extend naturally to @xmath433 . the ambient metric @xmath434 is a metric of signature @xmath435 on @xmath433 which satisfies the initial condition @xmath436 , is homogeneous in the sense that @xmath437 , and is an asymptotic solution of @xmath438 along @xmath419 . for @xmath4 odd , these conditions uniquely determine a formal power series expansion for @xmath434 up to diffeomorphism , but for @xmath4 even and @xmath439 , a formal power series solution exists in general only to order @xmath440 . 1 . by extending a density @xmath441 to a function @xmath442 homogeneous of degree @xmath443 on @xmath444 , applying @xmath445 , where @xmath446 denotes the laplacian in the metric @xmath434 , and restricting back to @xmath419 : @xmath447 . 2 . as the normalized obstruction to extending @xmath448 to a smooth function @xmath449 homogeneous of degree @xmath443 on @xmath450 , such that @xmath449 satisfies @xmath451 to infinite order along @xmath419 . for @xmath4 even , the condition @xmath7 is needed to ensure that the construction does not involve too many derivatives of the ambient metric . we remark that 2 . above can be restated in terms of non - smooth solutions : there are infinite order formal solutions whose expansion includes a @xmath452 term , and the invariant operator applied to @xmath137 can be characterized as a multiple of the restriction to @xmath419 of the coefficient of the @xmath452 term of a formal solution which equals @xmath137 on @xmath419 . let now @xmath0 be a cr manifold of dimension @xmath1 . we associate to @xmath0 its fefferman conformal manifold @xmath453)$ ] , of dimension @xmath3 . volume normalization gives a canonical identification between @xmath454 ( with usual scalar multiplication ) and the metric bundle of @xmath453)$ ] ( with dilations @xmath432 ) as @xmath396-principal bundles over @xmath2 . now a cr density @xmath455 may be viewed as a smooth function on @xmath454 homogeneous of degree @xmath20 in the sense that @xmath456 for @xmath457 . therefore @xmath10 may be regarded as the subspace of the space of conformal densities @xmath13 which satisfy @xmath458 where on the left hand side , @xmath459 denotes pull back by the isometry of @xmath460 given by multiplication by @xmath461 . _ proof of theorem [ ambient ] . _ the conformally invariant operator @xmath14 of @xcite satisfies @xmath462 . since multiplication by @xmath461 is an isometry of @xmath460 and @xmath14 is a natural operator , it follows that @xmath463 satisfies if @xmath137 does . therefore @xmath14 induces an operator from @xmath10 to @xmath464 . define @xmath465 . the cr invariance of @xmath19 follows from the conformal invariance of @xmath14 . using the formulae in @xcite for the connection of the fefferman metric , one can see without difficulty that any component of an iterated covariant derivative of @xmath187 with respect to the levi - civita connection of @xmath403 has a universal expression in terms of pseudohermitian covariant derivatives of @xmath137 and the pseudohermitian metric , torsion and curvature . since @xmath14 is a natural differential operator associated to pseudo - riemannian manifolds , it follows that @xmath19 is a natural differential operator associated to pseudohermitian manifolds in the sense that it also has such a universal expression . it is easily seen that the principal part of @xmath19 agrees with that of @xmath18 . the conformal operators @xmath14 have been shown in @xcite and @xcite to be self - adjoint ; the argument in @xcite is formal and valid in general signature . from this , the self - adjointness of @xmath19 is an easy consequence : a nondegenerate hypersurface @xmath474 inheirits a cr structure . this was the setting for fefferman s original construction of the ambient metric and conformal metric in @xcite , @xcite . the ambient metric is defined above on @xmath475 , which we may identify with @xmath476 over a neighborhood @xmath477 of @xmath0 , since @xmath478 is the restriction of @xmath479 . introduce a variable @xmath480 , which is interpreted as a fiber coordinate on @xmath481 relative to the trivialization defined by @xmath482 . so the ambient metric is defined on @xmath483 . fefferman showed that @xmath0 has a smooth defining function @xmath484 which solves @xmath485 , where @xmath486 here the subscripts denote coordinate derivatives and the levi form @xmath487 has signature @xmath80 , @xmath81 . the function @xmath488 satisfies @xmath489 where @xmath490 . the hermitian matrix @xmath491 is therefore nondegenerate and defines a khler metric @xmath434 of signature @xmath492 which is approximately ricci flat along @xmath0 . this metric is clearly homogeneous of degree 2 and it is shown in @xcite that its restriction to @xmath493 ( fefferman s original definition of the fefferman metric ) agrees with the definition above for the fefferman metric of the pseudohermitian form @xmath494 . therefore the khler metric @xmath434 is the ambient metric associated to @xmath0 . we remark that in this formulation , the pseudohermitian form @xmath494 is not general : the condition @xmath495 on @xmath0 means that the closed form @xmath496 is volume normalized with respect to @xmath76 . this forces conditions on @xmath76 : if @xmath497 then @xmath76 is pseudo - einstein ( @xcite ) ; the corresponding condition for @xmath35 is given in @xcite . invariance under rescaling @xmath76 is replaced by invariance under biholomorphic changes of coordinates . for @xmath0 a hypersurface in @xmath45 , a section of @xmath10 can be regarded as a function on @xmath498 homogeneous of degree @xmath20 in @xmath499 . the invariant operators @xmath19 are therefore realized as the obstruction to extending such a function to be homogeneous and harmonic with respect to the laplacian @xmath500 in the khler metric @xmath434 , or equivalently by applying @xmath501 to an arbitrary homogeneous extension . we next show how this realization of the operators @xmath19 can be reformulated in terms of the khler metric @xmath502 on @xmath503 . if @xmath484 solves @xmath495 exactly , then @xmath403 is the khler - einstein metric of cheng - yau @xcite . we begin by considering the relationship between the laplacians of @xmath504 and @xmath403 for a general metric of the form , not necessarily assuming that @xmath484 is an approximate solution of @xmath495 . we continue to use capital indices @xmath505 which run between @xmath306 and @xmath506 , so for example @xmath507 lower case indices @xmath508 will be raised and lowered using @xmath509 and its inverse . as observed in @xcite , the choice of defining function @xmath484 determines near @xmath0 a @xmath84 vector field @xmath510 and a scalar function @xmath511 ( called the transverse curvature in @xcite ) , as follows . observe first that since @xmath512 is nondegenerate , there is a unique direction in @xmath513 which is orthogonal to @xmath514 with respect to @xmath515 . therefore , near @xmath0 there is a unique @xmath84 vector field @xmath510 satisfying @xmath516 for some uniquely determined smooth function @xmath511 , and we have @xmath517 [ laplprop ] when acting on functions which are homogeneous of degree @xmath20 with respect to @xmath499 , we have @xmath518 here @xmath519 denotes the khler laplacian @xmath520 and similarly for @xmath521 . our first observation is a matrix factorization of @xmath504 given by : @xmath522 taking inverses and multiplying out gives @xmath523 from which we obtain @xmath524 we want to rewrite the coefficients of the last three terms . define @xmath525 so that @xmath526 . then @xmath527 is nondegenerate on @xmath514 , so we may consider its inverse there , extend to annhilate @xmath528 , and thus obtain a tensor @xmath529 defined by the relations @xmath530 and @xmath531 . now gives @xmath532 , so @xmath533 . from this we conclude that @xmath534 extends smoothly across @xmath535 , and that @xmath536 substituting into and using homogeneity gives . * remarks . * the factorization also gives a direct relation between the metrics @xmath504 and @xmath403 : @xmath537 . $ ] note that @xmath538 is the @xmath504-length squared of the @xmath84 euler field @xmath539 . define differential operators in a neighborhood of @xmath0 in @xmath45 by @xmath540 by the observation in the proof of proposition [ laplprop ] , @xmath541 has coefficients smooth near @xmath0 . it follows from proposition [ laplprop ] that extending a section of @xmath10 to be harmonic with respect to @xmath542 is the same as extending the corresponding function on @xmath0 to be annihilated by @xmath541 . therefore we conclude : [ obschar ] if @xmath543 , the invariant operator @xmath19 is the ( normalized ) obstruction to extending a smooth function on @xmath0 to a smooth solution of @xmath544 , where @xmath484 is chosen to solve @xmath545 . in the setting of proposition [ obschar ] , the obstruction , or compatibility operators , were studied in @xcite , @xcite , @xcite . in @xcite , the operators @xmath19 were calculated explicitly for the sphere and the heisenberg group with their usual defining functions , which solve @xmath548 exactly . of course in this flat situation , the ambient metric is invariantly defined to all orders , so the invariant operators @xmath19 exist without the restriction @xmath16 . the result is that each @xmath19 is a product of various of the folland - stein operators @xmath549 on the heisenberg group , or the analogous operators of geller on the sphere . one can explicitly check that for the heisenberg group , the operator @xmath19 in @xcite agrees with the operator @xmath550 derived via tractors for flat pseudohermitian structures in proposition [ flatgoody ] . in @xcite , a connection on a neighborhood of @xmath0 in @xmath45 associated to an arbitrary defining function @xmath484 was derived along with an explicit formula for @xmath521 in terms of this connection ( proposition 2.1 of @xcite ) . this together with the expressions for the curvature of this connection given in @xcite ( for commuting derivatives ) in principle enable explicit calculation of the obstruction operators in the curved case in terms of pseudohermitian curvature and torsion and their derivatives , and the transverse curvature @xmath511 and its derivatives . for general @xmath484 , the obstruction operator was calculated in @xcite for @xmath35 and @xmath34 to be : @xmath551 where all the quantities involved are those for the pseudohermitian structure induced by @xmath484 on @xmath0 . the cr invariance of this operator was observed in @xcite , from which one concludes that for @xmath35 , @xmath37 is given by this formula for any choice of pseudohermitian structure . direct calculation shows that the operator @xmath552 discussed after the proof of proposition [ flatgoody ] agrees with @xmath37 . when @xmath553 , the invariant operators can also be characterized as obstructions , or equivalently as log term coefficients , when solving @xmath554 . this is analogous to the corresponding characterization of the conformal operators @xmath14 given in @xcite . the resolvent @xmath555 has been studied in @xcite . if @xmath547 is a function on @xmath45 , then @xmath556 therefore , if @xmath557 and @xmath558 , then @xmath559 is the ( normalized ) obstruction to solving @xmath560 with @xmath561 smooth and @xmath562 on @xmath0 , where @xmath484 is chosen to solve @xmath545 . the relation is a direct calculation from . one expands the right hand side of applied to @xmath561 using the leibnitz rule and simplifies using , , , and the fact that @xmath563 this latter is a consequence of substituting into @xmath564 . we close with a brief discussion of cr @xmath46-curvature , considered also in @xcite . first recall branson s ( @xcite ) formulation of @xmath46-curvature in conformal geometry . for this discussion denote by @xmath567 the operator @xmath14 in dimension @xmath4 . fix @xmath568 . the construction of @xcite shows that the operator @xmath567 is natural in the strong sense that @xmath569 may be written as a linear combination of complete contractions of products of covariant derivatives of the curvature tensor of a representative for the conformal structure with covariant derivatives of @xmath137 , with coefficients which are rational in the dimension @xmath4 . also , the zeroth order term of @xmath567 may be written as @xmath570 for a scalar riemannian invariant @xmath571 with coefficients which are rational in @xmath4 and regular at @xmath572 . the @xmath46-curvature in even dimension @xmath4 is then defined as @xmath573 . an analytic continuation argument in the dimension then shows that under the conformal rescaling @xmath574 , we have @xmath575 . if @xmath0 is a pseudohermitian manifold , the @xmath46-curvature of its fefferman metric is invariant under rotations , so defines a function on @xmath0 which we denote by @xmath576 . it is then natural to define the cr @xmath46-curvature of @xmath577 by @xmath578 . observations above imply that @xmath579 is given by a universal formula in the pseudohermitian metric , torsion , curvature and their covariant derivatives . the transformation law for conformal @xmath46-curvature gives immediately that if @xmath580 , then @xmath581 for @xmath24 , the conformal @xmath46-curvature is given by @xmath582 , where @xmath583 denotes the scalar curvature . using lee s formulae for @xmath584 , @xmath583 , and @xmath585 , it is straightforward to calculate this for the fefferman metric ; one obtains that the @xmath46-curvature for 3-dimensional cr manifolds is given by : @xmath586 this quantity was introduced by hirachi @xcite , who showed that it gives the coefficient of the @xmath452 term in the szeg kernel . he also established the transformation law in terms of the operator by direct calculation . the role of @xmath46-curvature in cr geometry is not clear . in the conformal case , the total @xmath46-curvature @xmath587 is an interesting invariant of a compact conformal manifold . however , since the expression in is a divergence , it follows that we have @xmath588 for the corresponding integral for 3-dimensional cr manifolds . also , the @xmath46-curvature vanishes identically for a hypersurface in @xmath45 with pseudohermitian structure induced by a solution of @xmath589 . a proof of this fact using a characterization of @xmath46-curvature in terms of the ambient metric is given in @xcite . it can also be seen directly from the definition by noting that the zeroth order term @xmath590 for the conformal operator for the fefferman metric is a multiple of @xmath591 . however , already @xmath592 vanishes to second order at @xmath593 . , scalar pseudo - hermitian invariants and the szeg kernel on three - dimensional cr manifolds , in complex geometry , 1990 osaka conf . marcel dekker lecture notes in pure and applied math . * 143 * ( 1993 ) , 6776 . |
numerical models of the dynamically extended atmospheres of long period variable or mira stars have shown that their winds have a very simple , power law structure when averaged over the pulsation cycle . this structure is stable and robust despite the pulsational wave disturbances , and appears to be strongly self - regulated . observational studies support these conclusions . the numerical models also show that dust - free winds are nearly adiabatic , with little heating or cooling . however , the classical , steady , adiabatic wind solution to the hydrodynamic equations fails to account for an extensive region of nearly constant outflow velocity . an important process or group of processes is missing from this solution . since gas parcels moving out in the wind are periodically overrun by pulsational waves , we investigate analytic solutions which include the effects of wave pressure , heating , and the resulting entropy changes . in the case of dust - free winds we find that only a modest amount of wave pressure is needed to derive an analytic model for a steady , constant velocity , locally adiabatic outflow . wave pressure is represented with a term like that in the reynolds turbulence equation for the mean velocity . the waves damp relatively quickly with radius , as a result of the work they do in driving the mean flow . although the pressure from individual waves is modest , the waves are likely the primary agent of the self - regulation of the dust - free winds . in dusty miras , the numerical models show the radiation pressure on grains and the subsequent momentum transfer to the gas , play the dominant roles in driving the wind , and wave pressure is not very important . in the models of the dusty wind , the gas variables also adopt a power law dependence on radius . heating is required at all radii to maintain this flow , and grain heating and heat transfer to the gas are significant . both hydrodynamic and gas / grain thermal feedbacks can transform the flow towards particular self - regulated forms . [ firstpage ] stars : late - type stars : mass loss hydrodynamics | astro-ph0406127 | the astrophysical theory of hydrodynamic winds began with the solar wind equations proposed by @xcite , also see @xcite . parker s model predicted that the pressure gradient between the sun s outer atmosphere and the interstellar medium would drive the wind . in subsequent years parker s solutions to the hydrodynamic equations have proven to be a powerful tool for studying stellar winds in general . ( the analogous @xcite solutions for spherical accretion problems have proven to be equally important . ) @xcite constructed shock heated models of the solar corona and wind , which can be regarded as parker winds with an additional heating term . however , bird s models were not designed to capture the highly nonlinear shock dynamics of long period variable star atmospheres ( henceforth lpvs ) . the large - amplitude , quasi - ballistic motions behind lpv shocks were not considered ( see @xcite and @xcite ) . nonetheless , a number of important results relevant to miras were anticipated , including : acoustic heating of the circumstellar gas , a dynamic balance between shock heating and expansion cooling in the wind , and self - regulation in the wind flow . bird proposed these as driving forces for the solar wind and corona , which are now believed to be the result of magnetohydrodynamic dissipation ( e.g. , the review of @xcite ) . at the present time , these processes appear more relevant to the warm extended atmospheres and winds of dust - free lpvs ( see the reviews of @xcite , @xcite for discussions of these structures ) . as will be described below , self - regulating processes are very important , but bird s conjecture that the self - regulation works to maintain a constant mach number throughout the flow is not strictly correct . hartmann and macgregor ( @xcite , @xcite , also see @xcite ) proposed an application of solar corona and wind models with alfvn wave heating and pressure to lpv winds . these processes are similar to the shock pressure and heating processes described below . the hartmann and macgregor models assumed an approximately hydrostatic corona , with waves viewed as a modest perturbation , in contrast to the strong shocks observed in lpvs . specifically , these are polytropic wind models , without shock entropy generation . both shock and alfvn waves may play a role in lpv winds , but we believe that the strong pulsational shocks are dominant , so we will not consider magnetic effects in this paper . because lpvs have very large amplitude pulsations , they might be expected to provide difficulties for smooth wind solutions of the hydrodynamic equations . indeed , the numerical models of @xcite , and @xcite show that the atmospheres of pulsating miras are highly dynamic . this is also true of all published numerical models of these and related stars , e.g. , @xcite , @xcite , @xcite , @xcite , @xcite , and @xcite . most of these models do not include non - lte radiative cooling , so in the remainder of this paper we will primarily refer to bowen s models , and the particular runs described below . see @xcite for a comparison of various models . ) in the numerical models , the density near the photosphere retains a roughly exponential decline , but departs from this to a power law dependence at density value that depends on the mass outflow rate . in the transition from exponential to power law decline , the dynamical lifting by shocks followed by ( in some cases ) radiation pressure on grains lifts and accelerates the material . in this flatter density profile , more gas has been lifted to large radii , and this along with radiation pressure on dust grains in some cases , provide the means for driving much enhanced winds . bowen s models show that the inner atmosphere is a complex region where shock acceleration and heating and cooling all play important roles ( see @xcite ) . on the other hand , the wind region at large radii has a simpler structure , and the gas variables averaged over pulsation period can be well approximated as simple power law functions of radius ( see figures 1 - 3 and discussion below ) . such simple profile forms suggest that it should be possible to construct simple analytic models in the tradition of parker and bird , at least for the smooth wind region . we will not consider lpv observations in detail in this paper , since our main goal is to understand the regularities revealed by numerical models . in addition , the constraints on theory provided by observation ( e.g. , spectra ) are indirect , and generally the comparison between theory and observation is best done with the aid of detailed numerical models ( see e.g. , @xcite , @xcite and references therein ) . nonetheless , some recent observations provide quite direct information about the gas variables in the winds and extended atmospheres of lpvs , and a brief mention of these provides a context for the subsequent work . first of all , a major assumption of both analytic and numerical models is that the lpv winds are approximately spherically symmetric . in spite of evidence for modest deviations from symmetry , or asymmetries in binary systems , there is much evidence from a variety of wavebands for approximate symmetry in most cases ( e.g. , @xcite ) . recent infrared interferometric observations of `` dust shells '' around miras are also generally consistent with spherically symmetric models , though not with uniform dust density distributions . the data are better fit by models including a few distinct shells where the dust emission is high ( see e.g. , @xcite , @xcite , @xcite , @xcite , and references therein ) . however , these observations probe a region located at a distance of only a few stellar radii , where conditions are described as dynamic and complex . in fact , the numerical models show that this is the region where shocks have grown to very nonlinear amplitudes , where dust forms if it is able to , and where the wind is just beginning to be accelerated ( the `` shock acceleration '' and `` dissipation '' zones of @xcite ) . thus , these observations generally support the picture provided by the models , and do not contradict the notion of a generally smooth wind structure . observations and models are in general agreement about basic wind parameters like the mass loss rate or flow velocity . from co observations ( e.g. , @xcite , @xcite , @xcite , @xcite , @xcite , and @xcite , also see @xcite for mid - infrared results ) we know that mass loss rates range from @xmath0 up to @xmath1 for miras in agreement with bowen s models ( @xcite , @xcite , & @xcite ) . above about @xmath2 the stars are generally classified as oh - ir sources . in this paper we are most interested in low mass loss rates , in relatively dust - free cases . the dust free models develop an approximately constant outflow velocity with a speed well below the escape speed , and this pattern persists over many stellar radii . in this region the flow is also subsonic . thus , this constant velocity is not the same as the coasting flow in standard wind models well beyond the sonic point . the observations also find low wind velocities ( of order 5 - 10 km / s ) and little evidence for changes in the flow velocity through the wind . for the miras co observations are able to probe the flow at large distances from the star , and so , provide important constraints on any model . interestingly , cepheid wind velocities also seem to be low relative to the escape speed ( @xcite ) . in the following sections we will derive analytic wind models , described by hydrodynamic equations which including approximate , averaged terms , like those in the reynolds turbulence equations , to represent the effects of nonlinear pulsational waves on the flow . we then discuss the self - regulatory feedbacks in these winds , and use several approaches to argue that these processes select constant velocity outflows over other possible wind solutions . the usual parker wind equations are derived from the steady , spherically symmetric hydrodynamic continuity and momentum equations , which can be written as , @xmath3 @xmath4 where the variables , @xmath5 , @xmath6 , @xmath7 , @xmath8 , @xmath9 , and @xmath10 denote the mass loss rate , total mass of the star , the density , pressure , wind speed , and the distance from the center of the star , respectively . classical steady wind solutions also assume a polytropic equation of state , @xmath11 , with constant k , @xmath12 . in this paper we will assume a constant rate of mass loss as well . two properties of the outer region of the atmosphere suggest that it can be described as approximately locally adiabatic : 1 ) the temperatures there are in a range where radiative cooling is inefficient , and 2 ) the shocks in this region are relatively weak . the latter feature is evident in figures 1 - 3 , which show density , velocity , and temperature profiles taken from the end of a very long run ( about 1000 pulsation cycles ) of bowen s modeling code , without dust . detailed descriptions of the atmospheric dynamics can be found in @xcite . here we merely note that the code numerically solves the one - dimensional , spherically symmetric , hydrodynamic equations with a @xcite type lagrangian method , including a conventional artificial viscosity algorithm for stabilizing strong shocks . a number of different chemical rate , momentum and energy sources are calculated explicitly for each gas element . these include : the atomic radiative cooling rates , the free electron abundance , energy changes due to ionization and recombination . they also include simple approximations to the radiative transfer , grain formation , radiation pressure on grains , and grain heating and cooling . the effects of most of these processes will not be considered in this paper , which focusses instead on understanding the global hydrodynamics . however , the numerical models provide a standard for comparison , as well as assurance that the analytic models are not unrealistic . the parameters of the numerical model shown in figures 1 - 3 are essentially the same as those of a dust - free , @xmath13 , `` standard '' model of @xcite , though the model shown here was computed with increased spatial resolution . the numerical models will be compared to analytic models below , especially in section 2.6 . the wind density profile shown in figure 1 has an approximately @xmath14 form , while the form of the corresponding temperature profile is nearly @xmath15 ( fig . 3 ) . note that beyond a radius of @xmath16 in the figures the model wind flow is not fully relaxed . ( indeed , fig . 3 suggests that the flow is not thermally relaxed beyond @xmath17 . ) there is , in fact , a polytropic solution to equations ( 1 ) and ( 2 ) , with a @xmath14 density profile , and a @xmath15 temperature profile . however , the value of the polytropic index in this model is @xmath18 . not only does this not correspond to the adiabatic value of @xmath19 expected in the absence of cooling and heating , it does not have any obvious physical meaning . we should note , however , that parker argued against values of @xmath20 on the grounds that larger values did not yield `` a solution beginning at low velocity close to the sun and extending outward to zero pressures at infinity '' ( @xcite , pg . he also noted the need for a heating source when @xmath21 . what about the adiabatic , @xmath19 wind solution ? to illustrate the problems with this classical solution we integrate equation ( 2 ) from an inner radius @xmath22 to an arbitrary outer radius r , use the ideal gas law and employ the definition of the adiabatic sound speed to derive the following bernoulli equation , @xmath23 the numerical models show that the sound speed c decreases with radius approximately as @xmath24 , as does the escape velocity @xmath25 . thus , according to the equation ( 3 ) , the flow velocity u must decrease with radius by comparable amounts . it does not ; bowen s models show constant velocity in the wind at all times ( see fig . 2 ) . we note also that , if all the wind flow lies on a single adiabat in the thermodynamic phase space , and has a @xmath14 density profile , then the ideal gas law implies that the temperature goes as @xmath26 . with the long run we have performed , which yields a large wind region , the numerical profile does appear shallower than an adiabatic one . maintaining a shallower temperature profile requires extra heating and momentum sources in the wind . in fact , a likely reason for discrepancies between numerical models and the adiabatic solution is that , although the pulsational shocks are weak in the wind region , their momentum , heat , and entropy inputs are not negligible . the question is how to capture the effects of waves in a relatively simple analytic description of the overall wind flow , averaged over many pulsation cycles ? the problem is like that of describing turbulent flows in which large fluctuations on many scales coexist with a mean flow and long - lived coherent structures . the characteristics of the regular , pulsational shocks , which propagate outward through the wind , are very different from the stochastic , fluctuations described in the theory of well - developed , homogeneous turbulence ( e.g. , @xcite , @xcite ) . nonetheless , in both cases we are dealing with disturbances on small spatial scales and short timescales , which we do nt necessarily need to resolve in order to describe their average effect on the flow . in fact , the fundamental idea of turbulence theory , that the flow can be divided into a mean and a fluctuating part , provides a very appropriate approach to generalizing the classical stellar wind equations . we can not simply adopt the averaging procedures used to derive the reynolds equations of turbulence theory , since these depend on the random nature of the fluctuations in homogeneous turbulence . we can , however , provide physical justifications for including similar averaged fluctuation terms in a set of phenomenological reynolds equations for mean wind flows . the numerical models show that each pulsational shock provides a compression and an outward boost to an individual gas element , and the gas element follows a quasi - ballistic free - fall trajectory behind the shock ( @xcite , @xcite , @xcite , and the review of @xcite ) . more precisely , for a periodic pattern of motion in the atmosphere , the material at larger r must have smaller outward postshock speeds . this picture is good deep in the atmosphere . if the postshock speed is slightly too high , the mass element encounters the next shock at slightly larger than it did the last one . this gives a small net outward motion , as is seen in the atmosphere around 1.5 - 2.5 stellar radii . where the flow dominates , the material still responds nearly ballistically , setting up a stable pattern relative to the mean flow speed and providing a small increment of momentum to the gas in each cycle . this allows us to separate the mean flow from the oscillations , which average approximately to zero ( see appendix a for details ) . to incorporate the effects of the shocks on the mean flow , an additional term is included in the ( inviscid ) momentum equation . with the approximation of the previous paragraph , this term can be written as the divergence of a reynolds stress , which in the spherically symmetric case is just the radial derivative of the mean square velocity fluctuation , or @xmath27 . physically , this is the ( one - dimensional ) wave pressure gradient . in the bernoulli equation ( 3 ) it contributes a @xmath28 term , like the sound speed term . the net effect of shock heating can be included as a term in the mean , steady state energy equation ( see appendix a ) , which can be written as , @xmath29 the variables , @xmath30 and @xmath31 denote the internal energy per unit gas mass , and the shock heating function , respectively . henceforth , we define u as the sum of the mean ( @xmath32 ) and fluctuating ( @xmath33 ) velocity components . in turbulence theory , we assume that this is an ensemble average over many nearly identical systems , distinguished by the details of their random fluctuations . in the present case the average is also assumed to be over all pulsational phases . similarly , @xmath34 and p generally have both mean and fluctuating parts , but we will not need to adopt special symbols for the separate parts ( except in appendix a ) . we re - emphasize that in this equation we neglect the effects of the interaction between wind material and the stellar radiation , the interaction with the radiation generated in different parts of the winds , and energy exchanges between gas and dust grains . the reynolds formalism provides an additional equation for the mean square fluctuating velocity ( see mccomb , sec . 1.3.2 ) . in the case of a steady , inviscid , constant mean velocity radial flow , this equation can be written , @xmath35 \nonumber\\ & & -2{\sigma}^2 \frac{\partial u}{\partial r } - 2\gamma.\end{aligned}\ ] ] the first two terms on the right - hand - side represent turbulent ( or wave ) energy diffusion by nonlinear couplings , which we expect to fall off quickly with radius in this case , and so will neglect them . ( this neglect is part of a closure approximation for the reynolds moment equation set . ) what remains is a relation between the local wave pressure gradient and the shock heating , @xmath36 ( where the last term is small in winds with nearly constant mean velocity ) . although we now have the required energy and momentum source terms and a relation between them , the equation set is not quite complete . we employ a locally , but not globally , adiabatic equation of state to describe the effects of shock heating in the outer atmosphere . physically , in each pulsation cycle , any parcel of gas goes through an irreversible ( non - closed ) cycle in a p - v ( pressure - specific volume ) phase diagram . a pulsational shock wave pushes a gas parcel along a hugoniot curve in the phase diagram , and off its initial adiabat ( e.g.,@xcite ) . radiative cooling at constant pressure behind the shock then moves it to lower specific volume ( i.e. , compresses it ) . downstream from the shock the parcel evolves along a new adiabat to higher specific volumes and lower pressures in the rarefaction region , until it is hit by the next shock . a possible laboratory analogue of this physics is provided by the reverberating shock cavities produced in the ( gun - type ) experiments ( see @xcite ) . for weak shocks the specific entropy changes are not great , but they accumulate . gas parcels farther out in the wind have experienced more of these cycles , and so , lie on adiabats that are farther from the initial one than those in the inner wind . we assume that the specific entropy of the mean flow is time - independent and a smooth function of radius , like the other gas variables . we believe this gradual entropy change is an important factor in making the wind region of bowen s dust - free mira models appear quite adiabatic , but with a constant outflow velocity , which is not characteristic of a classical adiabatic wind . we can write the locally adiabatic equation of state as , @xmath37 where @xmath19 . because gas parcels at different radii are on different adiabats , k is a function of the radius , rather than a constant . ( @xcite have also recently considered model keplerian disks with entropy gradients due to local heating . ) since the gas is locally adiabatic , the perfect gas relationship for @xmath30 is valid ( locally ) at any radius , @xmath38 this relation completes the equation set . we have introduced three new terms into the parker wind equations , which describe wave pressure , shock heating , and shock entropy production . none of these extra phenomenological terms are found in the gas equations that are the basis of published numerical models ( e.g. , @xcite ) . those models follow the time - dependent pulsation and shock phenomena explicitly . we believe all of the phenomenological terms are necessary to describe the locally adiabatic _ mean _ flow in pulsationally driven winds . we note that the classical , linear treatment of waves propagating into a stellar atmosphere indicates that long period waves should be reflected at the surface . pijpers and collaborators , described the effects of stochastic acoustic waves in detail in a series of papers ( @xcite , @xcite , @xcite , @xcite , and @xcite ) . they found that such waves can propagate in the context of a general outflow ( see @xcite ) . @xcite found that the power needed to maintain large amplitude photospheric oscillations is actually lower for long periods , and these induce nonlinear wave propagation ( shocks ) traveling out through the atmosphere . here we take the existence of transmitted waves as a given , and focus on the large - scale , nonadiabatic effects of mildly nonlinear waves in the wind region only . motivated by the observations , we begin by considering the simple case in which the mean wind velocity is constant , @xmath39 . in this case , equation ( 1 ) immediately gives @xmath7 , and the mean flow momentum ( eq . ( 2 ) ) can be written , @xmath40 where @xmath7 and p are now the reynolds average quantities . with the additional assumption that all three terms on the left - hand - side of this equation scale the same way with radius throughout the wind flow , we can readily obtain a solution to this and the energy equation . this assumption appears quite strong , but is physically reasonable for the following reason . the wind extends over a very large range of radii . the gas thermal energy and temperature do not fall off as rapidly with radius in dust - free models as expected in an adiabatic flow . this suggests a heat source operates throughout ; shock heating is the only available source . ( we note that photo - heating of grains will have a similar scaling with radius , but we assuming negligible grain populations for the present . ) this fact ties the thermal pressure to the shocks and suggests that it scales like the wave pressure . beyond this physical argument , we can briefly note the mathematical possibilities . first is the possibility that thermal pressure dominates in part of the wind and wave pressure elsewhere . in the former region we will have a parker thermal wind , with the observational difficulties mentioned above . in the wave dominated region , we will get the same scaling with radius , which derives from balancing the gravitational term . since shock compression does provide thermal heating , this limit is nt physically self - consistent . a second possibility is that the thermal and wave pressures have different functional forms whose sum contrives to balance the gravitational term . for example , they could both have the same power law form derived below , but modulated by oscillatory parts which have the same amplitude for both , but which are @xmath41 out of phase . any other solution of this type would have to be similarly fine - tuned . given that the wind extends over orders of magnitude in radial extent , this seems physically contrived . we will argue later that the solution below is also preferred by self - regulatory processes . specifically , we assume that @xmath42 for some constant a , then equation ( 9 ) becomes , @xmath43 now @xmath7 can be substituted from the continuity equation and the result can be integrated ( from a radius r to @xmath44 ) to get @xmath45 assuming @xmath7 and p go to zero at infinite radius . it is interesting that this equation has the form @xmath46 of a constant velocity parker wind , since @xmath47 , though we here view the solution as only locally adiabatic , rather than globally barotropic . next , we want to simplify the energy equation ( 4 ) in this constant velocity case . we need the following result in taking the mean of the term on the left - hand - side of equation ( 4 ) , @xmath48 where @xmath7 is used here as the total density ( mean plus fluctuating parts ) . this equation assumes that the mean of odd powers of the fluctuating velocity vanish , and that the correlation @xmath49 is negligible ( see appendix a ) . henceforth , we drop the brackets @xmath50 , and assume again that @xmath7 and p refer to the mean quantities . when we use this result , pull @xmath51 factors out of partial derivatives , and substitute for @xmath30 from equation ( 8) , the energy equation becomes , @xmath52 to further simplify we can substitute for p with equation ( 10 ) , for @xmath53 with equation ( 6 ) , use the substitution @xmath54 . the result reduces to , @xmath55 and yields in turn , @xmath56 the physical interpretation of the last equation is that the shock heating rate is doing work against the gravitational potential @xmath57 at each radius , on a flow timescale of about @xmath58 . it is reducing the effective gravity ( i.e. , the factor gm - a in eq . ( 10 ) ) , that must be overcome by flow down the thermal pressure gradient . combining equation ( 10 ) with equation ( 7 ) and the continuity equation gives the following expression for the variable @xmath59 in equation ( 7 ) . @xmath60 and that equation becomes , @xmath61 a generalized ( position - dependent ) polytropic relation . to better appreciate why k varies with radius , recall that the entropy ( per unit mass ) of a perfect gas is , @xmath62 so equation ( 7 ) implies @xmath63 . thus , k is directly related to entropy , and its radial dependence implies a mean radial entropy gradient . _ as any gas element flows outward , it is overtaken by shocks moving through the wind , and each shock increases the entropy of the element . _ at any fixed radius the entropy production is balanced by the outward transport . the value for the constant a above implies that the wave energy is about 45% of the gravitational potential energy at all radii . this implies that @xmath64 should be about half the value of the escape velocity at any radius . the velocity jitter due to shocks in bowen s dust - free numerical models is considerably less than this , generally less than a third of the escape velocity . however , it is not clear that this velocity jitter should be directly identified with @xmath64 , see section 2.6 below . we expect difficulties in capturing the full range of velocity variation in the models . most of this variation occurs across thin shocks , and these short wavelength variations are not well resolved numerically . the constant velocity wind described above is an especially simple solution to the equations . in this section we consider steady accelerating winds with radially dependent mean velocities . we restrict our consideration to power law similarity solutions . mathematically , we are essentially exploring a specific kind of variation of the previous solution , but self - similar winds are the most relevant physically . specifically , we assume @xmath65 , so @xmath66 . in this case , the mean momentum equation is , @xmath67 to solve this equation we will again assume a specific form for the wave pressure term , that is , @xmath68 where the first term is as in the constant velocity case , and the second term is designed to adjust the wave pressure in accordance with the flow acceleration . like a , the coefficient b is independent of radius , but does depend on the exponent @xmath69 . note that in the case @xmath70 , this expression does not coincide with the previous one for constant flow unless we add an additional term of -b . this term complicates the algebra of the model considerably , and since we usually have @xmath71 ( see below ) , we will neglect it for the moment . we could also add other power law terms to the expression for the wave pressure , or pursue a series solution for the wind variables . we believe this would only complicate the equations without adding much physical insight . an expression for the pressure can now be derived by integrating the momentum equation as before . we obtain , @xmath72 the function k(r ) can be derived by substituting equation ( 7 ) for p. next we substitute equation ( 18 ) ( with nonconstant u ) into the @xmath73 equation ( 5 ) ( including the term for the flow of energy from the mean field to the fluctuating velocity field , @xmath74 ) . this produces a rather complex expression for the heating , @xmath75.\ ] ] finally , the above expressions can be substituted into the energy equation to derive the following expressions for the coefficients a and b as functions of the exponents @xmath12 and @xmath69 by equating the coefficients of the equal powers of @xmath69 , @xmath76 } { \bigl[(2-\gamma ) - ( 2-\delta)(3+\delta)(\gamma-1)\bigr ] } gm,\ ] ] @xmath77 } { \bigr[5(\gamma-1)(2-\delta ) - 4\delta(2-\gamma)\bigr ] } { u_0 ^ 2}.\ ] ] the equations ( 20)-(22 ) specify the net heating required to maintain the power law velocity profile . they do not directly take into account the dependence of specific heating processes , like shock heating , on the gas quantities , but rather assume that such processes can be regulated to the form above . however , the expression for the heating profile ( eq . ( 20 ) ) is so complex that it seems unlikely that realistic shock heating and radiative cooling processes would generate it . this is in contrast to the simple heating profile of equation ( 13 ) for the constant velocity outflow , and suggests that the simpler form would be preferred ( as it is in bowen s numerical models ) . this question will be examined more carefully in the next section . before continuing , however , we note a couple of peculiarities of the equations above . first of all , in the limit of @xmath78 and @xmath19 ( locally ) , the expressions for a and b become , @xmath79 @xmath80 b is negative , and not directly proportional to @xmath69 . this means that in the accelerating wind , the wave pressure could be reduced at @xmath81 relative to the constant velocity flow . physically , it is more reasonable to simply identify @xmath82 as the inner radius of the wind ( see eq . ( 18 ) ) . another peculiarity is that if , as we expect , the heating is dominated by the first term in equation ( 20 ) , it becomes negative when @xmath69 is increased above a value of 1/2 , which implies a cooling process . this also seems unphysical . the simple appearance of the profiles of gas variables in bowen s dust - free numerical models suggests that the system relaxes to a self - similar solution , independent of the details of the shock heating , and subsequent expansion cooling . we have suggested above that one key to deriving this solution is the assumption that the gas in the wind is _ only locally adiabatic_. this assumption allows for entropy production in weak shocks , which gradually evolves a gas element through a sequence of adiabats as it moves outward in the wind . the thermal state of the gas determines the environment traversed by the shocks , which in turn are responsible for producing this state . this feedback in the shock heating mechanism eventually works to regulate the wind to the @xmath14 density profile . the process of self - regulation is well illustrated by the early relaxation that occurs in bowen s numerical models , which are started with small ( but increasing ) pulsation amplitudes , and with an exponential atmospheric density profile . as described in @xcite the traveling wave parts of the pulsations steepen into strong shocks as they propagate down the initial exponential atmospheric profile . these shocks launch gas parcels out on quasi - ballistic trajectories ( @xcite and @xcite ) to such large radii that they are not able to fall back to their starting points before being hit by the next shock . as a result , the initial profile is stretched out into power law form . this causes a feedback , the shock acceleration is decreased in the flatter pressure profile , and the ballistic launch velocities are moderated . thus , if the pressure profile is steeper than @xmath83 , then the shock amplitude grows , pushing material out , until it matches that form ( see appendix b ) . if the density gradient is shallower , however , the loss of energy in pushing the gas will lead to declining shock amplitudes , and it will not be possible to hold the shallow profile up against gravity . a pressure gradient that is neither too steep , nor too flat , also maintains the flat outflow velocity . the steady gas variable profiles that are eventually established with finite amplitude pulsations include another region between the exponential atmosphere where shocks grow nonlinearly , and the power law wind region . this is the shock dissipation region where the power law density profile flattens , but the temperature and pressure increase with radius to a maximum , forming a warm `` calorisphere '' and the inner boundary of the wind . see the review of @xcite and references therein for more details . the development of this region in the numerical models provides a clear example of the effects of shock dissipation when density and pressure profiles flatten . this qualitative discussion on relaxation processes , which was partially anticipated by @xcite , helps us understand why there is a preferred wind profile . however , it is not yet sufficiently precise to predict the form of that profile . next we will consider some more quantitative approaches to the problem . the relaxation to the simple , constant velocity flow is driven , in part , by the tendency of such a system towards a state of least dissipation , or minimum entropy production . such states often correspond to those of maximum entropy content . the theorem of minimum entropy production for steady , nonequilibrium states derives from work of helmholtz , though it has been generalized and used in a variety of applications in the last few decades ( see e.g. , @xcite , @xcite , @xcite ) . proofs of the different versions of this theorem involve substantial restrictions , e.g. , that the system is not too far from thermodynamic equilibrium , or that there are linear relations between the flow variables and their driving force or source terms ( like the familiar linear stress - strain relation ) , see @xcite . these theorems have not been used much in astrophysics , where there are many non - equilibrium systems , but where these systems are often highly time - dependent . on the other hand , the winds of long - period variables appear to provide an astrophysical situation where the theorem can be usefully applied . if pulsations begin with small amplitudes and build up steadily , then the wind can adjust to this changing driving , and never find itself too far from a set of local equilibria . moreover , equation ( 9 ) ( neglecting the last term ) appears to give the linear relation required of the dissipative quantities in some proofs ( see @xcite ) . the minimization of shock dissipation also seems to be in accord with the qualitative considerations of the previous subsection . we can apply the theorem by comparing the net production of the entropy in the accelerating wind models . the theorem implies that the wind with minimal entropy production is the preferred state . in principle , we need to consider the details of the dissipational processes , which is very complex in the present case . fortunately , it seems clear physically that the entropy production in the steady winds must be directly related to the shock heating function @xmath84 ( see eq . thus , for the purposes of making a simple estimate , we define a function , h , as follows , @xmath85 where @xmath82 and @xmath86 are adopted inner and outer bounds of the wind region . the coefficient @xmath87 is included merely to absorb that common factor from the continuity equation into the definition of h , which then has units of velocity squared . substitution for @xmath84 and integration yields , @xmath88 where the functions a and b are given by equations ( 21 ) and ( 22 ) with @xmath19 , and @xmath89 . note that the outer boundary of the wind , @xmath90 , will generally be a function of @xmath69 , which governs the density falloff . for example , we can make a simple estimate by assuming that the outer boundary is where the mean flow velocity u equals the escape velocity . in that case , we have , @xmath91 the dimensionless factor @xmath92 , which appears both in the ratio of coefficients in equation ( 26 ) @xmath93 and in equation ( 27 ) , is the primary similarity parameter of the problem . for convenience , we name it , @xmath94 now , using equations ( 21 ) , ( 22 ) , and ( 27 ) we can approximate h as , @xmath95 then we take the derivative with respect to @xmath69 , and set it equal to zero to find the dissipation extrema . the resulting expression is very complicated , and not worth recording here without some simplifying manipulations and approximations . as yet , the only extra approximation we have used in deriving the expression for h ( or its derivative ) is the outer boundary estimate of equation ( 27 ) . however , at this point , it is helpful to adopt the approximation that @xmath96 , i.e. , that the inner - edge wind velocity is much less than the inner - edge escape velocity . this assumption is clearly satisfied in the numerical models . since @xmath97 , and @xmath98 , this is equivalent to the assumption that @xmath99 already noted above . however , it is not wise to neglect the b - term in equation ( 26 ) above if @xmath69 is small and negative , because the term @xmath100 can be relatively large . next , consider limits on the magnitude of @xmath69 . when @xmath69 is positive , the first term in equation ( 29 ) is larger than the second , and for @xmath101 it is negative . then the dissipation h would be negative , which is unphysical . thus , we do not need to consider large , positive values of @xmath69 . similarly , when @xmath102 the first term makes h negative , and unphysical . for large negative values of @xmath69 the second term of equation ( 29 ) is negative and dominant . in sum , it seems that the physically relevant region is where the exponent @xmath103 . these approximations justify dropping terms of order @xmath104 or @xmath105 , and then the derivative of h is given by , @xmath106 + \nonumber\\ & & 2\bigl(2\delta^2 - 2\delta + 11\bigr ) \bigl(49\delta^2 - 140\delta + 100\bigr ) \times \nonumber\\ & & \bigl(-4\delta^2 - 8\delta + 5\bigr ) \biggl(\frac{z_0}{2}\biggr)^{\frac{-2\delta}{1 + 2\delta } } + \nonumber\\ & & 32\delta \bigl(1 - \delta\bigr ) \bigl(2\delta^2 - 2\delta + 11\bigr)^2 \times \nonumber\\ & & \bigl(-7\delta + 10\bigr ) \biggl(\frac{z_0}{2}\biggr)^{\frac{2\delta}{1 + 2\delta}}.\end{aligned}\ ] ] this equation can be solved as a quadratic in the variable @xmath107 as a function of @xmath69 . the value of @xmath104 itself is then obtained for each value of @xmath69 . we find that equation ( 30 ) yields no positive , real values of @xmath104 for positive values of @xmath69 less than about 1/3 . with the constraints cited above this means that there is no significant range of positive values of @xmath69 that give physical solutions . that is , solutions with wind velocity increasing with radius do not satisfy the least dissipation constraint . on the other hand , there are solutions to equation ( 30 ) for negative values of @xmath69 . the value of @xmath108 increases monotonically ( and @xmath109 decreases monotonically ) as @xmath69 goes from 0 to increasing negative values . if we make the reasonable requirement that the radial extent of the outflow with @xmath110 , that is @xmath111 , be greater than a few , then @xmath112 must be less than about 0.1 . as the magnitude of @xmath112 is decreased below 0.1 , @xmath109 rises very rapidly , so the wind is very large for such small values of @xmath69 . in sum , the least dissipation theorem , as applied to the shock heating function , has constrained the family of wind solutions . specifically , the parameter @xmath69 is limited to a small range for optimal winds . because the physically relevant values of @xmath69 have a very small magnitude all the members of this one - parameter family are nearly constant velocity winds . this result is not completely rigorous , though physically reasonable . it is worth emphasizing that because self - regulation is a global process , global constraints determine the wind structure . as an integral of the momentum equation , the bernoulli equation provides another global constraint on wind structure . for the accelerating winds with wave pressure described equation ( 17 ) , the corresponding bernoulli equation is , @xmath113 where , as in equation ( 3 ) , @xmath114 is the escape velocity , @xmath115 is the global effective ratio of specific heats , and the integration constant @xmath116 is a function of the wind acceleration exponent @xmath69 only . in the case of an accelerating wind we have expressions for all of the terms on the left - hand - side of this equation , except the sound speed term . using equation ( 9 ) or ( 19 ) above , we find , @xmath117.\ ] ] since for @xmath112 small , we have @xmath118 , and then , @xmath119 the value of the constant @xmath116 can be determined by evaluating the left - hand - side of the bernoulli equation at the inner edge . in bowen s numerical models of dust - free winds we find that , @xmath120 the models also indicate that @xmath121 , and that @xmath51 is much less than either of these . thus , the terms on the left - hand - side of equation ( 31 ) nearly cancel , and @xmath122 has a relatively small value . next , we compare an accelerating wind , with @xmath112 small , but nonzero , to the constant velocity wind having the same outflow velocity at the inner wind radius . we also assume that the inner radius and stellar mass are the same for both winds . taking the difference between the bernoulli equations for the two winds we obtain , @xmath123 to determine the difference between constants on the right - hand - side of this equation we evaluate it at @xmath124 . we substitute the result back into equation ( 35 ) , and solve this equation for the fractional difference in the mean flow velocity , as a function of r , @xmath125 \nonumber\\ & & - \frac{2b}{u_0 ^ 2 } \biggl[\biggl(\frac{r}{r_0}\biggr)^{2\delta } - 1\biggr].\end{aligned}\ ] ] because @xmath126 is generally much less than the escape velocity throughout the wind , except possibly the outermost parts , and because the magnitude of @xmath127 is comparable to @xmath126 , the first term on the right - hand - side of this equation will generally dominate the others . ( note that in the limit of small @xmath128 , the last term is also proportional to @xmath69 , so it does not exceed the first term . ) however , the sign of the first term is such that it would imply a falling u for a positive @xmath69 , and a rising u for negative @xmath69 , which contradicts the definition @xmath129 . the only alternative is @xmath70 . as before , this conclusion is not mathematically rigorous , because it is not always true that the first term on the right - hand - side of the above equation must dominate . for example , the parameter b goes to positive infinity as @xmath69 approaches a value of 10/7 . however , such cases are exceptional , and unphysical . in sum , we generally expect that _ thermal winds driven by nonlinear acoustic waves have constant outflow velocities . _ one remaining aspect of the dust - free winds to consider is what happens in the outermost regions ? for brevity , we will confine our discussion to the constant velocity wind . the escape velocity @xmath114 , the sound speed c ( eq.(32 ) ) , and the velocity dispersion associated with the waves , @xmath64 ( eq . ( 18 ) ) all decrease with as @xmath130 in this case . in the outer wind the constant flow velocity will eventually equal and exceed each of these velocities in turn . the effective @xmath64 is generally somewhat less than the sound speed c , so we might expect a transition from the constant velocity wind to a nearly adiabatic outflow before the material escapes . specifically , for the constant velocity wind with @xmath131 , the parameter a = 2gm/5 ( eq . ( 21 ) ) , and equations ( 18 ) and ( 32 ) yield , @xmath132 @xmath133 thus , we see that when @xmath51 equals @xmath134 , it very nearly equals c. however , these velocities are less than half the escape velocity , and there is no reason to expect an abrupt increase of c(r ) . that is , we do not expect a transition to supersonic flow through a parker critical point ( where @xmath135 ) . physically , as @xmath51 becomes comparable to @xmath64 and c , and all about equal to @xmath136 , we might expect successive pulsational waves to propel material above the local escape velocity , and thus free of the star . it appears that the outermost wind must be nonsteady . the result that the sound speed is about half the local escape velocity is consistent with the numerical models . however , the velocity amplitude of the pulsations is much less than that . as discussed above , this likely due in part to the limited spatial resolution of the models . moreover , the wave pressure used here could well represent the aggregate effect of several pulsational waves . ( in fact , it is possible for one shock to overtake and merge with another , though this process is not well - resolved in the numerical models . ) we can illustrate this last point with a more detailed comparison to the numerical model . we consider what wave pressure boosts are sufficient to push a gas element through a radius change of order unity , @xmath137 . the wave pressure drives the constant mean flow , and the time for the gas element to flow this distance is @xmath138 . the number of pulsational shocks passing through the gas element in this time is about , @xmath139 , where t is the pulsation period . we identify the net wave pressure with the sum of the pressures in each shock to get , @xmath140 . according to equation ( 37 ) the model predicts , @xmath141 , and so it also predicts , @xmath142 as a specific example we take @xmath143 , and @xmath144 . this is a radius where the numerical model shows that strong shocks present at smaller radii have settled into the steady wind . there are still significant velocity fluctuations well beyond this radius , but the adopted mean outflow velocity is a fair average for the whole wind region . with these values and a pulsation period of about one year , we compute the following values for the quantities defined in the previous paragraph , @xmath145 @xmath146 we set @xmath147 equal to this latter value at the adopted radius , use the scaling of the constant velocity outflow model ( @xmath148 ) , and add the mean outflow u to derive the lower dashed curve in figure 2 . this curve is generally comparable to the velocity jumps of the shocks in the inner wind , where these jumps are fairly well resolved . in fact , the velocity jumps appear to be somewhat larger than the curve in the inner wind , but this is reasonable since some of the shock energy will go into thermal heating as well as doing work against the star s gravity . in the present example the work per mass of gas moved is @xmath149 . the corresponding shock heating per gas mass , integrated along the path , is @xmath150 . the integrated work done by the wave pressure is about half what is needed . this confirms the impression given by the figure that there is not too much wave energy available for heating . in sum , in terms of the wind driving energetics , wave pressure and thermal pressure play comparable roles . a fraction of the thermal energy is replaced by shock dissipation , as required for an effective adiabatic index of 3/2 . if dust grains , atoms or molecules are able to absorb a fraction of the substantial mira radiative output , then radiation pressure can play a substantial role in driving or accelerating the wind . in the case of typical mira variables , the models of @xcite , and @xcite , show that models with significant radiation pressure on grains have higher mass loss rates and higher wind velocities . the wind regions are also much smaller in these models because the wind velocity surpasses the escape velocity at a smaller radii . the pulsational hydrodynamics is essentially frozen out in a dusty wind . there is , however , another regulatory process that limits the wind velocity and keeps the flow from over - refrigerating . in fact , the numerical models show that the gas temperature tracks the radiative equilibrium temperature quite closely , which implies a coupling of photons , grains and gas atoms . atom - grain collisions are the mechanism of both momentum and energy transfers ( @xcite ) . the gas flow is driven by the momentum transfer from each grain to individual atoms , and these are inevitably accompanied by energy transfers . the interaction will also decelerate the grain , and will also result in grain ablation if the relative velocity is too great . ablation will reduce the grain cross section , and the radiative driving . slow wind speeds allow more time for grain growth in dense regions , leading to more radiative driving . in particular , once the grains relax to a fixed size distribution , the fraction of the photons they absorb , and the momentum input from them , decrease as @xmath14 . in this case , two terms dominate the right - hand - side of a momentum equation , like equation ( 2 ) , the gravitational term and the radiative driving source term . these terms scale the same way with radius , allowing a constant velocity outflow solution . other solutions to the momentum equation are possible . however , because of the regulatory feedbacks , we expect that the process of grain growth towards an equilibrium size distribution should naturally be accompanied by wind acceleration to a constant terminal velocity . in summary , the numerical models indicate that the winds that develop in the extended atmosphere of long - period variables have a simple power law structure . the isentropic parker wind solution that matches this structure has a barotropic index of @xmath18 . in a real gas heating is required to maintain this index , but the parker solution does not provide information about the heating source . in the case of non - dusty miras , where radiation pressure is unimportant , we believe that the power law structure is generated and supported by shock waves which travel through the wind . since shocks dissipate energy and generate entropy , these winds have significant heat inputs and entropy gradients . in section 2 we presented phenomenological equations including the three relevant terms : wave pressure , wave heating and large - scale entropy gradients . we then studied a family of analytic thermal wind solutions to these equations with power law velocity profiles . these wind solutions generalize the classical isentropic parker solutions to cases in which the gas is only locally adiabatic . the simplest member of this family has a constant outflow velocity , and matches numerical models of non - dusty mira winds quite well . section 2 concluded with a discussion of why the constant outflow velocity solution may be preferred in nondusty miras . both the least dissipation theorem of linearized , nonequilibrium thermodynamics , and the bernoulli equation with reasonable physical constraints indicate that this is a preferred solution among the family of power law winds . this result provides reassurance of the basic correctness of bowen s numerical models , and the validity of mass loss estimates predicted by them . in section 3 we considered dusty winds in long - period variables . in this case the winds are driven by radiation pressure on dust grains , and the flow `` freezes out '' in the sense that thermal , wave - driven or turbulent pressures are not dynamically important . this allows the wind velocity to attain supersonic speeds without passing throught the critical point of classical thermal wind theory . nonetheless , bowen s numerical models show that there are significant couplings between the radiation field , an equilibrium distribution of dust grains , and the atomic gas . these couplings allow a regulated , constant velocity outflow to form and be maintained . we believe that the model of dust - free mira winds may provide a paradigm , and that gas dynamic solutions with acoustic wave pressure may have much more general application in astrophysics . specifically , such solutions will be relevant whenever there is substantial supersonic or magnetosonic ( velocity ) noise on small scales as compared to large scale velocities and thermal pressure gradients . laboratory examples are provided by the reverberating shock cavities described in @xcite and @xcite ) . other possible astrophysical examples include : 1 ) accretion disks with strong acoustic - convective turbulence , 2 ) galactic winds or outflows driven by continuing supernova shocks , and 3 ) the hot gas halos of galaxy clusters , containing galaxies moving at transonic or supersonic velocities , and into which there is continuing infall . the defining property is that the entropy or the function k have a nonzero gradient . we are grateful to s. owocki for helpful conversations . alard , c. et al . ( the macho collaboration ) 2001 , apj , 552 , 289 bird , g. a. 1964a , apj , 139 , 675 bird , g. a. 1964b , apj , 139 , 684 bird , g. a. 1964c , apj , 141 , 1455 bondi , h. 1952 , mnras , 112 , 195 bowen , g. h. 1988 , apj , 329 , 299 bowen , g. h. 1990 , in j. r. buchler , ed . , numerical modeling of nonlinear stellar pulsations , kluwer , dordrecht , p. 155 bowen , g. h. , & willson , l. a. 1991 , apj , 375 , l53 bujarrabal , v. , & alcolea , j. 1991 , a&a , 251 , 536 feuchtinger , m. u. , dorfi , e. a. , & hfner , s. 1993 , a&a , 273 , 513 fleischer , a. j. , gauger , a. , & sedlmayr , e. 1992 , a&a 266 , 321 fong , d. , meixner , m. , & shah , r. y. 2003 , apj , 582 , l39 glansdorff , p. , & prigogine , i. 1971 , thermodynamic theory of structure , stability , and fluctuations , wiley , new york groenewegen , m. a. t. , bass , f. , blommaert , j. a. d. l. , stehle , r. , josselin , e. , & tilanus , r. p. j. 1999 , a&as 140 , 197 guderley , g. 1942 , luftfahrtforschung , 19 , 302 hale , d. d. s. , et al . , 1997 apj , 490 , 407 hartmann , l. , & macgregor , k. b. 1980 , apj , 242 , 260 hartmann , l. , & macgregor , k. b. 1982 , apj , 257 , 264 hill , s. j. , & willson , l. a. 1979 apj , 229 , 1029 hfner , s. , & dorfi , e. a. 1997 , a&a , 319 , 648 hfner , s. , gautschy - loidl , r. , aringer , b. , & jorgensen , u. g. 2003 , a&a , 399 , 589 holmes , n. c. , m. ross , and w. j. nellis 1995 , phys . b , 52 , 15835 kahane , c. , & jura , m. 1994 , a&a , 290 , 183 kerschbaum , f. , & olofsson , h. 1998 , a&a , 336 , 654 klahr , h. h. , & bodenheimer , p. 2003 , apj , 582 , 869 koninx , j .- m. , & pijpers , f. p. 1993 , a&a , 265 , 183 landau , l. d. , & e. m. lifshitz 1959 , fluid mechanics , pergamon , new york lopez , b. , et al . , 1997 , apj , 488 , 807 macgregor , k. b. & charbonneau , p. 1997 , in j. r. jokipii , c. p. sonnett , & m. s. giampapa , eds . , cosmic winds and the heliosphere , university of arizona press , tucson , p. 339 mccomb , w. d. 1990 , the physics of fluid turbulence , clarendon , oxford monnier , j. d. , et al . 1997 , apj , 481 , 420 parker , e. n. 1958 , apj , 128 , 664 parker , e. n. 1963 , interplanetary dynamical processes , wiley interscience , new york parker , e. n. 1997 , in j. r. jokipii , c. p. sonnett , & m. s. giampapa , eds . , cosmic winds and the heliosphere , university of arizona press , tucson , p. 3 pijpers , f. p. 1993 , a&a , 267 , 471 pijpers , f. p. 1995 , a&a , 295 , 435 pijpers , f. p. & hearn , a. g. 1989a , a&a , 209 , 198 pijpers , f. p. & hearn , a. g. 1989b , a&a , 215 , 334 prigogine , i. 1980 from being to becoming : time and complexity in the physical sciences , w. h. freeman , san francisco richtmyer , r. d. & morton , k. w. 1967 , difference methods for initial - value problems , 2nd ed . , wiley new york sasselov , d. d. , & lester , j. b. 1994a , apj , 423 , 777 sasselov , d. d. , & lester , j. b. 1994b , apj , 423 , 785 sakurai , a. 1960 , comm . pure appl . math , 13 , 353 steffen , m. , szczerba , r. , & schoenberner , d. 1998 , a&a , 337 , 149 tej , a. , lanon , a. , scholz , m. , & wood , p. r. 2003 , astro - ph/0309689 , submitted to a&a tennekes , h. & lumley , j. l. 1972 , a first course in turbulence , mit press , cambridge weir , s. t. , a. c. mitchell , and w. j. nellis 1996 , phys . lett . , 76 , 1860 whitham , g. b. 1974 , linear and nonlinear waves , wiley , new york willson , l. a. 2000 , araa , 38 , 573 willson , l. a. , & hill , s. j. 1979 , apj , 228 , 854 willson , l. a. , struck , c. , & bowen , g. h. 1997 , in j. r. jokipii , c. p. sonnett , & m. s. giampapa , eds . , cosmic winds and the heliosphere , university of arizona press , tucson , p. 155 winters , j. m. , le bertre , t. , jeong , k. s .. helling , ch . , & sedlmayr , e. 2000 , a&a , 361 , 641 winters , j. m. , le bertre , t. , nyman , l .- a . , omont , a. , & jeong , k. s. 2002 , a&a , 388 , 609 woods , l. c. 1996 , thermodynamic inequalities in gases and magnetoplasmas , wiley , new york young , k. 1995 , apj , 445 , 872 in this appendix we give a brief , qualitative discussion of how the density and velocity fluctuations driven by pulsational shocks propagating through an lpv wind can be divided into a mean and a fluctuating part that averages to nearly zero over a pulsational period . we also discuss the derivation of the reynolds equations for this flow . to begin , consider the trajectories of a couple of adjacent gas elements relative to the local mean flow . we choose the gas elements that are sufficiently close together that the local mean outflow can be approximated as constant velocity , even if it is not globally constant . next , we divide the pulsational cycle into two parts . the first part begins when the gas element is hit by a shock . this impact boosts the fiducial element outward , and after a brief delay does the same to the neighboring element at slightly larger radius . they are pushed closer together in the shock compression , so both @xmath151 and @xmath69u @xmath152 , i.e. , greater than the values of the mean flow ( see e.g. , fig . 2 of @xcite . the fiducial gas element is closer to the star than its neighbor , and so , has a shorter free - fall time . as a result of this and the fact that it received its outward velocity burst first , it will decelerate ahead of its neighbor . this ballistic description is equivalent to the statement that the shock is followed by a rarefaction wave . we define the beginning of the second part of the cycle as beginning at the moment when the velocity is reduced to the mean flow value , and continuing until the next shock impact . the density is reduced to the mean value at about the same time . thus , during this second interval we have , @xmath151 , and @xmath69u @xmath153 . because the pulsational shocks are not very strong in the wind region , the magnitudes of @xmath151 , and @xmath69u are not large , and the cycle is nearly symmetric in the sense that the two intervals are about equal to half a cycle . thus , the net density and velocity fluctuations over one cycle are small , i.e. , of higher order than @xmath69r / r , the fractional radius change in that time . this justifies the reynolds procedure of writing each gas quantity as a sum of a mean and a fluctuating part . e.g. , for the velocity @xmath154 , and for the other fluid variables , @xmath155 , @xmath156 , @xmath157 , etc . , where the primes denote the fluctuating parts . the reynolds equations are then derived by substituting these variables into the hydrodynamical equations and averaging over a time greater than the pulsational period . it is assumed that the mean flow still satisfies the hydrodynamical equations . then , subtracting the mean flow equation from the corresponding equation for mean plus fluctuating parts yields an equation for the fluctuating part . ( see section 1.3 of @xcite , which the following discussion parallels . ) for example , the spherically symmetric steady state continuity equation is , @xmath158 the mean flow continuity equation is , @xmath159 ( see eq . ( 1 ) ) , and the fluctuation continuity equation is , @xmath160 on averaging this last equation we find , @xmath161 so we have , @xmath162 the steady velocity equation is , @xmath163 where in contrast to equation ( 2 ) we have included a mean stress gradient , or wave pressure , term ( e.g. , @xcite ) . after we subtract the mean flow version of this equation , we get the following equation for the fluctuating velocity ( before averaging ) , @xmath164 as before , all of the first order terms will vanish on averaging , leaving only thermal pressure and density terms at second order , which we assume are negligible . equation ( 4 ) of section 2.2.2 is the total energy equation . writing each variable in terms of its mean and fluctuating parts , and averaging with the assumption that odd order moments average to zero , yields equation ( 5 ) . with the additional approximations described in that section , this equation reduces to equation ( 6 ) . the various approximations provide a simple closure of the reynolds moment equations in this application , but they do so via the assumption that dissipation in the wind formation region has taken the system to a relaxed state characterized by small fluctuations . therefore , the equations do not provide a valid description of the unrelaxed dynamics at the onset of pulsations , nor of the strong shock zone shown by numerical models to exist at smaller radii ( see fig . 1 - 3 ) . as described in the text , the numerical models show that the extended atmospheres and winds of lpvs relax to power law density and pressure profiles , and in fact , the main result of this paper is on how these profiles can be accounted for with the aid of shock pumping and mean entropy gradients . these conclusions are founded on the assumption that shocks propagate down the relaxed power law pressure gradient without significantly accelerating ( or decelerating ) . in this appendix we briefly review some classical results that demonstrate the existence of profiles for constant shock propagation . these results do not seem to be well - known in the astrophysical literature , though they are summarized in the text of @xcite . the first result concerns the outward propagation of a spherically symmetric shock wave in a constant density medium . the wave is not necessarily a strong blast as assumed in the well - known sedov - taylor solution . an approximate similarity solution to this more general case was presented by @xcite , also see section 6.16 of @xcite . specifically , in the spherical adiabatic case the post - shock velocity is found to vary as , density gradients affect shocks in a manner similar to geometric convergence or divergence , and a related result of @xcite , described in section 8.2 of @xcite is useful . that is , the velocity of a planar adiabatic shock on a density gradient , @xmath166 , varies as , |
we discuss the three - body decays , , , and ( ) of the lighter top squark within the minimal supersymmetric standard model . we give the complete analytical formulas for the decay widths and present a numerical study in view of an upgraded tevatron , the cern lhc , and a future lepton collider demonstrating the importance of these decay modes . | hep-ph9812230 | the search for supersymmetry ( susy ) @xcite plays an important rle in the experimental program at the colliders lep2 and tevatron . it will be even more important at future colliders , e.g. an upgraded tevatron , lhc , an @xmath6 linear collider or a @xmath7 collider . therefore many phenomenological studies have been carried out in recent years ( see e.g. @xcite and references therein ) . within the supersymmetric extensions of the standard model ( sm ) the minimal supersymmetric standard model ( mssm ) @xcite is the most investigated one . the mssm implies that every sm fermion has two spin 0 partners called sfermions @xmath8 and @xmath9 . in general sfermions decay according to @xmath10 where @xmath11 and @xmath12 denote neutralinos and charginos , respectively . here we assume that the lightest neutralino is the lightest supersymmetric particle ( lsp ) . owing to large yukawa couplings the sfermions of the third generation have a quite different phenomenology compared to those of the first two generations ( see e.g. @xcite and references therein ) . the large yukawa couplings imply a large mixing between @xmath8 and @xmath9 and large couplings to the higgsino components of neutralinos and charginos . this is in particular the case for the lighter top squark @xmath5 because of the large top quark mass @xcite . the large top quark mass also implies the existence of scenarios where all two - body decay modes of @xmath5 ( e.g. @xmath13 ) are kinematically forbidden at tree - level . in these scenarios higher order decays of @xmath5 become relevant : @xcite : @xmath14 where @xmath15 denotes @xmath16 . in @xcite it has been shown that decays into sleptons are dominating over the decays into @xmath17 if they are kinematically allowed . however , they have used the approximation : @xmath18 , @xmath19 ( @xmath20 ) , @xmath21 . in @xcite it has been shown that for small @xmath22 the decay @xmath0 in general dominates over @xmath23 whereas for large @xmath22 their branching ratios can be of comparable size . in this paper we present the complete formulas for the three - body decays which are so far missing in the literature . we also perform a numerical analysis for the mass range of an upgraded tevatron , the lhc , and a future lepton collider including the possibility that all of the above decay channels are simultaneously open . in particular it turns out that the inclusion of the bottom and tau yukawa couplings @xmath24 and @xmath25 is important . this paper is organized as follows : in sect . ii we fix our notation and give the analytical expressions for the decay amplitudes together with the relevant parts of the mssm lagrangian . in sect . iii we present our numerical results for the branching ratios of the three - body decays in scenarios accessible either at the tevatron run ii , lhc , or a future lepton collider . our conclusions are drawn in sect . the analytical formulas for the squared amplitudes are listed in appendix [ appa ] , and appendix [ appb ] gives the various couplings . in this section we fix our notation and we give the lagrangian relevant for the calculation of the decay widths . moreover , we present the analytical formulas for the matrix elements and generic formulas for the decay widths . the complete formulas for the latter are rather lengthy and are listed in appendix [ appa ] . the parameters relevant for the following discussion are @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath22 , @xmath30 , @xmath31 , @xmath32 , @xmath33 , and @xmath34 . @xmath35 and @xmath27 are the @xmath36 and @xmath37 gaugino masses , for which we assume the gut relation @xmath38 . @xmath29 is the parameter of the higgs superpotential , @xmath28 the mass of the pseudoscalar higgs boson , and @xmath39 where @xmath40 denotes the vacuum expectation value of the higgs doublet @xmath41 . @xmath30 , @xmath31 and @xmath32 are soft susy breaking masses for the squarks , @xmath33 and @xmath34 are trilinear higgs squark couplings , and @xmath42 is the generation index . the mass matrix for sfermions in the @xmath43 basis has the following form @xcite : @xmath44 with @xmath45 and @xmath46 where @xmath47 is a generation index ( @xmath48 ) which will be suppressed in the following . the mass eigenstates @xmath49 and @xmath50 are related to @xmath51 and @xmath52 by : @xmath53 with the eigenvalues @xmath54 the mixing angle @xmath55 is given by @xmath56 the sneutrino mass is given by : @xmath57 the part of the lagrangian , which is needed for the calculation of the three - body decay widths , is given by : @xmath58 where @xmath59 . the various couplings are given in appendix [ appb ] . the formula for the decay width @xmath60 has already been given in @xcite . therefore , we give only the corresponding matrix element @xmath61 : @xmath62 v ( p_{{\tilde \chi}^0_1 } ) \epsilon_{\mu } ( p_w ) \nonumber \\ & & + \ , g^2 \sum^2_{j=1 } \bar u(p_b ) \left [ l^{\tilde t}_{1j } p_r + k^{\tilde t}_{1j } p_l \right ] \frac { \not \ ! p_{{\tilde \chi}^+_j } - m_{\tilde{\chi}^+_j } } { p_{{\tilde \chi}^+_j}^2 - m^2_{\tilde{\chi}^+_j } - i m_{\tilde{\chi}^+_j } \gamma_{\tilde{\chi}^+_j } } \nonumber \\ & & \hspace{20 mm } * \left [ o^l_{1j } { } ' p_l + o^r_{1j } { } ' p_r \right ] \gamma^{\mu } v(p_{{\tilde \chi}^0_1 } ) \epsilon_{\mu } ( p_w ) \nonumber \\ & & - \frac{g^2}{\sqrt2 } \bar u ( p_b ) \gamma^{\mu } p_l \frac { \not \ ! p_t + m_t } { p_{t}^2 -m_{t}^2 - i m_t \gamma_{t } } \left [ b^{\tilde{t}}_{11 } p_l + a^{\tilde{t}}_{11 } p_r \right ] v ( p_{{\tilde \chi}^0_1 } ) \epsilon_{\mu } ( p_w ) \label{eq : tfistbwchi}\end{aligned}\ ] ] in @xcite also the formula for @xmath63 @xcite has been rewritten in the notation used here . in fig . [ stbhgchi ] we show the feynman diagrams for the decay @xmath64 . the matrix element @xmath65 for this decay is given by : @xmath66 v ( p_{{\tilde \chi}^0_1 } ) } { p_{{\tilde b}_j}^2 - m^2_{{\tilde b}_j } - i m_{{\tilde b}_j } \gamma_{\tilde{b}_j } } \nonumber \\ & & - \ , g^2 \sum^2_{j=1 } \frac { \bar u(p_b ) \left [ l^{\tilde t}_{1j } p_r + k^{\tilde t}_{1j } p_l \right ] [ \not \ ! p_{{\tilde \chi}^+_j } - m_{\tilde{\chi}^+_j } ] \left [ q^l_{1j } { } ' p_l + q^r_{1j } { } ' p_r \right ] v(p_{{\tilde \chi}^0_1 } ) } { p_{{\tilde \chi}^+_j}^2 - m^2_{\tilde{\chi}^+_j } - i m_{\tilde{\chi}^+_j } \gamma_{\tilde{\chi}^+_j } } \nonumber \\ & & + \frac{g^2 } { \sqrt2 m_w } \frac { \bar u ( p_b ) \left [ m_b \tan \beta p_l + m_t \cot \beta p_r \right ] [ \not \ ! p_t + m_t ] \left [ b^{\tilde{t}}_{11 } p_l + a^{\tilde{t}}_{11 } p_r \right ] v ( p_{{\tilde \chi}^0_1 } ) } { p_{t}^2 -m_{t}^2 - i m_t \gamma_{t } } \nonumber \\ \label{eq : tfistbhchi}\end{aligned}\ ] ] the decay width is given by @xmath67 with @xmath68 given in appendix [ appa ] . alternatively @xmath5 can decay into sleptons : @xmath69 , @xmath70 , @xmath71 , and @xmath72 . these decays are mediated through virtual charginos . the feynman graphs are similar to the second one in fig . [ stbhgchi ] , where one has to replace the higgs boson by a slepton and the neutralino by the corresponding lepton . note , that the decays into @xmath73 and @xmath74 are negligible because their couplings to the charginos are proportional to @xmath75 and @xmath76 , respectively . in the case of decays into sneutrinos and leptons the matrix elements @xmath77 have the generic form : @xmath78 \left [ \not \ ! p_{{\tilde \chi}^+_j } - m_{\tilde{\chi}^+_j } \right ] \left [ l^{\tilde \nu_l}_{j } p_r + k^{\tilde \nu_l}_{j } p_l \right ] v(p_l ) } { p_{{\tilde \chi}^+_j}^2 - m^2_{\tilde{\chi}^+_j } - i m_{\tilde{\chi}^+_j } \gamma_{\tilde{\chi}^+_j } } , \label{eq : tfistbsneut}\end{aligned}\ ] ] whereas for the decays into sleptons and neutrinos we get : @xmath79 \left [ \not \ ! p_{{\tilde \chi}^+_j } - m_{\tilde{\chi}^+_j } \right ] \left [ l^{\tilde l}_{kj } p_l \right ] v(p_{\nu_l } ) } { p_{{\tilde \chi}^+_j}^2 - m^2_{\tilde{\chi}^+_j } - i m_{\tilde{\chi}^+_j } \gamma_{\tilde{\chi}^+_j } } \ , . \label{eq : tfistbslept}\end{aligned}\ ] ] in both cases the decay width is given by @xmath80 the explicit expressions for @xmath81 , @xmath82 , and @xmath83 are given in appendix [ appa ] . in this section we present our numerical results for the branching ratios of the higher order decays of @xmath5 . here we consider scenarios where all two - body decays at tree - level are kinematically forbidden . we have fixed the parameters as in @xcite to avoid colour breaking minimas : we have used @xmath84 , @xmath85 , @xmath22 , and @xmath29 as input parameters in the top squark sector . for the sbottom ( stau ) sector we have fixed @xmath86 and @xmath87 ( @xmath88 , and @xmath89 ) as input parameters . for simplicity , we assume that the soft susy breaking parameters are equal for all generations . note , that due to @xmath37 invariance @xmath90 appears in both up- and down - type squark mass matrices . in the sbottom ( stau ) sector the physical quantities @xmath91 , @xmath92 , and @xmath93 ( @xmath94 , @xmath95 , and @xmath96 ) obviously change with @xmath29 and @xmath22 . in fig . [ brst3cosa](a ) and ( b ) we show the branching ratios of @xmath5 as a function of @xmath85 . we have restricted the @xmath85 range such that @xmath97 tev to avoid color / charge breaking minima . the parameters and physical quantities are given in tab . [ tabbrst3cosa ] . the slepton parameters have been chosen such that the sum of the masses of the final state particles are @xmath98 gev . in fig . [ brst3cosa](a ) we show @xmath99 , @xmath100 , @xmath101 ) + @xmath102 , and @xmath103 + @xmath104 + @xmath105 . here we have not included the possibility of the decay @xmath106 because with this parameter set there exists no value of @xmath28 which simultaneously allows this decay and fulfills the condition @xmath107 gev @xcite ( we have used the mssm formula for the calculation of @xmath108 including 1-loop corrections as given in @xcite ) . however , we will discuss this decay later on . we have summed up those branching ratios for the decays into sleptons that give the same final state , for example : @xmath109 note , that in the above cases the assumption @xmath110 implies @xmath111 . therefore , the sleptons can only decay into the corresponding lepton plus @xmath112 except for a small parameter region where the decay into @xmath113 is possible . however , this decay is negligible due to kinematics in that region . the branching ratios for decays into @xmath114 or @xmath115 are practically the same as those into @xmath116 or @xmath117 . for this set of parameters @xmath100 is @xmath118 independent of @xmath85 and therefore negligible . near @xmath119 @xmath99 is almost @xmath120 because the @xmath5-@xmath121-@xmath122 coupling @xmath123 vanishes . we have found that the decay @xmath124 is dominated by the @xmath125-quark exchange . in many cases the interference term between @xmath125 and @xmath126 is more important than the @xmath126 exchange . moreover , we have found that the contribution from sbottom exchange is in general negligible . in fig . [ brst3cosa](b ) the branching ratios for the various decays into sleptons are shown . for small @xmath22 sleptons couple mainly to the gaugino components of @xmath121 . this leads to @xmath127 because @xmath128 . the decays into sneutrinos are preferred by kinematics while the decay into @xmath129 is suppressed by the same reason ( table [ tabbrst3cosa ] ) . moreover , the matrix elements eqs . ( [ eq : tfistbsneut ] ) and ( [ eq : tfistbslept ] ) for the decays into charged and neutral sleptons have a different structure in the limit @xmath130 : @xmath131 this leads to different decay widths even in the limit of equal slepton masses . in figs . [ brst3tana](a ) and ( b ) we show the branching ratios as a function of @xmath22 for @xmath132 and the other parameters as above . for small @xmath22 the decay into @xmath124 is the most important one . the branching ratios for the decays into sleptons decrease with increasing @xmath22 except for the decay into @xmath133 . this results from : ( i ) for increasing @xmath22 the gaugino component of @xmath121 decreases while its mass increases , ( ii ) the masses of the sleptons increase with increasing @xmath22 , except @xmath94 which decreases , and ( iii ) the @xmath134 yukawa coupling increases . these facts lead to the dominance of @xmath135 for large @xmath22 as can be seen in fig . [ brst3tana](b ) . in addition , the decay into @xmath136 gains some importance for large @xmath22 because its width is proportional to the bottom yukawa coupling in the approximation of @xcite . the assumption that no two - body decays be allowed at tree level implies that @xmath137 . therefore , one expects an increase of @xmath99 if @xmath84 increases , because the decay into @xmath138 is dominated by the @xmath125 exchange whereas for the decays into sleptons @xmath121 exchange dominates . this is demonstrated in figs . [ brst3cospmu](a ) and ( b ) where we have fixed @xmath139 gev . here also the decay into @xmath140 is possible . however , this channel is in general suppressed by kinematics . we have not found any case with @xmath141 gev while @xmath142 gev @xcite . these general features still hold if @xmath22 increases as can be seen in figs . [ brst3tanb](a ) and ( b ) . here we have fixed @xmath143 . in accordance with the discussion above , the decay into @xmath135 gains importance with increasing @xmath22 . note , that for large @xmath22 @xmath144 decreases since @xmath108 increases due to radiative corrections . however , there are scenarios where the decay @xmath106 becomes important . this can be seen in fig . [ brst3md ] where the branching ratios are shown as a function of @xmath145 for @xmath146 gev , @xmath147 , and the other parameters as in tab . [ tabbrst3cospmu ] . at the lower end of the @xmath145 range we get @xmath148 gev . moreover , @xmath91 is approximately @xmath149 leading to an enhancement of this width . we have found that contrary to the case @xmath124 for the decays @xmath106 sbottom exchange can be important . this is a consequence of the different spin structure of the corresponding matrix elements ( eqs . ( [ eq : tfistbwchi ] ) and ( [ eq : tfistbhchi ] ) ) and because of the large bottom yukawa coupling . the decrease in @xmath144 for @xmath150 gev is mainly due to the fact that @xmath108 grows with increasing @xmath145 . finally , we want to discuss a scenario which is within the reach of an upgraded tevatron . here we refer to the examples of @xcite . in general the decays into sleptons clearly dominate when they are kinematically allowed ( except the case when the couplings of the top squark to the lighter chargino nearly vanishes ) . in the case @xmath151 gev , @xmath152 , @xmath153 gev , @xmath154 gev , @xmath155 gev , @xmath156 gev and @xmath157 ( scenario b of table i in @xcite ) we obtain : @xmath158 , @xmath159 , @xmath160 , @xmath161 , and @xmath162 . the order of magnitude is independent of @xmath29 and @xmath85 because the lighter chargino is mainly gaugino - like in the parameter space where the decay @xmath124 is possible ( see fig . 2b of @xcite ) . for increasing @xmath22 we have found a similar behaviour as in figs . [ brst3tana ] and [ brst3tanb ] : dominance of the decay @xmath163 , an increase of @xmath164 and a decrease of all other decay channels . we have calculated the three - body decays @xmath106 , @xmath165 , and @xmath166 ( @xmath20 ) including all terms proportional to @xmath167 , @xmath168 , and all yukawa couplings . we have compared these decays with @xmath164 @xcite and @xmath124 @xcite . these decays are competitive in that part of the parameter space accessible at an upgraded tevatron , the lhc or a future lepton collider where the tree level two - body decays @xmath169 and @xmath170 are kinematically forbidden . we have found that for @xmath171 gev the decays into sleptons dominate due to kinematics . in the range 200 gev @xmath172 gev all the decays mentioned above compete with each other . the branching ratios depend crucially on the coupling of @xmath5 to @xmath121 , implying that one can get information on the mixing angle of the top squarks , once the chargino properties are known . for heavier top squark masses the decay into @xmath138 , mainly proceeding via a virtual @xmath125 quark , is in general the most important one . in addition , we have found that for small @xmath22 the decays into sneutrinos are more important than the decays into charged sleptons . this is a result of the different spin structures of the corresponding matrix elements . for large @xmath22 the decay into the lighter stau becomes important due to the large tau yukawa coupling ( implying also a smaller stau mass ) . the decay into @xmath140 is kinematically suppressed because the existing mass bounds on the neutral higgs bosons also imply a lower bound on @xmath108 . however , in scenarios where radiative corrections decrease @xmath108 and where at the same time @xmath173 we have found branching ratios of the order of 30% . the large variety of possible three - body decay modes implies the chance to determine the properties of @xmath5 also when higher order decays are dominant . clearly , a detailed monte carlo study will be necessary to see how the different channels can be separated . i thank a. bartl , h. eberl , t. gajdosik , s. kraml , and w. majerotto for many helpful discussions and the inspiring atmosphere . i am grateful to j.w.f . valle for the kind hospitality and the pleasant atmosphere at the departamento de fsica terico . this work was supported by the `` fonds zur frderung der wissenschaftlichen forschung '' of austria , project no . p10843-phy , and by the eec under the tmr contract erbfmrx - ct96 - 0090 . in the subsequent sections the formulas for the decay widths are listed which have been omitted in sec . [ sec : threebody ] . the decay width is given by @xmath175 with @xmath176 \nonumber \\ & & + ( a_{31 } + a_{32 } s ) j^0_{tt}(m^2_{{\tilde t}_1 } + m^2_{h^+ } + m^2_b + m^2_{{\tilde \chi}^0_1 } - m^2_{{\tilde \chi}^+_1 } - s , \gamma_{{\tilde \chi}^+_1 } m_{{\tilde \chi}^+_1 } \nonumber \\ & & \hspace{33 mm } , m^2_{{\tilde t}_1 } + m^2_{h^+ } + m^2_b + m^2_{{\tilde \chi}^0_1 } - m^2_{{\tilde \chi}^+_2 } - s , \gamma_{{\tilde \chi}^+_2 } m_{{\tilde \chi}^+_2 } ) \nonumber \\ & & + ( a_{33 } + a_{34 } s ) j^1_{tt}(m^2_{{\tilde t}_1 } + m^2_{h^+ } + m^2_b + m^2_{{\tilde \chi}^0_1 } - m^2_{{\tilde \chi}^+_1 } - s , \gamma_{{\tilde \chi}^+_1 } m_{{\tilde \chi}^+_1 } \nonumber \\ & & \hspace{33 mm } , m^2_{{\tilde t}_1 } + m^2_{h^+ } + m^2_b + m^2_{{\tilde \chi}^0_1 } - m^2_{{\tilde \chi}^+_2 } - s , \gamma_{{\tilde \chi}^+_2 } m_{{\tilde \chi}^+_2 } ) \nonumber \\ & & + \ , a_{34 } \ , j^2_{tt}(m^2_{{\tilde t}_1 } + m^2_{h^+ } + m^2_b + m^2_{{\tilde \chi}^0_1 } - m^2_{{\tilde \chi}^+_1 } - s , \gamma_{{\tilde \chi}^+_1 } m_{{\tilde \chi}^+_1 } \nonumber \\ & & \hspace{17 mm } , m^2_{{\tilde t}_1 } + m^2_{h^+ } + m^2_b + m^2_{{\tilde \chi}^0_1 } - m^2_{{\tilde \chi}^+_2 } - s , \gamma_{{\tilde \chi}^+_2 } m_{{\tilde \chi}^+_2 } ) \ , , \end{aligned}\ ] ] @xmath177 \ , , \end{aligned}\ ] ] @xmath178 \ , , \end{aligned}\ ] ] @xmath179 @xmath180 \ , , \nonumber \\\end{aligned}\ ] ] @xmath181 \right\ } \ , .\end{aligned}\ ] ] the integrals @xmath182 are : @xmath183 with @xmath184 . their integration range is given by @xmath185 where @xmath186 and @xmath187 are the usual mandelstam variables . note , that @xmath188 appears in the entries of the integrals @xmath189 and @xmath190 because the chargino is exchanged in the @xmath191-channel in our convention . the coefficients are given by : @xmath192 \nonumber \\ & & - \left ( ( k^{\tilde t}_{11})^2 ( q^l_{11}{}')^2 + ( l^{\tilde t}_{11})^2 ( q^r_{11}{}')^2 \right ) m^2_{\tilde{\chi}^+_1 } \left(m^2_b+m^2_{{\tilde \chi}^0_1 } \right ) \ , , \\ % a_{12 } & = & 4 \ , k^{\tilde t}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' q^r_{11 } { } ' m_b m_{{\tilde \chi}^0_1 } + \left ( ( k^{\tilde t}_{11})^2 ( q^r_{11}{}')^2 + ( l^{\tilde t}_{11})^2 ( q^l_{11}{}')^2 \right ) \left ( m^2_b+m^2_{{\tilde \chi}^0_1 } \right ) \nonumber \\ & & + 2 \ , k^{\tilde t}_{11 } l^{\tilde t}_{11 } \left ( ( q^l_{11}{}')^2 + ( q^r_{11}{}')^2 \right ) m_b m_{\tilde{\chi}^+_1 } + 2 \ , q^l_{11 } { } ' q^r_{11 } { } ' \left ( ( k^{\tilde t}_{11})^2 + ( l^{\tilde t}_{11})^2 \right ) m_{{\tilde \chi}^0_1 } m_{\tilde{\chi}^+_1 } \nonumber \\ & & + \left ( ( k^{\tilde t}_{11})^2 ( q^l_{11}{}')^2 + ( l^{\tilde t}_{11})^2 ( q^r_{11}{}')^2 \right ) m^2_{\tilde{\chi}^+_1 } \ , , \\ % a_{13 } & = & 4 \ , k^{\tilde t}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' q^r_{11 } { } ' m_b m_{{\tilde \chi}^0_1 } + 2 \ , q^l_{11 } { } ' q^r_{11 } { } ' \left ( ( k^{\tilde t}_{11})^2 + ( l^{\tilde t}_{11})^2 \right ) m_{{\tilde \chi}^0_1 } m_{\tilde{\chi}^+_1 } \nonumber \\ & & + \left ( ( k^{\tilde t}_{11})^2 ( q^r_{11}{}')^2 + ( l^{\tilde t}_{11})^2 ( q^l_{11}{}')^2 \right ) \left(2 \ , m^2_b+2 \ , m^2_{{\tilde \chi}^0_1 } + m^2_{h^+}+m^2_{{\tilde t}_1}\right ) \nonumber \\ & & + 2 \ , k^{\tilde t}_{11 } l^{\tilde t}_{11 } \left ( ( q^l_{11}{}')^2 + ( q^r_{11}{}')^2 \right ) m_b m_{\tilde{\chi}^+_1 } \ , , \\ % a_{14 } & = & - ( k^{\tilde t}_{11})^2 ( q^r_{11}{}')^2 - ( l^{\tilde t}_{11})^2 ( q^l_{11}{}')^2 \ , , \end{aligned}\ ] ] @xmath193 \nonumber \\ & & - 2 \ , \left(k^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{12 } l^{\tilde t}_{11 } q^r_{11 } { } ' q^r_{12}{}'\right ) m_b m_{\tilde{\chi}^+_1 } \left(m^2_b+m^2_{{\tilde t}_1}+2 \ , m^2_{{\tilde \chi}^0_1 } \right ) \nonumber \\ & & - 2 \ , \left(k^{\tilde t}_{12 } l^{\tilde t}_{11 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{11 } l^{\tilde t}_{12 } q^r_{11 } { } ' q^r_{12}{}'\right ) m_b m_{\tilde{\chi}^+_2 } \left(m^2_b+m^2_{{\tilde t}_1}+2 \ , m^2_{{\tilde \chi}^0_1 } \right ) \nonumber \\ & & - 2 \ , \left(k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^l_{11 } { } ' q^l_{12 } { } ' + l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^r_{11 } { } ' q^r_{12}{}'\right ) m_{\tilde{\chi}^+_1 } m_{\tilde{\chi}^+_2 } \left(m^2_{{\tilde \chi}^0_1 } + m^2_b \right ) \ , , \\ % a_{32 } & = & 2 \left(l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{12 } { } ' q^r_{11 } { } ' + k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^l_{11 } { } ' q^r_{12}{}'\right ) m_{{\tilde \chi}^0_1 } m_{\tilde{\chi}^+_1 } \nonumber \\ & & + 4 \ , \left(k^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{12 } { } ' q^r_{11 } { } ' + k^{\tilde t}_{12 } l^{\tilde t}_{11 } q^l_{11 } { } ' q^r_{12}{}'\right ) m_b m_{{\tilde \chi}^0_1 } \nonumber \\ & & + 2 \ , \left(k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^l_{12 } { } ' q^r_{11 } { } ' + l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{11 } { } ' q^r_{12}{}'\right ) m_{{\tilde \chi}^0_1 } m_{\tilde{\chi}^+_2 } \nonumber \\ & & + 2 \ , \left(l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^r_{11 } { } ' q^r_{12}{}'\right ) \left ( m^2_b+m^2_{{\tilde \chi}^0_1 } \right ) \nonumber \\ & & + 2 \ , \left(k^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{12 } l^{\tilde t}_{11 } q^r_{11 } { } ' q^r_{12}{}'\right ) m_b m_{\tilde{\chi}^+_1 } \nonumber \\ & & + 2 \ , \left(k^{\tilde t}_{12 } l^{\tilde t}_{11 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{11 } l^{\tilde t}_{12 } q^r_{11 } { } ' q^r_{12}{}'\right ) m_b m_{\tilde{\chi}^+_2 } \nonumber \\ & & + 2 \ , \left(k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^l_{11 } { } ' q^l_{12 } { } ' + l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^r_{11 } { } ' q^r_{12}{}'\right ) m_{\tilde{\chi}^+_1 } m_{\tilde{\chi}^+_2 } \ , , \\ % a_{33 } & = & 2 \left(l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{12 } { } ' q^r_{11 } { } ' + k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^l_{11 } { } ' q^r_{12}{}'\right ) m_{{\tilde \chi}^0_1 } m_{\tilde{\chi}^+_1 } \nonumber \\ & & + 4 \ , \left(k^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{12 } { } ' q^r_{11 } { } ' + k^{\tilde t}_{12 } l^{\tilde t}_{11 } q^l_{11 } { } ' q^r_{12}{}'\right ) m_b m_{{\tilde \chi}^0_1 } \nonumber \\ & & + 2 \ , \left(k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^l_{12 } { } ' q^r_{11 } { } ' + l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{11 } { } ' q^r_{12}{}'\right ) m_{{\tilde \chi}^0_1 } m_{\tilde{\chi}^+_2 } \nonumber \\ & & + 2 \ , \left(l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^r_{11 } { } ' q^r_{12}{}'\right ) \left(2 \ , m^2_b+2 \ , m^2_{{\tilde \chi}^0_1 } + m^2_{h^+}+m^2_{{\tilde t}_1 } \right ) \nonumber \\ & & + 2 \ , \left(k^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{12 } l^{\tilde t}_{11 } q^r_{11 } { } ' q^r_{12}{}'\right ) m_b m_{\tilde{\chi}^+_1 } \nonumber \\ & & + 2 \ , \left(k^{\tilde t}_{12 } l^{\tilde t}_{11 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{11 } l^{\tilde t}_{12 } q^r_{11 } { } ' q^r_{12}{}'\right ) m_b m_{\tilde{\chi}^+_2 } \ , , \\ % a_{34 } & = & - 2 \ , \left(l^{\tilde t}_{11 } l^{\tilde t}_{12 } q^l_{11 } { } ' q^l_{12 } { } ' + k^{\tilde t}_{11 } k^{\tilde t}_{12 } q^r_{11 } { } ' q^r_{12}{}'\right ) \ , , \end{aligned}\ ] ] @xmath194 \nonumber \\ & & + b^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' m_t \left [ \left ( m^2_{h^+ } m^2_{{\tilde t}_1 } - m^2_b m^2_{{\tilde \chi}^0_1 } \right ) \cot \beta - m^2_b \left ( m^2_b + m^2_{{\tilde t}_1 } + 2 \ , m^2_{{\tilde \chi}^0_1 } \right ) \tan \beta \right ] \nonumber \\ & & + a^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' m_{{\tilde \chi}^0_1 } \nonumber \\ & & \hspace{2 mm } * \left [ m^2_b \left ( m^2_{h^+}+m^2_{{\tilde t}_1}-m^2_{{\tilde \chi}^0_1 } -m^2_b \right ) \tan \beta - m^2_t \left ( 2 \ , m^2_b + m^2_{{\tilde \chi}^0_1 } + m^2_{h^+ } \right ) \cot \beta \right ] \nonumber \\ & & + a^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^l_{11 } { } ' m_{\tilde{\chi}^+_1 } m_b m_{{\tilde \chi}^0_1 } \left [ \left(m^2_{h^+}-m^2_b \right ) \tan \beta - 2 \ , m^2_t \cot \beta \right ] \nonumber \\ & & + b^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' m_{{\tilde \chi}^0_1 } m_b m_t \nonumber \\ & & \hspace{2 mm } * \left [ \left(m^2_{{\tilde t}_1}+m^2_{h^+}-m^2_b -m^2_{{\tilde \chi}^0_1 } \right ) \cot \beta - \left ( m^2_{h^+}+2 \ , m^2_b + m^2_{{\tilde \chi}^0_1 } \right ) \tan \beta \right ] \nonumber \\ & & + b^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^r_{11 } { } ' m_t m_{{\tilde \chi}^0_1 } m_{\tilde{\chi}^+_1 } \left [ \left(m^2_{h^+}-m^2_b \right ) \cot \beta - 2 \ , m^2_b \tan \beta \right ] \nonumber \\ & & - a^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' m_b \nonumber \\ & & \hspace{2 mm } * \left [ \left(m^2_b m^2_{{\tilde \chi}^0_1 } - m^2_{{\tilde t}_1 } m^2_{h^+ } \right ) \tan \beta + m^2_t \left ( m^2_b + m^2_{{\tilde t}_1}+2 \ , m^2_{{\tilde \chi}^0_1 } \right ) \cot \beta \right ] \nonumber \\ & & - a^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^r_{11 } { } ' m_{\tilde{\chi}^+_1 } \left [ m^2_b \left(m^2_{{\tilde \chi}^0_1}-m^2_{{\tilde t}_1 } \right ) \tan \beta + m^2_t \left(m^2_b + m^2_{{\tilde \chi}^0_1 } \right ) \cot \beta \right ] \bigg\ } \ , , \\ % b_{12 } & = & \frac{\sqrt2 \,}{m_w } \bigg\ { b^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^l_{11 } { } ' m_{\tilde{\chi}^+_1 } m_b m_t \tan \beta + b^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' m^2_b m_t \tan \beta \nonumber \\ & & + a^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' m^2_t m_{{\tilde \chi}^0_1 } \cot \beta + b^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' m_{{\tilde \chi}^0_1 } m_b m_t \tan \beta \nonumber \\ & & + a^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^r_{11 } { } ' m_{\tilde{\chi}^+_1 } m^2_t \cot \beta + a^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' m_b m^2_t \cot \beta \bigg\ } \ , , \\ % b_{13 } & = & - \frac{\sqrt2 \,}{m_w } \bigg\ { b^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^l_{11 } { } ' m_{\tilde{\chi}^+_1 } m_b m_t \cot \beta - a^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' m^2_t m_{{\tilde \chi}^0_1 } \cot \beta \nonumber \\ & & + b^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' m_t \left [ \left ( m^2_{{\tilde \chi}^0_1 } + m^2_{h^+}+m^2_{{\tilde t}_1}+m^2_b \right ) \cot \beta - m^2_b \tan \beta \right ] \nonumber \\ & & + a^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^l_{11 } { } ' m_{\tilde{\chi}^+_1 } m_b m_{{\tilde \chi}^0_1 } \tan \beta - b^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' m_{{\tilde \chi}^0_1 } m_b m_t \tan \beta \nonumber \\ & & + b^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^r_{11 } { } ' m_{\tilde{\chi}^+_1 } m_t m_{{\tilde \chi}^0_1 } \cot \beta + a^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^r_{11 } { } ' m^2_b m_{\tilde{\chi}^+_1 } \tan \beta \nonumber \\ & & + a^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' m_b \left [ \left(m^2_b + m^2_{{\tilde \chi}^0_1 } + m^2_{h^+ } + m^2_{{\tilde t}_1 } \right ) \tan \beta - m^2_t \cot \beta \right ] \bigg\ } \ , , \\ % b_{14 } & = & \frac{\sqrt2 \,}{m_w } \left ( b^{\tilde{t}}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' m_t \cot \beta + a^{\tilde{t}}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' m_b \tan \beta \right ) \ , , \end{aligned}\ ] ] @xmath195 \ , , \\ % c_{112 } & = & - 2 \ , c^h_{{\tilde t}_1 { \tilde b}_1 } \bigg [ \left(b^{\tilde b}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' + a^{\tilde b}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' \right ) m_b + \left(a^{\tilde b}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' + b^{\tilde b}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' \right ) m_{{\tilde \chi}^0_1 } \nonumber \\ & & \hspace{16 mm } + \left(b^{\tilde b}_{11 } k^{\tilde t}_{11 } q^l_{11 } { } ' + a^{\tilde b}_{11 } l^{\tilde t}_{11 } q^r_{11 } { } ' \right ) m_{\tilde{\chi}^+_1 } \bigg ] \ , , \\ % c_{113 } & = & -2 \ , c^h_{{\tilde t}_1 { \tilde b}_1 } \bigg [ \left(b^{\tilde b}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' + a^{\tilde b}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' \right ) m_b + \left(a^{\tilde b}_{11 } l^{\tilde t}_{11 } q^l_{11 } { } ' + b^{\tilde b}_{11 } k^{\tilde t}_{11 } q^r_{11 } { } ' \right ) m_{{\tilde \chi}^0_1 } \bigg ] \ , , \nonumber \\\end{aligned}\ ] ] @xmath196 \nonumber \\ & & \hspace{5 mm } + 2 \ , \left ( ( a^{\tilde{t}}_{11})^2 + ( b^{\tilde{t}}_{11})^2 \right ) m^2_b m^2_t \left ( m^2_{{\tilde t}_1 } - m^2_{{\tilde \chi}^0_1 } \right ) \bigg ] \ , , \\ % d_2 & = & \frac{m^2_t}{2 \ , m^2_w } \left ( ( a^{\tilde{t}}_{11})^2 m^2_t \cot^2 \beta + ( b^{\tilde{t}}_{11})^2 m^2_b \tan^2 \beta \right ) \ , , \\ % d_{3 } & = & \frac{-1}{2 m^2_w } \bigg [ 2 \left ( ( a^{\tilde{t}}_{11})^2 + ( b^{\tilde{t}}_{11})^2 \right ) m^2_b m^2_t + 2 a^{\tilde{t}}_{11 } b^{\tilde{t}}_{11 } m_t m_{{\tilde \chi}^0_1 } \left ( m^2_b \left ( 2 + \tan^2 \beta \right ) + m^2_t \cot^2 \beta \right ) \nonumber \\ & & \hspace{12 mm } - \left ( ( a^{\tilde{t}}_{11})^2 m^2_b \tan^2 \beta + ( b^{\tilde{t}}_{11})^2 m^2_t \cot^2 \beta \right ) \left(m^2_{h^+}+m^2_{{\tilde t}_1 } \right ) \bigg ] \ , , \\ % d_{4 } & = & - \frac{(a^{\tilde{t}}_{11})^2 m^2_b \tan^2 \beta + ( b^{\tilde{t}}_{11})^2 m^2_t \cot^2 \beta } { 2 \ , m^2_w } \ , , \end{aligned}\ ] ] @xmath197 \nonumber \\ & & \hspace{21 mm } + b^{\tilde{b}}_{11 } b^{\tilde{t}}_{11 } m_b m_t \left [ \left ( m^2_{{\tilde t}_1 } - m^2_{{\tilde \chi}^0_1 } \right ) \cot \beta - \left(m^2_b+m^2_{{\tilde \chi}^0_1 } \right ) \tan \beta \right ] \nonumber \\ & & \hspace{21 mm } - a^{\tilde{b}}_{11 } a^{\tilde{t}}_{11 } \left [ m^2_t \left ( m^2_b+m^2_{{\tilde \chi}^0_1 } \right ) \cot \beta + m^2_b \left(m^2_{{\tilde \chi}^0_1 } -m^2_{{\tilde t}_1 } \right ) \tan \beta \right ] \nonumber \\ & & \hspace{21 mm } - b^{\tilde{b}}_{11 } a^{\tilde{t}}_{11 } m_b m_{{\tilde \chi}^0_1 } \left [ 2 \ , m^2_t \cot \beta + \left ( m^2_b - m^2_{h^+ } \right ) \tan \beta \right ] \bigg\ } \ , , \\ % e_{12 } & = & - \frac { \sqrt2 \ , c^h_{{\tilde t}_1 { \tilde b}_1}}{m_w } \left ( a^{\tilde{b}}_{11 } a^{\tilde{t}}_{11 } m^2_t \cot \beta + b^{\tilde{b}}_{11 } b^{\tilde{t}}_{11 } m_b m_t \tan \beta \right ) \ , , \\ % e_{13 } & = & \frac { \sqrt2 \ , c^h_{{\tilde t}_1 { \tilde b}_1}}{m_w } \bigg ( a^{\tilde{b}}_{11 } b^{\tilde{t}}_{11 } m_t m_{{\tilde \chi}^0_1 } \cot \beta + b^{\tilde{b}}_{11 } b^{\tilde{t}}_{11 } m_b m_t \cot \beta \nonumber \\ & & \hspace{17 mm } + \ , a^{\tilde{b}}_{11 } a^{\tilde{t}}_{11 } m^2_b \tan \beta + b^{\tilde{b}}_{11 } a^{\tilde{t}}_{11 } m_b m_{{\tilde \chi}^0_1 } \tan \beta \bigg ) \ , , \end{aligned}\ ] ] @xmath198 \ , , \\ % f_{12 } & = & ( c^h_{{\tilde t}_1 { \tilde b}_1})^2 \left((a^{\tilde{b}}_{11})^2+(b^{\tilde{b}}_{11})^2 \right ) \ , , \end{aligned}\ ] ] @xmath199 \ , , \nonumber \\ \\ % f_{32 } & = & 2 \ , c^h_{{\tilde t}_1 { \tilde b}_1 } c^h_{{\tilde t}_1 { \tilde b}_2 } \left ( a^{\tilde{b}}_{11 } a^{\tilde{b}}_{12 } + b^{\tilde{b}}_{11 } b^{\tilde{b}}_{12 } \right ) \ , .\end{aligned}\ ] ] one gets the remaining coefficients by replacements : + ( 135,175 ) ( 43,153)@xmath200 ( 106,144)@xmath201 ( 65,161)@xmath202 ( 76,153)@xmath203 ( 106,155)@xmath122 ( 43,93)@xmath200 ( 108,98)@xmath202 ( 107,83)@xmath201 ( 75,94)@xmath204 ( 68,101)@xmath122 ( 43,33)@xmath200 ( 65,42)@xmath201 ( 108,37)@xmath202 ( 77,32)@xmath125 ( 108,20)@xmath122 ( 30,10 ) |
we consider two independent binary i.i.d . random strings and of equal length and the optimal alignments according to a symmetric scoring functions only . we decompose the space of scoring functions into five components . two of these components add a part to the optimal score which does not depend on the alignment and which is asymptotically normal . we show that when we restrict the number of gaps sufficiently and add them only into one sequence , then the alignment score can be decomposed into a part which is normal and has order and a part which is on a smaller order and tends to a tracy - widom distribution . adding gaps only into one sequence is equivalent to aligning a string with its descendants in case of mutations and deletes . for testing relatedness of strings , the normal part is irrelevant , since it does not depend on the alignment hence it can be safely removed from the test statistic . | 1703.05788 | optimal alignments are among the main two components used by most dna - alignment algorithms which try to find similar substrings in dna - sequences . because of its great practical importance , these problems have received a tremendous amount of attention . let us explain the key points from the dna perspective . dna sequencing is one of the most important areas of study in biology . this is still rather expensive and very complex . suppose now that in a certain animal population , a certain gene is detected . this gene is probably likely to appear in the dna of other animals too , but in a slightly different form . thus , instead of repeating the same costly experiments on other animals to find the gene , they just search for a dna - segment which looks similar to the gene in the first population . this is why often dna - strings in different animals are compared in order to find places that are related . for more details on the analysis of dna we refer the reader to @xcite and in particular the iid assumption which appears also in this paper is explained on page 14 of the @xcite . here is a more concrete example . assume @xmath4 is a hypothetical dna of an ancestor with two descendants , one with the dna string @xmath5 ( which lost the @xmath6 and the @xmath7 ) and another with the dna equal to @xmath8 ( which had an @xmath9 mutated into a @xmath6 ) . we align these strings ( with gaps ) , as follows : @xmath10 in the analysis of the dna one has access to the dna of the descendants only . in terms of the alignments , one is interested in the _ historic alignment_. in our example , we would have that the alignment @xmath11 is given by : @xmath12 and we call this the historic alignment . the letters which disappeared in one descendant are now represented by gaps . which alignment with gaps corresponds to the evolution history of the strings is not known apriori . in other words , in our example , we know the strings @xmath13 and @xmath8 , but not `` the true historic '' alignment given here by . to guess which alignment is historic , the natural attempt is to find an alignment which maximizes similarity . for this , one starts with a scoring function @xmath14 which measures similarity between letters . ( here @xmath15 is the alphabet augmented by one symbol for the gap ) . the _ total score _ of an alignment is then the sum of the scores of the aligned symbol pairs . in alignment , the alignment score @xmath16 is equal to @xmath17 where @xmath7 stands for a gap . an alignment @xmath18 , which for given strings @xmath19 and @xmath20 maximizes the alignment score @xmath21 is called _ an optimal alignment_. the score of an optimal alignment is called _ optimal alignment score _ and we will denote it by @xmath22 where the maximum above is taken over all alignments with gaps of @xmath19 and @xmath20 . which alignment is optimal depends ultimately on the scoring function used . often , the scoring function biologists work with is a log likelihood . for this , one assumes models of letters evolving independently of their neighboring letters . one takes as scoring function @xmath23 , the logarithm of the probability that an ancestor letter evolves into an @xmath9 in one descendant and into a @xmath24 in the other descendant . here @xmath9 and @xmath24 are any two letters from the alphabet under consideration , or also the gap symbol . with this choice of scoring function , the optimal alignment becomes the maximum - likelihood estimate for the historic alignment . a high optimal alignment score indicates relatedness of the two dna - strings . thus in testing for relatedness of dna - strings one usually takes as test statistics the optimal alignment score . the significance of such a test depends then on the order of magnitude of the fluctuation of the optimal alignment score . it has been a long open question whether for two independent i.i.d . strings of length @xmath2 , the fluctuation of the optimal alignment score is of order @xmath25 or @xmath26{n}$ ] . ( see @xcite and @xcite ) . in this paper , we think we can see why these two different conjectures co - exist at the same time . we show that for two independent i.i.d strings of length @xmath2 the optimal alignment score ( depending on the parameters ) can contain a normal component with fluctuation of order @xmath25 . this component is not relevant for finding the optimal alignment score , because it does not depend on the alignment . _ when testing for relatedness of dna - strings one should remove this normal component , since it is only noise . this can be done by modifying the scoring function . once we `` cut out this normal component '' , the question that remain is what is the asymptotic distribution of the remainder ? is it tracy - widom ? we ca nt prove it in full generality , however we check this for the special case where the number of gaps is a small power of the length of the strings and the gaps are allowed only into one string . _ _ we also exhibit a scoring function with an alignment score which is of the order @xmath25 but not normal , nor tracy - widom . this , should then show that any claim that the order @xmath27=o(n)$ ] automatically implies asymptotic normality is wrong ! ( here @xmath28 designates the optimal alignment score of two i.i.d . random strings of length @xmath2 . ) to prove normality , one needs a more detailed knowledge of the path structure of the optimal alignments ! _ the optimal alignment can be viewed as a special case of last passage percolation with correlations . to our knowledge , many of the proofs for models of first and last passage percolation , where the asymptotic distribution is known , go back to proving asymptotic equivalence with the following functional of brownian motions : consider @xmath29 a sequence of i.i.d . standard brownian motions . also consider when @xmath30 goes to infinity , the functional @xmath31+[b^2(t_2)-b^2(t_1)]+\ldots+[b^k(t_k)-b^k(t_{k-1}))]\ ] ] which properly rescaled becomes tracy - widom @xcite . now , in our case , we have a similar situation , where instead we have a sequence of _ dependent _ identically distributed standard brownian motions @xmath32 . however , the correlation structure is rather simple . precisely , there exists a brownian motion @xmath33 and i.i.d brownian motions @xmath29 so that @xmath34 for all @xmath35 . hence , instead of independent brownian motions we have correlated ones and the maximum given in can be written as @xmath36+[w^2(t_2)-w^2(t_1)]+\ldots+[w^k(t_k)-w^k(t_{k-1}))]=\\&= v(1)-v(0)+\max_{0\leq t_1<t_2<\ldots < t_k\leq 1 } [ b^1(t_1)-b^1(0)]+[b^2(t_2)-b^2(t_1)]+\ldots+[b^k(t_k)-b^k(t_{k-1}))].\end{aligned}\ ] ] hence , we get a normal component plus a term on a smaller scale which is indeed tracy - widom . this is the reason why we can prove in the current paper , under a very special situation ( adding gaps only into one string and keeping their number a small power of @xmath2 ) , that in the limit we obtain a normal plus tracy - widom on a smaller scale . however , the normal part is often irrelevant for the optimal alignment problem . note here that the normal component @xmath37 does not dependent on the choice of @xmath38 ( which in what follows , correspond to alignments ) . again , the normal component is in some sense irrelevant for finding the optimal alignment . we should leave it out because it is only added noise . we consider two random i.i.d . strings @xmath39 and @xmath40 . in the sequel , @xmath28 denotes the optimal alignment score of @xmath0 and @xmath1 . chvatal - sankoff proved @xcite that @xmath41/n$ ] converges to a constant @xmath42 , but @xmath42 is not known in most cases . in @xcite , @xcite we have good confidence intervals for @xmath43 . we have been able to prove the order @xmath44=o(n)\ ] ] in many situations @xcite,@xcite , @xcite , @xcite . this proves the conjecture of waterman @xcite ( for certain specific distributions of the letters and scoring functions ) , but is different from what chavatal - sankoff have conjectured . the order conjectured by chavatal - sankoff is the order which one would expect knowing that @xmath45 is also the weight of the heaviest path from @xmath46 to @xmath47 in an lpp - formulation of our problem . ( in the physics literature , @xcite , the order for the standard deviation of the heaviest path from @xmath46 to @xmath47 in a vast class of lpp - models is conjectured to be @xmath48 not @xmath49 . for more details see also @xcite ) . let us explain what lpp is and how our optimal alignment can be formulated as an lpp - problem with correlated weights . in lpp , one considers an oriented graph @xmath50 with a random weight function @xmath51 of the edges . let @xmath52 . a path from @xmath19 to @xmath20 in the graph @xmath50 is then defined as a sequence of vertices : @xmath53 where @xmath54 and @xmath55 for all @xmath56 . the total weight of the path @xmath53 is given by : @xmath57 an optimal path from @xmath19 to @xmath20 is then a path which maximizes the total weight among all paths from @xmath19 to @xmath20 . + now , our optimal alignment problem can be formulated as an lpp on the integer lattice with the optimal alignment score @xmath28 being the weight of the path of maximal weight from @xmath46 to @xmath47 . for this we take the set of vertices @xmath58 to be @xmath59 and the edges to go always one to the right , one up or diagonally up to the next vertex . the weight for horizontal and vertical edges is minus the gap penalty . for the edge @xmath60 the weight is @xmath61 . with this setting when we align a letter with a gap , this corresponds to moving one unit vertically or horizontally . aligning @xmath62 with @xmath63 corresponds to moving along the edge @xmath64 . the optimal path then defines an optimal alignment where for every edge @xmath60 contained in the optimal path , aligns @xmath62 with @xmath63 . the transversal fluctuation measures how far the optimal path from @xmath46 to @xmath47 deviates from the diagonal . + let us give an example . take the two strings @xmath65 and @xmath66 . consider the following alignment with gaps : @xmath67 we can represent this alignment as a path on the integer points in two dimension . we are always allowed to go to the left by one unit or up by one unit or diagonal to the next integer point . when we go horizontal or vertical , this means that we align a letter with a gap . when we go diagonal this means that we align two letters with each other . below , we show how we represent the alignment given in by a path on the two dimensional integer lattice : @xmath68 to visualize , the path represented by the last table above , draw short segments between the points : @xmath69 to @xmath70 , then @xmath70 to @xmath71 , .... and up to a segment from @xmath72 to @xmath73 . the path is then given by this sequence of segments . so , the first such segment from @xmath69 to @xmath70 is diagonal , and corresponds to aligning the two @xmath74 s with each other . the score - contribution for this is @xmath75 . then , the segment from @xmath70 to @xmath71 is diagonal and corresponds to aligning the @xmath76 s . its score contribution is @xmath77 . the segment from @xmath71 to @xmath78 is vertical and hence corresponds to aligning a letter with a gap . in the current case , it is the first @xmath79 in @xmath80 which gets aligned with a gap , and the score for this is the gap penalty @xmath81 . then @xmath82 is again diagonal and corresponds to aligning @xmath2 with @xmath2 , with score @xmath83 . then , the segment @xmath84 is horizontal and corresponds to aligning @xmath85 in @xmath86 with a gap , which contributes to minus a gap penalty to the total score . the final segment @xmath87 is diagonal , and corresponds to aligning the last @xmath79 of @xmath86 with the last @xmath79 of @xmath80 . score is then @xmath88 . in other words , the horizontal or vertical moves contribute minus a gap penalty each , whilst the diagonal moves contribute to the score by aligning the corresponding letters . in this example , we aligned only identical letters . this is not always the case . in dna - analysis , when letters can mutate we also align similar letters with each other , not just identical . formally we can view an alignment with gaps of the strings @xmath89 and @xmath90 as a couple consisting of two increasing integer sequences . the sequence @xmath18 @xmath91 and the sequence @xmath92 @xmath93 are non - negative integers . here @xmath94 is any integer non - negative number . the alignment with gaps defined by @xmath95 is then the alignment which aligns @xmath96 with @xmath97 for all @xmath98 and aligns all remaining letters with a gap . the alignment score is then @xmath99 where is @xmath100 represents the total gap penalty . ( the penalty for one given gap being @xmath101 ) . in our current example , we have the number of aligned letter pairs is @xmath102 . furthermore , we find : @xmath103 , whilst @xmath104 . the path representing the alignment given by @xmath105 is the path containing all the segments : @xmath106 for @xmath98 and additionally a minimum amount of vertical and horizontal unit length integer segments so as to make this a connected path . + first passage percolation ( fpp ) and lpp are part of a vast area of statistical physics @xcite which is concerned with random growth models for which physicists expect some universality properties . more specifically , one considers growth of a cluster where material is being attached randomly on the surface of a nucleus . there are many fundamental questions open for decades , such as the universality of the fluctuation exponents . many theoretical physicists , crystallographers and probabilists have worked on this and many conjectures have been formulated @xcite , as for example that the fluctuation should behave like @xmath48 and the transversal fluctuation should be or order @xmath107 ( see kpz - conjecture in @xcite ) . these are difficult conjectures and rigorous results were obtained only under some special circumstances of lpp models , like longest increasing subsequence @xcite , lpp on @xmath108 with exponential or geometric waiting times @xcite and the case of percolation in a narrow tube @xcite . to our knowledge , the best results show only a fluctuations of less than @xmath109 @xcite and above @xmath110 @xcite . we have been able to prove our results for several longest common subsequence ( lcs ) and optimal alignments ( oa ) models . + lcs are a special case of oa . for two strings @xmath89 and @xmath111 a common subsequence of @xmath19 and @xmath20 is a sequence which is a subsequence of @xmath19 and at the same time a subsequence of @xmath20 . a lcs of @xmath19 and @xmath20 is a common subsequence of @xmath19 and @xmath20 of maximum length . the length of the lcs of @xmath19 and @xmath20 can be viewed as an optimal alignment score of @xmath19 and @xmath20 . for this we take @xmath112 to be @xmath70 if @xmath113 and @xmath114 otherwise . on top the gap penalty should be taken @xmath69 . with this choice , the length of the lcs is also equal to the optimal alignment score . + our results together with the results of baik , deift and johansson @xcite , are among the few rigorous results determining the exact order of the fluctuation exponents for some non - trivial fpp or lpp related 2-d models . in their work , as in ours , the fluctuation is of order @xmath48 and the rescaled asymptotic limiting law is tracy - widom . another interesting paper is for the tasep model , by borodin , ferrari , praehofer and sasamoto @xcite . there they do not use the same approach of reducing the problem to formula and , and it is not the gue which appears in the limit , but goe . other methods used to establish the solvability of lpp models are the vertex operator and fock space formalism put foward by okounkov @xcite in the study of schur process . our paper is organized as follows . in section [ s:2 ] , we introduce the main decomposition of the scoring function and show examples where the limiting distribution of the optimal alignment score is normal alongside with other examples which are non - normal . in the next section , namely section [ s:3 ] , we restrict the number of gaps to a certain number @xmath30 which is finite and fixed and allow the gaps only in one sequence . here we give a general description of the optimal alignment in terms of multiple dependent random walks and outline an argument which supports the conjecture that under some conditions , if @xmath30 is replaced by @xmath115 with small @xmath116 , then the ( rescaled ) limiting distribution of the optimal alignment is tracy - widom . we can not prove this statement rigorously in general , but in the next section , namely section [ s:4 ] , we prove this for the key scoring function @xmath117 under the assumption that the number of gaps is @xmath118 for small @xmath116 and the second sequence of letters is distributed normally rather than @xmath119 . the main rigorous result is stated in theorem [ t:1 ] and appears in section [ s:4 ] . section [ s:5 ] is dedicated to technical results while the last one , section [ s:6 ] contains a few simulations . in this section we discuss the general decomposition of the scoring function in a natural basis . the point of this is that some of the scoring functions are in some sense artificial because the optimal score does not depend on the alignment . this will give a normal behavior in the limit for the optimal score . we consider symmetric scoring for a binary alphabet . we do not allow alignments of gaps with gaps , so the scoring functions we consider are given by a @xmath120 matrix , with one irrelevant entry , namely @xmath121 here @xmath7 stands for a gap . in our case we will put @xmath122 and we will always assume that the above matrix is symmetric hence @xmath123 and @xmath124 , whilst @xmath125 . the linear space of such @xmath120 symmetric matrices is generated by the following five elements @xmath126 and @xmath127 note that the matrices @xmath128 , @xmath129 and @xmath130 are orthogonal to each other ( with respect to the canonical inner product given by @xmath131 ) . also , the matrices @xmath132 and @xmath133 concern only with the gap penalty . thus , any scoring function we will consider can be written as @xmath134 first , it is important to note that the part @xmath135 gives the same alignment score for any alignment . this is so , because @xmath128 and @xmath129 can both be viewed as linear additive functions in the following way . for a given a function @xmath136 , take the linear scoring function @xmath6 defined by : @xmath137 for all @xmath138 . we also assume that @xmath139 . clearly for such a `` linear '' scoring function , with value @xmath139 , the alignment score for any alignment @xmath11 does not depend @xmath11 . indeed , it is trivial to check that @xmath140 furthermore , the expression on the right side of the above equality is a linear function of a binomial random variable . now , in our case , @xmath129 is of the above type with the function @xmath74 given by @xmath141 thus , we find that @xmath142 where @xmath143 denotes the total number of @xmath144 s in both strings @xmath0 and @xmath1 combined whilst @xmath145 denotes the number of @xmath146 s in both strings combined . noting that @xmath147 , we finally find that @xmath148 here @xmath143 is a binomial with parameter @xmath149 and @xmath150 , consequently , @xmath151=var[n_n^a]=2np_{a}(1-p_{a}).\ ] ] similarly , the alignment score according to @xmath128 does not depend on the alignment and is also equal to @xmath2 . since the alignment scores according to @xmath128 and @xmath129 do not depend on the alignment we find for any scoring function @xmath152 that the optimal alignment score can be decomposed as follows : @xmath153 according to the central limit theorem , @xmath143 is asymptotically normal on the scale @xmath25 and thus we can conclude that @xmath154 where @xmath155 . now , since the part @xmath156 does not depend on the alignment it has no affect on which alignment is optimal . also , it is just noise added for testing if sequences are related or not . this noise can be removed , since it is known . equivalently , we can just take @xmath157 as scoring function instead of @xmath152 . in this section we present a simple example for which the optimal alignment score , properly rescaled , converges to a distribution which is not normal . interestingly , the scaling is still the same as in the case of the previous section , namely , @xmath25 . as mentioned , for scoring functions of the type @xmath158 , the alignment score does not depend on the alignment @xmath11 of the strings @xmath159 and @xmath160 . furthermore , the limiting distribution of the optimal alignment score is normal on the scale @xmath25 since @xmath161 where @xmath162 is the total number of @xmath144 s in the concatenated string @xmath163 . since @xmath162 is binomial with parameter @xmath149 and @xmath164 , the above optimal score , properly rescaled converges to the normal . thus we can interpret the scoring function with @xmath165 as degenerate and hence not relevant for practical purposes . next we consider the following scoring function @xmath166 if @xmath167 designate the number of @xmath144 s in @xmath39 and @xmath168 designate the number of @xmath144 s in @xmath40 , then , the optimal alignment score according to the scoring function @xmath169 is simply equal to the minimum of @xmath167 and @xmath170 , namely , @xmath171 this is so because aligning a @xmath146 with a @xmath146 gives a negative score , whilst aligning something with a gap gives @xmath69 . so , the optimal alignment is going to align all @xmath146 s with gaps . since an @xmath144 aligned with an @xmath144 scores positively , the optimal alignment is going to align a maximum number of @xmath144 s with each other . + note that @xmath167 and @xmath168 have the same expectation and both tend asymptotically to be normal when re - scaled by @xmath172 . hence , we get @xmath173}{\sqrt{np(1-p)}}= \min\left\{\frac{n_a^x - e[n_a^x]}{\sqrt{np(1-p ) } } , \frac{n_a^y - e[n_a^y]}{\sqrt{np(1-p)}}\right\ } \approx\min\{\mathcal{n}_1,\mathcal{n}_2 \}\ ] ] where @xmath174 and @xmath175 are two standard normal independent of each other , which is certainly not normal . thus , we have a non - normal limiting distribution , but still we have a fluctuation on a scale @xmath25 ! notice also the following interesting fact , namely that there are many different alignments , which are very far away from each other . indeed , say that @xmath176 . then , the optimal alignments can be characterized as follows . choose a subset of @xmath144 s in @xmath0 with cardinality equal to @xmath177 . these @xmath144 s are aligned with gaps . then align all the remaining @xmath144 s in @xmath0 with @xmath144 s from @xmath1 . finally , align all the @xmath146 s with gaps . now , real life scoring functions should not have @xmath178 too small , since in practical situations , like genetic evolution , there are usually not too many gaps . this reflects the fact that there are not too many deletes from an ancestors dna . in the next section we show that if we look for the optimal alignment under the constrain of the total number of gaps being small , then we get that @xmath179 can be decomposed into a normal part on the scale @xmath25 plus a smaller part whose limiting distribution is non - normal . in this section , under the assumptions that the number of gaps is finite and fixed , we present a general outline of an argument which should be sufficiently convincing that the limiting behavior of the optimal alignment is normal in the scale @xmath25 plus a tracy - widom on a smaller scale . let us assume that we only consider alignments with a fixed number of gaps equal to @xmath30 , where @xmath30 is a fixed constant . to simplify the problem we will allow the gaps only in one sequence . let @xmath30 be a constant and let @xmath180 and @xmath181 be the two random strings of letters . thus , we will align exactly @xmath30 letters from @xmath1 with gaps , but all the letters of @xmath0 will be aligned with letters of @xmath1 . we will consider the optimal alignment of @xmath0 and @xmath1 under the constrain that there are exactly @xmath30 gaps all inserted in @xmath182 . the corresponding optimal score will be denoted by @xmath183 . to specify an alignment we simply need to specify which @xmath30 letters of @xmath1 get aligned with gaps , thus we need to specify a sequence of @xmath30 increasing integers @xmath184 . the vectors consisting of these integers then define our alignment , namely , align @xmath185 with a gap for all @xmath186 . let @xmath187 denote the random walk defined recurrently on @xmath188 by @xmath189 and then : @xmath190 for all @xmath191 and for @xmath192 $ ] . ( to simplify the notation we assume that the sequence of the @xmath62 s is a double infinite sequence of i.i.d . variables @xmath193 in the problem which we will consider only a finite number of @xmath62 s will have @xmath194 and their contribution will be negligible ) . for example , the first random walk @xmath195 is then defined by @xmath196 similary , the second random walk @xmath197 is defined as follows : @xmath198 + consider the alignment score under the constrain that the @xmath199-th gap is aligned with @xmath200 , for all @xmath201 , where @xmath202 is a given increasing sequence of integers . the alignment with exactly @xmath30 gaps is uniquely defined by this constrain and its alignment score is : @xmath203 recall that we assume a zero gap penalty . let us give an example . take the string @xmath204 and @xmath205 we could align @xmath206 and @xmath207 with a gap . then we have @xmath208 gaps and @xmath209 whilst @xmath210 . so , this defines the @xmath71-gap - alignment @xmath11 given by : @xmath211 hence , the alignment score with a @xmath69-gap penalty is equal to : @xmath212 + let us denote by @xmath183 the optimal alignment score under the constrain that the alignment contains exactly @xmath30 gaps all aligned with letters from @xmath1 . since the gap penalty is @xmath69 the formula for @xmath183 is then given by @xmath213 hence , the maximum is obtained by choosing the increasing sequence @xmath214 which maximizes the following sum : + take the increment of the first random walk on @xmath215 $ ] , then we add the increment of the second random walk on @xmath216 $ ] and keep on going until the increment of the last random walk on @xmath217 $ ] has been added . next , we are going to put the different one - dimensional random walks @xmath187 together into one vector : @xmath218 in this manner @xmath219 becomes a @xmath220-dimensional `` correlated '' random walk . thus we have @xmath221 and @xmath222 if we subtract the drift and rescale by @xmath25 , the multidimensional correlated random walk @xmath223 converges to a @xmath220 dimensional brownian motion and we refer to @xcite for more details . this means , that holding @xmath30 fixed , we can find a @xmath224 which goes to @xmath69 as @xmath2 goes to infinity so that there exists a @xmath220-dimensional brownian motion @xmath225 for which @xmath226 } \left| \frac{\vec{r}(tn)-e[\vec{r}(tn)]}{\sqrt{n}}-\vec{b}(t)\right|\leq \delta_n,\ ] ] holds with probability @xmath227 where @xmath228 goes to @xmath69 as @xmath229 . here the brownian motion @xmath230 depends on @xmath2 , but of course , its distribution does not . if @xmath231 $ ] is not an integer then we take for @xmath232 the linear interpolation of @xmath233 . let us introduce a small modification of @xmath183 . this modification will be denoted by @xmath234 and is equal to @xmath235 clearly with this definition , @xmath236 note that for all the random walks @xmath187 , the drift is the same and equal to @xmath237 $ ] . now , we can rewrite @xmath238 using the unbiased random walk @xmath239 \right)\right)\;\;+\;\;n\cdot e[s(x_1,y_1)].\ ] ] assume next that inequality holds . then , we can approximate the unbiased random walks on the right side of by components of a brownian motion according to . since the sum contains @xmath240 terms , the approximation error will be at most @xmath241 ( if we assume to hold ) and thus @xmath242}{\sqrt{n}}-l^k(\vec{b } ) \right|\leq 2\delta_n(k+1),\ ] ] where @xmath243 refers to the maximum @xmath244 we can now use inequality together with to find @xmath245}{\sqrt{n}}-l^k(\vec{b } ) \right|\leq 2\delta_n(k+1)+ \frac{2c\cdot k}{\sqrt{n}}.\ ] ] the component of our brownian motions @xmath230 could be correlated . to see this , let s calculate the covariance between the first two random walks , that is between @xmath246 and @xmath247 . the idea is that most terms of the sum corresponding to the random walk are not correlated , however the closer ones are correlated . to the point , take for example the first term @xmath248 in the sum @xmath249 since , the @xmath62 s and @xmath250 s are all independent of each other , an expression of the type @xmath61 can only be correlated with @xmath251 if either @xmath252 or @xmath253 . therefore , in the sum @xmath254 there are only two terms which can have non - zero correlation to @xmath255 . these terms are @xmath256 and @xmath257 . this analysis holds true for every term in the sum for the first random walk . only in the very last term in the sum on the right side of there is a term for which there are not two terms in the sum on the right side of correlated with it . hence , the covariance between and is about @xmath149 times constant , where the constant is given as the covariance between @xmath251 and @xmath256 . let us expand on this calculation a little bit by looking at the covariance terms @xmath258 we can work out the same calculation for all pairs @xmath259 of random walks @xmath187 and @xmath260 provided @xmath261 . the covariance is always about @xmath262 . hence , the multi - dimensional brownian motion in our approximation has a covariance matrix with all non - diagonal entries equal to @xmath263 . hence , we find @xmath264 for @xmath261 and @xmath265=t\cdot var[s(x_1,y_1)].\ ] ] this is a simple covariance structure since there is one common factor contained in all the brownian motions @xmath266 if @xmath263 is non - zero . in other words , we find that there exists a multidimensional brownian motion @xmath267 with all components independent and having variance @xmath268=t \left(var[s(x_1,y_1)]-2cov(s(x_1,y_1),s(x_0,y_1))\right)\ ] ] so that @xmath269 where @xmath270 represents the @xmath220-dimensional identity matrix and @xmath33 is a one dimensional brownian motion which is independent of @xmath271 . ( to understand more in detail how we get the @xmath58 and the @xmath272 consider this : simply take a brownian motion @xmath58 with @xmath69 expectation starting at @xmath69 and independently a collection of brownian motions @xmath273 which among themselves are i.i.d . also , the @xmath274 s have all @xmath69 expectation and start at @xmath69 . you can now chose all these brownian motion , so that @xmath275 have the same correlation structure then the collection @xmath276 . then , because you deal with gaussian process you have the same distribution . then you can make a perfect coupling between the @xmath275 and the @xmath276 since they have the same distribution ) . we also have @xmath277=2tcov(s(x_1,y_1),s(x_1,y_2)).\ ] ] hence , we find that @xmath278 where @xmath279 it is known ( see @xcite ) , that the rescaled @xmath280 goes asymptotically to tracy - widom , more precisely , @xmath281 where convergence is in law as @xmath282 . hence , there is a sequence of random variable @xmath283 which goes to @xmath69 in probability as @xmath282 , and a sequence of variables @xmath284 each having same tracy - widom distribution , so that @xmath285 we can now rewrite expression using and get @xmath286}{\sqrt{n}}-(v(1)-v(0)+l^k(\vec{w } ) ) \right|\leq 2\delta_n(k+1)+ \frac{2c\cdot k}{\sqrt{n}}\ ] ] which , with the help of equation , becomes @xmath287 - 2\sqrt{nk}}{\sqrt{n}}-(v(1)-v(0)+ \frac{f_{tw}^k}{k^{1/6 } } ) \right|\leq 2\delta_n(k+1)+ \frac{2c\cdot k}{\sqrt{n}}+\frac{\mathcal{e}_k}{k^{1/6}}.\ ] ] for the above inequality to be useful , we need the right side of it to be smaller than @xmath288 . this will depend on the value of @xmath224 . now , we want to let @xmath30 go to infinity at a slower rate then @xmath2 . most papers on closeness of random walk and brownian motion are for fixed dimension . also , in our case , any @xmath30 consecutive components of the multi - dimensional random walk @xmath289 are correlated . however , still assuming that the approximations hold for fixed dimension and correlated multidimensional random walk , we could take @xmath224 to be @xmath290 whilst @xmath291 for @xmath292 a constant not depending on @xmath2 or @xmath30 . in this case for the inequality to be useful , we would need @xmath293 to be of smaller order than @xmath288 . this is the case when @xmath294 , with @xmath116 small enough . in this section , we are going to treat the case where the number of gaps @xmath30 increases to infinity with @xmath2 and the simplest such scenario is the case of @xmath295 for some small @xmath296 which we assume is not depending on @xmath2 . again , we look for the best alignment score under all alignments with the prescribed number of gaps being @xmath30 and the gaps appear only in the sequence of @xmath0 s , thus we allow gaps to align with @xmath1 s . the highest alignment score among all alignments which align exactly @xmath30 letters of @xmath40 with gaps , but no letter of @xmath297 with gaps , is again denoted by @xmath298 when the number of gaps is fixed , in determining which alignment is optimal , the penalty for gaps does not matter much . indeed , the gap penalty will affect the optimal alignment only by a term of order @xmath115 . moreover , if we assume that the penalty of aligning a letter with a gap does not depend on the letter , we can simply assume that the gap penalty is actually equal to @xmath69 . therefore , the only part of the scoring functions which matters is the @xmath299 symmetric matrix @xmath300 the space of all such @xmath299 symmetric matrices constitute a @xmath78 dimensional space where a basis is formed by the matrices @xmath301 in addition , if we define two scoring matrices to be _ equivalent _ if the optimal alignments are the same , then it is easy to see that for any arbitrary scoring function @xmath169 , the scoring itself @xmath169 and @xmath302 are equivalent . thus the main task is to understand what happens under alignments with the scoring function @xmath303 . thus we assume for the rest of the section that the scoring function @xmath169 is equal to @xmath130 , so that : @xmath304 now , in the case considered here of a two dimensional alphabet , that scoring function can be viewed as a product scoring function by observing that if we introduce a numeric code for the letters , namely take @xmath144 to be @xmath305 and @xmath146 to be @xmath306 . then we find that @xmath307 using this in , the key observation is that the part @xmath33 has covariance given in formula by @xmath308=2cov(s(x_1,y_1),s(x_1,y_2)),\ ] ] which in the present case becomes @xmath309=cov[x_1y_1,x_1y_2]=e[x_1 ^ 2y_1y_2]-e[x_1y_1]\cdot e[x_1y_2]=\\ & = e[x_1 ^ 2]e[y_1]e[y_2]-e[x_1]e[y_1]\cdot e[x_1]e[y_2]=0\end{aligned}\ ] ] due to independence and mean @xmath69 of @xmath310 . hence , the term @xmath33 disappears from inequality for this choice of the scoring function . next , we notice that for a one dimensional `` uncorrelated '' random walk , the best coupling of the walk with a brownian motion in can be attained with @xmath311 ( see @xcite ) . with this choice of @xmath224 , the term @xmath312 in is of order @xmath313 , thus is much smaller than @xmath314 which is of order @xmath315 . however , since we have to correlate a large number of random walks ( exactly @xmath30 and not independent ) , we expect @xmath224 to be larger than @xmath316 . a different look at the dominating term on the right hand side of , gives that @xmath317 which is enough ( at least heuristically ) to guarantee that the rescaled optimal score is asymptotically tracy - widom . in order to deal with the approximation of the random walk by brownian motions , we need to get good couplings of the random walk with brownian motions this is what we want to do next . we should also quickly mention , that at this stage , it should be heuristically clear why we should expect tracy - widom as limiting distribution as soon as the constant @xmath296 is small enough . indeed , when @xmath296 is very small , then @xmath224 should be as `` close as we want '' to the formula for the one dimensional random walk with independent increments . now , for the one - dimensional random walk , it is known to allow a coupling to a brownian motion with @xmath311 . again , the only problem is to find a good coupling of the multidimensional random walk @xmath223 to a multidimensional brownian motion . the random walk @xmath223 has @xmath220 components . furthermore , there is dependency among the increments of the random walk . each step ( increment ) of the random walk , can depend on @xmath30 steps before and after . hence , @xmath223 is independent of @xmath318 for all @xmath319 . the main question is thus how well can a random walk with dependencies be approximated by brownian motion ? ( see @xcite for the one dimensional case , @xcite for severeal dimensions and @xcite for overviews . ) is it possible to find such a coupling when the dimension of the random walk grows to infinity ? this is delicate as the constants involved depend on the dimension and thus are not very useful in our case . unfortunately we can not treat both problems simultaneously and it seems highly technical to deduce such a result in great generality . instead , we present here a simple approach for a simplified situation . first , we use the fact that our scoring function is a product . furthermore , _ for the rest of the paper we also assume the variables @xmath320 to be i.i.d . normal with @xmath321 instead of binary variables . the variables @xmath322 are still binary variables in this section with @xmath323 _ again , since @xmath324=0 $ ] the component @xmath37 in the equation dissapears . the normal assumption on @xmath1 s is in place here to simplify the calculations . we believe that macroscopically it amounts to the same as taking these variables binary . the fact that the variables @xmath320 are standard normal instead of binary has a very important consequence on the calculation side , namely , the optimal alignment still tries to align the maximum number of letter with the same sign and also will try to get that those @xmath250 s which get aligned with an @xmath62 of opposite sign to be small in absolute value . the key strategy of the proof is to condition on the sequence @xmath0 and exploit the normal structure of the @xmath1 . to set up the ground , let @xmath325 and let @xmath326 be a non - random binary vector . we are going to work conditional on @xmath327 . now , consider our random walk @xmath219 in the current case where @xmath328 . we have @xmath329 where @xmath330 and @xmath331 looking at the above expression for the random walk @xmath219 , we see that the steps are dependent but once we condition the @xmath62 s to be constants , i.e. @xmath327 and @xmath320 to be i.i.d @xmath332 , the @xmath333 s become independent normal vectors . hence , conditional on @xmath327 , we get that the multidimensional random walk @xmath219 has independent normal increment with expectation @xmath69 . this random walk is not yet a brownian motion restricted to discrete times but we are going to show that we can construct such a brownian motion which is a good approximation . in order to do this _ we consider at first the random walk @xmath223 restricted to the times @xmath334 which are a multiple of @xmath335=\text{time - mesh size for coupling the random walk}\ ] ] _ with some small @xmath336 which will be determined later and @xmath337 $ ] denoting the integer part of @xmath19 . thus , we will consider @xmath338 where to simplify notation we assume @xmath2 to be a multiple of @xmath188 , though we can formally replace @xmath2 by @xmath339 with @xmath340 $ ] and run essentially the same argument . using the independence of the increments conditional on @xmath327 , we calculate the covariance matrix of the position of the random walk at time @xmath188 in the following way : @xmath341= cov\left[\sum_{i=1}^j\delta r_i|x = x\right]= \sum_{i=1}^jcov[\delta r_i|x = x]\\ & = \sum_{i=1}^j \left(\begin{array}{c } x_{i}\\ x_{i-1}\\ x_{i-2}\\ ... \\ x_{i - k+1 } \end{array } \right)\cdot ( x_{i},x_{i-1},x_{i-2},\ldots , x_{i - k+1}).\end{aligned}\ ] ] the key fact here is that with very high probability the last sum in the above equation is approximately @xmath342 where @xmath270 is the @xmath220 identity matrix . now , using this , and the fact that @xmath342 is the covariance matrix of a standard @xmath220-dimensional brownian motion at time @xmath188 we will show that there is a coupling between the random walk and a standard brownian motion such that @xmath343 is close to the brownian motion position at time @xmath188 . moreover the argument can be extended to include also the increments of the random walk . more precisely , we show that the coupling can be chosen such that the increments @xmath344 are also close to the increment of the coupled brownian motion . to construct such a coupling , we will build a brownian motion @xmath345 by coupling at the times @xmath346 at first . more precisely , we couple @xmath343 with @xmath347 , then the increment @xmath348 with the increment @xmath349 and then we continue coupling the next increment @xmath350 with @xmath351 and so on and each step we do this independently of the construction at the previous steps . recall that we do this conditionally on @xmath327 . the key is that the construction at each step is a consequence of lemma [ coupling ] in the next section [ howto ] where we show how to couple a normal random variable having covariance close to a multiple of the identity . essentially the main assertion here is that if @xmath352 is a normal vector such that @xmath353 then we can find a normal vector @xmath354 of i.i.d normal components @xmath355 such that @xmath356 . here we use the operator norm @xmath357 for a matrix and measure the closeness in the operator norm , in particular all components of @xmath358 have variance at most @xmath359 . once all the increments @xmath360 are defined , we define @xmath361 for @xmath362 as the sum @xmath363 conditional on @xmath327 , the pieces @xmath364 for @xmath365 are independent . since @xmath366 depend only on , this implies that conditional on @xmath327 , the sequence of random vectors @xmath367 constitutes a multidimensional standard brownian motion restricted to the times @xmath368 . we can for instance use the construction of levy for the brownian motion to extend these vectors to a brownian motion such that the increments are given exactly by @xmath360 for @xmath369 . we do not look for other details about the brownian motion during the times @xmath370 $ ] , we will simply use some basic estimates for the maximum of brownian motions between these time intervals . for instance using the reflection principle we can show that @xmath371}|\vec{c}(u)-\vec{c}(s)|\ge \lambda)\le const(k+1)e^{-\lambda^{2}/(2(t - s))},\quad \forall \lambda>0.\ ] ] in particular what this yields is that for some constant @xmath372 independent of @xmath2 and @xmath116 @xmath373}|\vec{c}(u)-\vec{c}(s)|\ge \log(n)\sqrt{j})\le cn^{\alpha}e^{-\log^{2}(n)/2}\ ] ] thus the deviation of the brownian motion on each of these intervals is very well controlled . now define a few events related to our coupling which will take care of the main estimates . * let @xmath374 be the event that none of the components of the standard multidimensional brownian motions @xmath361 goes further than @xmath375 during any of the time intervals @xmath370\subset [ 0,n]$ ] for @xmath376 that is let @xmath377 be the event that @xmath378}|c^l(ij+s)-c^l(ij)|\leq \log(n)\sqrt{j}\ ] ] for all @xmath379 . set @xmath380 * let @xmath381 be the event that the multidimensional random walk @xmath223 does not move further than @xmath375 during any of the time intervals @xmath370\subset [ 0,n]$ ] . hence , let @xmath382 be the event @xmath378}|r^l(ij+s)-r^l(ij)|\leq \log(n)\sqrt{j}\ ] ] for all @xmath379 . let @xmath383 * the event @xmath384 takes care of the fact that the covariance matrix of @xmath385 for all @xmath379 is close to @xmath342,were @xmath270 denotes the @xmath386 identity matrix . more specifically , let thus @xmath387 be the event that @xmath388-ji|\leq c\cdot k\sqrt{j},\ ] ] where @xmath389 is a constant which will be defined later . similarly for any integer @xmath79 so that @xmath390\subset[0,n]\ ] ] holds , let @xmath391 be the event that @xmath392- ji|\leq c\cdot k\sqrt{j}.\ ] ] finally let @xmath393 be the event @xmath394 * the next event @xmath395 assures that the random walk @xmath223 and the brownian motion @xmath361 are close to each other at times which are multiples of @xmath188 . more specifically , let @xmath396 be the event that the @xmath199-th component of @xmath397 at times which are multiples of @xmath188 is bounded as follows : @xmath398 + let @xmath399 be intersection of these for all @xmath199 , precisely : @xmath400 now , we summarize the picture we have obtained so far . rescale the random walk and the brownian motion @xmath401 by a factor @xmath25 and the time by a factor @xmath2 by setting for @xmath402 $ ] @xmath403 by scaling the brownian motion , we know that @xmath230 is again a brownian motion . when the events @xmath399 , @xmath404 and @xmath405 all hold , inequality is satisfied for the multi - dimensional brownian motion @xmath406\rightarrow\mathbb{r}^k$ ] with @xmath224 given as @xmath407 where again , the number of gaps @xmath30 and the time step are given by @xmath408 we can now go back to inequality [ gissv4 ] , to find that : @xmath409 - 2\sqrt{nk}}{\sqrt{n}}- \frac{f_{tw}^k}{k^{1/6 } } \right|\leq \\ & \leq 9 \max\{\frac{k\log(n)\sqrt{j}}{\sqrt{n } } , \frac{k\log(n)\sqrt{k}}{j^{1/4}}\}+ + \frac{\mathcal{e}_k}{k^{1/6}}\end{aligned}\ ] ] holds with probability at least @xmath227 where @xmath410 recall also , that @xmath411 designates a sequence of random variables which go to @xmath69 in probability . the sequence @xmath412 designates on the other hand a sequence of random variables which are all identically distributed according to tracy - widom . + we eliminated the term @xmath413 by increasing the coefficient in front of @xmath414 . furthermore , from inequality [ turkish ] , we obtain the important conclusion that in order to show that the re - scaled optimal alignment score with @xmath118 imposed gaps goes to tracy widom we only need to prove prove two things : * i : * @xmath228 goes to zero , hence @xmath415 as @xmath416 , which we solve in section [ highprob ] . * ii : * the term @xmath417 in the bound in [ turkish ] is of order less than @xmath288 . this is the place where we choose @xmath116 and @xmath418 . we have to verify that there exists a choice @xmath296 and @xmath336 making expression [ reference ] less than order @xmath419 . in other words , we want @xmath420 which is satisfied for @xmath2 large enough as soon as @xmath421 this is of course possible if and only if @xmath422 which leads to @xmath423 . thus , for any @xmath424 , we have convergence of the re - scaled optimal alignment score to tracy - widom . hence , we are able to prove our result provided there are less than order @xmath425 gaps in the optimal alignment ! let us formulate this fact in the form of a theorem . let @xmath426 be a constant . let @xmath322 be an i.i.d . sequence of binary variables so that @xmath427 and let @xmath320 be an i.i.d . sequence of standard normal random variables . let @xmath130 denote the scoring function defined by @xmath428 for any two letters @xmath429 of our binary - alphabet . let @xmath430 $ ] . then , the optimal alignment score with constrain to have @xmath30 gaps and properly rescaled as @xmath431 - 2n^{1/2+\alpha/12}}{n^{1/2-\alpha/6}}\ ] ] converges in law to tracy - widom . multiplying by @xmath432 , we find that that @xmath433 - 2\sqrt{nk}}{\sqrt{n}}- f_{tw}^k \right|\leq \\ & \leq k^{1/6}\cdot 9 \max\{\frac{k\log(n)\sqrt{j}}{\sqrt{n } } , \frac{k\log(n)\sqrt{k}}{j^{1/4}}\}+ \mathcal{e}_k\end{aligned}\ ] ] holds with probability at least @xmath227 , in the next section it is proven that @xmath228 goes to @xmath69 as @xmath229 . ( recall also inequality ) . furthermore , since @xmath434 , we have from that @xmath435 goes to @xmath69 as @xmath229 . hence , since also @xmath283 go to @xmath69 in probability , we have that @xmath436 - 2\sqrt{nk}}{\sqrt{n}}- f_{tw}^k \right|\ ] ] goes to @xmath69 in probability , when @xmath416 . hence , since @xmath284 always has tracy - widom distribution , it follows , that @xmath437 - 2\sqrt{nk}}{\sqrt{n}}\ ] ] converges in law to tracy - widom as @xmath229 . but , when we replace @xmath30 by @xmath438 in expression [ k1/6 ] , we obtain expression from our theorem . so , we just finished proving that converges to tracy - widom in law . * remark . * we believe that taking the variables @xmath320 binary instead of standard normal our theorem should holds for @xmath296 small enough . however , the proof for approximating the multidimensional random walk @xmath439 by a multidimensional brownian motion will be much more difficult . the intricate part is very similar to passing from the clt to berry - essen type clt theorem is when there are non - finite correlations or dimension goes to infinity at the same time as @xmath2 ( see @xcite , @xcite ) . now we got tracy - widom limiting distribution for the binary case , by using the scoring function @xmath130 and fixing the number of gaps . if instead we use a scoring function @xmath440 then we get an additional normal term on the scale @xmath25 . indeed , @xmath128 and @xmath129 do not affect which alignment is optimal , thus , @xmath441 as argued before , @xmath442 depends only on the number of @xmath144 s and @xmath146 s in each of the sequences @xmath0 and @xmath1 . it is even a linear functional of these numbers of occurrences and hence asymptotically normal on the scale @xmath25 . on , the other hand , according to our theorem , we have that after rescaling @xmath443 is asymptotically tracy - widom on the scale @xmath444 , ( provided @xmath445 and @xmath294 ) . thus , there is a larger normal part and tracy - widom on a lower scale . but , for optimal alignment of dna sequences , one should always set @xmath446 , since the part @xmath135 does not influence which alignment is optimal . in that case when @xmath446 , the limiting distribution is purely tracy - widom . _ the main question to ask here is what happens if we allow the number of gaps to be of order @xmath447 which we leave as an open problem here . _ the main coupling result is contained in the following lemma . [ coupling ] let @xmath448 be a @xmath449-dimensional normal random vector with expectation @xmath69 . let @xmath450 and @xmath451 . assume that latexmath:[\[\label{e : cov } @xmath449-dimensional normal random vector @xmath453 with covariance matrix @xmath342 and expectation @xmath69 so that @xmath454|\le\epsilon.\ ] ] in particular all the components of @xmath455 have variance at most @xmath359 . the main idea is to construct a normal vector simply by rescaling the vector @xmath448 . the construction uses the fact that if @xmath9 is a linear transformation from @xmath456 into itself , then @xmath457 is again a normal random variable of covariance matrix @xmath458 . in this framework we take @xmath459 and @xmath460 clearly , we now have @xmath461 furthermore , @xmath462=cov[(i - j^{1/2}\sigma^{-1/2})\vec{v}]=(i - j^{1/2}\sigma^{-1/2})\sigma(i - j^{1/2}\sigma^{-1/2})=(\sigma^{1/2}-j^{1/2})^{2}.\ ] ] taking the eigenvalues of @xmath463 to be @xmath464 , the condition reads as @xmath465 on the other hand since @xmath466 , we also know have that @xmath467 . thus the conclusion is reached from the fact that the eigenvalues of @xmath468 are @xmath469 and the following chain of inequalities @xmath470 [ aj ] we have that @xmath393 has probability close to one . more precisely , assuming @xmath471 , then for any constant @xmath472 , the following holds @xmath473 consider the sum on the right side of . instead of taking the full sum , we are going to split the sum in the following form @xmath474 where for simplicity we assume that @xmath188 is a lot larger than @xmath30 and that @xmath188 is a multiple of @xmath30 . since for each @xmath475 , the sequence @xmath476 are iid when @xmath79 runs from @xmath70 to @xmath477 . the covariance matrix of these vectors is the @xmath220-dimensional identity matrix @xmath270 . so , @xmath478 is an estimate of the covariance matrix of @xmath479 . hence , expression [ sumo ] is approximately equal to @xmath270 and @xmath480 now , according to @xcite the precision of such an approximation in operator norm is the square root of the dimension times the square root of the sample size . more precisely , we have @xmath481 here , @xmath482 is a constant which does not depend on @xmath188 . hence , for the full sum , this means @xmath483-ji|\geq ck\sqrt{j}\right ) \leq ( k+1)e^{-j / k+1},$$ so , we have proven that $ $ p((a^{n}_{l})^{c})\leq ( k+1)\exp^{-j / k+1}.\ ] ] hence , since @xmath484 and since @xmath485 we have @xmath486 since , @xmath487 and @xmath294 with @xmath488 it follows that the dominant term in the bound on the right side of [ merde ] is @xmath489 which is an exponential decaying function in a fractional power of @xmath2 . the next lemma shows that when the event @xmath393 holds , then @xmath399 has high probability we have that @xmath490 in particular , @xmath491{}0 $ ] . we work conditional on @xmath327 . we have that when the event @xmath393 holds , then @xmath492 where @xmath270 is a @xmath220 dimensional identity matrix . so , our coupling of @xmath493 to @xmath494 is such that the covariance matrix of the difference has norm at most @xmath495 . thus , conditional on @xmath327 , and assuming @xmath393 holds , we get latexmath:[\[\label{differences } this implies that the @xmath199-th component of the difference @xmath497 has a variance less or equal to @xmath495 . but conditional on @xmath327 , we have that the differences are normal , independent with expectation @xmath498 . taking the @xmath199-th component of @xmath499 at the times @xmath500 is thus like taking a standard brownian motion at the times @xmath501 where @xmath502,\\ & t_2=var[\vec{r}(2j)-\vec{c}(2j)]\\ & \ldots\end{aligned}\ ] ] since the variance increases at each step by at most @xmath503 , we get that @xmath504 so , the maximum of the @xmath199-th component of the difference @xmath505 restricted to the times @xmath506 is thus stochastically bounded by the maximum of a standard brownian motion during the time interval @xmath507\right].\ ] ] the standard brownian motion starting at the origin . by the reflection principle for brownian motion , this distribution function can be bounded by the distribution function of the standard brownian motion at time @xmath508.\ ] ] we have just shown that conditional on @xmath327 and when @xmath393 holds , then @xmath509 where the last inequality was obtained by using the lemma [ boundstnormal ] below . we have just proven that @xmath510 we have that @xmath511 for all @xmath2 large enough . let @xmath512 be the event that : @xmath378}|c^l(ij+s)-c^l(ij)|\leq \log(n)\sqrt{j}\ ] ] so , we have @xmath513}b_t\geq \log(n)\sqrt{j}\ ] ] where @xmath514 designates a standard brownian motion starting at the origin in a one dimensional space . by the reflection principle , we get that the right side of [ fnc ] is equal to @xmath515 where @xmath516 designates a standard normal . using lemma [ boundstnormal ] , we find that the right side of [ groslolo ] can be bounded by @xmath517 so that @xmath518 now @xmath519 where in the last interesection above @xmath199 ranges over @xmath520 and @xmath79 ranges over @xmath521 . hence , we get @xmath522 where the dominant term is @xmath523 which beats any power of @xmath2 . note that @xmath524 is just a power of @xmath2 . we have that @xmath525 for all @xmath2 large enough . same as for @xmath404 . so we leave it to the reader . [ boundstnormal]let @xmath516 be a standard normal . let @xmath526 . then , we have @xmath527 in the pictures below we have simulated the optimal alignment with length @xmath2 running from @xmath70 to @xmath528 and each time we compute the estimated standard deviation ( of 100 simulations from a @xmath529 sequence ) . the gaps are allowed only in one sequence and the number of gaps is restricted to be either a fixed number , a number @xmath118 or a proportional number to @xmath2 . line and the solid black curve is @xmath48 . as it is clear from this picture , the standard deviation of the optimal alignment is suggested to be of size @xmath48 and not @xmath49 . ] for @xmath530 . as it is clear from this picture , the standard deviation of the optimal alignment is suggested to be of size @xmath48 and not @xmath49 . ] for @xmath530 . as it is clear from this picture , the standard deviation of the optimal alignment is suggested to be of size @xmath48 and not @xmath49 . ] we want to thank the reviewers of this paper for the comments which improved the paper . r. pemantle and y. peres . planar first - passage percolation times are not tight . in _ probability and phase transition ( cambridge , 1993 ) _ , volume 420 of _ nato adv . _ , pages 261264 . kluwer acad . publ . , dordrecht , 1994 . |
background : : mean - field methods based on an energy density functional ( edf ) are powerful tools used to describe many properties of nuclei in the entirety of the nuclear chart . the accuracy required on energies for nuclear physics and astrophysics applications is of the order of 500 kev and much effort is undertaken to build edfs that meet this requirement . purpose : : the mean - field calculations have to be accurate enough in order to preserve the accuracy of the edf . we study this numerical accuracy in detail for a specific numerical choice of representation for the mean - field equations that can accommodate any kind of symmetry breaking . method : : the method that we use is a particular implementation of 3-dimensional mesh calculations . its numerical accuracy is governed by three main factors : the size of the box in which the nucleus is confined , the way numerical derivatives are calculated and the distance between the points on the mesh . results : : we have examined the dependence of the results on these three factors for spherical doubly - magic nuclei , neutron - rich , the fission barrier of and isotopic chains around . conclusions : : mesh calculations offer the user extensive control over the numerical accuracy of the solution scheme . by making appropriate choices for the numerical scheme the achievable accuracy is well below the model uncertainties of mean - field methods . | 1509.00252 | the self - consistent mean - field approach , based on an energy density functional ( edf ) , is a tool of choice to study nuclei in any region of the nuclear chart @xcite . it allows to calculate the properties of the ground state but also of alternative configurations , like shape isomers , or to follow the behaviur of a nucleus along rotational bands or along fission paths . often , one is not directly interested in the total binding energy of a specific nucleus but by its evolution along a series of isotopes or isotones , which can signal structural changes for given neutron or proton numbers . motivated by the needs of the nuclear physics and astrophysics communities , large efforts are underway to push the predictive power of nuclear mass models well below the 500kev level . to reach this goal , the protocols used to adjust the edf s parameters have been revisited . in particular , methods are being developed @xcite to quantify the statistical uncertainty of these parameters . however , besides the errors on observables due to these uncertainties , there is also a numerical error due to the way the scmf equations are solved . one needs to verify that the numerics does not introduce errors that are larger than the maximum error tolerated for mass models . more importantly , these errors should not vary too rapidly from one nucleus to the other , to avoid a spurious behavior of mass differences . the numerical methods used to solve the mean - field equations can be classified according to the way the single - particle wave functions are represented : by coordinate space techniques or by a basis expansion . coordinate space techniques represent the single - particle wave functions in a discretized , finite volume . several discretization techniques exist , utilizing finite difference formulas @xcite , fourier transformations @xcite , b - splines @xcite , wavelets @xcite and the lagrange mesh method @xcite . the second family of numerical representations involves expanding the single - particle wave functions on some chosen ( finite ) set of basis states . usually these basis states are harmonic oscillator ( ho ) eigenstates , although the details often vary . while the origin of numerical errors is quite different for both families of representations , the type of edf does not seem to influence the accuracy of the methods much . the three main families ( relativistic edfs , zero - range skyrme edfs or finite - range edfs ) require similar numbers of basis states to achieve a similar precision ( see e.g. @xcite ) . in what follows , we will limit ourselves to the study of zero - range skyrme edfs . it is the aim of this paper to study the numerical accuracy of a specific implementation of coordinate space techniques : representation on a three - dimensional cartesian mesh of equidistant points . we will focus on two specific techniques : finite - difference ( fd ) formulas and the lagrange mesh ( lm ) method , which are the ones implemented in our codes . as far as we can infer from the tests published in the literature , the accuracy obtained with the other techniques mentioned above is similar to the one obtained within our lm scheme . most of the information relative to the tools that we have developed has been presented for the particular implementation made in the code ` ev8 ` @xcite . more involved implementations have also been used , which differ from ` ev8 ` only in that they impose less symmetries on the nucleus . the presence of these symmetries in general allows for the reduction of the dimension of the problem , e.g. in ` ev8 ` it allows for the reduction of the space by a factor of 1/8 . the article is organized as follows : first we define precisely the quantities that will be used to characterize the accuracy of a mean - field calculation . next , we review the basic ingredients needed to define wave functions on a cartesian mesh and to calculate derivatives and integrals in this representation . we then discuss the main sources of numerical errors : the size of the box in which the nucleus is confined and the step size of the mesh . we discuss the numerical accuracy that can be achieved by comparing energies and radii of doubly magic nuclei with those obtained with a spherical code . finally , we check the convergence of energies , radii and the multipole moments of deformed nuclei by comparing results obtained with decreasing mesh discretization lengths . a mean - field configuration is characterized by its energy , its rms radius and by multipole moments . in this section we define these quantities whose dependence on the mesh parameters will be studied . for a time - reversal invariant system as assumed here , the total energy is composed of the kinetic energy , the skyrme energy describing the strong interaction in the particle - hole channel , the pairing energy , the coulomb energy and a center - of - mass correction @xcite @xmath3 for the parameterizations used throughout this article , the skyrme edf takes the form of the sum of various bilinear combinations of the isoscalar ( @xmath4 ) and isovector ( @xmath5 ) local densities @xmath6 , kinetic densities @xmath7 and spin - current densities @xmath8 , @xmath9 , @xmath10 , @xmath11 , @xmath12 , @xmath13 \ , \rho_t^2 + c^{\rho^{\alpha}}_t \rho_0^\alpha \ , \rho_t^2 + c^\tau_t \ , \rho_t \ , \tau_t \nonumber \\ & & + c^{\delta \rho}_t \ , \rho_t \ , \delta \rho_t + c^{\nabla \cdot j}_t \ , \rho_t \nabla \cdot { { \bf j}}_t \nonumber \\ & & - c^{t}_t \sum_{\mu , \nu } j_{t , \mu \nu } \ , j_{t , \mu \nu } \big ) \ , , \end{aligned}\ ] ] with coupling constants as defined in ref . @xcite . the kinetic energy just depends on the kinetic density @xmath14 of protons and neutrons . while the skyrme and kinetic energies are local functionals of the densities , the direct coulomb energy is a nonlocal functional of the proton density @xmath15 @xmath16 compared to the other terms contributing to the total energy , the exact calculation of the coulomb exchange energy is orders of magnitude more costly as it is a functional of the complete nonlocal one - body density matrix . as a consequence , the local slater approximation that is of similar numerical cost as the skyrme energy is used instead @xmath17 the pairing energy contribution to the energy is : @xmath18 where the @xmath19 are anti - symmetrized matrix elements of the pairing interaction and the @xmath20 are cutoff factors , both of which are specified in appendix [ app : pairing ] . the expression for the c.m . correction , which is not relevant for our discussion , can be found in ref . @xcite . as in @xcite , the dimensionless multipole moments @xmath21 are related to the matrix elements of the multipole operators @xmath22 by @xmath23 where @xmath24 . when @xmath25 is omitted we imply it to be zero . another set of observables , related to the density profile of the nucleus , are the mean - square ( ms ) radii , root - mean - square ( rms ) radii and the isotopic shifts . the ms radius of the proton ( @xmath26 ) , neutron ( @xmath27 ) , and total density distribution is defined as @xmath28 \ , r^2 \ , .\end{aligned}\ ] ] the root - mean - square ( rms ) radii are then the square root of the corresponding mean - square radius . similarly , we will present results for the isotope shift of charge radii that are calculated as the difference between the proton ms radius of an isotope with @xmath29 neutrons and a reference isotope with @xmath30 neutrons @xmath31 without any corrections . assuming a 3-dimensional cartesian mesh , a function @xmath32 is represented by the tensor @xmath33 of its values at the collocation points @xmath34 @xmath35 a mesh can be defined in several ways , depending on the choice of the collocation points . for example , the origin of the coordinate system and the boundaries of the box can be included as collocation points or not . different choices can also be made for the boundary conditions at the edge of the box . to set up the self - consistent mean - field equations , one has to vary the edf with respect to @xmath33 . this requires to define prescriptions to calculate derivatives and integrals from the values of @xmath33 on the mesh . several choices for derivatives have been explored over the years . the most straightforward possibility to set - up a coordinate - space representation of the self - consistent mean - field equations is provided by the finite - difference method , a widely - used tool to solve partial differential equations @xcite . in such a scheme , the derivatives are calculated with @xmath36-point finite - difference formulas , and the integrals are obtained by summing up the integrand at the mesh points multiplied by a suitable volume element . there are three factors that determine the accuracy that can be achieved with the finite - difference method . one is the overall resolution scale provided by the mesh spacing ; decreasing the distance between mesh points improves the accuracy . second , the higher the order of the finite - difference formulas used for a given mesh spacing the better the accuracy . in both cases , however , better accuracy means also increase of the numerical cost . third , there are internal inconsistencies introduced by the method itself . for example , taking twice the numerical first - order derivatives of a given function is not equivalent to applying the numerical second - order derivatives . also , the numerical derivatives are not the inverse of the numerical integration . it is only for very small step sizes well below 0.1fm that these internal inconsistencies become irrelevant . while such small step sizes can be easily handled in spherical 1d codes @xcite , the required storage is prohibitive in axial 2d and cartesian 3d codes . in addition , such step sizes are much smaller than what can be expected to be the physically relevant resolution scale , see for example the arguments brought forward in ref . @xcite . several other schemes have been developed in the past with a better consistency between derivation and integration . for instance , derivatives have been defined through a fourier transformation to momentum space @xcite , which is equivalent to the assumption that the functions on the mesh can be developed into a set of plane waves . in this method , the derivatives are quasi - exact for a given resolution of the mesh , and first- and second - order derivatives internally consistent . similar ideas have been developed in quantum chemistry under the label of _ discrete variable representation _ ( dvr ) @xcite . a similar formalism that provides an internal consistent scheme for derivatives and integrals is the lagrange - mesh method that we will sketch in the following section . the idea underlying the lagrange - mesh method is that for each gauss quadrature one can construct a set of basis functions for which orthogonality and completeness relations are exactly fulfilled when evaluated with the given quadrature @xcite . this additional condition makes the lm method a special case of the slightly less rigorous concept of dvr @xcite . lagrange meshes have been constructed for a multitude of different geometries and used for a wide range of applications , see @xcite and references therein . we will use here the case of an equidistant 3d cartesian mesh . its three directions are separable in the formalism , such that the presentation of the principles of the method for one dimension is sufficient . the underlying basis of a one - dimensional cartesian equidistant lagrange mesh is constructed as the set of functions @xmath37 whose orthogonality relations are exact when evaluated with a simple @xmath38-point rectangular quadrature rule , sometimes called _ midpoint rule _ , @xcite @xmath39 where @xmath40 is the distance between the collocation points located at @xmath41 and where @xmath42 and @xmath43 are the boundaries of the numerical box @xmath44 = [ -n dx , + n dx]$ ] . a convenient representation of the @xmath38 basis functions @xmath37 are plane waves of the form @xmath45 where @xmath46 is the length of the numerical box and where @xmath47 , @xmath48 , , @xmath49 . the real part of the @xmath50 is symmetric , has nodes on the boundaries of the box and a maximum at the origin , whereas their imaginary part is skew - symmetric , consequently has a node at the origin , and maxima on the boundaries of the box . this also implies that @xmath51 . the @xmath52 form a complete set of functions to describe any function on the mesh points @xmath53 note that the box size @xmath54 is not a multiple of the wavelength of the basis functions . instead , twice the box size is an odd multiple of the wavelengths that take the values @xmath55 , @xmath56 , @xmath57 , @xmath58 . both the real and imaginary parts of all plane waves in eq . are non - zero at all mesh points . as recalled in ref . @xcite , in cartesian dvr and lm coordinate - space methods where the derivatives are defined through an expansion in plane waves , the analysis of a calculation s infrared and ultraviolet cutoffs introduced by the basis is straightforward . this has to be contrasted with the much more involved analyses required when working with an ho basis @xcite . it has also been argued in ref . @xcite that a dvr or lm representation of the nuclear many - body problem covers the relevant part of the phase space with a much smaller number of basis states than required by an ho basis . in practice , however , ho bases typically used for self - consistent mean - field calculations are much smaller than the typical number of mesh points used in the same kind of calculation . for a box with 20 points in every direction the number of linearly independent states is 64000 , to be compared with a harmonic oscillator expansion with 20 shells , which contains 14168 states . while the basis functions @xmath59 of eq . are useful to discuss the mathematical properties of the lm method , the actual coordinate representation then employs the set of @xmath38 lagrange interpolation functions @xmath60 obtained as @xcite @xmath61 } { \sin \big[\frac{\pi}{dx}\frac{x - r \ , dx}{2n}\big ] } \ , .\end{aligned}\ ] ] by construction , the lagrange interpolation functions have the property to be equal to one at the mesh point @xmath62 , and zero at all others , @xmath63@xcite . when developed into the lagrange functions , any function @xmath64 on the mesh @xmath65 is then simply represented by its values @xmath66 at the @xmath38 mesh points . the lagrange functions are smooth and infinitely derivable . they can be used to define matrices representing the first and second derivatives of functions discretized through . ] eq . @xmath67}{\sin^2\left[\pi ( i - j)/(2n)\right ] } } & \text{for $ i \neq j$ } , \\ { \displaystyle - \frac{\pi^2}{3dx^2 } \left(1-\frac{1}{(2n)^2}\right ) } & \text{for $ i = j$. } \end{array } \right.\end{aligned}\ ] ] the first derivative of any function @xmath64 on the mesh is obtained by multiplying the @xmath68 matrix @xmath69 by the vector @xmath70 @xmath71 and similar for the second derivatives . note that the derivative matrices have the property @xmath72 by construction @xcite , which is not the case for finite - difference formulas . as the derivatives of eqs . and correspond to full @xmath68 matrices , their application is more time consuming than finite - difference derivatives that correspond to a sparse band matrix . the full cartesian 3d representation of a function @xmath73 is then provided by @xmath74 where the number of discretization points does not have to be the same in each direction . in that case , the derivative matrices in eqs . and have to be set - up separately for each direction . as pointed out in refs . @xcite , a variational calculation using a dvr or lm derivatives delivers very precise values for the total energy in spite of the individual matrix elements being much less accurate . in what follows , we will illustrate that this property implies very accurate total energies while separate terms of the skyrme edf are less well represented . in addition , we will show that using a lm results in a variational calculation . unless explicitly stated , we have used the sly4 parameterization . to explore the dependence of the numerical accuracy on the edf , we have in addition tested a representative set of skyrme parameterizations , as listed in appendix [ app : forces ] . in the next two sections , we present calculations for doubly - magic spherical nuclei @xmath75ca , @xmath76sn and @xmath77pb , the neutron - rich nucleus @xmath0ne , cd , sn and te isotope chains and the fission path of @xmath1pu . it is worth noting that we only included pairing for the isotopic chains and for @xmath1pu , see appendix [ app : pairing ] for details . in all other cases , pairing has been neglected . in appendix [ app : constants ] we comment on the precise physical constants used during our calculations . the accuracy of a coordinate space calculation is limited by the size of the box , the discretization length @xmath40 and the way derivatives and integrals are calculated . in order to properly judge these effects we employ two ways of analyzing results . for spherical nuclei we can compare our 3d results with a one - dimensional spherical code that also represents the single - particle wave functions in coordinate space . because of spherical symmetry , we can use extremely fine discretizations and the results can thus be considered exact to very high precision . for this purpose we use ` lenteur ` @xcite as a reference . for deformed nuclei , we no longer have access to such a comparison . here we have to resort to looking at 3d results as a function of both box size and mesh spacing : we compare results in small boxes with large mesh spacing to results in very large boxes with very fine mesh spacing . obtained in calculations using different combinations of formulas for the derivatives ( see text ) . the differences are taken with respect to the total energy obtained with ` lenteur ` @xcite . ] lddddd energy ( mev ) & & & & & + kinetic + c.m . correction & 3908.548 & 3880.647 & 3872.035 & 3867.506 & 3866.190 + direct coulomb & 831.801 & 829.241 & 828.433 & 828.004 & 827.876 + coulomb exchange & -31.415 & -31.319 & -31.289 & -31.273 & -31.269 + @xmath78 & -22749.578 & -22510.048 & -22435.391 & -22395.941 & -22384.379 + @xmath79 & 1368.924 & 1343.506 & 1335.611 & 1331.436 & 1330.206 + @xmath80 & 14812.851 & 14631.848 & 14575.413 & 14545.584 & 14536.832 + @xmath81 & 323.771 & 318.147 & 316.431 & 315.539 & 315.287 + @xmath82 & -99.730 & -97.595 & -96.917 & -96.554 & -96.445 + total skyrme energy & -6343.762 & -6314.142 & -6304.854 & -6299.937 & -6298.501 + total energy & -1634.828 & -1635.574 & -1635.675 & -1635.700 & -1635.703 + lddddd energy ( mev ) & & & & & + kinetic + c.m . correction & 3866.323 & 3866.165 & 3866.182 & 3866.182 & 3866.190 + direct coulomb & 827.922 & 827.889 & 827.882 & 827.878 & 827.876 + coulomb exchange & -31.269 & -31.268 & -31.268 & -31.268 & -31.269 + @xmath78 & -22384.936 & -22384.188 & -22384.322 & -22384.320 & -22384.379 + @xmath79 & 1330.300 & 1330.193 & 1330.201 & 1330.200 & 1330.206 + @xmath80 & 14537.174 & 14536.691 & 14536.789 & 14536.787 & 14536.832 + @xmath81 & 315.238 & 315.275 & 315.284 & 315.284 & 315.287 + @xmath82 & -96.430 & -96.444 & -96.445 & -96.445 & -96.445 + total skyrme energy & -6298.657 & -6298.473 & -6298.493 & -6298.494 & -6298.501 + total energy & -1635.678 & -1635.687 & -1635.696 & -1635.700 & -1635.703 + the numerical cost of using lm derivatives is much higher than the fd alternative . to control the computational time , three options have been considered : they differ by the way derivatives are calculated during the mean - field iterations and after convergence . the first option ( fd+fd ) has been used in the first applications of the codes @xcite where derivatives were exclusively calculated by fd . the second one ( fd+lm ) is the most used one for more than 20 years : fd derivatives are used during the iterations but the energies are recalculated after convergence by lm formulas . finally , in the last option ( lm+lm ) , the lm formulas are used during the iterations and after convergence . in practice , we use a seven - point difference formula for the first order and a nine - point formula for the second order derivatives when employing fd formulas . it has been shown earlier in ref . @xcite that this provides an efficient compromise in terms of overall speed and precision . figure [ fig : newoldcomparison ] illustrates the accuracy on the total energy obtained using these three options . the lm + lm choice is by far the most accurate . as it can be seen in table [ tab : edfcompl ] , the result obtained with a mesh size of 1.0 fm differs by only 25 kev from the ` lenteur ` result . the fd+lm option is less accurate , but already sufficient for most applications , with an error of around 100 kev for @xmath83 fm . it is better by nearly an order of magnitude than the fd+fd choice . results presented in the following have been obtained with the fd+lm option , except otherwise stated . both the fd+lm and lm+lm calculations underestimate the binding energy , as it should be for a variational calculation . this is due to the fact that the single - particle wave functions are expanded on a complete and closed basis for given box size and mesh discretization length , see eq . . increasing the box size and/or decreasing the mesh discretization length enlarge the accessible subspace of the hilbert space @xcite and lead to a monotonous convergence of the energy . by contrast , such a basis can not be defined for the fd+fd option for which the calculation systematically overestimates the binding energy of @xmath77 . the same applies to mesh calculations with fourier derivatives , as can be deduced from the convergence analyses in refs . while for a given @xmath40 the overall accuracy of the binding energy found there is very similar to the one we find for lm+lm calculations , the energy does not converge monotonically when decreasing @xmath40 . while the use of lm derivatives after having used fd ones during the iterations ( fd+lm ) is sufficient to obtain an upper bound of the total energy since any wave function discretized on a mesh can be expanded on the lm basis , the errors on the various individual terms of the skyrme edf can be very large , as can be seen in table [ tab : edfcomp ] . while the total energy varies by slightly less than one mev when @xmath40 is decreased from @xmath84 fm to @xmath85 fm , the variation in the kinetic energy is of the order of 40 mev , counterbalanced by a similar change in the skyrme energy . the situation for the lm+lm scheme is shown in table [ tab : edfcompl ] . it indicates a similar effect , but on a much smaller scale : the total energy varies by 20 kev while the kinetic energy varies by roughly 150 kev . when performing symmetry restoration and configuration mixing by the gcm , a high level of accuracy is required to avoid buildup numerical noise while solving the hill - wheeler - griffin equation this calls for the use of lm derivatives in these calculations , as done since our first applications @xcite . the first requirement of a coordinate space calculation is that the box in which the nucleus is confined is large enough to avoid any spurious effect due to the truncation of the wave functions . the influence of the box size on the total energy for three spherical nuclei is represented in fig . [ fig : boxsizesphericals ] . the same mesh size @xmath86 fm is used in all calculations while the number of discretization points is varied , changing thus the volume of the box . the calculation in the largest box , using 23 points , is taken as a reference . the errors decrease quickly when the box size is enlarged . if one requires that the error is smaller than a kev , we see that taking boxes with half - sides of @xmath87 fm for @xmath75 , @xmath88 fm for @xmath76 and @xmath89 fm for @xmath77 is sufficient . since the numerical effort required for @xmath75ca is very low we opted to use a slightly larger half - side of @xmath90 fm in order to further increase our accuracy to about 0.1 kev . points for the three nuclei @xmath75 , @xmath76 and @xmath77 . in all cases , the step size is equal to 1.0 fm . ] cdddd nuclei & & & & + @xmath75 & 26 & 26 & 26 & 26 + @xmath76 & 30.8 & 30.8 & 46.8 & 46.8 + @xmath77 & 40 & 40 & 60 & 60 + @xmath91 & 40 & 40 & 60 & 60 + @xmath1 & 40 & 60 & 80 & 120 + similar analyses have been performed for all nuclei considered in this paper . since several nuclei in the isotopic chains around @xmath92 are deformed , we have performed all calculations with the same box size as @xmath77 . this choice allows us to calculate all isotopes with the same numerical conditions . the box dimensions are summarized in table [ tab : boxsize ] . the columns @xmath93 and @xmath94 indicate the size of the box in which the coulomb problem is solved . for every system , the box size was varied for fixed @xmath40 until the energy did not change by more than @xmath95 kev , with the exception of the @xmath1 for which this limit was of 1 kev . a non - ambiguous comparison between calculations performed with different mesh discretizations @xmath40 can only be achieved when the volume of the box is conserved . this is realized by determining the value of @xmath40 in such a way that the box has the same size for each number of mesh points . for different values of @xmath40 . calculations were initialized with nilsson - model single - particle wave functions . the box length is @xmath96 fm . ] calculated on a 3d mesh with different approximations in the calculation of the coulomb energy and that obtained with ` lenteur ` . the three lines correspond to using 3- , 5- and 7-point fd formulas for the calculation of the laplacian in the poisson equation . ] decreasing the mesh size improves the accuracy . however , this has a price in computing time . first , to keep the same box size requires to increase the number of discretization points . a second factor increasing the computing time is that the time step of the imaginary - time - step method @xcite implemented in the codes @xcite has to be decreased with decreasing mesh size . this considerably slows down the convergence . in fig . [ fig : convergence ] we show the evolution of the error on the total energy relative to ` lenteur ` during the iterations for the nucleus @xmath75 for different mesh discretizations @xmath40 . the most accurate result after 100 iterations is obtained with @xmath86 fm . gaining an order of magnitude of accuracy after convergence requires to carry out roughly 100 more iterations for the step sizes represented on the figure . the direct coulomb energy requires a special treatment because of its long range . one of the spatial integrations in eq . can be eliminated through the calculation of the coulomb potential of the protons , which satisfies the electrostatic poisson equation @xmath97 where @xmath98 is the square of the elementary charge . when solving this equation , boundary conditions need to be imposed at the edge of the box.these can be easily constructed when recalling that at large distances the potential is entirely determined by the multipoles of the nucleus charge distribution @xmath99 . expanding the coulomb potential on spherical harmonics and keeping terms up to , the coulomb potential outside the box is approximated by @xmath100 which provides the boundary condition for the numerical solution of eq . . the direct coulomb energy is then calculated as @xmath101 as for the nuclear part of the energy , the accuracy of the electrostatic potential , obtained by solving eq . , is limited by three factors : the size of the box , the mesh discretization length @xmath40 and the way derivatives are calculated . a suitable box size for the coulomb problem has to be larger than for the skyrme edf . this is a direct consequence of the long range of the coulomb force . to make negligible the contributions to the boundary conditions of terms higher than @xmath102 , see , one has to calculate the coulomb potential in a box larger than the one used for the nuclear part of the interaction . typical values are given in table [ tab : boxsize ] . for light nuclei such as @xmath75 , no extra points for coulomb need to be added , while the box has to be significantly enlarged for heavier systems in the @xmath76 and @xmath77 regions . for the calculation of the fission barrier of heavy nuclei such as @xmath1 up to very large deformations , the coulomb box size has to be two times larger than the one needed for the skyrme edf to obtain the same nuclear accuracy on all the energies . the laplacian in eq . has to be approximated on the mesh in such a way that the accuracy on the coulomb energy is similar to the one of the other terms in the edf . we show in fig . [ fig : coulcomparison ] the gain in accuracy on the total energy of @xmath77 obtained by going from a three - point to a seven - point fd formula for the laplacian . already a five - point formula brings the required accuracy and is used in all other calculations reported here . one can easily understand that a lower - order finite - difference formula than the one used to calculate the kinetic energy is sufficient for the laplacian in eq . : the typical length scale of the variation of the coulomb potential is much larger than the scale on which the wave functions vary . the final factor for the accuracy of the coulomb solution is the mesh discretization length @xmath40 . as the effect of the coulomb term is already incorporated in all of the applications , we will not discuss it separately . provided that the box size is large enough , the main factor determining the accuracy of our implementation of mesh calculations is the discretization length . in fig . [ fig : dxsphericals ] the energy difference with respect to ` lenteur ` results is plotted for three doubly - magic spherical nuclei , as a function of the mesh discretization @xmath40 for a representative set of skyrme parameterizations . it is remarkable that the interactions are grouped according to their effective mass ( see appendix [ app : forces ] for the actual values ) : interactions with larger effective mass @xmath103 give systematically more accurate results than interactions with smaller ones . this property is related to the term @xmath104 term of the skyrme edf in eq . that , in our experience , is the least well represented on a mesh . since the magnitude of this term increases when the effective mass decreases , the accuracy obtained for a given mesh size deteriorates for lower effective mass . one can see that the accuracy obtained with a mesh discretization as large as @xmath86 fm is lower than @xmath84 mev for @xmath77 . the energy difference decreases to a few hundred kev for @xmath83 fm and to a few kev for @xmath105 fm . note that a similar accuracy for @xmath105 fm was found for a 2-dimensional code based on splines @xcite . to obtain an agreement between the spherical code ` lenteur ` and our 3-dimensional codes below the @xmath106 kev level would require to increase the box size but also to make the codes more similar . for a nucleus with a binding energy larger than 1 gev , this implies a relative discrepancy better than @xmath107 and there are several sources of differences in the codes that can play a role , none of which is easy to control . let us now study the convergence properties of our numerical scheme for the fission path of @xmath1 . our motivation is twofold : @xmath1 is a frequent benchmark for models that describe fission @xcite but also for numerical algorithms @xcite . the energy curve of this nucleus presents two minima at prolate deformations , the ground state and a fission isomer . in fig . [ fig : pulandscape ] , we show the variation of the energy with deformation . calculated with @xmath105 fm . the regions where the deformation is axial , triaxial or axial and reflection asymmetric are indicated in blue , green and red respectively . ] with @xmath108 and @xmath109 fm and those corresponding to @xmath105 fm . ] the box used for these calculations has the same size for all discretizations , as indicated by table [ tab : boxsize ] . when the left - right symmetry is broken , the number of points along the @xmath12 direction is doubled . we have performed calculations with four different mesh discretization , @xmath86 , 0.82 , 0.69 and 0.60 fm and tested the convergence as a function of @xmath40 by taking the difference with respect to the results obtained with @xmath105 fm . for each value of @xmath40 , the energy at each deformation is the energy relative to the prolate ground state . the energy curve obtained with @xmath105 fm is shown in fig . [ fig : pulandscape ] . the topography obtained for other values of @xmath40 is the same . shapes are triaxial in the vicinity of the first barrier , whereas everywhere else they remain axial . at deformations smaller than the one of the fission isomer the configurations are reflection symmetric , whereas at larger deformations they are increasingly asymmetric . we will use this curve as a reference to determine the accuracy of the calculations carried out for other values of @xmath40 . for each @xmath40 , the ground state energy is taken as the zero of the energy . the results are shown fig . [ fig : puerror ] . the properties of the minimum are summarized in table [ tab : pustates ] . the error decreases roughly by an order of magnitude by going from @xmath86 to 0.82 fm and from @xmath110 to 0.69 fm . at @xmath86 fm the error is of the order of a few times 100 kev , with a rather large oscillation . for a mesh discretization of @xmath111 fm , the error becomes lower than 100 kev ( except in the vicinity of the spherical configuration where it reaches 150 kev , but this configuration is very excited ) and is quite acceptable for the calculation of energy curves . decreasing still the discretization to 0.69 fm reduces the error to values around a few times 10 kev at most . some published results allow for a comparison between the accuracies of mesh calculations and of calculations using an expansion on an ho basis . et al . _ @xcite have performed calculations on an axial mesh using b - splines and on ho bases either spherical or deformed , with 20 oscillator shells in both cases . the accuracy obtained in @xcite on a mesh of @xmath112 fm seems very similar to the one we obtain . the use of a spherical ho basis is rather unreliable , with an error larger than 1 mev already for the excitation energy of the fission isomer and that quickly increases to several mev at larger deformations . for an axial oscillator basis , the results are similar to those that we obtain with a mesh size of 0.82 fm up to the first barrier but the accuracy deteriorates rapidly for larger deformations , being of several hundreds of kev at the deformation corresponding to the fission isomer . similar results can be found in @xcite for @xmath113 and in @xcite for @xmath114 . as a number of shells significantly larger than 20 is numerically prohibitive , one either has to resort to a two - center oscillator basis or one has to construct a suitable subspace within a much larger one - center ho basis by carefully selecting the low - lying single - particle states . the former option is developed in ref . @xcite whereas the latter has been used during the construction of the unedf1 parametrization @xcite , where the lowest 1771 basis states out of a basis of 50 ho shells has been kept . the accuracy obtained in this way for the excitation energy of the fission isomer is of the order of 100 kev . as a comparison , the experimental excitation energy of the fission isomer that can be found in the literature is 2.4 @xmath115 0.3 mev@xcite , approximately 2.8 mev @xcite and 2.25 @xmath115 0.20 mev@xcite . in the light of these error bars , a numerical accuracy of 100 kev is sufficient for the adjustment of an edf . however , from the published results by pei _ et al . _ @xcite , it can be estimated that the numerical error on the fission barrier height is a few times these @xmath116 kev . similar results have been obtained in the case of the rmf method @xcite . for @xmath117 fm . the reference calculation was performed in a box with @xmath118 . ] in different box sizes with @xmath83 fm . ] with respect to @xmath76 for different sn isotopes and different mesh sizes . ] the rms radius is intimately linked to the radial density distribution of a nucleus . one can expect that it is particularly sensitive to the box size for nuclei with a large excess of neutrons . tests have been performed for the very neutron - rich nucleus @xmath0 by varying the box size for a fixed mesh discretization @xmath83 fm . to avoid any ambiguity in the calculation , pairing has been omitted . the results are presented in fig . [ fig : ne34 ] , where we show the difference in total rms radius as a function of the box size for a representative set of edf parameterizations . for the size of the box recommended for @xmath75 in table [ tab : boxsize ] , the number of points is 16 for a mesh size of 0.8 fm . it leads to an error of the order of @xmath119 fm for most interactions , the results being slightly less accurate for sv - min . for smaller boxes , the accuracy of radii is lower and depends on the interaction . in fig . [ fig : ne34densityprofile ] the radial profile of the total density of @xmath0 is plotted as a function of the box size . the distortion of the density in the smallest box is large and demonstrates that half the box size has to be larger than 8.0 fm . in all other boxes , the exponential tail of the density distribution is well described , up to the point before the last one . for a box size around 12 fm , the density is well described up to a decrease of the central density by six order of magnitudes . the confinement in a volume is less evident in an expansion on a basis than in a mesh calculation , but it is also present . while oscillator basis functions extend to infinity , they are in practice strongly localized by their gaussian form factor . if one takes its classical turning point as a measure for the extension of a ho state , one obtains , for @xmath77 and 20 oscillator shells , a value for the turning point that varies from 14 fm for @xmath120 to 16 fm for @xmath121 . to increase the value of this turning point to 20 fm would require to use 28 oscillator shells for @xmath120 . this effect of confinement by an oscillator basis has been put in evidence in ref . @xcite for the case for @xmath122 . for comparison , the experimental uncertainty on rms charge radii for the ne isotopes ( up to @xmath123 ) varies from @xmath124 close to stability to @xmath125 for exotic isotopes @xcite . it is interesting to note that the numerical accuracy of a mesh mean - field calculation has a similar level ( provided the box is large enough ) , but that the model already introduces uncertainties on the rms radii that are at least one order of magnitude larger @xcite . in fig . [ fig : radii ] , we compare the total rms radii calculated with decreasing mesh sizes to those obtained with ` lenteur ` for three spherical nuclei @xmath75 , @xmath76 and @xmath77 . the agreement is already very satisfying for a large mesh size of 1.0 fm , with one order of magnitude gained in accuracy when decreasing the mesh size to 0.8 fm , which is the usual value of production calculations . an interesting feature that can not be deduced from fig . [ fig : radii ] is that all of the parameterizations , with the exception of unedf0 , always produce an rms radius that is smaller than the ` lenteur ` result . in fig . [ fig : snshift ] , we present the isotopic shifts @xmath126 for a range of even - even sn nuclei , the reference being @xmath76 . all curves almost exactly coincide . this demonstrates that the isotopic shifts are quite reliable even with coarse meshes . similar results are obtained for cd , xe and te isotopes . fm for cd ( a ) , sn ( b ) , and te ( c ) isotopes . in ( d ) we plot the two - neutron separation energies of all of these isotope chains as a reference.,title="fig : " ] fm for cd ( a ) , sn ( b ) , and te ( c ) isotopes . in ( d ) we plot the two - neutron separation energies of all of these isotope chains as a reference.,title="fig : " ] fm for cd ( a ) , sn ( b ) , and te ( c ) isotopes . in ( d ) we plot the two - neutron separation energies of all of these isotope chains as a reference.,title="fig : " ] fm for cd ( a ) , sn ( b ) , and te ( c ) isotopes . in ( d ) we plot the two - neutron separation energies of all of these isotope chains as a reference.,title="fig : " ] to put into evidence changes of nuclear structure with nucleon number , one often uses mass filters that are computed by taking specific differences between the binding energies of neighboring nuclei . the simplest filter is the two - nucleon separation energy that is defined as the energy difference between two isotopes ( or isotones ) whose nucleon number differs by two . in fig . [ fig : s2neutron ] , we show the evolution of the two - neutron separation energies , @xmath127 , of even - even nuclei for three neighboring isotopic chains when the mesh size @xmath40 is decreased . for each discretization @xmath40 we have plotted the difference of the @xmath127 values to the one obtained at @xmath128 fm . even with a mesh size as large as @xmath86 fm , the accuracy on the @xmath127 is already better than 100 kev , which is small enough for most applications . the mesh size used in most of our published applications , @xmath83 fm leads to an accuracy better than 10 kev . in the bottom panel , the two - neutron separation energies of the three isotope chains are plotted for four values of @xmath40 . the curves can not be distinguished using a scale adapted to the variation of @xmath127 as a function of the neutron number . this result is in strong contrast with respect to some published calculations using an expansion on an oscillator basis @xcite , where special algorithms have to be devised to smooth numerical irregularities that can be of the order of few hundred kev . quadrupole moment as a function of neutron number for a series of te isotopes . ] the dimensionless ground - state quadrupole moments @xmath129 of even - even te isotopes are shown in fig . [ fig : tebeta ] . differences between the curves corresponding to different values of @xmath40 are tiny and not significant . similar results were obtained for the cd and sn isotopes . we now examine how the multipole moments of @xmath1 along the fission path are affected by the mesh size . in figs . [ fig : pu240octu ] and [ fig : pu240hexa ] we show the octupole and hexadecapole moments , respectively , in the region of the fission path where parity is broken . similar results obtained for the axial and triaxial cases are not shown . in tables [ tab : pustates ] and [ tab : puiso ] we show the multipole moments of the ground state and fission isomer of @xmath1 for the different mesh discretizations as obtained by unconstrained calculations . along the fission path for @xmath1 . ] for the parity breaking configurations along part of the fission path for @xmath1 . ] from figs . [ fig : pu240octu ] and [ fig : pu240hexa ] we see that the overall sequence of shapes along the fission path is robust with respect to the mesh spacing . the fission path is already precisely defined at the coarsest mesh ( @xmath86 fm ) we used . a single exception can be seen at the onset of octupole deformation ; in the vicinity of this point , however , the energy surface is very flat in @xmath130 direction . on a smaller scale , the multipole moments do vary as a function of the mesh discretization . this is best visible in tables [ tab : pustates ] and [ tab : puiso ] . since our method hinges on the variation of the total energy in eq . , there is no guarantee that the values of the multipole moments converge in a predictable way . it is , however , reassuring to see that the typical variation of these moments is of the order of a few percent to at most about 10 percent . the larger variations present themselves in the higher - order @xmath131 and @xmath132 moments . these are more difficult to resolve on coarse meshes because of the high number of nodes their associated legendre polynomials have . ddcccdd & & & & & & + 1.0 & -1801.909 & 0.288 & 0.160 & 0.043 & -0.002 & -0.003 + 0.849 & -1802.770 & 0.289 & 0.163 & 0.045 & 0.001 & 0.001 + 0.739 & -1802.929 & 0.292 & 0.164 & 0.046 & 0.002 & 0.002 + 0.653 & -1802.969 & 0.290 & 0.165 & 0.046 & 0.002 & 0.001 + ddccccc & & & & & & + 1.0 & -1796.929 & 0.832 & 0.494 & 0.344 & 0.279 & 0.255 + 0.849 & -1797.950 & 0.840 & 0.510 & 0.367 & 0.303 & 0.278 + 0.739 & -1798.099 & 0.847 & 0.528 & 0.388 & 0.319 & 0.268 + 0.653 & -1798.123 & 0.841 & 0.516 & 0.375 & 0.312 & 0.259 + as a function of mesh discretization @xmath40 . only neutron single - particle levels within 1.5 mev of the fermi energy are shown . ] in fig . [ fig : pulevels ] we show the evolution of the neutron single - particle levels within 1.5 mev of the fermi energy in the ground state of @xmath1 as a function of the mesh spacing @xmath40 . while slight shifts of the position of the levels are observed as a function of the mesh size , the largest error at @xmath86 fm is of the order of 100 kev . one can also note that the level ordering within the parity subspaces is the same for all values of @xmath40 . a similar dependence on box parameters is found for the proton states and the lighter nuclei studied here . the aim of this paper was to study the numerical accuracy of the solution of the self - consistent mean - field equations using a discretization on a 3-dimensional cartesian coordinate - space mesh . three elements permit to control its numerical accuracy . the first one is the method used to calculate derivatives . using lagrange - mesh derivatives leads to much more accurate results than finite - difference formulas for derivatives . in addition , a cartesian lagrange mesh corresponds to a representation in a closed subspace of the hilbert space , such that it always provides an upper bound to the binding energy that becomes tighter when adding points outside a given box or when decreasing the distance of mesh points in a given box . neither is the case for finite - difference derivatives . however , we have shown quantitatively that the accuracy of a calculation that uses finite - difference formulas during the iterations can be significantly improved upon by recalculating the edf at convergence with lagrange - mesh derivatives . again , this procedure provides us with an upper bound of the energy , thus restoring the variational character of the calculation . using lagrange derivatives during the iterations allows to still improve the accuracy on energies but at the cost of at least doubling the computing time . the second element on which mesh calculations depend is the size of the box in which the nucleus is confined . the examples of doubly - magic nuclei and neutron - rich @xmath0 illustrate that results for energies and densities are already stable at small box sizes . thirdly , the quality of the results depends on the mesh size with errors on energies that are almost independent on the number of neutrons and protons and on the shape of the nucleus . a mesh size @xmath83 fm guarantees an accuracy that is in general better than 100 kev , which corresponds to a relative accuracy of less than a tenth of percent , even for lighter nuclei . decreasing the mesh size to 0.7 fm permits to gain nearly an order of magnitude and to reach an accuracy that is well below all the uncertainties of the mean - field model . one can summarize these results by concluding that a mesh technique as implemented in our codes is flexible ( it can accommodate any kind of symmetry breaking ) , robust ( the accuracy can be controlled by an adequate choice of the three elements mentioned above ) and that it can be very accurate if needed . the positive aspect of our numerical scheme is that using a mesh size of 0.8 fm , as used in most of our past applications ensures an accuracy better than 100 kev on energies and reliable shape properties for nuclei of any mass . our study has been focussed on the solution of the mean - field equations and we have not touched the description of pairing correlations . there has already been a study of this problem by terasaki _ et al . _ it should be revisited today to take into account new developments . however , the problem is not exclusively a problem related to the way the mean - field equations are solved . the description of single - particle states well above the fermi energy is probably very different when using a discretization on a mesh or an expansion on an oscillator basis . the authors are grateful to many collaborators that have contributed to the codes used in this work , both with constructive criticism and sometimes their involvement with coding , in particular , p. bonche , h. flocard , j. meyer , j. dobaczewski , b. gall , n. tajima , j. terasaki , b. avez , b. bally and v. hellemans . special thanks to k. bennaceur for providing us with ` lenteur ` , b. bally for critical reading of the manuscript and g. scamps for pointing out a typographical error in the corrigendum to our ref . @xcite . this work has been supported in part by the european union s seventh framework programme ensar under grant agreement 262010 and by the belgian office for scientific policy under grant no . pai - p7 - 12 . we here give some remarks on the interactions used throughout the text . * sly4 and sly5 @xcite : these parameterizations were used as intended . * t22 and t65 @xcite : these parameterizations were used as intended . * unedf0 @xcite : while this parametrization was adjusted with a non - zero pairing interaction , we used it without any pairing . * sv - min @xcite : this parametrization was adjusted with non - equal nucleon masses @xmath133 . ` lenteur ` does not handle this option , so we used instead the average value @xmath134 as nucleon mass . * ski3 @xcite : this parameteriation was adjusted with the inclusion of a perturbative two - body c.m . correction , an option not included in ` lenteur ` . the values of their isoscalar effective mass are listed in table [ tab : effmass ] . the density - dependent pairing interaction we used for the isotope chains and the fission path of @xmath1pu is defined by @xcite : @xmath135 \ , \delta ( { { \bf r}}-{{\bf r } } ' ) \ , , \ ] ] where @xmath136 is the isoscalar density at @xmath137 . the parameters take the values @xmath138 , @xmath139 and @xmath140 . in addition , this interaction was supplemented by two cutoffs , one above and the other below the fermi energy , in order to eliminate the basis - size dependence of the total energy . they are defined by two fermi functions @xmath141^{-1/2 } \big [ 1 + \text{e}^{(\epsilon_k - \lambda_q + \delta \epsilon_q)/\mu_q } \big]^{-1/2 } \ , , \ ] ] where @xmath142 is the fermi energy , @xmath143 the single - particle energy of the single - particle state @xmath144 and we chose @xmath145 , and @xmath146 for protons and neutrons . by default , the physical constants used in our calculations are the following @xcite : @xmath147 whenever possible , we used the value of @xmath148 that was used during the adjustment of the parametrization . it might seem superfluous to completely specify the physical constant used , but the results of our calculations depend on the precise values of these constants . in particular , the level of agreement between ` ev8 ` and ` lenteur ` described in sect . [ sec : totale ] is only attainable when these codes use exactly the same numerical values for the physical constants . in fact , significant errors can be introduced when the values of the physical constants are slightly changed . the seemingly innocuous value of @xmath148 plays in fact a very important role . figure [ fig : hbm ] shows ` lenteur ` calculations for the spherical nuclei @xmath75 , @xmath76 and @xmath77 with sly4 . every point was calculated by slightly changing the value of @xmath148 from @xmath149 , the sly4 value . we see that using a value for @xmath148 that is not consistent with the value used during the fit of the edf can lead to an error of several mev on the total energy . @xmath148 is after all the proportionality constant of the kinetic energy in eq . . typical values that have been used over the years vary at least between 20.73 and @xmath150 . if the exact values of the physical constants used during the adjustment of a given parametrization are not available , then one can not reliably compare the results with experimental data . in this case , one can not judge the predictive power of this parametrization . similar concerns arise for the parameters of the skyrme interactions . the energy obtained in our calculations is more sensitive to some skyrme parameters than to others , but the close agreement observed in sect . [ sec : totale ] is not obtainable without carefully checking that the skyrme parameters are completely consistent across codes . that this is not trivial can be concluded from fig . [ fig : sigmadependence ] . there we plotted the relative difference in energy found by ` lenteur ` between modified versions of the sly4 functional and the correct sly4 . the interaction parameters are the same for every point , save for the density dependence parameter @xmath151 in eq . . there are only very few parameterizations for which the value of @xmath151 corresponds to a terminating decimal , for example sv - min for which @xmath153 . for the large majority of parameterizations the value of @xmath151 is either @xmath154 or , as in the case of sly4 , @xmath155 . both of these correspond to a repeating decimal number , whose numerical representation might differ from code to code . using @xmath156 in a calculation with sly4 corresponds to a rounding error of @xmath157 , which introduces an error of the total binding energy of @xmath75 of a few tens of kev . it can clearly be seen that a limited representation of @xmath151 implies a roundoff error that has a visible effect on the energy . this kind of error shows up when comparing ` lenteur ` and ` ev8 ` results and for this reason we conclude that relative errors smaller than @xmath158 become meaningless . similar analyses can be made for the other interaction parameters , including the values of physical constants used to fit the interaction . j. c. pei , m. v. stoitsov , g. i. fann , w. nazarewicz , n. schunck , and f. r. xu , deformed coordinate - space hartree - fock - bogoliubov approach to weakly bound nuclei and large deformations , phys . c * 78 * , 064306 ( 2008 ) . m. v. stoitsov , j. dobaczewski , w. nazarewicz , and p. ring , axially deformed solution of the skyrme - hartree - fock - bogoliubov equations using the transformed harmonic oscillator basis ( ii ) ` hfbtho ` v2.00d : a new version of the program , comp . * 167 * , 43 ( 2005 ) . w. ryssens , v. hellemans , m. bender , and p .- h . heenen , solution of the skyrme - hf+bcs equation on a 3d mesh , ii : a new version of the ` ev8 ` code , comp . . comm . * 187 * , 175 ( 2015 ) ; comp . . comm . * 190 * , 231(e ) ( 2015 ) . s. a. coon , m. i. avetian , m. k. g. kruse , u. van kolck , p. maris , and j. p. vary , convergence properties of ab initio calculations of light nuclei in a harmonic oscillator basis , phys . c * 86 * , 054002 ( 2012 ) . heenen , p. bonche , j. dobaczewski , h. flocard , s. j. krieger , j. meyer , j. skalski , n. tajima , and m. s. weiss , proc . of the international workshop on nuclear structure models , oak ridge ( u.s.a . ) , march 1992 , world scientific , singapore , p. 3 ( 1993 ) . heenen , p. bonche , j. dobaczewski , and h. flocard , generator - coordinate method for triaxial quadrupole dynamics in sr isotopes . ( ii ) results for particle - number projected states , nucl . * a561 * , 367 ( 1993 ) . n. schunck , d. duke , h. carr and , a. knoll , description of induced nuclear fission with skyrme energy functionals : static potential energy surfaces and fission fragment properties , phys . c * 90 * , 054305 ( 2014 ) . j. libert , m. girod , and j .- delaroche , microscopic descriptions of superdeformed bands with the gogny force : configuration mixing calculations in the @xmath160 mass region , phys . c * 60 * , 054301 ( 1999 ) . m. kortelainen , j. mcdonnell , w. nazarewicz , p .- reinhard , j. sarich , n. schunck , m. v. stoitsov , and s. m. wild , nuclear energy density optimization : large deformations , phys . c * 85 * , 024304 ( 2012 ) . m. hunyadi , d. gassmann , a. krasznahorkay , d. habs , p. g. thirolf , m. csatls , y. eisermann , t. faestermann , g. graw , j. gulys , r. hertenberger , h. j. maier , z. mt , a. metz and , m. j. chromik , excited superdeformed @xmath161 rotational bands in @xmath162-vibrational fission resonances of @xmath1 , phys . lett . * b505 * , 27 ( 2001 ) . a. blazkiewicz , v. e. oberacker , a. s. umar , and m. stoitsov , coordinate space hartree - fock - bogoliubov calculations for the zirconium isotope chain up to the two - neutron drip line , phys . rev c * 71 * , 054321 ( 2005 ) . e. chabanat , p. bonche , p. haensel , j. meyer , and r. schaeffer , a skyrme parametrization from subnuclear to neutron star densities . nuclei far from stabilities nucl . a635 * , 231 ( 1998 ) ; nucl . phys . * a643 * , 441(e ) ( 1998 ) |
we study the central hyperplane arrangement whose hyperplanes are the vanishing loci of the weights of the first and the second fundamental representations of restricted to the dual fundamental weyl chamber . we obtain generating functions that count flats and faces of a given dimension . this counting is interpreted in physics as the enumeration of the phases of the coulomb and mixed coulomb - higgs branches of a five dimensional gauge theory with 8 supercharges in presence of hypermultiplets transforming in the fundamental and antisymmetric representation of a gauge group as described by the intriligator - morrison - seiberg superpotential . | 1508.03038 | the theory of hyperplane arrangements has deep connections with many areas of mathematics , such as combinatorics , representation theory , algebraic geometry , algebraic topology , singularity theory , and the theory of hypergeometric functions . one beautiful application of the use of hyperplane arrangements in algebraic geometry is the proof of brieskorn s conjecture by deligne @xcite in 1972 . real hyperplane arrangements serve as the motivation for the axioms that define oriented matroids . see @xcite for detailed reviews of the theory of hyperplane arrangements . real hyperplane arrangements have been used recently in the physics literature to analyze the fiber structure and the network of flop transitions between different resolutions of elliptically fibered calabi - yau threefolds @xcite . compactifications of m - theory on calabi - yau threefolds form an elegant bridge between the study of supersymmetric gauge theories in five dimensional spacetime and certain types of highly structured hyperplane arrangements defined by a lie algebra @xmath2 and a representation @xmath3 of @xmath2 . these hyperplane arrangements have the peculiarity of being defined not in the full affine space as they are restricted to the dual fundamental weyl chamber of the lie algebra @xmath2 . [ def : incgeogeneral ] let @xmath2 be a reductive lie algebra over @xmath4 , let @xmath5 be a split , real form of a cartan subalgebra of @xmath6 and let @xmath7 be a representation of @xmath2 . we denote by @xmath8 the real hyperplane arrangement consisting of the kernels of the weights of @xmath7 restricted to a dual fundamental weyl chamber in @xmath5 . because all cartan subalgebras are conjugate and all fundamental weyl chambers canonically related by the weyl group action , the incidence geometry @xmath8 is independent of the choice of @xmath5 and of a dual fundamental weyl chamber in @xmath5 . in the rest of the paper , we call the dual fundamental weyl chamber " the _ weyl chamber_. an arrangement of hyperplanes stratifies the ambient space into open polyhedral cones called _ faces_. a non - empty intersection of a finite number of hyperplanes of the arrangement is called a _ flat_. a @xmath9-face ( resp . @xmath9-flat ) is a face ( resp . a flat ) that generates a @xmath9-dimensional linear space . faces of maximum dimension are called _ a classical combinatorial problem asks how many @xmath9-flats and @xmath9-faces a given hyperplane arrangement contains . zaslasky @xcite enumerated chambers and bounded chambers of an affine hyperplane arrangement using the mbius function of its semi - lattice of flats . the purpose of this paper is to count the numbers of @xmath9-faces and @xmath9-flats of hyperplane arrangements @xmath10 where @xmath11 is the vector representation of @xmath0 ( the first fundamental representation ) and @xmath12 is its second exterior power ( the second fundamental representation ) . since our arrangements are restricted to the weyl chamber , we can not apply zaslasky s result directly . however , the highly symmetric nature of the hyperplane arrangements under consideration allows us to study them via their extreme rays . we also analyze the geometries @xmath13 when @xmath2 is not simply laced . the case of @xmath14 is discussed in @xcite . the counting problem solved in this paper is formulated purely in terms of representation theory . however , its motivation comes from the geometry of string theory . in this subsection , we review the motivation . the reader uninterested in the interface of string theory and algebraic geometry can safely ignore this subsection . hyperplane arrangements of the form @xmath15 arise naturally in the study _ coulomb branch _ of a gauge theory in five - dimensional supersymmetric theory with eight supercharges @xcite . such theories contain different types of particles organized into representations of the supersymmetric algebra called _ supermultiplets_. we will consider only vector multiplets and hypermultiplets . when a gauge theory is defined in a five dimensional supersymmetric theory , the vector multiplets transform in the adjoint representation while the hypermultiplets transform in some representation @xmath3 of @xmath2 . the hypermultiplets can become massive through the higgs mechanism along hyperplanes passing through the origin of the weyl chamber of @xmath2 . these hyperplanes partition the weyl chamber into different connected regions called _ coulomb phases _ of the gauge theory , which are characterized by a cubic prepotential depending on real coordinates @xmath16 of the weight space . intriligator - morrision - seiberg computed the prepotential of a five dimensional gauge theory for a theory in which hypermultiplets transform in the representation @xmath17 of a lie algebra @xmath2 . the result is the real function @xcite : @xmath18 depending on the real vector @xmath19 in the weight space . we only wrote the terms of @xmath20 that are relevant for our purposes . the variable @xmath21 runs through the simple roots of @xmath2 , and @xmath22 through the weights of the representation @xmath7 under which the hypermultiplets transform . the cubic absolute values generate singularities , which correspond to hyperplanes along which new massless particles are generated . a well - defined phase of the coulomb branch of the theory is a connected region in which the quantities @xmath23 and @xmath24 each take fixed signs . we fix the sign of @xmath23 by requiring @xmath19 to be in the fundamental weyl chamber . the condition that @xmath25 for all @xmath26 restricts us inside a specific chamber of the arrangement @xmath27 . the counting of flats is connected to the enumeration of mixed _ coulomb - higgs branches _ of the gauge theory . a five dimensional gauge theory with 8 supercharges can be geometrically engineered by a compactification of m - theory on a calabi - yau threefold @xcite . when the calabi - yau threefold is elliptically fibered , its singular fibers determine the lie algebra @xmath2 and the matter representation @xmath3 of the gauge theory . a smooth elliptic fibration with a rational section is a resolution of a singular weierstrass model , and flop transitions join pairs of crepant resolutions of the weierstrass model . it is conjectured that the network of ( partial ) crepant resolutions of the weierstrass model is isomorphic to the adjacency graph of the faces of the hyperplane arrangement @xmath15 . more precisely , it is conjectured that each crepant resolution corresponds to a unique chamber of @xmath15 @xcite , and that two crepant resolutions are connected by a flop if and only if the intersection of the corresponding chambers is a face of codimension one . their common face is then conjectured to correspond to a unique partial resolution of the weierstrass model . see @xcite for examples of explicit matchings between ( partial ) resolutions and faces of a hyperplane arrangements @xmath15 . there are subtleties in presence of non - trivial @xmath28-factorial terminal singularities , see for example @xcite . a proof of this conjecture will likely rely heavily on ideas from mori s program along the lines of @xcite . the intriligator - morrision - seiberg prepotential @xmath29 can also be obtained geometrically @xcite . in the case of @xmath30 , the gauge group is @xmath1 , and therefore the relevant elliptic fibrations each have a mordell - well group of rank one and a singular fiber of type i@xmath31 over a divisor of the discriminant locus . models for elliptic fibrations with a mordell - weil group of rank one are discussed in @xcite and @xcite . elliptic fibrations for geometric engineering of @xmath1 gauge theories have been extensively studied recently in the physics literature ( see for example @xcite and references therein ) . we are now ready to discuss the enumerations performed in this paper . we denote the number of @xmath9-faces of @xmath32 by @xmath33 and collect these numbers in the following formal power series ( generating function ) : @xmath34 the following theorem enumerates faces of @xmath30 . it is proved in sec : genfuncfaces . [ thm : facecount ] the generating function counting faces of @xmath35 by dimension is given by the following rational function : @xmath36 thm : facecount implies a simple recurrence relation for @xmath33 . [ cor : facerecurse ] the sequence @xmath33 obeys the recurrence relation : @xmath37 for @xmath38 . we can also obtain formulas for the generating polynomials @xmath39 g(s , t ) = \sum_{k=0}^n g(n , k)t^k.\ ] ] [ cor : facegenpoly ] the generating polynomials @xmath40 are given by @xmath41 cor : facegenpoly yields the following values of @xmath40 : .3 cm from the above formulae , we can read off the value of @xmath33 for @xmath42 as the coefficient of @xmath43 in @xmath40 : @xmath44 we can also extract formulae for @xmath33 for @xmath9 small and @xmath45 small from thm : facecount . [ cor : facesmalllarge ] we have @xmath46 and @xmath47 cor : facesmalllarge yields the following formulae for @xmath33 and @xmath48 for @xmath49 : @xmath50 thm : facecount and corollaries [ cor : facerecurse ] , [ cor : facegenpoly ] , and [ cor : facesmalllarge ] are proved in sec : genfuncfaces . we denote the number of @xmath9-flats of @xmath32 by @xmath51 and collect these numbers in the following formal power series ( generating function ) : @xmath52 the following theorem enumerates flats of @xmath30 . [ thm : flatcount ] the generating function counting flats of @xmath35 by dimension is given by the following rational function : @xmath53 thm : flatcount is proven sec : genfuncflats . it implies a simple recurrence relation for @xmath51 , the analogue of cor : facerecurse for flats . [ cor : flatrecurse ] the sequence @xmath51 obeys the recurrence relation @xmath54 for @xmath55 . the taylor expansion of thm : flatcount to order 10 in @xmath56 is : from the above formula , we can read off the value of @xmath51 for @xmath42 as the coefficient of @xmath57 in @xmath58 : @xmath59 the form of the series @xmath58 does not allow simple explicit formulae for @xmath60 or for @xmath51 as in corollaries [ cor : facegenpoly ] and [ cor : facesmalllarge ] , respectively . thm : flatcount is proved in sec : genfuncflats . ( 1,1 ) & ( 1,2 ) & ( 1,3 ) & & ( 1,n ) + & ( 2,2 ) & ( 2,3 ) & & ( 2,n ) + & & ( 3,3 ) & & ( 3,n ) + & & & & + & & & & ( n , n ) a face is determined by assigning signs to @xmath61 for all values of @xmath62 such that @xmath63 . these weights can be organized in a right - justified young tableau as in figure [ fig : signtableau ] . it is important to realize that not all sign patterns are allowed . since we are in the weyl chamber , we have @xmath64 using the identities @xmath65 one can prove the following simple sign rules called _ sign flows _ in the physics literature @xcite : the signs in a tableau corresponding to a chamber satisfy the following conditions . * all the boxes above or on the left of a box with positive entries are also positive . * all the boxes below or on the right of a box with negative entries are also negative . a tableau satisfying these two rules corresponds to a unique chamber of @xmath66 . however , a different notation for chambers will be more convenient for dealing with extreme rays . a different notation for chambers will be more convenient for dealing with extreme rays . the sign rules implies that the positive entries of each row are next to each other and start on the left border of the table . it is therefore efficient to denote a given chamber by the numbers of positive entries on each row . the sign rules is then automatically satisfied if these numbers form a decreasing sequence @xmath67 , where we denote by @xmath68 the number of positive entries on the @xmath69th row of the tableau . we do nt count the rows that do not have any positive entries . an entry of the tableau located on the @xmath69th row and the @xmath70th column ( with @xmath71 ) is positive if and only if @xmath72 and @xmath73 . it is negative otherwise . this justifies the following definition . [ def : subsettochamber ] for @xmath74 $ ] , define a face @xmath75 of @xmath35 as the subset of @xmath5 on which @xmath76 we can equivalently write a subset @xmath77 $ ] as a _ characteristic vector _ @xmath78 , where @xmath79 and @xmath80 . a characteristic vector @xmath79 defines a subset @xmath81 $ ] such that @xmath82 ( resp . @xmath83 ) if and only if @xmath84 ( resp . @xmath85 ) . the notation of def : subsettochamber allows us to state the following classification of chambers and extreme rays . [ thm : chamberstructure ] the chambers and extreme rays of @xmath30 satisfy the following properties . a. [ assert : chamberbiject ] the map @xmath86 defines a bijection from @xmath87}$ ] to the set of chambers of @xmath35 . in particular , @xmath35 has @xmath88 chambers . b. [ assert : chamberraysprelim ] the extreme rays of @xmath75 are generated by the vectors @xmath89 where @xmath90 where @xmath91 where @xmath91 counts the elements of @xmath92 that are greater or equal to @xmath93 , and the vectors @xmath94 non - negatively span @xmath75 . in particular , the geometry @xmath30 is simplicial . the enumeration of chambers of @xmath30 announced in thm : chamberstructureassert : chamberbiject recovers the count of phases of @xmath1 with fundamental and antisymmetric matter from the physics literature @xcite . theorem [ thm : chamberstructure ] is proven in sec : extremerayschambers . suppose that @xmath95 with @xmath96 . a direct calculation shows that the box diagram associated to @xmath75 has @xmath97 plusses in the @xmath69th row for @xmath98 and all minus signs in the @xmath69th row for @xmath99 . see tables [ tab : example.gl2 ] and [ tab : example.gl3 ] for examples . c | l|c| l sign tableau & subset of @xmath100 $ ] & characteristic vector & interior points + ' '' '' & @xmath101 & @xmath102 & @xmath103 + ' '' '' & @xmath104 & @xmath105 & @xmath106 + ' '' '' & @xmath107 & @xmath108 & @xmath109 + ' '' '' & @xmath110 & @xmath111 & @xmath112 l | c| c | l sign tableau & subset of @xmath100 $ ] & characteristic vector & interior points of the chamber + ' '' '' - & - & - + & - & - + & & - & @xmath110 & @xmath113 & @xmath114 + ' '' '' + & - & - + & - & - + & & - & @xmath107 & @xmath115 & @xmath116 + ' '' '' + & + & - + & - & - + & & - & @xmath104 & @xmath117 & @xmath118 + ' '' '' + & + & + + & - & - + & & - & @xmath119 & @xmath120 & @xmath121 + ' '' '' + & + & + + & + & - + & & - & @xmath122 & @xmath123 & @xmath124 + ' '' '' + & + & - + & + & - + & & - & @xmath101 & @xmath125 & @xmath126 + ' '' '' + & + & + + & + & + + & & - & @xmath127 & @xmath128 & @xmath129 + ' '' '' + & + & + + & + & + + & & + & @xmath130 & @xmath131 & @xmath132 in light of thm : chamberstructure , we can introduce additional structure on the set of extreme rays of @xmath30 . we will equip the set of extreme rays with a partial order and use the combinatorics of the resulting partially ordered set ( poset ) to study faces and flats in @xmath30 . the following definitions describe the posets that will be relevant to us . [ def : quarterplaneposet ] the _ discrete quarter plane poset _ is the set @xmath133 endowed with the cartesian order induced by the usual order of the set @xmath134 of non - negative integers : @xmath135 this is a graded poset with grading function @xmath136 . we call @xmath137 the _ level _ of @xmath138 . [ def : posetforextrays ] we denote by @xmath139 the subset of the discrete quarter plane that consists of points at level less or equal to @xmath140 . @xmath139 is a poset with the order induced by the cartesian order defined above . we denote by @xmath141 the poset @xmath142 where @xmath143 is greater than all the elements of @xmath139 . we denote by @xmath144 the poset @xmath139 with the origin removed : @xmath145 we are now ready to relate @xmath144 to the set of extreme rays of @xmath30 . [ def : vfn ] define a function @xmath146 by @xmath147 the following theorem follows directly from thm : chamberstructure . [ thm : raystruture ] the extreme rays of @xmath35 satisfy the following properties . a. the function @xmath148 defines a bijection from @xmath144 to the set of extreme rays of @xmath35 . b. the set of extreme rays that lie in a chamber @xmath75 are the extreme rays @xmath149 where @xmath150 . we will characterize faces and flats by the extreme rays they contain . for that reason we define a function which returns the extreme rays lying in a given subset of the weyl chamber . [ def : raysoperator ] define a function @xmath151 from @xmath152 to the power set of the set of extreme rays of @xmath30 as follows . for @xmath153 let @xmath154 be the set of extreme rays of @xmath30 that lie in @xmath155 . the following theorem , proven in sec : extremeraysfaces relates faces in @xmath30 to the combinatorics of @xmath144 . [ thm : characterization.faces ] for all @xmath9 , the function @xmath156 induces a bijection from the set of @xmath9-faces of @xmath30 to the set of @xmath9-chains in @xmath144 . here , we say that @xmath157 is a @xmath9-chain if @xmath92 is a chain and @xmath158 . to state an analogue of thm : characterization.faces for flats , we will need to define structures called _ ensembles _ that will play the role of chains . [ def : ensemble ] an _ ensemble _ is the restriction to @xmath144 of a union @xmath159 \right)\cap { \mathbb{e}^\ast}_n\ ] ] of intervals @xmath160 $ ] of @xmath141 satisfying the following four conditions : 1 . @xmath161 2 . @xmath162 for @xmath163 ; 3 . @xmath164 for @xmath165 ; and 4 . @xmath166 or @xmath167 . we say that @xmath168 is a _ @xmath9-ensemble _ if @xmath169 , so that @xmath9 counts the number of distinct levels of elements of @xmath168 . figure [ fig : ensembleoneeg ] . shows one example of an ensemble . 2 -ensemble in @xmath170 . * figure ( a ) represents the @xmath171-ensemble @xmath92 which is the restriction to @xmath172 of the union of intervals @xmath173 \cup [ ( 3,3),(4,4 ) ] \cup [ ( 5,5),\infty]$ ] . theorem [ thm : characterization.flats ] guarantees that @xmath174 is the set of extreme rays of a unique 10-flat in @xmath175 . figure ( b ) illustrates a maximal chain of @xmath92 , which corresponds to a @xmath171-face of @xmath175 . ] -ensemble in @xmath170 . * figure ( a ) represents the @xmath171-ensemble @xmath92 which is the restriction to @xmath172 of the union of intervals @xmath173 \cup [ ( 3,3),(4,4 ) ] \cup [ ( 5,5),\infty]$ ] . theorem [ thm : characterization.flats ] guarantees that @xmath174 is the set of extreme rays of a unique 10-flat in @xmath175 . figure ( b ) illustrates a maximal chain of @xmath92 , which corresponds to a @xmath171-face of @xmath175 . ] [ thm : characterization.flats ] for all @xmath9 , the function @xmath156 induces a bijection from the set of @xmath9-flats of @xmath30 to the set of @xmath9-ensembles of @xmath144 . theorem [ thm : characterization.flats ] is proven in section [ sec : extremeraysflats ] on page . denote by @xmath11 the vector representation of @xmath176 and of @xmath177 . recall that the representation @xmath178 of @xmath177 factors as @xmath179 , where @xmath180 is irreducible . let @xmath181 ( resp . @xmath182 ) denote the unique 26-dimensional ( resp . 7-dimensional ) irreducible representation of @xmath183 ( resp . @xmath184 ) . we are now ready to discuss the geometry @xmath185 when @xmath2 is non - simply laced simple lie algebra . [ thm : nonsimplylaced ] suppose that @xmath186 and suppose that @xmath2 has rank @xmath140 . the geometry @xmath185 is a cone over a simplex @xmath187 . in particular , there is 1 chamber and the number of @xmath9-faces ( resp . @xmath9-flats ) is @xmath188 . there is only one other geometry @xmath13 when @xmath2 is simple and non - simply laced and @xmath7 is a direct sum of minuscule and quasi - minuscule representation . this is the geometry @xmath189 it is mentioned in ( * ? ? ? * section 6.3 ) that @xmath190 and @xmath191 have no chambers . this is in contradiction with the conclusion of thm : nonsimplylaced . a direct calculation shows that all the weights of @xmath7 are scalar multiples of the roots of @xmath2 in the cases discussed in the theorem . therefore , the hyperplanes @xmath192 for @xmath22 a weight of @xmath7 coincide with vanishing loci of roots of @xmath2 . the theorem then follows from the fact that the weyl chamber is simplicial . . * the geometry contains 4 chambers , 5 half - lines and the origin . there is one 2-flat ( the half - plane below the line @xmath193 ) , three 1-flats ( the interior walls @xmath194 , @xmath195 , @xmath196 ) , and one 0-flat ( the origin ) . ] in sec : prelims , we recall basic facts regarding the lie algebra @xmath0 and from the theory of hyperplane arrangements . in sec : extrayschmfaces , we prove theorems [ thm : chamberstructure ] , [ thm : raystruture ] , and [ thm : characterization.faces ] . in sec : facecount , we prove thm : facecount , and in sec : flatcount , we prove theorems [ thm : flatcount ] and [ thm : characterization.flats ] . m.e . is grateful to shu - heng shao and shing - tung yau for helpful discussions . m.e . and a.n . thank william massey and all the organizers of the 20th conference for african american researchers in mathematical sciences ( caarms 20 ) where this started . is supported in part by the national science foundation ( nsf ) grant dms-1406925 elliptic fibrations and string theory " . is supported by the harvard college research program . in sec : prelimliealg , we fix notation for the weyl chamber of @xmath0 . in sec : hyparrange , we recall some facts from the theory of hyperplane arrangements . a detailed list of symbols is presented in tab : notations . let @xmath197 denote the cartan subalgebra of @xmath0 consisting of the diagonal matrices . we identify @xmath5 with @xmath198 via the isomorphism @xmath199 the positive roots of @xmath0 are @xmath200 . we define the open weyl chamber of @xmath0 as the set of vectors @xmath201 forming a non - increasing sequence @xmath202 and denote by @xmath203 the closure of @xmath204 . define @xmath205 . the weights of the vector representation are @xmath206 for @xmath207 and the weights of the antisymmetric representation are @xmath61 for @xmath208 . when @xmath209 , the hyperplane @xmath210 is the vanishing locus of the weight @xmath206 of the vector representation of @xmath0 . when @xmath211 , the hyperplane @xmath212 is the vanishing locus of the weight @xmath61 of the antisymmetric representation of @xmath0 . by definition , each chamber is a convex rational polyhedral cone ( finite intersection of rational half spaces ) . in the following we will refer to a convex rational polyhedral cone simply as `` cone '' . face _ is either a chamber or the intersection of a chamber with a supporting hyperplane . there are finitely many faces , and each face is a lower dimensional cone . the chambers cover @xmath155 and form a _ fan_. a fan is a finite collection of cones such that for each cone all its faces are in the fan and the intersection of two cones is a face of each @xcite . by classical theorems of weyl and minkowski , a cone is finitely generated if and only if it is a finite intersection of closed linear half spaces ( see ( * ? ? ? * theorem 1.3 ) ) . it follows that each chamber of @xmath32 admits a system of generators . a cone of dimension @xmath140 is said to be _ simplical _ if it is non - negatively spanned by @xmath140 linearly independent vectors . a face of dimension one which is a half - line is called an _ extreme ray_. a _ facet _ is a face of co - dimension one . the roots of the lie algebra @xmath0 define a hyperplane arrangement known as the _ braid arrangement _ @xcite , which has @xmath213 chambers , namely the weyl chambers of @xmath214 . the braid arrangement is also called the coxeter arrangement of type @xmath214 . the weights of the fundamental representation define an arrangement of hyperplanes known as the _ boolean arrangement _ @xcite . the weights of the antisymmetric representation @xmath215 define an arrangement of hyperplanes known as the _ threshold arrangement _ @xcite . altogether , the roots of @xmath0 and the weights of the first two fundamental representations define the coxeter arrangement of type @xmath216 @xcite . in sec : extremerayschambers , we prove theorems [ thm : chamberstructure ] and [ thm : raystruture ] , and in sec : extremeraysfaces , we prove thm : characterization.faces . given a vector @xmath217 , we denote by @xmath218 ( resp . @xmath219 ) the vector obtained from @xmath220 by rearranging its components in non - increasing ( resp . non - decreasing ) order . the idea of the proof is to derive for each @xmath78 an explicit linear isomorphism @xmath221 from @xmath198 to @xmath5 such @xmath222 . the elementary basis vectors of @xmath198 then map under @xmath221 to the set of extreme rays of @xmath223 . define @xmath224 the absolute values of the coordinates @xmath225 of a point @xmath226 are pairwise distinct and non - zero , because @xmath227 and @xmath228 for all @xmath208 . after reordering , we can write the set @xmath229 as a strictly increasing vector @xmath230 so that @xmath231 for all @xmath69 because @xmath232 there exists a unique sequence @xmath233 of signs such that @xmath234 for all @xmath235 . we see immediately that @xmath236 defines a continuous function from @xmath237 to @xmath238 . a direct calculation shows that @xmath239 and that @xmath240 . we now prove part assert : chamberbiject of the theorem . suppose that @xmath241 . consider the vector @xmath242 note that the absolute values of the components of @xmath243 are @xmath244 . we see immediately that @xmath245 for all @xmath78 , so that @xmath246 . because @xmath247 and @xmath246 , the face @xmath223 contains a point of @xmath204 not lying in any hyperplane of @xmath30 . it follows that @xmath223 is a chamber of @xmath30 for all @xmath248 . it remains to prove that @xmath249 defines a bijection from @xmath238 to the set of chambers of @xmath30 . consider a chamber @xmath250 of @xmath30 and let @xmath251 be an interior point . note that @xmath240 . we see immediately that @xmath252 because @xmath253 is a chamber of @xmath30 . suppose that @xmath254 . it follows that @xmath243 and @xmath255 lie in the same connected component of @xmath237 . because @xmath236 is continuous , we obtain that @xmath256 so that @xmath249 is injective . part assert : chamberbiject follows . we now prove part assert : chamberraysprelim . consider a chamber @xmath223 of @xmath30 . let @xmath257 denote the subset of @xmath100 $ ] corresponding to @xmath258 . note that @xmath259 with @xmath260 . hence , @xmath223 is non - negatively spanned by @xmath261 . it follows that @xmath94 are the extreme rays of @xmath223 , as desired . recall the definition of a chain in a poset . a _ * chain * _ in a poset is a subset of the poset in which any two elements are comparable . the _ length _ of a chain is its size , and a _ @xmath9-chain _ is a chain of length @xmath9 . it follows from thm : chamberstructure that @xmath262 induces a bijection from the set of chambers of @xmath30 to the set of @xmath140-chains in @xmath144 . to prove thm : characterization.faces , we will exploit the simpliciality of the chambers of @xmath30 , which was also proved in thm : chamberstructure . suppose that @xmath250 is a @xmath9-face of @xmath30 . consider a chamber @xmath249 containing @xmath250 . because @xmath249 is simplicial ( by thm : chamberstructure ) , @xmath250 must be the non - negative span of @xmath9 extreme rays of @xmath249 . in particular , we obtain that @xmath263 is a subset of @xmath264 of size @xmath9 . because @xmath265 is an @xmath140-chain in @xmath144 and @xmath266 is a subset of @xmath265 of size @xmath9 , it follows that @xmath266 is a @xmath9-chain in @xmath144 . therefore , @xmath262 is a function from the set of @xmath9-faces of @xmath30 to the set of @xmath9-chains of @xmath144 . the fact that every face is non - negatively spanned by the extreme rays that it contains ensures that @xmath262 induces an injection from the set of @xmath9-faces of @xmath30 to the set of @xmath9-chains in @xmath144 . it remains to prove that @xmath262 is surjective . suppose that @xmath267 is a @xmath9-chain . let @xmath250 denote the non - negative span of @xmath268 . consider an @xmath140-chain @xmath269 containing @xmath270 which clearly exists , and the corresponding chamber @xmath249 ( which satisfies the property that @xmath271 ) . it follows from the simpliciality of @xmath249 ( thm : chamberstructure ) that @xmath250 is a sub - face of @xmath249 of dimension @xmath9 and that @xmath272 . therefore , every @xmath9-chain in @xmath144 can be expressed in the form @xmath273 for some @xmath9-face @xmath250 of @xmath30 . the theorem follows . in sec : prelimonposet , we describe some basic properties of the posets @xmath139 and @xmath144 . in sec : genfuncfaces , we prove thm : facecount . given two points @xmath274 and @xmath275 in @xmath144 , such that @xmath276 , consider the rectangle with vertical and horizontal edges whose ne - sw diagonal is the segment @xmath277 . the lattice points of this rectangle are exactly the points of the interval @xmath278 $ ] in @xmath144 . see figures [ fig : posetsn ] on page for examples of intervals in @xmath144 . the discrete quarter - plane @xmath133 is equipped with a grading given by the level function . the origin is the only point of level zero . denote by @xmath279 the set of points of level @xmath69 of @xmath133 . the level sets @xmath279 are all anti - chains and form a partition of @xmath133 . see fig : levelset for an example . note that @xmath280 for all @xmath281 , the interval @xmath282 in @xmath139 ( or in @xmath144 ) is isomorphic to @xmath144 as a graded poset . the isomorphism is induced by the translation by @xmath274 in @xmath133 : the map @xmath283 defines an isomorphism from @xmath284 to @xmath282 . 3 and @xmath144 . * ] and @xmath144 . * ] and @xmath144 . * ] in order to compute the generating function @xmath285 we first derive a recurrence relation for @xmath33 . [ prop : facerecursion ] the integers @xmath33 satisfy the recurrence relation @xmath286 for @xmath287 , where @xmath288 for @xmath289 and for @xmath290 $ ] . we prove prop : facerecursion by tracking the minimum of each @xmath9-chain and interpreting the @xmath291 largest elements of the chain as @xmath292-chain in @xmath293 with @xmath294 . the proposition is clear for @xmath295 . assume for the remainder of the proof that @xmath296 . denote by @xmath297 the set of @xmath9-chains in @xmath144 with minimum @xmath274 . suppose that @xmath298 . recall that translation by @xmath299 in @xmath133 defines an isomorphism from @xmath282 to @xmath300 . we see immediately that the process of removing @xmath274 from a @xmath9-chain and translating by @xmath299 ( i.e. , the function defined by @xmath301 ) defines a bijection from @xmath297 to the set of @xmath292-chains in @xmath302 . it follows that @xmath303 . note in particular that @xmath304 for @xmath305 . recall that @xmath144 contains @xmath306 elements of level @xmath307 for @xmath308 . grouping by level , we obtain that @xmath309 as desired . we need to introduce the correction term @xmath310 in order to deal with the case of @xmath295 . after some algebra , we can derive the expression for @xmath311 announced in thm : facecount . translating prop : facerecursion in to the language generating functions yields the linear equation @xmath312 where we used the identity @xmath313 the theorem follows by solving the linear equation for @xmath311 . corollaries [ cor : facerecurse ] , [ cor : facegenpoly ] and [ cor : facesmalllarge ] follow from thm : facecount by elementary algebraic manipulations . in particular , cor : facerecurse can be derived by noting that the coefficient of @xmath57 in @xmath314 vanishes for @xmath315 due to thm : facecount . cor : facegenpoly can be derived by computing a partial fraction decomposition for the expression for @xmath311 given in thm : facecount . on the other hand , cor : facesmalllarge is proved by regarding @xmath311 as the sum of geometric series with ratio @xmath316 . in sec : extremeraysflats , we prove thm : characterization.flats , and in sec : genfuncflats , we prove thm : flatcount . thm : characterization.flats will be derived from the following two lemmas . the first asserts that @xmath317 is a flat whenever @xmath318 is an ensemble , and the second asserts that every ensemble is of the form @xmath317 for some flat @xmath318 . to prove lem : flatyieldsensemble , we exploit the the fact that the set of extreme rays of a flat is the intersection of the sets of extreme rays of the hyperplanes containing it . recall that the set of extreme rays of @xmath319 @xmath320 corresponds in the poset @xmath144 to the union of intervals @xmath321\cup [ \varphi(i , j)+(1,1),\infty],\ ] ] where @xmath322 notice that @xmath323 with equality if and only if @xmath324 when @xmath209 , the second interval is empty and @xmath325 is a weight of the fundamental representation @xmath11 . the first interval is empty when @xmath326 and completely contained in the @xmath327-axis ( reps . @xmath220-axis ) when @xmath328 ( resp . @xmath329 ) . pedagogical examples are presented in fig : flats.hyperplane . * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] * . the set of extreme rays of a hyperplane @xmath330 ( @xmath71 ) corresponding to a union of intervals @xmath331\cup [ \varphi(i , j)+(1,1),\infty]$ ] in the poset @xmath139 with @xmath332 . for extreme values ( @xmath328 , @xmath333 , @xmath329 , @xmath334 , @xmath335 ) one of the two intervals can be contained in one of the axis , shrink to a point , or disappear altogether . the points @xmath336 and @xmath337 are the only points of @xmath144 not correspond to any extreme ray of any hyperplane @xmath330 . [ fig : flats.hyperplane ] ] consider a @xmath9-flat @xmath318 and denote by @xmath168 the set of extreme rays of @xmath318 . we will prove that @xmath168 is an ensemble . suppose that @xmath338 note that @xmath339 using the structure of extreme rays of hyperplanes @xmath319 and the identity @xmath340\cap[c , d]=[a\vee c , b\wedge d],\ ] ] we see immediately that @xmath341 can be minimally expressed in the form @xmath342 \cup [ a_m , b_m].\ ] ] we claim that this expression exhibits @xmath168 as an ensemble . condition ( 1 ) is satisfied because @xmath343 condition ( 2 ) follows from the minimality of the expression for @xmath168 . note that for all @xmath344 there exists @xmath345 such that @xmath346 and @xmath347 . in particular , we have @xmath348 for all @xmath349 , which is condition ( 3 ) . note that either @xmath350 for some @xmath345 for @xmath351 . condition ( 4 ) follows from the fact that @xmath323 for all @xmath352 . therefore , @xmath168 is an ensemble . it remains to prove that @xmath168 is a @xmath9-ensemble . note that because @xmath168 is a union of intervals , the number of distinct levels of elements of @xmath168 is the same as the length of any maximal chain in @xmath168 . suppose that @xmath168 is an @xmath353-ensemble and that @xmath354 is an @xmath353-chain . thm : characterization.faces ensures that there exists an @xmath353-face @xmath250 whose set of extreme rays is @xmath355 . because @xmath250 is a non - negatively spanned by @xmath355 and @xmath318 is non - negatively spanned by @xmath268 , we obtain that @xmath356 . it follows that @xmath357 . consider a @xmath9-face @xmath356 , which exists because @xmath318 is a @xmath9-flat and every flat is the union of the ( finitely many ) faces it contains . we see immediately that @xmath358 . thm : characterization.faces guarantees that @xmath266 is a @xmath9-chain in @xmath144 , from which it follows that @xmath359 . hence , we have @xmath360 , so that @xmath168 is a @xmath9-ensemble . the lemma follows . [ lem : getspecialflats ] given two points of @xmath361 and @xmath362 of @xmath141 such that @xmath363 , there exists a flat @xmath318 of @xmath30 such that @xmath364\cup [ a , \infty]\right ) \cap { \mathbb{e}^\ast}_n = \operatorname{rays}(l).\ ] ] \cup [ a,\infty]\cap { \mathbb{e}^\ast}_n$ ] * with @xmath365 in @xmath141 , @xmath366 , and @xmath367 . the three figures illustrate how to construct a flat @xmath318 whose set of extreme rays @xmath317 corresponds to the ensemble @xmath368 \cup [ a,\infty]\cap { \mathbb{e}^\ast}_n$ ] . in every case , @xmath318 is the intersection of two hyperplanes . there are three cases to consider : ( case 1 ) @xmath369 ( case 2.a ) @xmath369 and @xmath370 , ( case 2.b ) @xmath371 and @xmath372 . ] \cup [ a,\infty]\cap { \mathbb{e}^\ast}_n$ ] * with @xmath365 in @xmath141 , @xmath366 , and @xmath367 . the three figures illustrate how to construct a flat @xmath318 whose set of extreme rays @xmath317 corresponds to the ensemble @xmath368 \cup [ a,\infty]\cap { \mathbb{e}^\ast}_n$ ] . in every case , @xmath318 is the intersection of two hyperplanes . there are three cases to consider : ( case 1 ) @xmath369 ( case 2.a ) @xmath369 and @xmath370 , ( case 2.b ) @xmath371 and @xmath372 . ] \cup [ a,\infty]\cap { \mathbb{e}^\ast}_n$ ] * with @xmath365 in @xmath141 , @xmath366 , and @xmath367 . the three figures illustrate how to construct a flat @xmath318 whose set of extreme rays @xmath317 corresponds to the ensemble @xmath368 \cup [ a,\infty]\cap { \mathbb{e}^\ast}_n$ ] . in every case , @xmath318 is the intersection of two hyperplanes . there are three cases to consider : ( case 1 ) @xmath369 ( case 2.a ) @xmath369 and @xmath370 , ( case 2.b ) @xmath371 and @xmath372 . ] * case 1 : @xmath369 . consider the flat @xmath375 of @xmath30 . a direct calculation shows that @xmath364 \cup [ a , \infty]\right ) \cap { \mathbb{e}^\ast}_n = \operatorname{rays}\left(l\right).\ ] ] * case 2 : @xmath376 . write @xmath377 . consider the flat @xmath378 of @xmath30 . a direct calculation shows that @xmath364 \cup [ a , \infty]\right ) \cap { \mathbb{e}^\ast}_n = \operatorname{rays}\left(l\right).\ ] ] consider an ensemble given by @xmath379\cup [ a_1 , b_1]\cup \cdots \cup [ a_k , b_k]\right).\ ] ] define the intervals @xmath380\cup [ a_{i+1 } , \infty ) \text { for } 0 \le i \le k-1\\ d_k&= ( 0 , b_k].\end{aligned}\ ] ] lem : getspecialflats guarantees that each @xmath381 is the set of extreme rays of a flat @xmath382 . consider the flat @xmath383 a direct calculation shows that @xmath384 as desired . lem : flatyieldsensemble ensures that @xmath151 induces a function from the set of @xmath9-flats of @xmath30 to the set of @xmath9-ensembles in @xmath144 . because every flat is non - negatively spanned by its extreme rays , the induced function is injective . it remains to prove that the induced function is surjective . let @xmath168 be a @xmath9-ensemble . lem : ensembleyieldsflat guarantees the existence of a flat @xmath250 whose set of extreme rays is @xmath168 . the fact that @xmath250 is a @xmath9-flat follows from lem : flatyieldsensemble and the fact that @xmath168 is a @xmath9-ensemble . therefore , the induced function is surjective , which implies the theorem . a subset @xmath267 is a @xmath9-ensemble if and only if @xmath387 is a @xmath9-pseudo - ensemble ( necessarily with starting point 0 ) . in particular , the function defined by @xmath388 induces a bijection from the set of @xmath9-ensembles in @xmath144 to the set of @xmath9-pseudo - ensembles in @xmath139 with starting point @xmath389 . for all @xmath390 the set @xmath391 is the unique @xmath392-pseudo - ensemble in @xmath139 . denote by @xmath393 the number of @xmath9-pseudo - ensembles in @xmath139 for @xmath287 . define @xmath394 for @xmath289 and note that @xmath394 for @xmath395 and for @xmath396 . the following two lemmas illustrate the role of @xmath393 in counting ensembles : thm : flatcount will follow from the recurrence relations presented in the lemmas by routine algebraic manipulations involving generating functions . the idea of lem : rhonkrecursion is to divide into cases based on the starting point of a pseudo - ensemble . after translating the starting point to @xmath389 and removing @xmath389 , we obtain an ensemble . the lemma is clear for @xmath399 . assume for the remainder of the proof that @xmath400 . denote by @xmath401 the set of @xmath9-pseudo - ensembles in @xmath139 with starting point @xmath274 . recall that translation by @xmath299 induces an isomorphism from @xmath282 to @xmath302 . we see immediately that the composite of translation by @xmath299 and the removal of @xmath389 ( i.e. , the function @xmath402 ) induces a bijection from @xmath401 to the set of @xmath9-ensembles in @xmath302 . hence , we have @xmath403 recall that @xmath144 contains @xmath306 elements of level @xmath307 for @xmath308 . grouping by level , we obtain that @xmath404 as desired . we need to introduce the correction term @xmath405 in order to deal with the case of @xmath399 . the proof of lem : hnkrecursion is similar . we instead divide into cases based on the level of @xmath406 , the maximum of the first interval in the definition of an ensemble . translating @xmath407 to @xmath408 and intersecting with @xmath139 , we obtain a pseudo - ensemble . the lemma is clear for @xmath409 and for @xmath410 . assume for the remainder of the proof that @xmath411 . denote by @xmath412 the set of @xmath9-ensembles in @xmath144 whose first interval is @xmath413 $ ] . the hypothesis that @xmath414 ensures that @xmath415 , while the definition of an ensemble guarantees that @xmath415 for @xmath416 . recall that translation by @xmath299 induces an isomorphism from @xmath282 to @xmath302 . recall that translation by @xmath417 induces an isomorphism from @xmath418 to @xmath419 . we see immediately that the composite of translation by @xmath417 and intersection with @xmath133 ( i.e. , the function @xmath420 ) induces a bijection from @xmath297 to the set of @xmath421-pseudo - ensembles in @xmath422 . hence , we have @xmath423 the proof of cor : flatrecurse is similar to that of cor : facerecurse . in particular , cor : flatrecurse can be derived by noting that the coefficient of @xmath57 in @xmath428 vanishes for @xmath55 due to thm : flatcount . p. orlik and h. terao , http://dx.doi.org/10.1007/978-3-662-02772-1 [ _ arrangements of hyperplanes _ ] , vol . 300 of _ grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences]_. springer - verlag , berlin , 1992 . http://dx.doi.org/10.1007/978-3-662-02772-1 . g. m. ziegler , http://dx.doi.org/10.1007/978-1-4613-8431-1[_lectures on polytopes _ ] , vol . 152 of _ graduate texts in mathematics_. springer - verlag , new york , 1995 . http://dx.doi.org/10.1007/978-1-4613-8431-1 . m. esole , s .- h . shao , and s .- t . yau , `` singularities and gauge theory phases , '' http://arxiv.org/abs/1402.6331[arxiv:1402.6331 [ hep - th ] ] . m. esole , s .- h . shao , and s .- t . yau , `` singularities and gauge theory phases ii , '' http://arxiv.org/abs/1407.1867[arxiv:1407.1867 [ hep - th ] ] . h. hayashi , c. lawrie , d. r. morrison , and s. schafer - nameki , `` box graphs and singular fibers , '' http://dx.doi.org/10.1007/jhep05(2014)048 [ _ jhep _ * 1405 * ( 2014 ) 048 ] , http://arxiv.org/abs/1402.2653[arxiv:1402.2653 [ hep - th ] ] . t. zaslavsky , `` facing up to arrangements : face - count formulas for partitions of space by hyperplanes , '' _ mem . amer . math . soc . _ * 1 * no . issue 1 , 154 , ( 1975 ) vii+102 . m. esole , s. g. jackson , r. jagadeesan and a. g. nol , `` incidence geometry in a weyl chamber ii : @xmath429 , '' arxiv:1601.05070 [ math.rt ] . s. ferrara , r. minasian , and a. sagnotti , `` low - energy analysis of m and f theories on calabi - yau threefolds , '' http://dx.doi.org/10.1016/0550-3213(96)00268-4[_nucl . * b474 * ( 1996 ) 323342 ] , http://arxiv.org/abs/hep-th/9604097[arxiv:hep-th/9604097 [ hep - th ] ] . k. a. intriligator , d. r. morrison , and n. seiberg , `` five - dimensional supersymmetric gauge theories and degenerations of calabi - yau spaces , '' http://dx.doi.org/10.1016/s0550-3213(97)00279-4[_nucl.phys._ * b497 * ( 1997 ) 56100 ] , http://arxiv.org/abs/hep-th/9702198[arxiv:hep-th/9702198 [ hep - th ] ] . m. esole and s .- t . yau , `` small resolutions of su(5)-models in f - theory , '' http://dx.doi.org/10.4310/atmp.2013.v17.n6.a1[_adv . * 17 * ( 2013 ) 11951253 ] , http://arxiv.org/abs/1107.0733[arxiv:1107.0733 [ hep - th ] ] . k. matsuki , `` weyl groups and birational transformations among minimal models , '' http://dx.doi.org/10.1090/memo/0557[_mem . soc . _ * 116 * no . 557 , ( 1995 ) vi+133 ] . http://dx.doi.org/10.1090/memo/0557 . d .- e . diaconescu and r. entin , `` calabi - yau spaces and five - dimensional field theories with exceptional gauge symmetry , '' http://dx.doi.org/10.1016/s0550-3213(98)00689-0[_nucl . phys . _ * b538 * ( 1999 ) 451484 ] , http://arxiv.org/abs/hep-th/9807170[arxiv:hep-th/9807170 [ hep - th ] ] . m. esole , s. jackson , r. jagadeesan , and a. nol . incidence geometry in a weyl chamber ii : @xmath429 http://arxiv.org/abs/1601.05070[arxiv:1601.05070 [ math - rt ] ] . m. esole and s .- h . shao , `` m - theory on elliptic calabi - yau threefolds and 6d anomalies , '' http://arxiv.org/abs/1504.01387[arxiv:1504.01387 [ hep - th ] ] . d. r. morrison and d. s. park , `` f - theory and the mordell - weil group of elliptically - fibered calabi - yau threefolds , '' http://dx.doi.org/10.1007/jhep10(2012)128[_jhep_ * 10 * ( 2012 ) 128 ] , http://arxiv.org/abs/1208.2695[arxiv:1208.2695 [ hep - th ] ] . m. esole , m. j. kang , and s .- t . yau , `` a new model for elliptic fibrations with a rank one mordell - weil group : i. singular fibers and semi - stable degenerations , '' http://arxiv.org/abs/1410.0003[arxiv:1410.0003 [ hep - th ] ] . v. braun , t. w. grimm , and j. keitel , `` geometric engineering in toric f - theory and guts with u(1 ) gauge factors , '' http://dx.doi.org/10.1007/jhep12(2013)069[_jhep_ * 1312 * ( 2013 ) 069 ] , http://arxiv.org/abs/1306.0577[arxiv:1306.0577 [ hep - th ] ] . m. cvetic , d. klevers , and h. piragua , `` f - theory compactifications with multiple u(1)-factors : constructing elliptic fibrations with rational sections , '' http://dx.doi.org/10.1007/jhep06(2013)067[_jhep_ * 1306 * ( 2013 ) 067 ] , http://arxiv.org/abs/1303.6970[arxiv:1303.6970 [ hep - th ] ] . m. kuntzler and s. schafer - nameki , `` tate trees for elliptic fibrations with rank one mordell - weil group , '' http://arxiv.org/abs/1406.5174[arxiv:1406.5174 [ hep - th ] ] . c. lawrie , s. schafer - nameki , and j .- wong , `` f - theory and all things rational : surveying u(1 ) symmetries with rational sections , '' http://arxiv.org/abs/1504.05593[arxiv:1504.05593 [ hep - th ] ] . c. mayrhofer , e. palti , and t. weigand , `` u(1 ) symmetries in f - theory guts with multiple sections , '' http://dx.doi.org/10.1007/jhep03(2013)098[_jhep_ * 1303 * ( 2013 ) 098 ] , http://arxiv.org/abs/1211.6742[arxiv:1211.6742 [ hep - th ] ] . d. r. morrison and w. taylor , `` sections , multisections , and u(1 ) fields in f - theory , '' http://arxiv.org/abs/1404.1527[arxiv:1404.1527 [ hep - th ] ] . w. fulton , _ introduction to toric varieties _ , 131 of _ annals of mathematics studies_. princeton university press , princeton , nj , 1993 . the william h. roever lectures in geometry . |
we present an overview of a multi - wavelength survey of the shapley supercluster ( ssc ; z.05 ) covering a contiguous area of 260mpc including the supercluster core . the project main aim is to quantify the influence of cluster - scale mass assembly on galaxy evolution in one of the most massive structures in the local universe . the shapley supercluster survey ( shass ) includes nine abell clusters ( a3552 , a3554 , a3556 , a3558 , a3559 , a3560 , a3562 , as0724 , as0726 ) and two poor clusters ( sc1327- 312 , sc1329 - 313 ) showing evidence of cluster - cluster interactions . optical ( ) and near - infrared ( ) imaging acquired with vst and vista allow us to study the galaxy population down to m + 6 at the supercluster redshift . a dedicated spectroscopic survey with aaomega on the anglo - australian telescope provides a magnitude - limited sample of supercluster members with 80% completeness atm + 3 . we derive the galaxy density across the whole area , demonstrating that all structures within this area are embedded in a single network of clusters , groups and filaments . the stellar mass density in the core of the ssc is always higher than , which is the cosmic stellar mass density for galaxies in the local universe . we find a new filamentary structure (mpc long in projection ) connecting the ssc core to the cluster a3559 , as well as previously unidentified density peaks . we perform a weak - lensing analysis of the central 1deg field of the survey obtaining for the central cluster a3558 a mass of , in agreement with x - ray based estimates . [ firstpage ] galaxies : evolution galaxies : clusters : general galaxies : clusters : individual : a3552 , a3554 , a3556 , a3558 , a3559 , a3560 , a3562 , as0724 , as0726 , sc1327 - 312 , sc1329 - 313 galaxies : photometry galaxies : stellar contents gravitational lensing : weak | 1407.4628 | it is well established that the properties of galaxies are correlated with the environment ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? . at @xmath9 galaxy populations in rich clusters are dominated by ellipticals , s0s and at lower masses dwarf ellipticals ( des ) with few , if any , star - forming spirals @xcite . cluster galaxies have not always been as inactive as they are at the present epoch . @xcite showed that the fraction of blue ( star - forming ) galaxies among cluster members increases from almost zero in the local universe to @xmath020% by @xmath10 , while recent spitzer / herschel surveys have confirmed large numbers of starburst galaxies in clusters to @xmath11 and beyond ( e.g. * ? ? ? * ) . while some of this rapid evolution of cluster galaxies can be explained by the cosmic @xmath12 decline in star formation among field galaxies since @xmath11 ( e.g. * * ) , cluster populations have shown an accelerated evolution in star formation ( sf ) over the last four billion years , resulting in an overall @xmath13 reduction in the total star formation rates ( sfrs ) per unit halo mass since @xmath14 @xcite . empirically , clusters accrete gas - rich , star - forming spirals at @xmath15.51.0 and then contribute to transforming them somehow into the passive s0s and des of local clusters . several mechanisms affecting the galaxy properties and dependent upon the environment have been proposed and investigated in detail and all of them serve to kinematically disturb spiral galaxies and/or transform their structural properties and/or deplete their reservoirs of gas , and so quench star formation . these physical processes include gravitational and tidal interactions amongst galaxies @xcite , between galaxies and the cluster gravitational field @xcite , galaxy mergers @xcite , group - cluster collisions @xcite , ram - pressure @xcite and viscous stripping @xcite , evaporation @xcite and ` starvation ' @xcite . since these mechanisms are characterized by different time - scales and efficiencies which depend , in turn , on the properties of both the galaxies themselves ( e.g. their stellar masses , morphologies ) and their environment , they can affect the galaxy properties in different characteristic ways @xcite . examples of this are the different effects of tidal and hydrodynamical interactions . the ram pressure exerted by the hot and dense intracluster medium ( icm ) can effectively remove the cold gas supply , truncating the gas disc and quenching star formation , in massive cluster galaxies passing through the centre of rich clusters in about one crossing time ( @xmath16yr ) , while the interstellar gas of a massive galaxy may never be completely stripped if the galaxy moves on a tangential orbit or is member of a poor cluster . repeated high - velocity encounters ( harassment ) of cluster galaxies can destroy the fragile disc of dwarf galaxies , but not significantly affect the structure of a giant galaxy . the development of large spectroscopic surveys such as the sloan digital sky survey ( sdss ) , plus the availability of panoramic far ultra - violet far infra - red ( fuv fir ) data from the _ galex _ and _ spitzer _ space telescopes have allowed the impact of environment on sf to be quantified in unprecedented detail . in these studies , however , the environment and its characteristics are usually treated as ` static parameters ' . while the galaxies move across the cluster , field and filament environments ( experiencing different ram pressures , encountering other galaxies and possibly being involved in merging ) , the cluster potential well , the galaxy density as well as the icm are usually not considered ` time - dependent ' . of course , at first pass this approach is an unavoidable reduction on complexity . on the other hand in a hierarchical universe with the assembly of the structures , the galaxies evolve and move , tending towards denser regions with time , while the environments change too , thus what we actually observe is _ galaxy evolution in an evolving environment_. the most massive structures in the local universe are superclusters , which are still collapsing with galaxy clusters and groups frequently interacting and merging , and where a significant number of galaxies are encountering dense environments for the first time . the relative dynamical immaturity of superclusters and the presence of infalling dark matter halos make them ideal laboratories to test the predictions of hierarchical mass assembly models , and in particular on galaxy evolution . superclusters are not so rare systems in the universe ( see * ? ? ? the observations of superclusters is often considered a challenge to the hierarchical structure formation paradigm since such extreme dense structures , but also voids , are not reproduced by the @xmath17-body simulations . @xcite pointed out that the reason of this discrepancy can be due to the method used to assess the probability of finding such events in the distribution of cold matter . they proposed a new technique to analyse an ensemble of @xmath17-body simulations in a volume equal to that of the two - degree field galaxy redshift survey ( 2dfgrs , * ? ? ? * ) where the probability to find peculiar structures ( overdense and underdense ) was thus estimated to be @xmath02 per cent and not null . superclusters allow in principle both to study dynamical processes such as cluster - cluster collisions and group - cluster mergers and to sample different environments from cluster cores to filaments and fields . furthermore , within a dynamically active and locally dense structure the probability to observe evidence of environmental effects on galaxy evolution is dramatically enhanced making these systems a sort of _ magnifying glass _ to identify the different physical mechanisms which transform the properties of galaxies . in order to address all these aspects , a careful selection of the target is fundamental . the ideal structure should map different environments with evidence of cluster - cluster interactions . finally , in order to study in detail the galaxy properties a resolution @xmath18kpc is required in a wide range of galaxy mass down to the dwarf regime where such galaxies are not quenched by internal processes , but are more susceptible to environmental transformations . with all this in mind , we have undertaken a study of the shapley supercluster which is the largest conglomeration of abell clusters in the local universe . the shapley supercluster survey ( shass ) will map a 23deg@xmath2 region ( @xmath0 260mpc@xmath19 ) of the shapley supercluster at @xmath20 , containing filaments and embedded galaxy groups which form a dynamically - bound network connecting nine abell and two poor clusters , in order to identify the primary locations ( groups , filaments , clusters ) and mechanisms for the transformation of spirals into the s0s and des . among the observational studies of superclusters , the stages project @xcite addressed in particular the study of galaxy evolution in the abell901(a , b ) supercluster at redshift @xmath210.165 while the orelse survey ( lubin et al . @xcite , but see also mei et al . @xcite ) searched for structures on scales greater than 10mpc at higher redshifs ( [email protected] ) with the aim to investigate the properties of member galaxies . the comparison between the stages , orelse and other similar studies and shass will be then unavoidable and useful to trace the evolution with redshift , although we notice that the area ( in mpc@xmath2 ) of shass is a factor 10 that of the stages survey and was chosen to map the filaments connecting the abell clusters . the optical survey vst - access in four bands collected at the eso vlt survey telescope ( vst ) represents the core of this multi - band project . the infrared coverage with wise , the dedicated spectroscopic survey with aat / aaomega and the shapley - vista survey , together with other proprietary data , provide the fundamental data - set to achieve the scientific goals of shass . in this article we will give an overview of the project and present the first results . in sect . [ objec ] the motivations and main scientific objectives of the project are discussed . the target is described in sect . the characteristics and strategy of the survey are described in sect . [ survey ] . details of the data reduction and analysis of the data quality are given in sects . [ dr ] and [ da ] , respectively . the first results concerning the characterization of the environment are presented in sect . [ results ] , where we derive the galaxy density across the whole shass region and the underlying dark matter distribution for the central 1deg@xmath2 field which allows to estimate the mass of the galaxy cluster a3558 . in sect . [ sum ] we present the summary and conclusions . throughout the paper we adopt a cosmology with @xmath23=0.3 , @xmath24= 0.7 , and h@xmath25=70kms@xmath26mpc@xmath26 . according to this cosmology 1arcsec corresponds to 0.941kpc at @xmath27=0.048 and the distance modulus is 36.66 . the magnitudes are given in the ab photometric system . in @xmath28cdm cosmological models structure formation occurs hierarchically , such that the most massive halos corresponding to galaxy clusters form latest , doubling their masses on average since @xmath29 @xcite , and are also the most dynamically immature . preferentially located at the nodes of the complex filamentary web , clusters are continually accreting dark matter halos containing individual galaxies or galaxy groups . on average , 50% of galaxies in local clusters have been accreted since @xmath10 @xcite , of which 40% are within groups @xcite . the process of central galaxies becoming satellites , as their host halos are accreted into more massive halos , has been shown to strongly affect their evolution , with satellite galaxies more likely to be quenched than central galaxies of the same stellar mass ( e.g. * ? ? ? * ) . star formation is suppressed within galaxy groups , with the fraction of star - forming galaxies declining steadily with increasing group mass ( at fixed stellar mass ) and proximity to the group centre @xcite . galaxies which are quenched within such groups and later accreted into clusters are described as having been ` pre - processed ' . numerical simulations and theoretical studies predict the effect of cluster - cluster mergers on galaxy properties . @xcite claimed that merging of galaxy clusters may induce sf due to the increase of the external pressure of the icm compressing the cold gas in the cluster galaxies . this results in a population of starburts or post - starburst galaxies having the same age and so dating the merging event . the spatial distribution of these star - forming galaxies is expected to differ from the overall distribution of the other cluster galaxies . similarly , for comparable group and cluster halo densities gravitational shocking as the group enters the cluster could temporarily increases its mass and pull the group members into a denser , more compact system inducing galaxy mergers @xcite . simulations also showed that starbursts induced by galaxy mergers can be amplified by a factor @xmath30 @xcite if they take place in the tidal field of cluster / group of galaxies . @xcite studied the excess population of radio galaxies in the cluster a2125 at @xmath27=0.25 , mostly located in groups outside the cluster core and with radio luminosities indicating the sf as the main mechanism responsible of the radio emission . the authors related this observational evidence with the ongoing major cluster - cluster merger and explained the increased sf as due to the variation of the tidal field experienced by the member galaxies during a cluster - cluster interaction , probably being close to the core passage . studying a system of merging clusters , @xcite observed an excess of star - forming galaxies aligned along the bridge of galaxies connecting a3158 to the a3125/a3128 complex , suggestive of merger - induced sf . in addition , they found that the fraction of radio - loud sources is lower with respect to a global cluster environment and similar to that measured by @xcite in the cluster a3558 which is considered as an example of a late merger stage . they suggested that in the cluster cores radio emission from active galactic nuclei ( agns ) is suppressed in the late stages of cluster mergers . the suppressed radio emission observed in both bright and faint cluster galaxies has been also associated to cluster merger by @xcite . they explained their findings with the disruption of cool cores , i.e. the gas supply to the central agn for the bright galaxies , while the low radio loudness and fainter cluster galaxies have been affected by the enhanced ram pressure when crossing the shock front between the merging clusters . further observational support of the important role that cluster mergers play in triggering the evolution of cluster galaxies is provided by @xcite studying the merging cluster a2744 . they identified three rare ` jellyfish ' galaxies located in close proximity to the icm features associated with a merging subcluster and its shock front . their interpretation is that the sf knots detected in the tails of stripped gas are due to the rapid increase in pressure experienced by the galaxies interacting with the shock . hence , the observations seem to confirm that galaxy / group accretion and cluster - cluster merger affect the cluster galaxies and that the galaxy properties can be sensitive indicators of the merger stage . the colour - density relation is observed in galaxy clusters well beyond @xmath32 ( e.g. * ? ? ? * ) and the fraction of star - forming galaxies steadily increases with cluster - centric radius . nevertheless even at 3 - 5@xmath33 r@xmath31 , the fraction of star - forming galaxies is below that seen in the field ( e.g. * ? ? ? * ) suggesting that the environment plays a role inducing transformations well outside @xmath32 . by means of hydrodynamical cosmological simulations @xcite investigated the trend of cold and hot gas contents and sf as function of clustercentric distance in clusters and groups of galaxies . they found a large - scale trend with cold and hot gas contents and sf increasing with the clustercentric distance , but approaching the values of the field galaxy sample only at @xmath34 ( corresponding to 10mpc for a massive cluster ) . moreover , the sf in low - mass cluster galaxies ( m=10@xmath35m@xmath36 ) still shows a significant discrepancy even at this large radius with respect to that of the field galaxies . according to the authors , three main causes can explain the observed trends : i ) ` pre - processing ' of galaxies within infalling groups ; ii ) ` overshooting ' for those galaxies that are not falling in for the first time ( also called ` backsplash ' galaxies ) ; iii ) ram - pressure stripping . of course , the three mechanisms act differently depending on the galaxy and host halo mass and the clustercentric distance , but in any case the radial trend of the hot gas can not be explained solely by overshooting and pre - processing . a direct interaction with the host group / cluster is also required and their simulations suggest that ram - pressure stripping can strip the hot gas from low- and high - mass galaxies out to several times @xmath32 , although it can not affect the cold gas except for the low - mass galaxies . the authors also claimed that tidal interaction does not play an important role in the gas removal , but it should be an important factor in the morphological transformation of these galaxies . s0s differ from normal spirals by their higher bulge luminosities rather than fainter disks @xcite , disfavouring simple rps or starvation mechanisms ( but see * ? ? ? instead , mechanisms such as merging or harassment are capable of channelling material to a central bulge , sufficient to produce the higher central mass densities seen in cluster spirals and ultimately the stellar phase densities found in s0s @xcite . this channeling of material will also likely fuel rapid growth of the central supermassive black hole and trigger a period of nuclear activity . deep observations are needed to reveal tidal streams or ` fans ' , characteristic of recent merging activity @xcite . in - situ transformation via merging in clusters is strongly suppressed due to the high encounter velocities . low - velocity encounters and mergers should be much more frequent in galaxy groups , leading to the suggestion that the bulge growth required to form s0s occurs primarily via pre - processing in galaxy groups , which are subsequently accreted into clusters @xcite . moreover , the s0 fractions of @xmath37 groups match those seen in clusters at the same redshifts , and are much higher than found in the field @xcite . strong nuclear activity which is linked to bulge growth is also found to be effectively suppressed in cluster galaxies , with the distribution of x - ray agn in rich clusters revealing them to be an infalling population with high relative velocities @xcite . this confirms that the transformation of galaxies often occurs ( or begins at least ) well outside the cluster cores , in groups or as infalling galaxies enter the virialized region for the first time @xcite . hints about the distribution of galaxy luminosity , colours and morphology in the filaments have been given by @xcite , who found that brighter galaxies have a less filamentary distribution than the fainter ones . with respect to the morphology , the early - type galaxies are concentrated in the vicinity of the nodes , while spiral galaxies are sparsely distributed across the filaments . @xcite observed that the fraction of starburst galaxies in the filaments around the cluster a1763 is twice that in other cluster regions as detected by _ spitzer _ and suggested that filaments are a sort of _ galaxy reservoir _ for clusters . by studying a sample of supercluster filaments in the two - degree field galaxy redshift survey , @xcite observed a sudden enhancement in sf in faint dwarf galaxies outside the cluster virial radius . they interpreted their findings with a close interaction and/or harassment with other infalling galaxies along the same filaments . another possible explanation is the very first interaction that these galaxies experienced with the icm inducing tidal shocks and then a burst of sf . actually , the dominant processes that quench sf in galaxies depend crucially on the galaxy mass . dwarf galaxies , given their shallow potential wells , should be more susceptible to environmental processes such as tidal / ram - pressure stripping ( e.g. * ? ? ? indeed their sf histories are completely defined by their local environment , since passive des are only found as satellites within massive halos ( i.e. cluster , group or massive galaxy ; * ? ? ? * ; * ? ? ? the formation of many cluster des is often rather recent , manifest by young stellar ages ( @xmath38gyr ) and the significant populations of starburst and post starburst dwarf galaxies in the outskirts of local clusters @xcite . ram - pressure stripping ( rps ) , as originally proposed by @xcite requires , in principle , the presence of a dense icm . thus , its effect would be limited to cluster cores where the gas discs of massive spirals are rapidly truncated . nevertheless , with a 3d hydrodynamical simulation , @xcite showed that rps may extend to poorer environments for low - mass galaxies which , thanks to their lower escape velocities , are easier to strip . @xcite , by means of high resolution 2d hydrodynamical simulations , demonstrated that ram pressure effects can be observed over a wide range of icm conditions . in particular , in high density environments rps severely truncates the gas disc of l@xmath5 galaxies , while in low density environments , where moderate ram pressure is foreseen , their gas disc is clearly disturbed and bent @xcite . the gas discs of these galaxies can be truncated to 15 - 20kpc in the first 20200myr of rps . @xcite identified a bright ( l@xmath39l@xmath5 ) barred spiral galaxy 1mpc from the centre of the rich cluster a3558 in the shapley supercluster core , which is strongly affected by rps . integral - field spectroscopic observations revealed ongoing gas stripping in the form of one - sided extraplanar ionized gas along the full extent of the disk , simultaneously with a starburst triggered by gas compression along the leading edge of the galaxy . the galaxy is estimated to be being subjected to weak - moderate ram pressure , as defined by @xcite . this adds a piece of evidence to the fact that rps is acting more efficiently on the galaxy interstellar medium ( ism ) than previously foreseen and also outside of the cluster cores as also observed in the virgo cluster by @xcite . this new understanding of the rps supports the view that this mechanism is the principal transformation process to quench sf in spirals although probably helped by other processes affecting the structure of the galaxies . @xcite drew a similar conclusion analysing a sample of 182 disk galaxies in the cluster system abell901/902 . the fraction of galaxies showing asymmetric gas rotation curves , and thus probably affected by rps @xcite , turned out to be 75 per cent higher in the clusters than in the field , the majority of them being morphologically undisturbed . although these galaxies seemed preferentially located at low clustercentric radii , they also observed a population of dusty and red spirals in the cluster outskirts , which could be also affected by ism - icm interactions . the relevance of rps for quenching sf in cluster galaxies is further complicated by the fact that a non - homogeneous icm is expected for non virialized merging and post - merging clusters , thus the clustercentric distance is not the only parameter describing the icm density , i.e. the ram - pressure strength . our study is based on characterizing a dynamically active environment in a multifaceted way and disentangling its effects on galaxy evolution . it has the following main objectives . - : : to investigate the role of cluster - scale mass assembly on the evolution of galaxies mapping the effects of the environment in the cluster outskirts and along the filaments with the aim to identify the very first interactions between galaxies and their environment . - : : to identify and measure signs of ongoing transformation in galaxies belonging to a complex structure with the goal of improving our comprehension of what drives their star - formation quenching and structural modification . - : : to obtain detailed maps of the dark matter and baryonic matter distributions ( galaxies , icm ) , combining weak lensing , x - ray and dynamical analyses . - : : to quantify the variation in the stellar mass fractions going from cluster cores to groups , by comparing the near - infrared light distribution with the dark matter maps and dynamical masses . - : : to build up a multi - band homogeneous data - set on this area of the sky made of sub - kiloparsec resolution imaging and magnitude - limited spectroscopy , thus providing the community with a solid background for studies of the shapley supercluster . to address the above objectives we will explore the global and internal properties of galaxy populations extending outside the cluster / group virial radius and aim for an accurate characterization of the environment . this will be defined through galaxy density , dark matter distribution , dynamical substructure , and icm properties . the different quantifications of the environment will allow us to disentangle the effects of local and large - scale density , cluster and group merging , dynamical state and mass of the host system on the properties of galaxies in different ranges of mass . recent studies investigated the effects of the local and large - scale environment on galaxy evolution , mainly considering groups of galaxies . some of these works defined the environment through the luminosity - density only ( e.g. * ? ? ? * ) , or associate galaxy groups to a large - scale environment identified by a smoothed luminosity field ( e.g. * ? ? ? * ) , or are limited to massive galaxies ( e.g. * ? ? ? therefore , although there is a general agreement that the group environment is important for the evolution of the group members , a study of global and internal properties of galaxy population in groups and , in general , in different structures dynamically bound in a supercluster is still lacking . for instance , @xcite found that both local ( distance from the centre ) and global ( mass ) group environments play a role in quenching star formation , while @xcite show that global group - scale mechanisms linked to the presence of a hot gas halo are dominant in quenching sf in group galaxies rather than purely density related processes . in the following we briefly outline how we will derive the quantities necessary for our study ( see also sect . [ survey ] ) . as the main classification scheme of galaxies we will adopt the concentration , asymmetry and clumpiness parameters @xcite , complemented by the m20 and gini coefficients @xcite . this parameter set ( hereafter ` cas+mg ' ) has proven to robustly link the internal light distribution of galaxies to their formation and evolution ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) . besides being effective in separating the different morphological types and tracing star formation , the cas+mg scheme is particularly sensitive to their recent interaction or merging activities , making it the ideal tool to obtain a census of galaxies whose structure appears disturbed by the environment , which is crucial for our project . for a selected sample of ` normal ' galaxies ( as determined through the cas+mg parameters ) it will be possible to quantify the relevance of ( pseudo-)bulges and bars as a function of the environment by surface photometry fitting . in fig . [ gals ] we show examples of ssc galaxies in the magnitude range from m*-1 to m*+4 ( @xmath40mag ) . for galaxies in this magnitude range we plan to perform classical ( visual ) morphological classification , that will be used to calibrate the cas+mg parameters against the hubble sequence for our data . the signal - to - noise ratio ( snr ) and spatial resolution of vst images will instead allow us to obtain reliable cas+mg parameters up to m*+6 ( @xmath41mag ) , i.e. well into the dwarf galaxy regime ( see sect . [ optical ] ) . finally , aaomega and literature spectra of more luminous galaxies will provide spectral classification through line indices . galaxy global properties such as colours , stellar masses and star - formation rates will be derived . the contribution of both unobscured and obscured sfr need to be considered since dust - obscured cluster galaxies are common in merging clusters ( e.g. * ? ? ? * ) . the @xmath42band luminosity will provide a star - formation rate ( sfr ) indicator once calibrated using the multi wavelength data already available in the shapley supercluster core ( see sect . [ cd ] ) , by obtaining correlations between luminosity , dust extinction , and metallicity . the available mid - infrared data will provide a robust and independent indicator of obscured star formation ( sect . [ cd ] ) , and also allow us to identify agn via their unusually red w1-w2 colours ( e.g. * ? ? ? * ) . agn will also be identified from their emission - line ratios or broad emission lines from the aaomega spectra , or their x - ray emission in the supercluster core where xmm data are available . the depths of our survey are conceived to reach the necessary accuracy of all of the above quantities , as will be outlined in the next sections . finally , we will not only measure statistically the environmental effects on galaxy properties on such large scales , but also to ` catch in the act ' the direct interaction of supercluster galaxies with their environment into the surrounding large - scale structure . the latter is possible only using imaging with a sub - kiloparsec resolution and follow - up integral - field spectroscopy for a few individual cases ( e.g. * ? ? ? the ` remote cloud of galaxies in centaurus ' first identified by @xcite is one of the richest superclusters in the nearby universe , consisting of as many as 25 abell clusters @xcite in the redshift range [email protected] @xcite . the first spectroscopic study confirming the existence of the ` centaurus supercluster ' was carried out by @xcite , then it was re - discovered by ( * ? ? ? * their @xmath44-region ) as a cluster overdensity in the ( * ? ? ? * aco ) cluster catalogue and identified by ( * ? ? ? * his shapley concentration ) as an excess in galaxy number counts in the uk schmidt telescope sky survey plates . the shapley supercluster ( hereafter ssc ) because of its peculiar richness and location , lying in the direction of the dipole anisotropy of the cosmic microwave background ( cmb ) , was investigated as responsible of at least a fraction of the local group acceleration ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) . to assess the role of this structure in the acceleration of the local group a robust estimate of its mass is required . this gave a boost to spectroscopic and x - ray observations of the supercluster devoted to i ) map the whole structure and measure the density contrast in galaxy number counts and mass / number of associated clusters ; ii ) investigate its dynamical state and iii ) derive the underlying mass distribution . a detailed review of the properties of the shapley supercluster based on previous investigations will be given elsewhere ( merluzzi et al . , in preparation ) , here we will describe the key characteristics of the ssc - i.e. those features making the ssc , and above all its central region , _ a magnifying glass _ to investigate the effect of the environment and mass assembly on galaxy evolution . by analysing a spectroscopic sample across a region of 15deg in diameter centred on the dominant cluster a3558 , @xcite concluded that the supercluster has a ` cigar - shape with the eastern side being the closest to us ' ( but see also * ? ? ? * ) and from the flattened geometry they suggested that it is not spherical and virialized . the complex morphology of the ssc comprises a main body at @xmath4515000kms@xmath26 together with walls / filaments / arms of galaxies connecting the three main systems of interacting clusters ( the a3558 , a3528 and a3571 complexes ) as well as a foreground structure connecting the ssc to the hydra - centaurus supercluster ( @xmath454000kms@xmath26 ) and hints of a background structure at @xmath4523000kms@xmath26(for details see * ? ? ? * ; * ? ? ? the main plane of the ssc ( @xmath4514800kms@xmath26 ) extends 10x20deg@xmath19 ( see * ? ? ? is this vast and complex structure gravitationally bound ? in this case it would be the most massive bound structure known in the universe . applying a spherical collapse model , @xcite found that the ssc is gravitationally collapsing at least in its central region within a radius of 8@xmath46mpc centred on a3558 including 11 aco clusters ; the very inner region , associated with the massive clusters , is likely in the final stages of collapse . x - ray observations of clusters in the ssc confirmed the overdensity : a factor 10 to 50 cluster overdensity @xcite and factor 3 baryon overdensity @xcite over a region of 60x80mpc@xmath19 . these results indicated that the structure is gravitationally bound at least in its central region . from the rosat all - sky survey in a region of 0.27sterad , @xcite measured the cluster number density which turned out to be more than an order of magnitude greater than the mean density of abell clusters at similar latitudes - mainly due to an excess of low - luminosity x - ray systems in the outskirts , suggesting that the supercluster is still accreting low - mass systems . @xcite analysing a mosaic of 15x20deg@xmath19 of rosat and einstein observatory x - ray observations centred on a3558 , concluded that the ssc core is dynamically bound within @xmath47mpc , approaching the maximum expansion before collapsing , i.e. the turnaround point . on the other hand , the dynamical analysis of @xcite proposed a scenario where the ssc has already reached its turnaround radius and the final collapse will happen in @xmath48gyr . using the x - ray cluster sample of @xcite and the spherical collapse model , @xcite investigated the dynamics of the ssc . their study showed that the ssc is not bound at a radius of 51mpc with an outer shell moving radially away from the centre , while the excess of mass becomes enough to bind a spherical region of @xmath010mpc radius . recently , @xcite ran @xmath17-body numerical simulations of the ssc and other superclusters in order to determine what portions of the superclusters were potentially gravitationally bound . the ssc showed the most extended bound structure among the other analyzed superclusters . in particular , a3554 , a3556 , a3558 , a3560 and a3562 have a large number of close encounters in their simulations , while an additional pair , a1736 and a3559 , has also some chance of being bound . in such a crowded environment it is very unlikely that galaxies have not been affected by cluster related processes such as rps , starvation and/or galaxy harassment and tidal interaction . one of the key objectives of shass is to constrain the dark matter distribution and mass over this whole region via a weak lensing analysis , and in conjunction with x - ray and dynamical analyses estimate the extent of the region that is gravitationally bound to the ssc core , and determine whether clusters such as a3559 are currently falling towards the supercluster core . we now zoom in to the very central dense region of ssc , namely the ` core ' , which consists of three abell clusters : a3558 , a3562 , and a3556 and two poor clusters ( sc1327 - 312 and sc1329 - 313 ) , resolved by @xcite by means of x - ray observations . the evolutionary stage of the ssc core ( sscc ) is still matter of debate . [ cols="<,<,<,<,<,^,^,^,<,<,<,<",options="header " , ] [ zp ] the vst images have been processed using the vst tube imaging pipeline @xcite , developed for the data produced by the vst . after applying the overscan correction and bias subtraction , we divide by the master flat - a normalized combination of the dome and twilight flats , in which the twilight flat is passed through a low - pass filter first . due to differences between the electronic amplifiers , the ccds do not have the same gain levels . in order to have the same photometric zeropoint ( zp ) for all the mosaic chips a gain harmonization procedure has been applied . the procedure finds the relative ccd gain coefficients which minimizes the background level differences in adjacent ccds . a further correction is needed for the light scattered by the telescope and instrumental baffling . the resulting uncontrolled redistribution of light adds a component to the background and the flat field will not be an accurate estimate of the spatial detector response . indeed , after flat - fielding , the image background will appear flat but the photometric response will be position - dependent . this error in the flat - fielding can be mitigated through the determination and application of the illumination correction ( ic ) map . the ic map is determined by comparing the magnitudes of photometric standard fields with the corresponding sdss dr8 psf magnitudes . the differences between the vst and sdss magnitudes are plotted vs. x and y axis in fig . [ ic ] before ( top panels ) and after ( bottom panels ) the ic . usually , the same ic map can be used for observations carried out on the time scale of a month . in the case of the _ i _ band it is required to correct for the fringe pattern due to thin - film interference effects in the detector of sky emission lines . also this is an additive component that must be subtracted . the fringing pattern is estimated using the @xmath49 ratio where _ superflat _ is obtained by overscan and bias correcting a sigma - clipped combination of science images and _ tflat _ is the sky flat . to this aim , the dithering amplitudes of the science frames used must be larger than the extended object sizes in the same images . this is mandatory to allow the sigma - clip procedure to efficiently remove such objects . the difference between the above ratio across the image and its model obtained with a surface polynomial fit is the fringing pattern . the fringe pattern is subtracted from the image applying a scale factor which minimizes the absolute difference between the _ peak _ and _ valley _ values ( maximum and minimum in the image background ) in the fringe corrected image . the relative photometric calibration among the exposures contributing to the final mosaic image is obtained comparing the magnitudes of bright unsaturated stars in the different exposures , i.e. minimizing the quadratic sum of differences in magnitude using scamp @xcite . the absolute photometric calibration is computed on the photometric nights comparing the observed magnitude of stars in photometric standard fields with sdss photometry . for those fields that are not observed in photometric but clear nights , we take advantage of the sample of bright unsaturated stars in the overlapping regions between clear and photometric pointings and , by using scamp , each exposure of the clear fields is calibrated onto the contiguous photometrically calibrated field . in table [ zp ] we give examples of the fitted values for the zp and colour term obtained using the photcal tool @xcite . in several cases , the photometric standard star fields were observed with insufficient span in airmass to do a suitable fit . the extinction coefficient was then taken from the extinction curve provided by eso . the errors listed in table [ zp ] are those of the fit . the actual errors for the zp is given by the r.m.s . of the detection among the different exposures of each pointing and turns out to be less than 0.03mag in all bands . -band magnitudes of stars in a standard field observed with vst and sdss magnitudes vs. x and y image axis . _ bottom _ : the same @xmath50mag after the ic correction has been applied.,width=317 ] the relative astrometry among the pointings is derived using scamp , while for the absolute astrometry we refer to the astrometric catalogue 2mass obtaining @xmath51arcsec astrometric accuracy in all bands . the image resampling , application of astrometric solution and co - addition is performed using the tool swarp @xcite which produces the final stacked image with a weighted average . at the stage of the co - addition the scale factors for relative and absolute photometric calibration are applied . once the mosaic images were obtained , we proceeded with the catalogue extraction and measured the photometric accuracy and completeness . details of shass catalogue extraction and release will be given in a companion article ( mercurio et al . in preparation ) . we used the software package sextractor @xcite estimating the background locally and using a gaussian filter for source detection . for each source we measured magnitudes in different apertures as well as kron and psf magnitudes . to distinguish between stars and galaxies , we adopted a progressive approach @xcite using the following parameters provided by sextractor : i ) the stellarity index to select point - like sources ; ii ) the half - flux radius as a measure of source concentration ; iii ) the new sextractor parameter which takes into account the difference between the model of the source and the model of the local psf ; iv ) the peak surface brightness above background ; v ) a final visual inspection for objects with ambiguous values of the stellarity index . we estimated the completeness magnitudes using both trends of the galaxy number counts and the method by @xcite . the catalogues turned out to be 100 per cent complete at the total magnitudes of 23.9 - 24.1 , 23.8 - 24.0 , 23.3 - 23.5 , 21.8 - 22.0mag in @xmath3 bands , respectively . the ranges of magnitudes are due to small differences in seeing among the vst fields . we measured the snr inside a 3arcsec aperture as function of magnitudes achieving snr=20 at @xmath3 at 22.8 , 23.3 , 22.8 , 21.7mag , respectively . snr@xmath0100 is reached at 21.1mag in @xmath52 band ( m@xmath5 + 6 ) as required . these depths enable us to study the galaxy population down to m@xmath5 + 6 . .,width=302 ] in this section we derive the first quantitative characterization of the environment in the survey area in terms of stellar mass density . we also derive the dark matter distribution in the very centre of the sscc including the cluster a3558 and the poor cluster sc1327 - 312 . to map the structure of the supercluster , and determine its extent in redshift space and across the plane of the sky , we take advantage of our redshift survey which allows us to demarcate the supercluster in redshift space as lying within the recession velocities 1130017000kms@xmath26 ( see fig . [ histo ] ) . these cuts select a supercluster sample of 2281 galaxies across the shass area . each galaxy was weighted ( @xmath53 ) according to the inverse probability of it having been observed spectroscopically . firstly , each galaxy which could have been targeted for spectroscopy ( @xmath54 , @xmath55 ) or was a spectroscopic member of the supercluster , was given an initial equal weight of 1.0 . for each of these galaxies lacking a redshift , its weight was transferred equally to its ten nearest neighbours with known redshift that also had the same priority level in our aaomega spectroscopic survey . this results in galaxies without redshifts having zero weight , while galaxies in regions where the spectroscopic survey is locally 50@xmath56 complete having weights of 2.0 . the transferring of weight only within priority levels ensures that we can account statistically for the systematic differences in spectroscopic completeness from one level to another , as well as mapping the local spatial variations in completeness . each galaxy @xmath57 belonging to the supercluster is represented by a gaussian kernel whose transverse width is iteratively set to @xmath58^{-1/2}$ ] , where @xmath59 is the geometric mean of the @xmath60 , and a fixed radial width of 700kms@xmath26 . each galaxy is normalized by the weight parameter @xmath53 to account for spectroscopic incompleteness . the transverse kernel width for each galaxy is initially set to @xmath61arcmin , and then iteratively adjusted to account for variations in the spatial density of galaxies , such that it typically encloses the 5 to 10 nearest neighbours of the galaxy , irrespective of its location within the supercluster . figure [ d_map ] shows the resulting density map in which each galaxy is further weighted by its w1 flux as a proxy for its stellar mass ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? the density map draws our attention to several features of the whole structure . a ) : : across the sscc the density of supercluster galaxies is always higher than 315mjydeg@xmath62 . this corresponds to @xmath63 based on the w1stellar mass calibration obtained for the ssc galaxies ( see end of sect . [ aao ] ) . assuming a depth of 5000kms@xmath26 or 71mpc the stellar mass density is @xmath64 . both the stellar mass estimates include a 12% correction to account for the low - mass galaxies with @xmath65 or w@xmath66 not targeted in our spectroscopic survey , based on the stellar mass function of @xcite . this is a mean overdensity of @xmath7 with respect to the cosmic total stellar mass density for galaxies in the local universe ( @xmath67 ) estimated by @xcite from the gama survey @xcite . we obtain a similar overdensity of 3540@xmath68 if we simply compare the observed w1 flux density of sscc galaxies with the average w1 flux density of @xmath69 galaxies within the same redshift range selected from the sdss dr7 ( 8032deg@xmath19 ) . although the sdss is @xmath52-band selected , it should be complete to w1=14.5 at @xmath20 , and only marginally incomplete to w1=15.0 . the higher density peaks in the sscc , corresponding to the cluster cores but also the groups and cluster sub - structures , are interconnected forming a single system . this implies that the galaxy distribution follows the hot gas distribution observed across the whole sscc ( e.g. * ? ? ? * ) . + . red diamonds ( @xmath70 ) and blue circles ( @xmath71 ) denote the correlation function for corrected galaxy ellipticities , @xmath72 and residual stellar ellipticities , @xmath73 after correcting the psf anisotropy . green triangles ( @xmath70 ) and magenta squares ( @xmath71 ) are the correlation function for raw galaxy ellipticities , @xmath74 and raw stellar ellipticities , @xmath75 before the correction . the separation radii are offset by @xmath76 . ] ( b ) : : there is clear evidence of a filament ( @xmath07mpc in projection ) heading north from the sscc , connecting it with a3559 , with w1 flux densities of @xmath7750mjydeg@xmath62 ( @xmath78 overdensities ) . @xcite qualitatively indicated a ` broad arm running north ' from the sscc , but in their description this feature should point to a3557a which is located nw of the shass region . in this direction we do not detect any clear overdensity . c ) : : several overdensities are detected across the density map . some of them can be associated to cluster dynamical substructures already identified by ( * ? ? ? * see fig . [ d_map ] ) , but other density peaks are also detected , i.e. that w of a3556 . the positions of the system centres as defined by x - ray peak and galaxy overdensity appear slightly different , as expected in dynamically active systems . the highest discrepancy between these determinations is found for as0724 , where the x - ray centre is located 35arcmin from the galaxy overdensity centre . notice that as0726 was not detected by rosat and its centre was derived by a previous spectroscopic survey @xcite , likely affected by incompleteness . our newly detected overdensity is actually located 16arcmin from the previous centre . d ) : : in a3560 two main density clumps are sharply defined together with a western clump . the northern clump can be related to the x - ray emission , elongated towards a3558 , detected in the very centre of the cluster which suggested a minor merger scenario between the main cluster and a group located @xmath79arcmin n of the cluster centre @xcite . also the galaxy distribution shows an asymmetric shape elongated in the n direction , but extending a factor two in projection with respect to the x - ray emission ( c.f . fig . 5 of * ? ? ? * ) , taking into account only the two denser clumps in our map . thus the group involved in the claimed merger could be significantly more massive than that previously identified in the apm catalogue . in a companion paper , we will assign galaxies to supercluster structures / overdensities . this will be achieved through the dynamical analysis enabling us to detect and measure the amount of cluster dynamical substructures , and to identify possible pre - merging clumps or merger remnants . in particular , we will i ) select cluster members ( e.g. * ? ? * ; * ? ? ? * ) ; ii ) derive the centre of the most significant peaks of each identified system applying the 2d adaptive kernel technique to galaxy positions ; iii ) detect possible subclumps and assign objects to groups @xcite ; iv ) identify the 3d substructures combining velocity and position information @xcite . we conducted a weak lensing analysis of the vst - access field 8 ( see fig . [ vst - access ] ) following @xcite as modified by @xcite . we measure image ellipticity of objects detected in the @xmath52-band data , @xmath80 , from the weighted quadrupole moments of the surface brightness . we then correct a psf anisotropy of galaxy ellipticities by the function of second - order bi - polynomials of the stellar anisotropy kernel . a cross correlation function of residual stellar ellipticities and the corrected galaxy ellipticities does not show overcorrection or insufficient correction as shown in fig . [ fig : xi_ge ] . then , a reduced distortion signal , @xmath81 , is estimated with a correction of isotropic smearing effect . we select background galaxies following @xcite and adopt a new method for minimizing the contamination of member galaxies extending to the colour - colour plane . we measure two colours , @xmath82 and @xmath83 , and fit the red sequence with a linear function . a colour offset is defined for each galaxy by @xmath84 , where @xmath85 and @xmath86 and ` es0 ' denotes the red sequence galaxies . we select background galaxies by @xmath87 which is the lower limit to allow @xmath88 contamination level . the number density of background source galaxies is @xmath89 , and their mean redshift is @xmath90 . the mass map is reconstructed as described in @xcite . the reduced shear is pixelized into a regular grid with a gaussian smoothing , @xmath91 $ ] . the fwhm resolution of the map is 11.7arcmin . the smoothed shear pattern is estimated with a gaussian kernel , @xmath92 , and a statistical weight @xmath93 for the @xmath94th galaxy , where @xmath95 is the rms error of the shear estimate and the softening constant variance . we choose @xmath96 as a typical value of the mean rms of @xmath95 . we next invert the smoothed shear field with the kernel @xcite in fourier space to obtain the projected mass distribution . the resulting mass map , @xmath97 , for field 08 of vst - access is shown in fig . [ fig : wlmap ] . the mass map is elongated along the east - west direction which is parallel to the large - scale filamentary structure . a main peak is associated with distributions of member galaxies and diffuse x - ray emission of a3558 . another possible clump , located e of a3558 , is likely associated with diffuse x - ray emission of the galaxy group sc1327 - 3136 . we compute the tangential distortion component , @xmath98 , with respect to the projected cluster - centric radius from the bcg of a3558 , shown in fig . [ gt_profile ] . the lensing signal @xmath99 is decreasing as the radius , @xmath100 increases , and becomes flat or increases at @xmath101arcmin . to understand this feature , we fit the two - dimensional shear pattern with two mass components : a3558 and the east clump ( e.g. * ? ? ? we use the navarro frenk & white ( nfw ) profile @xcite as the mass model represented by two parameters : @xmath102 which is the enclosed mass within a sphere of radius @xmath103 , and the halo concentration @xmath104 . here , @xmath103 is the radius inside of which the mean density is @xmath50 times the critical mass density , @xmath105 , at the redshift @xmath27 . since it is difficult to constrain the concentration for the east clump , we assume a mass - concentration relation @xcite . the centre of a3558 is fixed at the bcg position , while the position of the secondary mass component is treated as a free parameter . the model is described by five parameters in total . the best - fit m@xmath106 for a3558 is @xmath107m@xmath108 . the total lensing signal of the two components ( solid line in fig . [ gt_profile ] ) well describes the observed lensing signals . since a3558 and the east clump are embedded in the sscc , surrounding data are essential to further constrain the mass . , as a function of the projected cluster - centric radius from the bcg , is estimated by azimuthally averaging the measured galaxy ellipticities . a bump in lensing signals is found around @xmath109 . the profile is well described by two nfw components of a3558 and the east clump . the red solid , green dashed and blue dotted lines are the best - fit nfw profile for the total mass , a3558 and the east clump , respectively . _ bottom panel : _ the @xmath110 degree rotated component , @xmath111 , is consistent with a null signal . ] although this is a preliminary result , the derived dark matter mass can be compared with that evaluated from the xmm - newton data - set . to this aim , we computed the y@xmath112 mass proxy , defined as the product of gas mass m@xmath113 and average temperature @xmath114 @xcite . as described in @xcite , we inverted a gas mass profile from the radially average surface brightness of a3558 , then iterated about the y@xmath112-m@xmath106 scaling relation calibrated from hydrostatic mass estimates in a nearby sample of clusters observed with xmm - newton @xcite . note that the surface brightness of a3558 has been evaluated within an angular sector excluding the eastern filament connecting a3558 to sc1327 - 312 . this yielded estimates of gas mass m@xmath115m@xmath108 , average temperature @xmath116kev , and total mass m@xmath117m@xmath108 within r@xmath118kpc , which is consistent with the wl estimate . previous x - ray mass estimates based on rosat data gave for m@xmath106 : 8.7@xmath119m@xmath108 @xcite ; 6.09@xmath119m@xmath108 @xcite ; 8.4@xmath120m@xmath108 ( 30% uncertainty , * ? ? ? the dynamical mass derived by @xcite of a3558 and listed in table [ clusters ] turns out to be a factor 1.7 higher than the wl mass , but also a factor 3 higher than x - ray mass determination previously available as the authors stated in their work . @xcite computed the dynamical mass inside the virial radius obtaining m@xmath121=6.7@xmath119m@xmath108 . the shapley supercluster survey ( shass ) aims to assess the role of cluster - scale mass assembly on galaxy evolution searching for possible connections between the properties of the cosmological structures ( density , dynamical status , hot - gas content , dark and luminous matter distribution ) and those of the associated galaxies ( morphology , internal structure , star formation , nuclear activity ) . this requires that we extend the investigation from the cluster cores to their outskirts , to the infalling galaxies and groups along the filaments in a dynamically - bound network . the centre ( @xmath03deg radius ) of the shapley supercluster at @xmath122 is the optimal target to undertake such a study enclosing a massive and dynamically active structure showing signs of cluster - cluster mergers , enhancing the probability to observe evidence of environmental effects on galaxy evolution , but also providing an extraordinary variety of environments concentrated in a small survey volume . shass includes nine abell clusters ( a3552 , a3554 , a3556 , a3558 , a3559 , a3560 , a3562 , as0724 , as0726 ) and two groups ( sc1327 - 312 , sc1329 - 313 ) covering a region of @xmath0260mpc@xmath2 . the survey includes the following data - sets . - : : optical ( @xmath3 ) imaging acquired with the vlt survey telescope ( pi p. merluzzi ) provides a galaxy catalogue complete down to @xmath123 ( snr@xmath010 , limit for star / galaxy separation as measured from the collected data ) corresponding to @xmath0m@xmath5 + 8.3 at the supercluster redshift . the multi band @xmath124 catalogue is complete to @xmath0m@xmath5 + 7.1 and the @xmath42-band catalogue to @xmath0m@xmath5 + 6.7 . the achieved snrs allow to study the galaxy population down to @xmath0m@xmath5 + 6 and derive morphological parameters ( cas+mg , see sect . [ da ] ) in @xmath52 band to the same depth . the vst - access survey is ongoing . - : : near - infrared ( @xmath4 ) imaging acquired with the visible and infrared survey telescope for astronomy ( pi c. p. haines ) reaching the depth of @xmath125 ( @xmath0m@xmath5 + 7.9 at the supercluster redshift ) allows to study the galaxy population down to the magnitude limit of the optical catalogue . the survey started in april 2014 . - : : the spectroscopic survey with aaomega at the anglo - australian telescope ( pi p. merluzzi ) collected 4037 new redshifts across 21deg@xmath2 of shass . together with the already available redshifts the spectroscopic sample is now 80% complete down to @xmath126 ( @xmath0m@xmath5 + 3 at the supercluster redshift ) . - : : the above dedicated surveys are complemented by near / mid - infrared data from the wide - field infrared survey explorer ( wise ) in four bands ( w1-w4 ) 3.4 , 4.6 , 12 and 22@xmath127 m . over the shass area , we reach w1=16.96 ( vega magnitude ) and w3=11.12 mag ( 1.0mjy ) at snr=10 corresponding to @xmath0m@xmath5 + 5 and sfr of 0.46m@xmath108yr@xmath26 , respectively , for galaxies in the supercluster . for the central 2 - 3deg@xmath2 , xmm - newton , _ spitzer_/mips 24@xmath127 m and 70@xmath127 m , _ galex _ near - ultraviolet and far - ultraviolet data are also available , as well as targeted observations of single galaxies providing h@xmath44 imaging with the maryland - magellan tunable filter on the magellan - baade 6.5 m telescope at the las campanas observatory in chile and integral - field spectroscopy with wifes on the australian national university 2.3 m telescope at siding spring in australia . in this first article we derived the stellar mass density distribution based on supercluster members weighted by the w1 flux . this first quantitative characterization of the environment for the whole region covered by shass shows a clumpy structure both in the sscc and the surrounding clusters with several substructures , most of them already identified in previous works . all the clusters in the shass area are embedded in a common network . this was suggested , but only for the sscc where supercluster galaxies continuously populate the core following the gas filaments connecting the abell clusters as mapped by x - ray observations @xcite . we estimate the mean overdensity across the sscc being @xmath7 with respect to the cosmic total stellar mass density for galaxies in the local universe ( @xmath67 ) . some new substructures with respect to previous works have been identified in the shass density map such as those associated with a3560 n from the cluster centre in the direction of a3558 , towards which also the x - ray emission is elongated @xcite . the most important new feature is however the filament connecting the sscc and the cluster a3559 as well as the less pronounced overdensity extending from the sscc towards a3560 . the other environment indicator analyzed here is the dark matter distribution derived from the weak - lensing analysis of vst imaging . using this approach , we studied the central 1deg@xmath19 field including a3558 and sc1327 - 312 . the derived wl map shows that the dark matter is concentrated in two peaks which correspond to the rich cluster and the group , although the centres seem slightly offset with respect to the x - ray emission and the galaxy density . this can be due to the lower resolution of the wl map and/or ascribed to the complex dynamical state of the sscc . the estimated mass of a3558 is @xmath8 , consistent with the x - ray estimate of m@xmath128m@xmath108 . we notice that the wl mass determination will be improved extending the analysis to a larger region of the supercluster as planned in the shass project . we conclude pointing out that the vst imaging is the first ccd photometry covering homogeneously and continuously such a large portion of the shapley supercluster , with this depth and resolution ( corresponding to 0.75kpc at the supercluster redshift ) . taking advantage of the @xmath94-band and wise w1 photometry , the aaomega survey has been designed and carried out to obtain a magnitude - limited redshift sample , which was never achieved before ( e.g. * ? ? ? * ; * ? ? ? * ) . with these characteristics , shass will build up the first multi - band homogeneous data - set of a vast region of the shapley supercluster and provide a fundamental local counterpart to the supercluster surveys at higher redshifts . this work was conceived in the framework of the collaboration of the fp7-people - irses-2008 project access . based on data collected with i ) the eso - vlt survey telescope with omegacam ( eso programmes 088.a-4008 , 089.a-0095 , 090.a-0094 , 091.a-0050 ) and the eso - visible and infrared survey telescope for astronomy with vircam ( eso programme 093a-0465 ) at the european southern observatory , chile and ii ) anglo - australian telescope and 2df+aaomega at the australian astronomical observatory , australia ( opticon proposal 2013a/014 ) . the optical imaging is collected at the vlt survey telescope using the italian inaf guaranteed time observations . the research leading to these results has received funding from the european community s seventh framework programme ( fp7/2007 - 13 ) under grant agreement number 312430 ( opticon ; pi : p. merluzzi ) and prin - inaf 2011 ` galaxy evolution with the vlt surveys telescope ( vst ) ' ( pi a. grado ) . cph was funded by conicyt anillo project act-1122 . n.okabe is supported by a grant - in - aid from the ministry of education , culture , sports , science , and technology of japan ( 26800097 ) and by world premier international research center initiative ( wpi initiative ) , mext , japan . pm and gb would like to thank a. m. hopkins for his support during the spectroscopic observations at the australian astronomical observatory and the universidad de chile for the hospitality and support during their staying . pm thanks m. petr - gotzens for her support in the vst observations . the authors thank prof . l. campusano for his helpful comments to the manuscript . this publication makes use of data products from the wide - field infrared survey explorer , which is a joint project of the university of california , los angeles , and the jet propulsion laboratory / california institute of technology , funded by the national aeronautics and space administration . the author thank the anonymous referee for her / his constructive comments and suggestions . b. , mcintosh d. h. , barden m. , bell e. f. , rix h .- w . , borch a. , beckwith s. v. w. , caldwell j. a. r. , heymans c. , jahnke k. , jogee s. , koposov s. e. , meisenheimer k. , snchez s. f. , somerville r. s. , wisotzki l. , wolf c. , 2007 , apjs , 172 , 615 m. j. , lewis j. , hodgkin s. , bunclark p. , evans d. , mcmahon r. , emerson j. p. , stewart m. , beard s. , 2004 , in quinn p. j. , bridger a. , eds , optimizing scientific return for astronomy through information technologies vol . 5493 of society of photo - optical instrumentation engineers ( spie ) conference series , vista data flow system : pipeline processing for wfcam and vista . pp 411422 j. c. , gil de paz a. , zamorano j. , boissier s. , dale d. a. , prez - gonzlez p. g. , gallego j. , madore b. f. , bendo g. , boselli a. , buat v. , calzetti d. , moustakas j. , kennicutt jr . r. c. , 2009 , apj , 703 , 1569 |
measurements are presented of single and double - differential dijet cross sections in diffractive photoproduction based on a data sample with an integrated luminosity of . the events are of the type , where the hadronic system contains at least two jets and is separated by a large rapidity gap from the system , which consists of a leading proton or low - mass proton excitation . the dijet cross sections are compared with qcd calculations at next - to - leading order and with a monte carlo model based on leading order matrix elements with parton showers . the measured cross sections are smaller than those obtained from the next - to - leading order calculations by a factor of about . this suppression factor has no significant dependence on the fraction of the photon four - momentum entering the hard subprocess . ratios of the diffractive to the inclusive dijet cross sections are measured for the first time and are compared with monte carlo models . # 1#2#3#4#1 * # 2 * ( # 3 ) # 4 desy 10 - 043issn 0418 - 9833 + march 2010 * diffractive dijet photoproduction + in collisions at hera * h1 collaboration submitted to f.d . aaron , c. alexa , v. andreev , s. backovic , a. baghdasaryan , e. barrelet , w. bartel , k. begzsuren , a. belousov , j.c . bizot , v. boudry , i. bozovic - jelisavcic , j. bracinik , g. brandt , m. brinkmann , v. brisson , d. bruncko , a. bunyatyan , g. buschhorn , l. bystritskaya , a.j . campbell , k.b . cantun avila , k. cerny , v. cerny , v. chekelian , a. cholewa , j.g . contreras , j.a . coughlan , j. cvach , j.b . dainton , k. daum , m. dek , b. delcourt , j. delvax , e.a . de wolf , c. diaconu , m. dobre , v. dodonov , a. dossanov , a. dubak , g. eckerlin , v. efremenko , s. egli , a. eliseev , e. elsen , a. falkiewicz , l. favart , a. fedotov , r. felst , j. feltesse , j. ferencei , d .- j . fischer , m. fleischer , a. fomenko , e. gabathuler , j. gayler , s. ghazaryan , a. glazov , l. goerlich , n. gogitidze , m. gouzevitch , c. grab , a. grebenyuk , t. greenshaw , b.r . grell , g. grindhammer , s. habib , d. haidt , c. helebrant , r.c.w . henderson , e. hennekemper , h. henschel , m. herbst , g. herrera , m. hildebrandt , k.h . hiller , d. hoffmann , r. horisberger , t. hreus , f. huber , m. jacquet , x. janssen , l. jnsson , a.w . jung , h. jung , m. kapichine , j. katzy , i.r . kenyon , c. kiesling , m. klein , c. kleinwort , t. kluge , a. knutsson , r. kogler , p. kostka , m. kraemer , j. kretzschmar , a. kropivnitskaya , k. krger , k. kutak , m.p.j . landon , w. lange , g. latovika - medin , p. laycock , a. lebedev , v. lendermann , s. levonian , k. lipka , b. list , j. list , n. loktionova , r. lopez - fernandez , v. lubimov , a. makankine , e. malinovski , p. marage , ll . marti , h .- u . martyn , s.j . maxfield , a. mehta , a.b . meyer , h. meyer , j. meyer , s. mikocki , i. milcewicz - mika , f. moreau , a. morozov , j.v . morris , m.u . mozer , m. mudrinic , k. mller , th . naumann , p.r . newman , c. niebuhr , a. nikiforov , d. nikitin , g. nowak , k. nowak , j.e . olsson , s. osman , d. ozerov , p. pahl , v. palichik , i. panagoulias , m. pandurovic , th . papadopoulou , c. pascaud , g.d . patel , e. perez , a. petrukhin , i. picuric , s. piec , h. pirumov , d. pitzl , r. plaakyt , b. pokorny , r. polifka , b. povh , v. radescu , n. raicevic , a. raspiareza , t. ravdandorj , p. reimer , e. rizvi , p. robmann , r. roosen , a. rostovtsev , m. rotaru , j.e . ruiz tabasco , s. rusakov , d. lek , d.p.c . sankey , m. sauter , e. sauvan , s. schmitt , l. schoeffel , a. schning , h .- c . schultz - coulon , f. sefkow , r.n . shaw - west , l.n . shtarkov , s. shushkevich , t. sloan , i. smiljanic , y. soloviev , p. sopicki , d. south , v. spaskov , a. specka , z. staykova , m. steder , b. stella , g. stoicea , u. straumann , d. sunar , t. sykora , g. thompson , p.d . thompson , t. toll , t.h . tran , d. traynor , p. trul , i. tsakov , b. tseepeldorj , j. turnau , k. urban , a. valkrov , c. valle , p. van mechelen , a. vargas trevino , y. vazdik , m. von den driesch , d. wegener , e. wnsch , j. ek , j. zlek , z. zhang , a. zhokin , h. zohrabyan , and f. zomer | 1006.0946 | quantum chromodynamics ( qcd ) , as the gauge field theory of the strong interaction , reliably predicts scattering cross sections involving short distance partonic interactions . however , the vast majority of hadron - hadron scatterings take place through long - distance strong interactions , where no hard scales are present and perturbative qcd calculations are not possible . prominent among these soft interactions are diffractive processes , in which the interacting hadrons remain intact or dissociate into low mass hadronic systems via an exchange which has vacuum quantum numbers , often referred to as a ` pomeron ' @xcite . following the observation of diffractive @xmath63 collisions in which a hard scale is provided by high transverse momentum jets @xcite , it has become possible to describe some classes of diffractive processes in terms of partonic interactions @xcite . more recently , diffractive deep - inelastic scattering ( ddis ) processes at hera @xcite , of the type @xmath64 , have been studied in detail and have led to a new level of understanding of the properties and structure of the diffractive exchange . developing this microscopic description of diffraction in terms of qcd and parton dynamics is a step towards a more complete understanding of the strong interaction . in the framework of a collinear factorisation theorem @xcite for hard scattering in semi - inclusive processes such as ddis , diffractive parton distribution functions ( dpdfs ) may be defined . the dpdfs have similar properties to the standard parton distribution functions ( pdfs ) of the proton , but with the constraint that there be a leading proton present in the final state . this condition may be satisfied equivalently by the experimental signatures of either a leading proton @xcite or the presence of a large gap in the rapidity distribution of final state hadrons , separating an unobserved outgoing proton from the remainder of the hadronic final state @xcite . in various extractions using recent ddis data @xcite , the dpdfs have been found to be dominated by gluons . to good approximation they exhibit a ` proton vertex factorisation ' property , whereby they vary only in normalisation with the four - momentum of the final state proton , the normalisation being well modelled using regge phenomenology @xcite . given a knowledge of the dpdfs , perturbative qcd calculations are applicable to other ddis observables . such calculations have been successful in the prediction of jet @xcite and heavy quark @xcite production in ddis at hera . in both cases , next - to - leading order ( nlo ) qcd predictions using the dpdfs from @xcite describe the measured cross sections well . however , as has long been anticipated @xcite , dpdf - based predictions for hard diffractive processes such as dijet production in @xmath63 scattering fail by around an order of magnitude to describe the data @xcite . this factorisation breaking is generally attributed to absorptive corrections , corresponding to the destruction of the rapidity gap due to multiple interactions within a single event . such effects are possible in @xmath63 scattering , where a beam remnant is present , in contrast to the electron scattering case in ddis at hera . a diversity of models of the absorptive corrections has developed @xcite , several of which reproduce the approximate @xmath65 ` rapidity gap survival probability ' observed in single diffractive @xmath63 scattering . the issues of dpdf applicability and rapidity gap survival can be studied in @xmath6 collisions at hera in hard diffractive ` photoproduction ' , where the virtuality @xmath66 of the exchange photon is close to zero . under these circumstances , the photon can develop an effective partonic structure via @xmath67 fluctuations and further subsequent splittings @xcite . in a leading order picture , there are thus two classes of hard photoproduction : ` direct ' interactions , where the photon enters the hard scatter as a structureless object and ` resolved ' interactions , where the photon interacts via its partonic structure and only a fraction @xmath68 of its four - momentum participates in the hard subprocess . resolved photoproduction interactions can be further divided into a ` hadron - like ' contribution and an ` anomalous ' or ` point - like ' contribution , the latter arising from the inhomogeneous term in the dglap equation for the photon @xcite . interactions involving the hadron - like component resemble hadron - hadron scattering to a large extent and are therefore widely expected to exhibit gap destruction effects . the rapidity gap survival probability for these hadron - like processes has been estimated in a phenomenological model to be @xmath69 @xcite . the point - like contribution to photon structure is expected to be subject to smaller absorptive corrections than the hadron - like part @xcite . in a recent model @xcite a survival probability of around @xmath70 was obtained for diffractive dijet photoproduction , depending slightly on the jet transverse energies ( @xmath71 ) . previous h1 measurements of diffractive dijet photoproduction @xcite have found cross sections to be smaller than nlo theoretical predictions , suggesting rapidity gap survival probabilities of around @xmath72 with little dependence on @xmath68 . a recent zeus measurement at somewhat larger @xmath71 @xcite yielded a larger survival probability , compatible with unity . it has been proposed that this apparent discrepancy may be resolved if the rapidity gap survival probability depends on the scale of the hard interaction , an idea which is supported to some extent by data @xcite . neither h1 nor zeus data provide any evidence for the expected @xmath68 dependence of the rapidity gap survival probability . a measurement of the ratio of the diffractive to the inclusive dijet photoproduction cross sections was proposed in @xcite as a means of evaluating the gap survival probability . this ratio is expected to be relatively insensitive to the model of the photon parton densities and also offers cancellations of experimental systematics and higher order qcd corrections . a similar ratio was measured by the cdf collaboration @xcite as a means of extracting effective @xmath73 dpdfs for comparison with hera predictions and assessment of gap survival probabilities . this paper reports diffractive dijet photoproduction cross section measurements based on a positron - proton scattering data sample with luminosity about a factor three larger than that previously published by h1 @xcite . the larger sample makes double - differential measurements possible , giving greater detail on the dynamics of gap survival and allowing studies of the correlations between the kinematic variables . the hypothesis of an @xmath71 dependent rapidity gap survival probability is tested . the ratio of the diffractive to the inclusive dijet photoproduction cross sections is also extracted for the first time . figures [ fig : dirresfeyn ] ( a ) and ( b ) show leading order examples of direct and resolved diffractive dijet production . denoting the four - vectors of the incoming positron , the incoming proton and the exchanged photon as @xmath74 , @xmath75 and @xmath76 , respectively , the standard dis kinematics can be described in terms of the invariants @xmath77 here , @xmath78 is the square of the total centre of mass energy of the collision , @xmath79 is the photon virtuality , @xmath80 is the scattered positron inelasticity and @xmath81 is the fraction of the proton four - momentum carried by the quark coupling to the photon . these variables are related through @xmath82 and to the invariant mass @xmath83 of the photon - proton system by @xmath84 defining @xmath85 and @xmath86 to be the four - vectors of the two distinct final state systems , where @xmath3 may be either a proton or a low mass proton excitation , the diffractive kinematics are described by the variables @xmath87 here , @xmath88 and @xmath89 are the invariant masses of the systems @xmath2 and @xmath3 , @xmath90 is the squared four - momentum transfer at the proton vertex and @xmath91 is the fraction of the longitudinal momentum of the proton transferred to the system @xmath2 . with @xmath92 and @xmath93 being the four - momenta of the particles entering the hard scatter from the photon and proton , respectively ( @xmath94 in the direct photon case ) , the invariant mass @xmath95 of the dijet system and the fractional photon ( @xmath5 ) and pomeron ( @xmath96 ) longitudinal momenta entering the hard subprocess can be expressed as @xmath97 dijet electroproduction cross sections for @xmath98 can be calculated in a fixed order qcd approach , assuming qcd collinear factorisation and neglecting any rapidity gap destruction effects , as a convolution of partonic cross sections , photon pdfs and dpdfs , according to @xmath99 here , the hadronic system @xmath100 corresponds to the remainder of the system @xmath2 after removing the two jets . the sum runs over all partons @xmath101 and @xmath102 that contribute , @xmath103 is the equivalent flux of photons that emerge from the incoming lepton @xcite and @xmath104 are the photon pdfs ( @xmath105 in the direct photon case ) . the hard partonic cross sections are denoted @xmath106 , @xmath107 are the dpdfs of the proton and @xmath108 is the factorisation scale . in this analysis , the grv ho @xcite parton densities are used to describe the structure of the resolved photon . the h1 2006 fit b set is used for the dpdfs , obtained by the h1 collaboration in fits to inclusive ddis data @xcite . in the poorly constrained large @xmath96 region , previous ddis final state data @xcite have shown a clear preference for these dpdfs over the fit a set from @xcite . further dpdf sets from h1 ( h1 2007 fit jets ) @xcite and zeus ( zeus dpdf sj ) @xcite in which dijet data from ddis are used to improve the sensitivity to the large @xmath96 region , are also considered . all of these dpdf sets assume proton vertex factorisation , such that @xmath109 in the interpretation illustrated in figure [ fig : dirresfeyn ] , @xmath110 may be considered as a pomeron flux , parameterised using regge phenomenology in the dpdf sets used here . the @xmath111 factor then represents the parton densities of the pomeron . at relatively large values of @xmath112 a small contribution from a sub - leading meson exchange is required in the dpdf fits . this is taken into account by adding a second term of the same form as equation [ eq : regge factorization ] , but with different parton densities and a flux factor which is suppressed as @xmath113 . the dijet electroproduction cross sections ( equation [ eq : factorised cross section ] ) are calculated at nlo of qcd using the program of frixione _ et al_. ( fr ) @xcite adapted for diffractive applications as described in @xcite . in the fr program , the renormalisation and factorisation scales are set to be equal and both are taken here from the leading jet transverse energy , i.e @xmath114 . the nlo calculations are performed with the number of flavours fixed to 5 and the qcd scale parameter set to @xmath115 , corresponding to a 2-loop @xmath116 . the sensitivity of the calculated cross sections to the chosen @xmath117 and @xmath108 values is studied by varying both scales simultaneously by factors of @xmath72 and @xmath118 . the nlo calculations were cross - checked with the program written by klasen and kramer @xcite , which yields consistent results @xcite . a monte carlo ( mc ) simulation is used to correct the data for detector effects in obtaining cross sections at the level of stable hadrons . all mc samples are passed through a detailed simulation of the h1 detector response based on the geant program @xcite and are subjected to the same reconstruction and analysis algorithms as are used for the data . diffractive dijet photoproduction events are generated using the rapgap mc generator @xcite in the range @xmath119 gev@xmath17 with the minimum transverse momentum of the partons entering the hard subprocess set to @xmath120 . rapgap is based on leading order qcd matrix elements with dglap parton showers . the h1 2006 fit b dpdfs and the grv - g lo photon parton densities @xcite are used at a factorisation scale given by @xmath121 . a sub - leading meson exchange is included in the dpdf simulation , though its contribution is smaller than @xmath122 in the kinematic range covered here . the selection of diffractive events using the rapidity gap method ( section [ sec : selection ] ) yields a sample which is dominated by elastically scattered protons , but which also contains an admixture of events in which the proton dissociates to low @xmath123 states . the measurement is corrected to the region @xmath124 and @xmath125 ( section [ sec : xsec ] ) , using mc samples generated using the diffvm @xcite program , with and without proton dissociation , following the method described in @xcite . in order to estimate the small ( @xmath126 ) contribution to the data sample from non - diffractive events which pass the diffractive selection , the pythia mc generator @xcite is used in photoproduction mode ( @xmath127 ) . the pythia model is also used to correct the inclusive photoproduction dijet measurement for detector effects when extracting the ratio of diffractive to inclusive cross sections . to estimate the model uncertainties on the corrections to the inclusive dijet cross section , a further mc sample is obtained using the herwig generator @xcite . multiple interaction models are included in both pythia and herwig as described in section [ sec : mchad ] . with these settings , pythia provides a good description of the shapes of the uncorrected inclusive dijet distributions , with a normalisation slightly larger than that of the data . the herwig mc underestimates the normalisation of the cross section by about a factor of two , but gives an acceptable description of the shapes of the measured distributions . for comparison with the diffractive measurements , it is necessary to convert the calculated nlo parton - level cross sections to the level of stable hadrons by evaluating effects due to initial and final state parton showering , fragmentation , hadronisation and the influence of beam remnants . the rapgap mc is used to compute the required ` hadronisation correction ' factors to the diffractive dijet calculations . these factors are defined for each measured data point by @xmath128 they reduce the predicted cross sections by typically @xmath129 and are given for each data point in tables [ table1d ] and [ table2d ] . the shape of the @xmath5 distribution is most strongly affected by the hadronisation corrections , the main effect being the migration of some direct photon interactions , for which the cross section is large , to lower @xmath5 values , substantially increasing the prediction in the interval @xmath130 . resolved photon interactions in inclusive dijet production @xcite are poorly described by nlo calculations unless a model of multiple interactions ( mi ) within a single event is included in addition to the low transverse momentum beam remnant , qcd radiation and hadronisation contributions which are simulated in standard mc models . the pythia mc generator is used to investigate the influence of mis on inclusive dijet cross sections and hence on the diffractive - to - inclusive ratios . several tunes of the pythia model for multiple hard parton - parton scatterings are available . in the version used here ( tune a in @xcite ) , both the proton and the resolved photon have double - gaussian matter distributions . the minimum transverse momentum down to which secondary partonic scatterings are calculated depends on the impact parameter of the collision and is governed by a regularisation scale @xmath131 @xcite . multiple hard partonic interactions are not simulated in the herwig mc . instead , an empirically motivated soft multiple interaction model @xcite is used . the probability of such activity in a given event is set to @xmath132 . a detailed description of the h1 detector can be found elsewhere @xcite . here , a brief account is given of the detector components most relevant to the present analysis . the h1 coordinate system is defined such that the origin is at the nominal @xmath6 interaction point and the polar angle @xmath133 and the positive @xmath134 axis correspond to the direction of the outgoing proton beam . the region @xmath135 , which has positive pseudorapidity @xmath136 , is referred to as the ` forward ' hemisphere . the _ ep _ interaction point in h1 is surrounded by a central tracking region , which includes silicon strip detectors as well as two large concentric drift chambers . these chambers cover a pseudorapidity region of @xmath137 and have a transverse momentum resolution of @xmath138 . they also provide triggering information . the central tracking detectors are surrounded by a finely segmented liquid argon ( lar ) sampling calorimeter covering @xmath139 . its resolution is @xmath140 for electrons and photons and @xmath141 for hadrons , as measured in test beams @xcite . the central tracker and lar calorimeter are placed inside a large superconducting solenoid , which produces a uniform magnetic field of @xmath142 t. the backward region @xmath143 is covered by a lead - scintillating fibre calorimeter ( spacal ) with electromagnetic and hadronic sections . information from the central tracker and the lar and spacal calorimeters is combined using an energy flow algorithm to obtain the hadronic final state ( hfs ) @xcite . the hadronic energy scale is known to @xmath144 for this analysis @xcite . photoproduction events are selected by tagging positrons scattered through very small angles , corresponding to quasi - real photon emission , using a crystal erenkov calorimeter at @xmath145 ( electron tagger ) . the luminosity is measured via the bethe - heitler bremsstrahlung process @xmath146 , the final state photon being detected in another crystal calorimeter at @xmath147 . a set of drift chambers around @xmath148 comprise the forward muon detector ( fmd ) . the proton remnant tagger ( prt ) is a set of scintillators surrounding the beam pipe at @xmath149 . these detectors , used together with the most forward part of the lar , are efficient in the identification of very forward energy flow and are used to select events with large rapidity gaps near to the outgoing proton direction . the analysis is based on a sample of integrated luminosity @xmath0 , collected by h1 in 1999 and 2000 with proton and positron beam energies of @xmath150 and @xmath151 , respectively . the events are triggered on the basis of a scattered positron signal in the electron tagger and at least three high transverse momentum tracks in the drift chambers of the central tracker . the event inelasticity @xmath80 and hence the invariant mass @xmath83 of the photon - proton system are reconstructed using the scattered positron energy @xmath152 measured in the electron tagger according to @xmath153 where @xmath154 is the positron beam energy . the geometric acceptance of the electron tagger limits the measurement to @xmath155 and intermediate values of @xmath80 . the reconstructed hadronic final state objects ( section [ sec : h1 detector ] ) are subjected to the @xmath156 longitudinally invariant jet algorithm @xcite , applied in the laboratory frame with parameters @xmath157 and @xmath158 . to facilitate comparisons with the nlo calculations , different cuts are placed on the transverse energies @xmath121 and @xmath159 of the leading and next - to - leading jets , respectively . as well as these variables , the jet properties are studied in terms of the variables @xmath160 obtained from the laboratory frame pseudorapidities of the jet axes . with @xmath161 and @xmath162 denoting the four - momenta of the two jets , hadron level estimators of the dijet invariant mass and of @xmath68 are obtained from @xmath163 where the sums labelled ` _ hfs _ ' and ` _ jets _ ' run over all hadronic final state objects and those included in the jets , respectively . the diffractive event selection is based on the presence of a large forward rapidity gap . the pseudorapidity of the most forward cluster in the lar calorimeter with energy above @xmath164 mev is required to satisfy @xmath165 . the activity in the prt and the fmd is required not to exceed that typical of noise levels as obtained from randomly triggered events . these requirements ensure that the analysed sample is dominated by elastically scattered protons at small @xmath166 , with a small admixture of events with leading neutrons and low @xmath89 baryon excitations , collectively referred to here as ` proton dissociation ' contributions . the diffractive kinematics are reconstructed using @xmath167 where @xmath168 is the proton beam energy . a cut on @xmath91 is applied to ensure good containment of the system @xmath2 and to suppress sub - leading exchange contributions . a hadron level estimator for the momentum fraction @xmath96 is obtained using @xmath169 the kinematic range in which the diffractive dijet measurement is performed is specified in table [ tab : kinrange ] . the inclusive measurement phase space is defined by these conditions , with the requirements relaxed on the diffractive variables @xmath91 , @xmath170 , @xmath89 and @xmath90 . except where measurements are made explicitly as a function of @xmath171 , the @xmath172 region is excluded from the diffractive analysis . this improves the reliability of the comparison between data and theoretical predictions , since the dpdf sets used are not valid at the largest @xmath96 values . after applying all selection criteria , about @xmath173 out of roughly @xmath174 inclusive dijet photoproduction events are used in the diffractive analysis . a more detailed description of the analysis can be found in @xcite . @xmath175 & & @xmath176 + @xmath177 & & @xmath178 + @xmath179 & & @xmath180 + + @xmath181 & & @xmath182 + @xmath183 & & @xmath184 + the diffractive differential cross section is measured in each bin @xmath101 of a variable @xmath81 using the formula @xmath185 here , @xmath186 is the raw number of reconstructed events passing the selection criteria listed in section [ sec : selection ] and @xmath187 is the trigger efficiency , obtained by reference to an independently triggered sample and parameterised as a function of the multiplicity of charged particle tracks . the trigger efficiency averaged over the full measurement range is @xmath188 . the non - diffractive background contribution obtained from the pythia mc simulation is denoted @xmath189 and does not extend beyond the few percent level for any of the measured data points . the factor @xmath190 corrects the measurement for detector effects , including migrations between bins , to the level of stable hadrons . it is calculated from the rapgap mc and has an average value of @xmath191 for the diffractive analysis , most of the losses being due to the limited electron tagger acceptance of @xmath192 integrated over the measured @xmath80 range . the bin width is denoted @xmath193 , @xmath194 is the luminosity of the data sample and @xmath195 , evaluated using the diffvm mc , corrects the measurement to the chosen range of @xmath123 and @xmath90 ( table [ tab : kinrange ] ) . in the inclusive analysis the cross section is obtained analogously to equation [ xsecformula ] except for the @xmath196 and @xmath197 terms , which are not relevant . uncertainties are evaluated for all significant sources of possible systematic bias . these sources are summarised for the diffractive analysis below , together with their corresponding influences on the total diffractive cross section . [ [ energy - scale ] ] energy scale : + + + + + + + + + + + + + the energy scale of the hfs measurement is tested using the momentum balance constraint between the precisely reconstructed positron and the hfs in neutral current dis events . dedicated data and mc samples are analysed and found to agree to better than @xmath144 . the effect of a relative @xmath144 change in the energy of the hfs between the data and the mc is a @xmath198 shift to the total diffractive dijet cross section . this arises mainly from changes in the migration corrections across the minimum @xmath71 values and the maximum @xmath91 value of the measurement . the energy scale uncertainties are thus highly correlated between the bins of the differential cross section measurements . [ [ large - rapidity - gap - selection ] ] large rapidity gap selection : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + a fraction of the events in the kinematic range of the analysis ( table [ tab : kinrange ] ) give rise to hadronic activity in the forward detectors or at pseudorapidities beyond those allowed by the @xmath199 cut in the lar calorimeter . corrections for this inefficiency of the large rapidity gap selection are made using the rapgap mc simulation . the uncertainties in the correction factors are assessed through a study of forward energy flow in a sample of dijet photoproduction events with leading protons tagged in the h1 forward proton spectrometer @xcite . rapgap is found to describe these migrations to within 10% @xcite , which translates into a @xmath200 uncertainty on the measured total cross section and uncertainties which are correlated between bins of the differential distributions . [ [ proton - dissociation ] ] proton dissociation : + + + + + + + + + + + + + + + + + + + + the model dependence uncertainty on the proton dissociation correction factor ( @xmath197 in equation [ xsecformula ] ) is obtained by varying the elastic and proton dissociation cross sections and the proton dissociation @xmath89 and @xmath90 dependences in the diffvm mc samples , following @xcite . the largest effect arises from varying the ratio of the proton - elastic to the proton - dissociative cross sections between @xmath72 and @xmath118 . the resulting uncertainty on the measured cross section is @xmath201 . [ [ model - dependence ] ] model dependence : + + + + + + + + + + + + + + + + + the influence of the model assumptions on the acceptance and bin migration corrections ( @xmath202 in equation [ xsecformula ] ) , is determined in the diffractive analysis by varying the kinematic distributions in the rapgap simulation within the limits allowed by maintaining an acceptable description of the uncorrected data . the following variations are implemented by reweighting each mc event according to the value of generator level kinematic variables , leading to the quoted systematic uncertainties on the total cross section . * the @xmath91 distribution is reweighted by @xmath203 , leading to a @xmath204 uncertainty . * the @xmath171 distribution is reweighted by @xmath205 , leading to a @xmath206 uncertainty . * the @xmath207 distribution is reweighted by @xmath208 , leading to a @xmath209 uncertainty . * the @xmath121 distribution is reweighted by @xmath210 , leading to a @xmath211 uncertainty . * the @xmath90 distribution is reweighted by @xmath212 , leading to a @xmath213 uncertainty . * the @xmath80 distribution is reweighted by @xmath214 , leading to a @xmath215 uncertainty . [ [ electron - tagger - acceptance ] ] electron tagger acceptance : + + + + + + + + + + + + + + + + + + + + + + + + + + + a dedicated procedure external to this analysis is used to obtain the electron tagger acceptance @xcite . the integrated acceptance over the full @xmath80 range is known to @xmath216 , which affects the cross section normalisation . [ [ trigger - efficiency ] ] trigger efficiency : + + + + + + + + + + + + + + + + + + + the procedure for parameterising the trigger efficiency ( @xmath187 in equation [ xsecformula ] ) leads to a @xmath122 uncertainty . this covers the observed deviations of the parameterisation from the measured efficiencies as a function of all variables relevant to the analysis . this uncertainty is treated as being uncorrelated between data points . [ [ luminosity ] ] luminosity : + + + + + + + + + + + the measurement of the integrated luminosity has an uncertainty of @xmath217 . this translates directly into a @xmath217 normalisation uncertainty on the measured cross sections . [ [ non - diffractive - background ] ] non - diffractive background : + + + + + + + + + + + + + + + + + + + + + + + + + + + a @xmath218 normalisation variation is applied to the non - diffractive background contribution given by the pythia mc model ( @xmath196 in equation [ xsecformula ] ) . the effect of this change is correlated between the data points and leads to a @xmath219 uncertainty on the total cross section . [ [ forward - detector - noise ] ] forward detector noise : + + + + + + + + + + + + + + + + + + + + + + + fluctuations in the fmd noise , leading to losses in the large rapidity gap event selection , are evaluated for each run using randomly triggered events . the standard deviation in the run - by - run distribution of the correction factors is used to derive a @xmath220 normalisation uncertainty on the measured cross sections . noise in the prt detector is negligible . a similar procedure is followed to evaluate the systematic uncertainties in the inclusive dijet analysis . the uncertainties associated with the large rapidity gap selection and the model dependence are no longer relevant . instead , comparisons between the pythia and herwig mcs are used determine a @xmath221 model dependence uncertainty on the acceptance correction when integrated over the full phase space studied . the inclusive cross section systematics are dominated by a contribution at the @xmath222 level from the hfs energy scale uncertainty . however , when forming the ratio of diffractive to inclusive cross sections , this error source cancels to good approximation , the residual uncertainty being less than @xmath223 . the largest remaining contribution to the systematic uncertainty on the cross section ratio arises from the model dependence . the total systematic uncertainty on each data point is formed by adding the individual contributions in quadrature . in the figures and tables that follow , the systematic uncertainties are separated into two categories : those which are uncorrelated between data points ( the model dependence and trigger efficiency ) and those which lead to correlations between data points ( all other sources ) . cross sections are measured integrated over the full kinematic range specified in table [ tab : kinrange ] and also single- and double - differentially as a function of a variety of variables which are sensitive to the overall event structure , the hard subprocess and the presence of remnants of the virtual photon and the diffractive exchange . the measured differential cross sections , which correspond to averages over the specified measurement intervals , are given numerically in tables [ table1d ] and [ table2d ] , where the experimental uncertainties and hadronisation corrections applied to the nlo calculations are also listed . tables [ table1dratios ] and [ table2dratios ] contain the ratios of the measurements to the nlo calculations , obtained using the fr framework ( section [ nlo ] ) and the h1 2006 fit b dpdfs ( referred to in the following as fr fit b ) . the total diffractive dijet positron - proton cross section integrated over the full measured kinematic range is @xmath224 the ratio of this result to the corresponding fr fit b nlo prediction is @xmath225 where the statistical and systematic uncertainties originate from the measurement . the scale uncertainty corresponds to the effect of simultaneously varying the renormalisation and factorisation scales from their central values , @xmath114 , by a factor of two in either direction . this large ( @xmath226 ) scale uncertainty arises due to the relatively low @xmath121 range of this analysis , and is the limiting factor in the comparison between data and theory . the dpdf uncertainty is obtained using the method of @xcite , by propagating the eigenvector decomposition of the fit uncertainties . if the h1 2006 fit b dpdfs are replaced by the h1 2007 fit jets dpdfs , the result is @xmath227 , which is inside the quoted dpdf uncertainty . using zeus dpdf sj , a compatible result of @xmath228 is obtained . adding all uncertainties in quadrature , the ratio result in equation [ sigmatot : ratio ] implies at the @xmath229 level that the nlo qcd calculation , neglecting any gap destruction effects , yields a larger diffractive dijet photoproduction cross section than that measured . it confirms the result of a previous h1 analysis in a very similar kinematic range @xcite and is broadly as expected from theoretical calculations of rapidity gap survival probabilities @xcite . figure [ fig : sigma single ] shows the diffractive dijet cross section measured single - differentially in @xmath207 , @xmath121 , @xmath230 , @xmath171 , @xmath231 , @xmath232 , @xmath83 , @xmath95 and @xmath233 , in the phase space defined in table [ tab : kinrange ] . in figure [ fig : sigma single](d ) , the @xmath234 requirement is relaxed and the cross section measured for the region @xmath172 is shown without theoretical comparisons , since the dpdfs are not defined ( see section [ sec : selection ] ) . to allow a more detailed shape comparison between the data and the predictions , ratios of the measured differential cross sections to the fr fit b calculations are plotted in figure [ fig : sigma single ratio ] . these ratios may be taken as measurements of the dependence of the rapidity gap survival probability on the kinematic variables . figures [ fig : sigma single ] and [ fig : sigma single ratio ] show that the suppression by around a factor of @xmath4 of the data with respect to the fr fit b nlo calculations has at most a weak dependence on the kinematic variables . notably , within the uncertainties there is no dependence on @xmath207 ( figure [ fig : sigma single ratio](a ) ) , in contrast to theoretical predictions for the rapidity gap survival probability @xcite . the largest dependence of the central values of the measured ratios on any of the variables appears in the cross section differential in @xmath121 ( figure [ fig : sigma single ratio](b ) ) . although not well established by the current data , this dependence is compatible with previous data @xcite . the @xmath121 dependence is investigated further in section [ double : diff ] . the measured cross sections in figure [ fig : sigma single ] are also compared with a prediction obtained using the rapgap mc generator ( section [ sec : mc ] ) , which does not contain any model of rapidity gap destruction . the shapes of the measured cross sections are well described and the normalisation is only slightly lower than that of the data . however the scale uncertainty in this model is rather large and the same model undershoots diffractive dijet measurements in dis @xcite , where factorisation is expected to hold . in @xcite , the zeus collaboration presented an analysis of diffractive dijet photoproduction data with @xmath235 , which is most readily compared with the second and third @xmath121 intervals in figures [ fig : sigma single](b ) and [ fig : sigma single ratio](b ) . however , even for @xmath235 , a direct comparison between h1 and zeus data is not possible , since the zeus analysis covers a wider @xmath80 range and cuts on the second jet at an @xmath159 value of @xmath236 , larger than the value used here . an indirect comparison can be made on the basis of ratios of the data to nlo theoretical calculations using the h1 fit b dpdfs . zeus obtains a result of around @xmath237 for this ratio , which is compatible with the result for @xmath235 obtained here , within the large combined uncertainties . range are well within the experimental uncertainties alone . ] as discussed in detail in @xcite , the error bands on the dpdfs extracted from inclusive diffraction alone do not include uncertainties due to parton parameterisation choices and thus do not reflect the full uncertainties , particularly in the large @xmath171 region . to give a complementary indication of the possible range of variation , comparisons between the ratios obtained with the h1 2006 fit b dpdfs , the h1 2007 fit jets dpdfs and zeus dpdf sj fit are shown for a subset of variables ( @xmath207 , @xmath121 and @xmath171 ) in figure [ fig : ratiodpdfs ] . the zeus dpdfs lead to ratios which are uniformly @xmath238 larger than those obtained with h1 2006 fit b , with no strong dependence on any of the kinematic variables . the deviation of the h1 2007 fit jets result from the h1 2006 fit b result extends beyond the dpdf error band for @xmath239 , which is correlated with a somewhat stronger dependence of the ratio of data to theory on @xmath121 and a slightly different shape at low @xmath207 . according to the rapgap model , approximately half of the cross section in the kinematic range studied arises from each of the direct and resolved photon - induced contributions . the decomposition of photoproduction processes into direct and resolved interactions is not uniquely defined beyond lo . when modelling rapidity gap survival probabilities in the following , the resolved photon contribution is defined to correspond exactly to that which is calculated using the photon structure function . following the calculation using an absorptive model of a gap survival probability of @xmath69 for the hadron - like component of resolved photoproduction @xcite , previous h1 data @xcite were compared in @xcite with predictions in which the full resolved photon contribution was suppressed by this factor , the direct photon contribution being left unsuppressed . in a later analysis @xcite , this procedure was extended to nlo . the conclusions of these previous studies are confirmed in figures [ fig : sigma single res supp ] and [ fig : sigma single ratio res supp ] through a similar comparison of the current data with nlo calculations in which the resolved photon contribution is globally suppressed by a factor of 0.34 . the overall normalisation of this calculation is in good agreement with the data . however , the shapes of some of the differential distributions are not well reproduced . in particular , there is a variation by more than a factor of two in the ratio of data to theory as a function of @xmath207 ( figure [ fig : sigma single ratio res supp]a ) . the distinction between point - like and hadron - like resolved photon interactions recently developed in @xcite leads to a significantly weaker predicted suppression in the kinematic range of the current analysis . the data are compared with this refined ` kkmr ' model under the approximation of completely neglecting hadron - like resolved photon contributions , which , according to the authors , become dominant only for @xmath240 @xcite , beyond the range of the current analysis . the rapidity gap survival probabilities obtained in @xcite for point - like photon interactions using the grv ho photon pdfs are applied to all resolved photon interactions . interactions involving quarks and gluons from the photon are thus suppressed by factors of @xmath241 and @xmath242 , respectively . the quark - initiated contribution is dominant throughout the measured range , such that the rapidity gap survival probability in the model is approximately @xmath243 for resolved photon interactions and @xmath244 for direct photon interactions . figure [ fig : sigma single res supp ] shows the comparison between the measured single differential cross sections and the nlo qcd predictions , with the resolved photon contribution scaled according to the kkmr model . the corresponding ratios of data to theoretical predictions are shown in figure [ fig : sigma single ratio res supp ] . the overall normalisation of the kkmr - based calculation is larger than that of the data , but is compatible within the large uncertainties . many of the distributions studied are well described in shape ( @xmath121 , @xmath171 , @xmath232 , @xmath83 and @xmath95 ) . the data thus agree with the prediction @xcite that the @xmath121 dependence of the data / theory ratio flattens if the resolved photon contribution alone is suppressed . however , there remains a variation in the ratio of data to the kkmr model with @xmath207 and to a lesser extent with @xmath231 , @xmath91 and @xmath233 . a comparison of figures [ fig : sigma single ratio ] and [ fig : sigma single ratio res supp ] shows that the shapes of the differential cross sections are generally better described with a global suppression factor than with a survival probability applied to resolved photon interactions only . to study further the dynamics of rapidity gap suppression and their dependence on the nature of the photon interaction , cross sections are measured double differentially in two regions of @xmath207 , which are enriched with either resolved ( @xmath245 ) or direct ( @xmath246 ) photon processes . using the rapgap mc model with the grv - g lo photon pdfs , the @xmath245 region is estimated to contain @xmath247 resolved photon interactions integrated over the measurement region ( table [ tab : kinrange ] ) , with a @xmath248 direct photon contribution for @xmath246 . in figures [ fig : sigma et xg 1](a)-(c ) , measurements are presented of the double - differential dijet cross section @xmath249 for three @xmath121 ranges in the resolved and direct photon - enriched @xmath207 intervals . the data are compared with the fr fit b calculations and with the rapgap mc predictions . due to kinematic constraints , the resolved - enriched cross section at low @xmath207 falls most rapidly as @xmath121 increases . there is a suggestion that this @xmath121 dependence in the resolved - enriched region is stronger for the nlo qcd theory than for the data . in the direct - enriched high @xmath207 region , the cross section falls more slowly with @xmath121 and the dependence in the data is similar to that predicted by the nlo calculation . these features are illustrated further in figures [ fig : sigma et xg 1 ] ( d)-(e ) , where the ratios of the data to the nlo theory from figures [ fig : sigma et xg 1 ] ( a)-(c ) are presented as a function of @xmath121 in the resolved and direct photon - enriched @xmath207 regions , respectively . the significance of the @xmath121 dependence in the resolved - enriched region ( figure [ fig : sigma et xg 1 ] ( d ) ) is evaluated through a @xmath250 test . all uncertainties are taken into account in this procedure , though the main contribution comes from the statistical and uncorrelated systematic uncertainties on the data , the remaining uncertainties changing only the normalisation of the ratio to first approximation . a test of the hypothesis that there is no @xmath121 dependence yields a @xmath250 value of 1.36 , with two degrees of freedom , corresponding to an @xmath121 variation at the 73% confidence level . the suppression of the data relative to the nlo prediction in the direct - enriched large @xmath207 region is , within errors , independent of @xmath121 ( figure [ fig : sigma et xg 1 ] ( e ) ) . figures [ fig : sigma et xg 1 ] ( d)-(e ) thus indicate that any @xmath121 dependence of the data - to - theory ratio in figure [ fig : sigma single ratio](b ) is driven primarily by resolved photon interactions . an @xmath121 dependence of the gap survival probability is predicted in the kkmr model , due to variations in the size of the @xmath251 dipole produced by the point - like photon splitting , and hence in the absorptive correction . however the predicted effect is small ( @xmath252 as @xmath121 changes from @xmath253 to @xmath254 ) . figures [ fig : sigma single ratio](d)-(e ) also indicate that when @xmath121 becomes large , the suppression in the direct region may be stronger than that in the resolved region , which is not expected in any model . the large uncertainties permit statistical fluctuations in the data or small inadequacies in the theory as possible explanations . in figure [ fig : sigma zp xg 1 ] , the cross section is shown double differentially in @xmath171 and @xmath207 . the measured cross section is compared with the nlo theory as a function of @xmath171 in two bins of @xmath207 in figures [ fig : sigma zp xg 1 ] ( a)-(b ) and the ratios of data to theoretical predictions are shown in figures [ fig : sigma zp xg 1 ] ( c)-(d ) . the nlo calculations describe the measured shapes rather well , with no evidence for any variation of the suppression factor between any of the measurement ranges . the gap survival probability in a region where there are small or no remnants of either the photon or the diffractive exchange ( highest @xmath171 bin in figure [ fig : sigma zp xg 1]d ) is thus similar to that where both remnants are significant ( lowest @xmath171 bin in figure [ fig : sigma zp xg 1]c ) . this remains the case when the h1 2007 fit jets dpdfs are used in place of h1 2006 fit b. in both figures [ fig : sigma et xg 1 ] and [ fig : sigma zp xg 1 ] , the rapgap mc prediction gives a satisfactory description of the shapes of the double differential cross sections , the normalisation being slightly lower than that of the data . measurements of ratios of diffractive to inclusive dijet photoproduction cross sections have been proposed @xcite as a further test of gap survival issues . their potential advantages over straight - forward diffractive measurements lie in the partial cancellations of some experimental systematics and of theoretical uncertainties due to the photon structure and factorisation and renormalisation scale choices . the sensitivity to absorptive effects of diffractive - to - inclusive ratios is thus potentially superior to that of pure diffractive cross sections . for the ratio extraction presented here , inclusive dijet cross sections are measured using data collected in the same period as the diffractive sample . the experimental method and systematic error treatment for the inclusive case is described in section [ sec : procedure ] . it is identical to the diffractive measurement method , with the exception of the large rapidity gap requirements . at the relatively low transverse energies studied in the present analysis , underlying event effects have a large influence on jet cross sections in inclusive photoproduction @xcite . here , the pythia and herwig mc models are used to correct the inclusive data for detector effects , with mi included as described in section [ sec : mchad ] . the two models agree rather well on the corrections to be applied to the data . the average of the results with the two models is therefore used to calculate the corrections and the uncertainty . the latter is taken from the difference between the results with the two models and is relatively small ( @xmath221 when integrated over the full measured range ) . the ratios of diffractive to inclusive single - differential dijet cross sections are given numerically in table [ tabledifftoincl ] and are shown in figure [ fig : r ] as a function of @xmath207 , @xmath121 , @xmath231 , @xmath232 , @xmath95 and @xmath83 . due to the partial or complete cancellations of some error sources when forming the ratio , the correlated uncertainties are reduced compared with those for the diffractive distributions . since they give adequate descriptions of the diffractive and inclusive data , respectively , the rapgap and pythia mc models are used to assess the relative sensitivity of the diffractive - to - inclusive ratio to the gap survival and mi effects . with no mi effects included in the pythia model , the description of the inclusive data is poor and the ratio of rapgap to pythia exceeds the data by a factor of around 1.5 . as expected , this factor becomes smaller as @xmath121 increases . however , the shape of the prediction also differs from that of the ratio data for most of the other variables studied , in particular @xmath207 . the inclusion of the pythia mi model changes the predicted inclusive cross sections , and hence the ratios , substantially . the ratio of rapgap to pythia then gives an improved description of the shapes of the distributions . the mi effects alter the predicted ratio by a factor of 0.5 at low @xmath207 , where the resolved photon remnant is most important . as expected , there is little effect in the direct photon - dominated large @xmath207 region . the normalisation of the ratio of the models when mi are included is smaller than that of the data . this partially reflects the rapgap description of the diffractive data ( figure [ fig : sigma single ] ) and is partially due to an overshoot in the pythia description of the inclusive data . the fractional reduction in the predicted inclusive cross section when mi are introduced in the pythia model is comparable to the magnitude of the gap survival suppression factor in the diffractive data ( section [ sec : difdata ] ) . the uncertainties in modelling the mi are large and difficult to quantify . the precision with which gap survival issues can be unfolded from mi complications in the ratio of diffractive to inclusive data is correspondingly poor . therefore no strong conclusions can be drawn with our current understanding of mi , despite the relatively good precision of the data . single and double - differential cross sections are measured for diffractive dijet photoproduction and are compared with predictions based on nlo qcd calculations using different sets of dpdfs . ratios of the measured to the predicted differential cross sections are also studied . the total diffractive dijet cross section is overestimated by the nlo qcd theory by about a factor of two , which is consistent with previous h1 measurements @xcite . the shapes of the single - differential cross sections are well described when the h1 2006 dpdf fit b partons are used . a good overall description of the differential cross sections is obtained by applying a global suppression factor of @xmath255 to the nlo calculations . as in similar previous analyses @xcite , there is a suggestion of a dependence of the rapidity gap survival probability on @xmath121 , though the significance of this effect is not large . if only the resolved photon contribution in the calculation is suppressed by a factor of @xmath69 , as predicted for hadron - like resolved photon interactions @xcite , the overall normalisation of the nlo qcd prediction agrees well with the data . however , the description of the @xmath256 distribution , which best distinguishes direct from resolved photon interactions , becomes poor . if rapidity gap survival probabilities expected for point - like resolved photons are applied instead @xcite , the overall normalisation is acceptable and the @xmath121 dependence of the data is better described . however , the description of the @xmath256 dependence remains problematic . the analysis of the double - differential cross section @xmath249 indicates that the @xmath121 dependence of the data / theory ratio originates from the resolved photon - enriched region of @xmath207 . however , the data are also consistent with no dependence on @xmath121 for either of the @xmath207 regions studied . the ratio of the data to the nlo theory for the double - differential cross section @xmath257 is constant within errors throughout the region studied , indicating that the gap survival probability is insensitive to the presence or nature of remnants of either the photon or the diffractive exchange . measurements of the ratio of diffractive to inclusive single - differential cross sections are presented as a function of several variables . the influence of multiple interaction effects in the inclusive data is large in the kinematic range studied here . the large uncertainties in modelling these multiple interactions preclude strong conclusions about rapidity gap survival on the basis of these data , although a reasonable description of the ratios can be obtained with suitably tuned monte carlo models . we are grateful to the hera machine group whose outstanding efforts have made this experiment possible . we thank the engineers and technicians for their work in constructing and maintaining the h1 detector , our funding agencies for financial support , the desy technical staff for continual assistance and the desy directorate for support and for the hospitality which they extend to the non - desy members of the collaboration . we express our gratitude to a. kaidalov , v. khoze , m. klasen , g. kramer , a. martin and m. ryskin for many helpful discussions . a. brandt _ et al . _ [ ua8 collaboration ] , phys . b * 297 * ( 1992 ) 417 . g. ingelman and p. schlein , phys . b * 152 * ( 1985 ) 256 ; + a. donnachie and p. landshoff , phys . b * 191 * ( 1987 ) 309 [ erratum - ibid . b * 198 * ( 1987 ) 590 ] . m. derrick _ et al . _ [ zeus collaboration ] , phys . b * 315 * ( 1993 ) 481 ; + t. ahmed _ et al . _ [ h1 collaboration ] , nucl . b * 429 * ( 1994 ) 477 . j. collins , phys . d * 57 * ( 1998 ) 3051 [ erratum - ibid . d * 61 * ( 2000 ) 019902 ] [ hep - ph/9709499 ] . a. aktas _ et al . _ [ h1 collaboration ] , eur . phys . j. c * 48 * ( 2006 ) 749 [ hep - ex/0606003 ] . s. chekanov _ et al . _ [ zeus collaboration ] , nucl . b * 816 * ( 2009 ) 1 [ arxiv:0812.2003 ] . a. aktas _ et al . _ [ h1 collaboration ] , eur . j. c * 48 * ( 2006 ) 715 [ hep - ex/0606004 ] . a. aktas _ et al . _ [ h1 collaboration ] , jhep * 0710 * ( 2007 ) 042 [ arxiv:0708.3217 ] . s. chekanov _ et al . _ [ zeus collaboration ] , nucl . b * 831 * ( 2010 ) 1 [ arxiv:0911.4119 ] . a. martin , m. ryskin and g. watt , eur . phys . j. c * 44 * ( 2005 ) 69 [ hep - ph/0504132 ] . see for example p. collins , `` an introduction to regge theory & high energy physics '' , cambridge university press , cambridge ( 1977 ) . c. adloff _ et al . _ [ h1 collaboration ] , eur . j. c * 6 * ( 1999 ) 421 [ hep - ex/9808013 ] . c. adloff _ et al . _ [ h1 collaboration ] , eur . j. c * 20 * ( 2001 ) 29 [ hep - ex/0012051 ] ; + s. chekanov _ et al . _ [ zeus collaboration ] , eur . j. c * 52 * ( 2007 ) 813 [ arxiv:0708.1415 ] . a. aktas _ et al . _ [ h1 collaboration ] , eur . j. c * 50 * ( 2007 ) 1 [ hep - ex/0610076 ] ; + s. chekanov _ et al . _ [ zeus collaboration ] , nucl . b * 672 * ( 2003 ) 3 [ hep - ex/0307068 ] . j. collins , l. frankfurt and m. strikman , phys . b * 307 * ( 1993 ) 161 [ hep - ph/9212212 ] ; + a. berera and d. soper , phys . rev . d * 50 * ( 1994 ) 4328 [ hep - ph/9403276 ] ; + a. berera and d. soper , phys . d * 53 * ( 1996 ) 6162 [ hep - ph/9509239 ] . a. affolder _ et al . _ [ cdf collaboration ] , phys . rev . * 84 * ( 2000 ) 5043 . m. klasen and g. kramer , phys . d * 80 * ( 2009 ) 074006 [ arxiv:0908.2531 ] . y. dokshitzer , v. khoze and t. sjstrand , phys . b * 274 * ( 1992 ) 116 ; + j. bjorken , phys . d * 47 * ( 1993 ) 101 ; + e. gotsman , e. levin and u. maor , phys . b * 309 * ( 1993 ) 199 [ hep - ph/9302248 ] ; + e. gotsman , e. levin and u. maor , phys . b * 438 * ( 1998 ) 229 [ hep - ph/9804404 ] ; b. cox , j. forshaw and l. lnnblad , jhep * 9910 * ( 1999 ) 023 [ hep - ph/9908464 ] ; + v. khoze , a. martin and m. ryskin , eur . j. c * 18 * ( 2000 ) 167 [ hep - ph/0007359 ] ; + a. kaidalov , v. khoze , a. martin and m. ryskin , eur . j. c * 21 * ( 2001 ) 521 [ hep - ph/0105145 ] . a. kaidalov , v. khoze , a. martin and m. ryskin , phys . b * 559 * ( 2003 ) 235 [ hep - ph/0302091 ] . t. bauer , r. spital , d. yennie and f. pipkin , rev . * 50 * ( 1978 ) 261 [ erratum - ibid . * 51 * ( 1979 ) 407 ] ; + j. butterworth and m. wing , rept . prog . * 68 * ( 2005 ) 2773 [ hep - ex/0509018 ] . e. witten , nucl . b * 120 * ( 1977 ) 189 . a. kaidalov , v. khoze , a. martin and m. ryskin , phys . lett . b * 567 * ( 2003 ) 61 [ hep - ph/0306134 ] . a. kaidalov , v. khoze , a. martin and m. ryskin , eur . j. c * 66 * ( 2010 ) 373 [ arxiv:0911.3716 ] . m. klasen and g. kramer , eur . j. c * 38 * ( 2004 ) 93 [ hep - ph/0408203 ] . a. aktas _ et al . _ [ h1 collaboration ] , eur . j. c * 51 * ( 2007 ) 549 [ hep - ex/0703022 ] . s. chekanov _ et al . _ [ zeus collaboration ] , eur . j. c * 55 * ( 2008 ) 177 [ arxiv:0710.1498 ] . k. cerny , `` diffractive photoproduction of dijets in ep collisions at hera '' , proceedings of dis08 ( london ) , april 2008 , http://dx.doi.org/10.3360/dis.2008.69 . m. klasen and g. kramer , mod . a * 23 * ( 2008 ) 1885 [ arxiv:0806.2269 ] . v. budnev , i. ginzburg , g. meledin and v. serbo , phys . rept . * 15 * ( 1975 ) 181 . m. glck , e. reya and a. vogt , phys . d * 45 * ( 1992 ) 3986 ; + m. glck , e. reya and a. vogt , phys . d * 46 * ( 1992 ) 1973 . s. frixione , z. kunszt and a. signer , nucl . b * 467 * ( 1996 ) 399 [ hep - ph/9512328 ] ; + s. frixione , nucl . b * 507 * ( 1997 ) 295 [ hep - ph/9706545 ] ; + see also http://www.ge.infn.it/@xmath258ridolfi/. s. schtzel , `` measurements of dijet cross sections in diffractive photoproduction and deep - inelastic scattering at hera '' , ph.d . thesis , univ . heidelberg ( 2004 ) , h1 thesis 331 , available from http://www-h1.desy.de/publications/theses_list.html . k. cerny , `` tests of qcd hard factorization in diffractive photoproduction of dijets at hera '' , ph.d . thesis , charles university , prague ( 2008 ) , h1 thesis 493 , available from http://www-h1.desy.de/publications/theses_list.html . h. jung , rapgap version 3.1 , comput . commun . * 86 * ( 1995 ) 147 . b. list , a. mastroberardino , `` diffvm : a monte carlo generator for diffractive processes in @xmath6 scattering '' in a. doyle _ ( eds . ) , proceedings of the workshop on monte carlo generators for hera physics , desy - proc-1992 - 02 ( 1999 ) 396 . t. sjostrand , pythia version 6.4 , comput . commun . * 82 * ( 1994 ) 74 ; + t. sjostrand , s. mrenna and p. skands , jhep * 0605 * ( 2006 ) 026 [ hep - ph/0603175 ] . g. marchesini _ et al . _ , herwig version 6.5 , comput . phys . commun . * 67 * ( 1992 ) 465 . s. aid _ et al . _ [ h1 collaboration ] , z. phys . c * 70 * ( 1996 ) 17 [ hep - ex/9511012 ] ; + c. adloff _ et al . _ [ h1 collaboration ] , eur . j. c * 29 * ( 2003 ) 497 [ hep - ex/0302034 ] . i. abt _ et al . _ [ h1 collaboration ] , nucl . instrum . a * 386 * ( 1997 ) 310 ; + i. abt _ et al . _ [ h1 collaboration ] , nucl . instrum . meth . a * 386 * ( 1997 ) 348 ; + r. appuhn _ et al . _ spacal group ] , nucl . instrum . a * 386 * ( 1997 ) 397 . b. andrieu _ et al . _ [ h1 calorimeter group ] , nucl . instrum . a * 336 * ( 1993 ) 499 ; + b. andrieu _ et al . _ [ h1 calorimeter group ] , nucl . instrum . a * 350 * ( 1994 ) 57 . c. adloff _ et al . _ [ h1 collaboration ] , z. phys . c * 74 * ( 1997 ) 221 [ hep - ex/9702003 ] ; + m. peez , `` search for deviations from the standard model in high transverse energy processes at the electron - proton collider , hera '' , ph.d . thesis ( in french ) , university of lyon ( 2003 ) , h1 thesis 317 , available from http://www-h1.desy.de/publications/theses_list.html . s. catani , y. dokshitzer and b. webber , phys . b * 285 * ( 1992 ) 291 . s. schenk , `` energy flow in hard diffractive deep - inelastic scattering and photoproduction with a leading proton '' , diploma thesis , univ . heidelberg , germany , 2003 , available from http://www.physi.uni - heidelberg.de / publications/. t. ahmed _ et al . _ [ h1 collaboration ] , z. phys . c * 66 * ( 1995 ) 529 ; + s. aid _ et al . _ [ h1 collaboration ] , z. phys . c * 69 * ( 1995 ) 27 [ hep - ex/9509001 ] . w. giele _ et al . _ , `` report of the working group on quantum chromodynamics and the standard model for the workshop on physics at tev colliders '' , les houches , france , may 2001 , hep - ph/0204316 ( chapter 1 ) . newman and m. ruspa , proceedings of the hera and the lhc workshop , desy - proc-2009 - 02 ( 2009 ) 401 [ arxiv:0903.2957 ] . m. klasen and g. kramer , j. phys . g * 31 * ( 2005 ) 1391 [ hep - ph/0506121 ] . a. kaidalov , v. khoze , a. martin , m. ryskin , private communications . .bin averaged hadron level differential cross sections for diffractive dijet photoproduction . for each data point , the absolute statistical , uncorrelated and correlated systematic uncertainties and the hadronisation correction factors applied to the nlo calculations are given as @xmath259 , @xmath260 and @xmath261 and @xmath262 , respectively . [ cols="^,^,^,^,^,^,^,^ " , ] |
we consider the resonances of a quantum graph that consists of a compact part with one or more infinite leads attached to it . we discuss the leading term of the asymptotics of the number of resonances of in a disc of a large radius . we call a _ weyl graph _ if the coefficient in front of this leading term coincides with the volume of the compact part of . we give an explicit topological criterion for a graph to be weyl . in the final section we analyze a particular example in some detail to explain how the transition from the weyl to the non - weyl case occurs . | 1003.0051 | let @xmath1 be a finite compact metric graph . that is , @xmath1 has finitely many edges and each edge is equipped with coordinates ( denoted @xmath2 ) that identify this edge with a bounded interval of the real line . we choose some subset of vertices of @xmath1 , to be called _ external vertices _ , and attach one or more copies of @xmath3 , to be called _ leads _ , to each external vertex ; the point @xmath4 in a lead is thus identified with the relevant external vertex . we call the thus extended graph @xmath0 . we assume that @xmath0 has no `` tadpoles '' , i.e. no edge starts and ends at the same vertex ; this can always be achieved by introducing additional vertices , if necessary . in order to distinguish the edges of @xmath1 from the leads , we will call the former the _ internal edges _ of @xmath0 . in @xmath5 we consider the self - adjoint operator @xmath6 with the continuity condition and the kirchhoff boundary condition at each vertex of @xmath0 ; see section [ sec.b ] for the precise definitions . the metric graph @xmath0 equipped with the self - adjoint operator @xmath7 in @xmath5 is called the _ quantum graph_. we refer to the surveys @xcite for a general exposition of quantum graph theory . if the set of leads is non - empty , it is easy to show by standard techniques ( see e.g. ( * ? ? ? * lemma 1 ) ) that the spectrum of @xmath7 is @xmath3 . the operator @xmath7 may have embedded eigenvalues . the `` classical '' definition of resonances is [ def0 ] we will say that @xmath8 , @xmath9 , is a _ resonance _ of @xmath7 ( or , by a slight abuse of terminology , a resonance of @xmath0 ) if there exists a _ resonance eigenfunction _ @xmath10 , @xmath11 , which satisfies the equation @xmath12 on each edge and lead of @xmath0 , is continuous on @xmath0 , satisfies the kirchhoff s boundary condition at each vertex of @xmath0 and the _ radiation condition _ @xmath13 on each lead of @xmath0 . we denote the set of all resonances of @xmath7 by @xmath14 . any solution to on a lead @xmath15 satisfies @xmath16 ; the above definition requires that there exists a non - zero solution with all coefficients @xmath17 vanishing . it is easy to see that all resonances must satisfy @xmath18 ; indeed , if @xmath19 with @xmath20 is a resonance then the corresponding resonance eigenfunction is in @xmath5 , so @xmath21 is an eigenvalue of @xmath7 , which is impossible since @xmath22 . as we will only be interested in the asymptotics of the number of resonances in large disks , we exclude the case @xmath23 from further consideration . in the absence of leads , the spectrum of @xmath7 consists of non - negative eigenvalues and @xmath9 is a resonance if and only if @xmath24 and @xmath25 is an eigenvalue of @xmath7 . it is well known ( see e.g. @xcite ) that the above `` classical '' definition of a resonance coincides with the definition via exterior complex scaling ( see @xcite ) . in the complex scaling approach , the resonances of @xmath7 are identified with the eigenvalues of an auxiliary non - selfadjoint operator @xmath26 , @xmath27 . the _ algebraic multiplicity _ of a resonance is then defined as the algebraic multiplicity of the corresponding eigenvalue of @xmath26 . we discuss this in more detail in section [ sec.b ] , where we show that the multiplicity is independent of @xmath28 . in particular , we show ( in theorem [ thm.b3 ] ) that any @xmath24 , @xmath9 , is a resonance if and only if @xmath25 is an eigenvalue of @xmath7 and in this case the corresponding multiplicities coincide . we define the _ resonance counting function _ by @xmath29 with the convention that each resonance is counted with its algebraic multiplicity taken into account . note that the set @xmath14 of resonances is invariant under the symmetry @xmath30 , so this method of counting yields , roughly speaking , twice as many resonances as one would obtain if one imposed an additional condition @xmath31 . in particular , in the absence of leads , @xmath32 equals twice the number of eigenvalues @xmath33 of @xmath7 ( counting multiplicities ) with @xmath34 . this paper is concerned with the asymptotics of the resonance counting function @xmath32 as @xmath35 . we say that @xmath0 is a _ weyl graph _ , if @xmath36 where @xmath37 is the sum of the lengths of the edges of @xmath1 . if there are no leads then @xmath7 has pure point spectrum , resonances are identified with eigenvalues of @xmath7 and weyl s law may be proved by dirichlet - neumann bracketing . thus , every compact quantum graph is weyl in our sense . as we show below , in the presence of leads this may not be the case . we call an external vertex @xmath38 of @xmath0 _ balanced _ if the number of leads attached to @xmath38 equals the number of internal edges attached to @xmath38 . if @xmath38 is not balanced , we call it _ unbalanced_. our main result is [ th1 ] one has @xmath39 where the coefficient @xmath40 satisfies @xmath41 . one has @xmath42 if and only if every external vertex of @xmath0 is unbalanced . this theorem shows , in particular , that as the graph becomes larger and more complex the failure of weyl s law becomes increasingly likely in an obvious sense . the simplest example of a graph @xmath0 with a balanced external vertex occurs when exactly one lead @xmath43 and exactly one internal edge @xmath44 meet at a vertex . in this case , one can merge @xmath44 and @xmath43 into a new lead ; this will not affect the resonances of @xmath0 but will reduce @xmath45 . this already shows that @xmath0 can not be weyl . section [ sec.f ] discusses the second most simple example . our proof of theorem [ th1 ] consists of two steps . the first step is to identify the set @xmath14 of resonances with the set of zeros of @xmath46 , where @xmath47 is a certain analytic matrix - valued function . this identification is straightforward , but it has a subtle aspect : this is to show that the algebraic multiplicity of a resonance coincides with the order of the zero of @xmath46 . this is done in sections [ sec.c][sec.d ] by employing a range of rather standard techniques of spectral theory , including a resolvent identity which involves the dirichlet - to - neumann map . the function @xmath46 turns out to be an exponential polynomial . by a classical result ( theorem [ langer ] ) , the asymptotics of the zeros of an exponential polynomial can be explicitly expressed in terms of the coefficients of this polynomial . thus , the second step of our proof is a direct and completely elementary analysis of the matrix @xmath47 which allows us to relate the required information about the coefficients of the polynomial @xmath46 to the question of whether the external vertices of @xmath0 are balanced . this is done in section [ sec.e ] . resonance asymptotics of weyl type have been established for compactly supported potentials on the real line , a class of super - exponentially decaying potentials on the real line , compactly supported potentials on cylinders and laplace operators on surfaces with finite volume hyperbolic cusps in @xcite respectively . the proofs rely upon theorems about the zeros of certain classes of entire functions . likewise , our analysis uses a simple classical result ( theorem [ langer ] ) about zeros of exponential polynomials . the situation with resonance asymptotics for potential and obstacle scattering in euclidean space in dimensions greater than one and in hyperbolic space is more complicated and still not fully understood ; the current state of knowledge is described in @xcite . here we remark only that generically , the resonance asymptotics in the multi - dimensional case is not given by the weyl formula . we hope that theorem [ th1 ] can provide some insight to the multi - dimensional case . resonances for quantum graphs have been discussed in a recent publication @xcite . our paper has very little technical content in common with @xcite , in spite of their common theme . in section [ sec.f ] we consider the resonances of a particularly simple quantum graph which can be described as a circle with two leads attached to it . theorem [ th1 ] says that if the leads are attached at different points on the circle , the corresponding quantum graph is weyl , and if they are attached at the same point , we have a non - weyl graph . when the two points where the leads are attached move closer to each other and eventually coalesce , one observes the transition from the weyl to the non - weyl case . we study this transition in much detail . we show that as the two external vertices get closer , `` half '' of the resonances move off to infinity . in the course of this analysis , we also obtain bounds on the positions of individual resonances for this model . the same example was recently considered by exner and lipovsky @xcite subject to general boundary conditions that include the kirchhoff s boundary condition case as a singular limit . although some of their results are broadly similar to ours , none of our theorems may be found in @xcite . here we introduce the necessary notation , recall the definition of resonances via the complex scaling procedure and show that the resonances on the real axis coincide with the eigenvalues of @xmath7 . let @xmath48 be the set of all internal edges of @xmath0 ( i.e. the set of all edges of @xmath1 ) and let @xmath49 be the set of all leads ; we also denote @xmath50 . the term `` edge '' without an adjective will refer to any element of @xmath51 . for @xmath52 , we denote by @xmath53 the length of @xmath44 ; i.e. an edge @xmath54 is identified with the interval @xmath55 $ ] . let @xmath56 be the set of all vertices of @xmath0 , let @xmath57 be the set of all external vertices , and let @xmath58 ; the elements of @xmath59 will be called _ internal vertices_. the degree of a vertex @xmath38 is denoted by @xmath60 . the number of leads attached to an external vertex @xmath38 is denoted by @xmath61 ; we also set @xmath62 for @xmath63 . if an edge or a lead @xmath44 is attached to a vertex @xmath38 , we write @xmath64 . if two vertices @xmath65 are connected by one or more edges , we write @xmath66 . we denote by @xmath67 the graph @xmath0 with all the internal edges and vertices removed . we let @xmath68 and @xmath69 be the characteristic functions of @xmath1 and @xmath67 . let @xmath70 be a function such that the restriction of @xmath71 onto every edge is continuously differentiable . then for @xmath72 , we denote by @xmath73 the sum of the outgoing derivatives of @xmath71 at @xmath38 over all edges attached to @xmath38 . if @xmath38 is an external vertex , we denote by @xmath74 ( resp . @xmath75 ) the sum of all outgoing derivatives of @xmath71 at @xmath38 over all internal edges ( resp . leads ) attached to @xmath38 . let @xmath76 be the class of functions @xmath70 which are continuous on @xmath77 and such that for each external vertex @xmath38 the function @xmath78 approaches a limiting value ( to be denoted by @xmath79 ) as @xmath2 approaches @xmath38 along any internal edge and @xmath78 approaches another limiting value ( to be denoted by @xmath80 ) as @xmath2 approaches @xmath38 along any lead . for any finite set @xmath81 , we denote by @xmath82 the number of elements of @xmath81 . we will use the identity @xmath83 finally , we use the notation @xmath84 . the domain of the self - adjoint operator @xmath7 consists of all continuous functions @xmath70 such that the restriction of @xmath71 onto any @xmath86 lies in the sobolev space @xmath87 , and @xmath71 satisfies the kirchhoff boundary condition @xmath88 on every vertex @xmath38 of @xmath0 . for @xmath89 , let @xmath90 be the unitary operator which acts as identity on @xmath91 and as a dilation on all leads @xmath15 : @xmath92 note that we have @xmath93 for any @xmath89 . consider the operator @xmath94 this operator admits an analytic continuation to @xmath95 , which we describe below . for @xmath95 , the operator @xmath85 in @xmath5 acts according to the formula @xmath96 the domain of @xmath85 is defined to be the set of all @xmath70 which satisfy the following conditions : \(i ) the restriction of @xmath71 onto any @xmath86 lies in the sobolev space @xmath87 ; \(ii ) @xmath97 ; \(iii ) @xmath71 satisfies the condition @xmath88 at every internal vertex @xmath38 ; \(iv ) for any @xmath98 , one has @xmath99 in particular , @xmath100 is the operator called @xmath7 so far . for complex @xmath101 , the operator @xmath85 is in general non - selfadjoint . a standard straightforward computation shows that for any @xmath95 the operator @xmath85 is closed and @xmath102 the following theorem is standard in the method of complex scaling ; see @xcite : [ thm.b2 ] the family of operators @xmath85 , @xmath95 , is analytic in the sense of kato ( see e.g. ( * ? ? ? * section xii.2 ) ) and the identity @xmath103 holds . the essential spectrum of @xmath85 coincides with the half - line @xmath104 . let @xmath27 ; then the sector @xmath105 , @xmath33 , contains no eigenvalues of @xmath26 , and any @xmath33 in the sector @xmath106 is an eigenvalue of @xmath26 if and only if @xmath107 with @xmath108 . for completeness , we give the proof in section [ sec.d ] . as @xmath27 increases monotonically , the essential spectrum @xmath109 of @xmath26 rotates clockwise , uncovering more and more eigenvalues @xmath110 . these eigenvalues are identified with the resonances @xmath111 of @xmath7 via @xmath107 . if @xmath33 is an eigenvalue of @xmath26 , @xmath27 , @xmath112 , kato s theory of analytic perturbations implies that the eigenvalue and associated riesz spectral projection depend analytically on @xmath28 . noting ( [ a3 ] ) and using analytic continuation it follows that the algebraic multiplicity of @xmath110 is independent of @xmath28 . it is easy to see directly that the geometric multiplicity of @xmath110 is also independent of @xmath28 . the _ algebraic ( resp . geometric ) multiplicity _ of a resonance @xmath111 is defined as the algebraic ( resp . geometric ) multiplicity of the corresponding eigenvalue @xmath107 of @xmath26 . the geometric multiplicities of resonances will not play any role in our analysis . however , we note that for the schrdinger operator on the real line , resonances can have arbitrary large algebraic multiplicity @xcite , while their geometric multiplicity is always equal to one . this gives an example of resonances with distinct algebraic and geometric multiplicities . it would be interesting to see if one can have distinct algebraic and geometric multiplicities of resonances for quantum graphs in the situation we are discussing . we have nothing to say about this except for the case of the resonances on the real line : [ thm.b3 ] 1 . if @xmath24 , @xmath9 , is a resonance of @xmath7 then the algebraic and geometric multiplicities of @xmath111 coincide . 2 . any @xmath24 , @xmath9 , is a resonance of @xmath7 if and only if @xmath25 is an eigenvalue of @xmath7 and the multiplicity of the resonance @xmath111 coincides with the multiplicity of the eigenvalue @xmath25 . let @xmath113 be an eigenvalue of @xmath7 with the eigenfunction @xmath71 . if @xmath15 is a lead , then @xmath114 , @xmath115 , where @xmath116 . since @xmath117 , we conclude that @xmath118 and so @xmath119 on all leads . it follows that @xmath120 for all @xmath28 and @xmath121 . this argument proves that @xmath122 \2 . let @xmath123 , @xmath113 , @xmath27 . let us prove that @xmath71 vanishes identically on all leads . let @xmath107 , @xmath124 . on every lead , we have @xmath125 consider the difference @xmath126 integrating by parts , we get @xmath127 using the boundary condition and formula , we obtain @xmath128 by the definition of @xmath129 , we have @xmath130 . this yields that @xmath131 on all external vertices @xmath38 . by , it follows that @xmath71 vanishes identically on all leads . \3 . by combining the previous step of the proof with and we obtain @xmath132 . it follows that for any @xmath123 , @xmath113 , @xmath27 , we have @xmath133 and @xmath134 . this argument also proves that @xmath135 \4 . it remains to prove that if @xmath113 is an eigenvalue of @xmath26 , @xmath27 , then its algebraic and geometric multiplicities coincide . suppose this is not the case . then there exist non - zero elements @xmath136 such that @xmath137 and @xmath138 . by step 2 of the proof , @xmath139 vanishes on all leads . it follows that on all leads the function @xmath71 satisfies . next , since @xmath140 on @xmath1 , we have @xmath141 integrating by parts , we get @xmath142 consider the three terms in the r.h.s . of . the first term vanishes since @xmath137 . next , since @xmath143 on @xmath67 , we have @xmath144 for any @xmath145 . by the boundary conditions and for @xmath139 it follows that @xmath146 . thus , the second and third terms in the r.h.s . of also vanish . this contradicts . here we describe the resonances as zeros of @xmath46 , where @xmath47 is certain entire matrix - valued function . using this characterisation , we prove our main result . fix @xmath147 . let @xmath148 denote the space of all solutions @xmath149 to @xmath150 on @xmath0 without any boundary conditions . the restriction of @xmath151 onto any internal edge @xmath44 has the form @xmath152 , and the restriction of @xmath71 onto any lead @xmath43 has the form @xmath153 . thus , @xmath154 . let us describe in detail the set of all conditions on @xmath151 required to ensure that @xmath71 is a resonance eigenfunction . if @xmath155 denotes the restriction of @xmath71 to an edge @xmath44 , then we can write the continuity conditions at the vertex @xmath38 as @xmath156 where @xmath157 is an auxiliary variable . we also have the condition @xmath158 writing down conditions , for every vertex @xmath72 , we obtain @xmath159 conditions . our variables are @xmath160 , @xmath161 , @xmath162 , @xmath163 ; altogether we have @xmath164 variables . let @xmath165 , @xmath166 , @xmath167 , @xmath168 be the sequences of coordinates @xmath160 , @xmath161 , @xmath162 , @xmath163 of length @xmath169 , @xmath170 , @xmath170 , @xmath171 respectively , and let @xmath172 . we may write the constraints , in the form @xmath173 , where @xmath81 is an @xmath174 matrix . each row of @xmath81 relates to one of the constraints , and each constraint is of the form @xmath175 if the constraint is of the form , then @xmath176 and @xmath177 all contain a multiplicative factor @xmath178 which we eliminate before proceeding . the coefficient @xmath179 is @xmath4 , @xmath180 , or @xmath181 , and the coefficient @xmath182 is @xmath4 , @xmath180 , or @xmath183 . the coefficient @xmath184 is @xmath4 or @xmath185 , and the coefficient @xmath186 is @xmath4 or @xmath187 . we have not specified the order of the rows or columns of @xmath47 . however , the object of importance in the sequel is the set of zeros of @xmath46 , and the choice of the order of rows or columns of @xmath47 will not affect this set . as an example , let us display the matrix @xmath47 for a graph which consists of two vertices @xmath188 and @xmath189 , two edges @xmath190 and @xmath191 of length @xmath192 and @xmath193 and a lead attached at @xmath188 . in this case we have , denoting @xmath194 : @xmath195 although @xmath47 was defined above for @xmath147 , we see that all elements of @xmath47 are entire functions of @xmath8 . thus , we will consider @xmath47 as an entire matrix - valued function of @xmath111 . in sections [ sec.c][sec.d ] we prove [ thm.e1 ] any @xmath196 is a resonance of @xmath7 if and only if @xmath197 . in this case , the algebraic multiplicity of the resonance @xmath19 coincides with the order of @xmath19 as a zero of @xmath46 . the first part of this theorem is obvious : by the construction of the matrix @xmath81 , we have @xmath198 iff there exists a non - zero resonance eigenfunction @xmath199 . the part concerning multiplicity is less obvious . unfortunately , we were not able to find a completely elementary proof of this part . the proof we give in sections [ sec.c][sec.d ] involves a standard set of techniques of spectral theory of quantum graphs : a resolvent identity involving the dirichlet - to - neumann map and a certain trace formula . by theorem [ thm.e1 ] , the question reduces to counting the total multiplicity of zeros of the entire function @xmath46 in large discs . as it is clear from the structure of the matrix @xmath47 , its determinant is an exponential polynomial , i.e. a linear combination of the terms of the type @xmath200 , @xmath201 . thus , we need to discuss the zeros of exponential polynomials . exponential polynomials are entire functions @xmath202 , @xmath8 , of the form @xmath203 where @xmath204 are constants . the study of the zeros of such polynomials has a long history ; see e.g. @xcite and references therein . for more recent literature see @xcite . some of these results were rediscovered in @xcite , where they were used to analyze the spectra of non - self - adjoint systems of odes and directed finite graphs . the asymptotic distribution of the zeros of @xmath205 depends heavily on the location of the extreme points of the convex hull of the set @xmath206 . we are only interested in the case in which @xmath207 are distinct real numbers . we denote @xmath208 and @xmath209 . for @xmath210 we denote by @xmath211 the number of zeros of @xmath205 ( counting the orders ) in the disc @xmath212 . the following classical statement is from ( * ? ? ? * theorem 3 ) . [ langer ] let @xmath205 be a function of the form , where @xmath213 are non - zero complex numbers and @xmath207 are distinct real numbers . then there exists a constant @xmath214 such that all the zeros of @xmath205 lie within a strip of the form @xmath215 . the counting function @xmath211 satisfies @xmath216 as @xmath217 . here we prove the first part of the main theorem [ th1 ] . let @xmath218 . from the structure of @xmath47 it is clear that @xmath202 is given by where @xmath219 are real coefficients . by theorem [ langer ] , it suffices to prove that in the representation we have @xmath220 in order to prove , let us discuss the entries of @xmath47 in detail . for simplicity of notation we will not draw attention in our equations to the fact that all of the matrices below depend on @xmath111 . the matrix @xmath81 has some constant terms and some terms that are exponential in @xmath111 . the term @xmath221 can only appear in the column associated with the variable @xmath161 and the term @xmath222 can only appear in the column associated with the variable @xmath162 . the columns associated with the variables @xmath165 and @xmath168 contain only constant terms . since the determinant is formed from the products of entries of @xmath81 where every column contributes one entry to each product , we see that the maximum possible value for the coefficient @xmath207 in is attained when every column corresponding to the variable @xmath161 contributes the term @xmath221 and every column corresponding to @xmath162 contributes a constant term to the product . the maximal value of @xmath207 thus attained will be exactly @xmath223 . this proves the first inequality in . the second one is proven in the same way by considering the minimal possible value for @xmath207 . of course , the coefficients @xmath224 of the terms @xmath225 in the representation for @xmath226 may well happen to be zero . theorem [ th1 ] will be proven if we show that these coefficients do not vanish if and only if every external vertex of @xmath0 is unbalanced . in what follows , for an exponential polynomial @xmath205 with the representation we denote by @xmath227 the coefficient @xmath213 of the term @xmath228 , @xmath229 . before proceeding with the proof , we need to discuss a minor technical point . our definition of the matrix @xmath47 assumes that a certain orientation of all internal edges of @xmath0 is fixed . suppose we have changed the parameterisation of an internal edge @xmath44 by reversing its orientation . in other words , suppose that instead of the variable @xmath230 $ ] we decided to use the variable @xmath231 . we claim that this change will not affect the zeros of @xmath46 . indeed , let @xmath232 be the matrix corresponding to the new parametrization . the matrix @xmath232 corresponds to the parametrization of solutions @xmath151 on @xmath44 by @xmath233 instead of @xmath234 . we have @xmath235 and thus @xmath236 . assume that a particular external vertex @xmath38 of @xmath0 is balanced . below we prove that the coefficient @xmath237 vanishes . let us re - order the rows and columns of @xmath81 by reference to the vertex @xmath38 . we assume that @xmath238 internal edges and @xmath238 leads are attached to @xmath38 , @xmath239 . ( the case @xmath240 is trivial because one may then merge the lead with the edge to which it is connected . ) using the observation of section [ sec.e2a ] , we can choose an orientation of these internal edges so that they all end at @xmath38 ( i.e. @xmath38 is identified with the point @xmath241 of the intervals @xmath242 $ ] ) . let the first @xmath243 rows of @xmath81 be those relating to the conditions for the vertex @xmath38 and let the @xmath244st row be the one relating to the condition for the vertex @xmath38 . the ordering of the remaining rows does not matter . let the first @xmath243 columns be related to the variables @xmath245 and let the @xmath244st column be related to the variable @xmath160 ; these variables were all defined in section [ notation ] . the ordering of the remaining columns does not matter . we write @xmath81 in the block form @xmath246 where @xmath247 is a @xmath248 matrix . for example , in the case @xmath249 we have @xmath250 where @xmath251 . the determinant is the sum of the products of entries of @xmath81 where every column contributes one entry to each product . in order for the product to be of the type @xmath252 , each column corresponding to a variable @xmath161 must contribute the entry @xmath221 . thus , the constant entries of the columns corresponding to the variables @xmath161 are irrelevant to our question and can be replaced by zeros ; this will not affect the value of @xmath237 . noticing that the columns of @xmath253 corresponding to the variables @xmath254 and @xmath160 are all zeros , we conclude that @xmath255 by a general matrix identity , @xmath256 . finally , a simple row reduction shows that @xmath257 ; this is easy to see in the case of . thus , the coefficient @xmath237 vanishes . by , it follows that @xmath258 , as claimed . we note ( although this is not needed for our proof ) that @xmath259 both in the balanced and in the unbalanced case ; this will be clear from the next part of the proof . assume that all external vertices are unbalanced . we will prove that @xmath260 the proof uses the same reduction as in section [ sec.e3 ] , but the details are somewhat more complicated , since now we have to consider _ all _ external vertices . we label the external vertices by @xmath188, ,@xmath261 where @xmath262 . let @xmath263 denote the graph obtained from @xmath1 by adding all the leads of @xmath0 that have ends in the set @xmath264 , so that @xmath265 . let @xmath266 denote the constraint matrix @xmath81 corresponding to the graph @xmath263 and let @xmath267 . by the previous reasoning , the graph @xmath263 is weyl if and only if @xmath268 and @xmath269 . our claim follows inductively from the following statements : 1 . the graph @xmath1 is weyl . 2 . the coefficient @xmath270 is non - zero for all @xmath271 . 3 . for all @xmath271 , if @xmath272 then @xmath268 . item 1 holds because the operator @xmath7 on @xmath1 has discrete spectrum and no other resonances . the eigenvalues obey the weyl law by a standard variational argument using dirichlet - neumann bracketing . let us prove item 3 . we reorder the rows and columns of @xmath266 with reference to @xmath273 as in section [ sec.e3 ] . we assume that @xmath274 internal edges @xmath190, ,@xmath275 and @xmath238 leads @xmath276 , , @xmath277 are attached to @xmath273 , and @xmath278 . the first @xmath279 columns of @xmath266 are those relating to the variables @xmath280 , , @xmath281 ( associated with @xmath276 , , @xmath277 ) , @xmath282 , , @xmath283 ( associated with @xmath190, ,@xmath275 ) , and @xmath284 . the first @xmath279 rows of @xmath266 are those relating to the conditions and for the vertex @xmath273 . as in section [ sec.e3 ] , this allows us to write @xmath285 where @xmath286 is a @xmath287 matrix . writing the matrix @xmath288 in the same way with reference to _ the same vertex _ @xmath273 , we obtain @xmath289 where @xmath290 is a @xmath291 matrix . in other words , @xmath290 , @xmath292 , @xmath293 are the matrices @xmath286 , @xmath294 , @xmath295 with relevant @xmath238 rows and @xmath238 columns deleted . the deleted columns correspond to the variables @xmath280, ,@xmath281 , and the deleted rows correspond to the conditions associated with the leads @xmath276 , , @xmath277 . note that the matrix @xmath296 is the same in and . next , just as in the argument of section [ sec.e3 ] , we notice that @xmath297 finally , by a simple row reduction we obtain @xmath298 where @xmath299 . it follows that @xmath300 and @xmath301 differ by a non - zero coefficient @xmath302 . this proves item 3 . let us prove item 2 . here the argument follows that of the proof of item 3 , only instead of keeping track of the coefficient of @xmath303 we need to keep track of the coefficient of @xmath304 , and instead of the variables @xmath282, ,@xmath283 we consider the variables @xmath305 , , @xmath306 . instead of the coefficient @xmath307 in we get @xmath308 , which never vanishes ( even if @xmath273 is balanced ) . this proves our claim . in order to complete the proof of theorem [ th1 ] , it remains to provide the proof of theorem [ thm.e1 ] . theorem [ thm.c1 ] below provides an explicit formula for the difference @xmath309 in terms of the dirichlet - to - neumann map . this leads immediately to the trace formula , which is the key to our proof of theorem [ thm.e1 ] in section [ sec.d ] . the formulae obtained in this section are `` complex - scaled '' versions of resolvent identities well known in the theory of boundary value problems , see e.g. @xcite and references therein . throughout this section , we assume that the parameter @xmath147 is fixed . let @xmath148 be as defined in section [ sec.d1 ] and let @xmath310 . each @xmath311 determines a vector @xmath312 by restriction to @xmath56 . conversely , every @xmath313 arises from a function @xmath311 ; this can be seen by comparing @xmath314 with the number of constraints imposed by writing @xmath315 , @xmath72 . finally , the assumption @xmath147 implies that only one function @xmath311 corresponds to each set of values @xmath313 ( otherwise we would have a complex eigenvalue of the operator with dirichlet boundary conditions on all vertices ) . this shows that we may define the dirichlet - to - neumann map @xmath316 by @xmath317 where @xmath71 corresponds to @xmath165 as described above and @xmath318 was defined in section [ notation ] . this map is a well known tool in the spectral theory of boundary value problems and has also been used in quantum graph theory @xcite . given @xmath72 , let @xmath319 be the function in @xmath321 that satisfies @xmath322 the functions @xmath319 are given by the following explicit expressions . let @xmath323 , @xmath54 and identify @xmath44 with @xmath324 $ ] where @xmath38 corresponds to the point @xmath4 . then @xmath325=e . \label{c1}\ ] ] in the same way , if @xmath326 and @xmath38 is identified with the point @xmath4 , then @xmath327 if the dependence on @xmath111 needs to be emphasized , we will write @xmath328 instead of @xmath329 . [ extralemma ] if @xmath147 then the map @xmath330 is invertible . its matrix entries are given by @xmath331 where @xmath61 was defined in section [ notation ] . if @xmath332 , then the corresponding function @xmath333 satisfies the kirchhoff s boundary condition at every vertex , which implies that @xmath133 and @xmath334 . since @xmath335 and @xmath336 , this implies that @xmath337 . therefore @xmath330 is invertible . by the definition of @xmath319 , we have @xmath338 the formulae for the matrix entries are obtained by combining this with and . it follows from lemma [ extralemma ] that @xmath330 can be extended to a meromorphic function of @xmath8 whose poles are all on the real axis , and that for any @xmath339 one has @xmath340 in the calculations below the expressions @xmath341 will denote the matrix entries of @xmath342 . we will need a version of the functions @xmath319 pertaining to the `` complex - scaled '' operator @xmath85 . let @xmath147 and @xmath95 be such that @xmath343 . given @xmath344,we define @xmath345 by @xmath346 clearly , @xmath345 is a solution to the equation @xmath347 on every edge of @xmath0 . it is also straightforward to see that @xmath348 and @xmath345 satisfies the boundary condition on every external vertex . for @xmath97 , let us denote @xmath349 it is straightforward to see that @xmath350 where the l.h.s . depends on @xmath111 but not on @xmath101 . moreover @xmath351 let @xmath352 be the self - adjoint operator in @xmath5 defined by @xmath353 with a dirichlet boundary condition at every vertex of @xmath0 . given @xmath95 , we define the `` complex - scaled '' version of @xmath352 as follows ; @xmath354 is the operator acting in @xmath5 defined by @xmath355 with a dirichlet boundary condition at every vertex of @xmath0 . of course , @xmath354 splits into an orthogonal sum of operators acting on @xmath356 for all @xmath86 . we see immediately that in addition to its essential spectrum @xmath104 , the operator @xmath354 has a discrete set of positive eigenvalues with finite multiplicities . we set @xmath357 whenever the inverse operators exist . we denote by @xmath358 ( resp . @xmath359 ) , @xmath360 , the integral kernel of the resolvent @xmath361 ( resp . of @xmath362 ) . the fact that @xmath354 and @xmath85 coincide except for different boundary conditions at each of the @xmath363 vertices indicates that the difference of the two resolvents should have rank @xmath363 . our next theorem makes this explicit . formulae of this type are well known in the theory of boundary value problems ; see e.g.@xcite and references therein . in the context of graphs , similar considerations have been used in @xcite . [ thm.c1 ] for any @xmath147 and any @xmath95 , such that @xmath343 , we have @xmath364 for any @xmath360 . let @xmath365 be the operator in @xmath5 with the integral kernel given by @xmath366 we need to prove that @xmath365 is a bounded operator , that it maps @xmath5 into @xmath367 and that the identities @xmath368 hold true . first note that since @xmath345 decays exponentially on all leads , the boundedness of @xmath365 is obvious . next , using , one obtains @xmath369 . from here and by taking adjoints we see that is equivalent to @xmath370 which is with @xmath371 , @xmath372 instead of @xmath111 , @xmath101 . we note that @xmath147 , @xmath343 if and only if @xmath373 , @xmath374 . thus , follows from . \2 . it suffices to prove that for a dense set of elements @xmath149 , the inclusion @xmath375 and the identity @xmath376 hold true . let @xmath71 be from the dense set of all continuous functions compactly supported on @xmath0 and vanishing near all vertices of @xmath0 . let us check that the function @xmath377 belongs to @xmath367 . it is clear that the restriction of @xmath139 onto any edge @xmath44 of @xmath0 belongs to the sobolev space @xmath87 . thus , it suffices to check that @xmath139 belongs to @xmath76 and satisfies the boundary conditions and . denote @xmath378 . since @xmath379 , @xmath380 vanishes on all vertices . therefore @xmath380 lies in @xmath76 and satisfies at every external vertex @xmath38 . as mentioned in section [ sec.c3 ] , the functions @xmath381 also belong to @xmath76 and satisfy at every external vertex @xmath38 . thus , @xmath139 also has these properties . our next task is to prove that the boundary condition is satisfied for the function @xmath139 . suppose that @xmath71 is supported on a single edge , which we identify with @xmath324 $ ] . then the integral kernel of @xmath362 can be explicitly calculated , which gives @xmath382 similarly , if @xmath71 is supported on a lead @xmath3 , then a direct calculation shows that @xmath383 combining this , we see that for any @xmath384 we have @xmath385 using the last identity and , for any @xmath384 we get : @xmath386 and so the boundary condition is satisfied for @xmath139 . thus , @xmath387 , as required . \3 . it remains to note that the identity follows from the fact that @xmath388 is the resolvent of @xmath354 and the fact that @xmath345 satisfies the equation @xmath347 on every edge and lead of @xmath0 . the trace formula below results by calculating the traces of both sides of . since the r.h.s . of is a finite rank operator , the trace is well defined ; the fact that the value of does not depend on @xmath101 can be proved by complex scaling , but the direct proof is almost as easy . the identity below can be rephrased by saying that the ( modified ) perturbation determinant of the pair of operators @xmath85 , @xmath354 equals @xmath389 . statements of this nature ( for @xmath390 ) are well known in the theory of boundary value problems ; see e.g. @xcite and references therein . the key to our proof of theorem [ thm.e1 ] will be and lemma [ lma.d1 ] , in which @xmath46 and @xmath389 are related . [ thm.c2 ] for any @xmath147 and any @xmath95 , such that @xmath343 , we have @xmath391 in particular , the l.h.s . is independent of @xmath101 . theorem [ thm.c1 ] yields @xmath392 where @xmath393 we next compute the coefficients @xmath394 explicitly . if @xmath395 and @xmath396 then @xmath397 and so @xmath398 . if @xmath395 and @xmath399 then by @xmath400 and finally , @xmath401 \2 . noting that @xmath394 depend on @xmath111 but not on @xmath101 , a direct calculation using yields @xmath402 it follows that @xmath403 as required . given @xmath8 , we define @xmath404 let @xmath47 be the matrix defined in section [ sec.d1 ] . [ lma.d1 ] for any @xmath147 , we have the identity @xmath405 where the sign @xmath406 depends on the ordering of the rows and columns of the matrix @xmath47 . let us order the rows and the columns of @xmath47 in such a way that the first @xmath169 rows correspond to the conditions @xmath407 , and the first @xmath169 columns correspond to the variables @xmath165 . then @xmath47 can be written in the block form as @xmath408 where @xmath4 is the @xmath409 zero matrix and @xmath410 is a @xmath411 matrix . the elements of @xmath412 are @xmath4 or @xmath185 , the elements of @xmath413 are @xmath4 , @xmath414 , @xmath415 , and the elements of @xmath410 are @xmath4 , @xmath414 , or @xmath416 . for example , the matrix is written in this form . let us reorder the rows of @xmath410 in such a way that any two constraints associated with the continuity conditions at the two endpoints of the same edge follow one another . let us also reorder the columns of @xmath410 such that each variable @xmath162 follows the corresponding variable @xmath161 . for example , the block @xmath410 of the matrix after such reordering will be @xmath417 in general , after this reordering , @xmath410 assumes a block - diagonal structure with blocks either of size @xmath418 with elements @xmath419 or of size @xmath420 with the element @xmath185 . from here it follows that @xmath421 in particular , since @xmath147 , the matrix @xmath410 is invertible . \3 . by applying the schur complement method to one obtains @xmath422 let us prove that @xmath423 let @xmath424 and let @xmath425 . the vector @xmath426 represents a set of parameters @xmath166 , @xmath167 , @xmath168 . let @xmath151 be the solution with this set of parameters . the equation @xmath427 implies that the solution @xmath71 is continuous on @xmath0 and satisfies @xmath315 for any vertex @xmath38 . next , the coordinates of the vector @xmath428 are given by @xmath429 this shows that @xmath430 , as required . \4 . by combining one obtains @xmath431 which yields immediately . let @xmath147 and let @xmath68 and @xmath69 be defined as in section [ notation ] . clearly , @xmath432 is an orthogonal sum of resolvents of the operators @xmath433 on the intervals @xmath434 , @xmath54 , with the dirichlet boundary conditions . for each such operator we have that @xmath435 is trace class and @xmath436 summing over all edges , a direct calculation shows that @xmath437 is a trace class operator and @xmath438 \2 . let @xmath147 , @xmath439 . it is easy to see that the resolvent @xmath440 commutes with @xmath68 , @xmath69 and that @xmath441 therefore we have @xmath442 by combining theorem [ thm.c2 ] and , we obtain @xmath443 using lemma [ lma.d1 ] , we then obtain @xmath444 for all @xmath147 and @xmath439 . \3 . the r.h.s . of is a single - valued meromorphic function of @xmath8 . let @xmath445 be the l.h.s . of . for each fixed @xmath95 , the function @xmath445 is meromorphic in @xmath446 with the cut along the line determined by the condition @xmath447 . in other words , @xmath448 is meromorphic and single - valued in each of the two half - planes @xmath449 and @xmath450 . by the uniqueness of analytic continuation , for each @xmath101 the identity extends to all @xmath111 such that @xmath451 . let @xmath452 with the algebraic multiplicity @xmath453 and let @xmath27 with @xmath454 . then @xmath455 and so the identity with @xmath456 holds for all @xmath111 near @xmath19 . if @xmath168 is a sufficiently small circle with centre at @xmath19 , then the multiplicity @xmath457 equals the rank , or equivalently the trace , of the riesz spectral projection @xmath458 next , since the operator @xmath459 restricted to @xmath460 has no eigenvalues , the operator valued function @xmath461 is analytic for @xmath462 . it follows that @xmath463 by taking the trace of the difference of the last two equations and using we obtain @xmath464 therefore @xmath457 equals the order of the zero of @xmath46 at @xmath465 , as required . this theorem is well known and the proof is presented here for the sake of completeness . first note that by theorem [ thm.c1 ] , the difference of the resolvents of @xmath85 and @xmath354 is a finite rank operator . by weyl s theorem on the invariance of the essential spectrum under a relatively compact perturbation we obtain @xmath466 \2 . the fact that the family @xmath85 is analytic in the sense of kato follows again from theorem [ thm.c1 ] , since @xmath354 is analytic in the sense of kato and each of the functions @xmath345 is analytic in @xmath101 . the identity can be checked by a direct calculation . let @xmath108 and let @xmath71 be the corresponding eigenfunction . for any @xmath27 with @xmath467 , let @xmath468 be the function defined _ formally _ by @xmath469 . more precisely , we set @xmath470 on @xmath1 and @xmath471 for @xmath2 on any lead @xmath15 . by the choice of @xmath28 , we have @xmath472 and so @xmath473 . a straightforward inspection shows that @xmath474 and @xmath475 . conversely , let @xmath476 be an eigenvalue of @xmath26 for @xmath27 . write @xmath107 with @xmath472 . then , for the corresponding eigenfunction @xmath139 of @xmath26 we have @xmath477 on any lead of @xmath0 . a direct inspection shows that @xmath478 in the same sense as , where @xmath71 is a resonance eigenfunction . thus , @xmath108 and in particular , @xmath18 . it follows that @xmath479 . here we consider resonances of a particular simple graph @xmath480 , where @xmath481 $ ] is a certain geometric parameter . the graph @xmath480 was also considered in ( * ? ? ? * section 4 ) , but with different boundary conditions at the vertices . the graph @xmath480 is weyl for @xmath482 and non - weyl for @xmath483 . this section has two goals . the first one is to discuss the transition between the weyl and the non - weyl cases in order to throw new light on the failure of the weyl law . our second goal is to obtain rigorous bounds on the locations of individual resonances of @xmath480 , which was not addressed in @xcite . given @xmath484 , we consider the graph @xmath485 which consists of two vertices @xmath188 and @xmath189 and two edges @xmath486 $ ] , @xmath487 , and @xmath488 $ ] , @xmath489 . the vertex @xmath189 is identified with the point @xmath4 of @xmath190 and with the point @xmath4 of @xmath191 , and the vertex @xmath188 is identified with the point @xmath192 of @xmath190 and with the point @xmath193 of @xmath191 . thus , the graph @xmath485 is simply a circle with the circumference @xmath490 for all @xmath491 . we attach a lead @xmath276 at @xmath188 and a lead @xmath492 at @xmath189 and denote the thus extended graph by @xmath480 . geometrically , @xmath480 is a circle with two leads attached to it . finally , for @xmath483 , let @xmath480 be the circle of length @xmath493 with two leads attached at the same point . we will denote by @xmath494 the operator @xmath495 acting in @xmath496 subject to the usual continuity and kirchhoff s boundary conditions at the vertices @xmath188 and @xmath189 . by theorem [ th1 ] , the graph @xmath480 is weyl if and only if @xmath482 . at the same time , the graph @xmath497 can be regarded as the limit of @xmath480 as @xmath498 in an obvious geometric sense , so we need to explain what happens to resonances as @xmath499 . as we will see , roughly speaking , half of the resonances of @xmath494 move off to infinity as @xmath500 . we will obtain bounds on the curves along which the resonances move as @xmath491 increases from @xmath4 to @xmath185 . let us display the constraints corresponding to the graph @xmath480 ; the matrix @xmath501 will be built up of the rows corresponding to these constraints . we denote @xmath502 , @xmath503 . the constraints corresponding to the vertex @xmath188 are @xmath504 the first three constraints above are the continuity conditions , and the last one is the requirement that the sum of the outgoing derivatives vanishes . similarly , the constraints corresponding to the vertex @xmath189 are @xmath505 we list these constraints in the order @xmath506 , @xmath507 , @xmath508 , @xmath509 , @xmath510 , @xmath511 , @xmath512 , @xmath513 , and order the variables as @xmath282 , @xmath305 , @xmath514 , @xmath515 , @xmath280 , @xmath516 , @xmath517 , @xmath518 . this leads to the matrix @xmath519 the graph @xmath480 has a reflection symmetry with respect to the midpoints of @xmath190 and @xmath191 . this allows to decompose the space @xmath148 into the direct sum of the subspaces corresponding to even and odd functions with respect to this symmetry . we use this decomposition to represent the matrix @xmath501 in a block - diagonal form where the blocks correspond to the even and odd solutions . more precisely , let @xmath521 and @xmath522 be the matrices @xmath523 a straightforward calculation shows that @xmath524 . next , let @xmath525 ; the reader is invited to check that the matrix @xmath526 can be written as @xmath527 with blocks @xmath528 where we have used the notation @xmath529 , @xmath530 , @xmath503 . straightforward calculations of @xmath531 and @xmath532 now yield we will call the zeros of @xmath535 ( resp . of @xmath536 ) the even ( resp . odd ) resonances . it is not difficult to check that the resonance eigenfunctions which correspond to even / odd resonances are even / odd with respect to the symmetry of the graph @xmath480 . by theorem [ thm.b3 ] , the real even / odd resonances are actually eigenvalues of @xmath494 and therefore we will call them even / odd eigenvalues . finally , it is not difficult to check that the resonances of @xmath537 are given , as expected , by the zeros of @xmath538 . in fact , in this case we have @xmath539 and @xmath540 thus , the resonances of @xmath537 coincide with the solutions to @xmath541 , i.e. they are given by @xmath542 . by theorem [ thm.b3 ] , these resonances ( for @xmath9 ) coincide with the eigenvalues of @xmath537 and all of them have the multiplicity one . this shows that for @xmath483 we have the asymptotics with @xmath543 . \(i ) by an explicit calculation , @xmath550 where @xmath551 and @xmath552 we deduce that @xmath553 ( ii ) we start by observing that @xmath554 where @xmath555 if @xmath556 and @xmath557 then @xmath558 therefore @xmath559 we deduce that @xmath560 and @xmath561 these bounds imply that @xmath562 if @xmath563 , which yields the theorem immediately . it follows that all odd resonances are located in the rectangles @xmath564 the following statement , in combination with rouchet s theorem , shows that each of the rectangles @xmath565 contains exactly one odd resonance of algebraic multiplicity one for all @xmath481 $ ] . [ cequals0 ] if @xmath566 there is a resonance of algebraic multiplicity one at @xmath567 for every odd @xmath544 and an eigenvalue of multiplicity one at @xmath568 for every non - zero even @xmath544 . there is also a resonance of algebraic multiplicity one at @xmath23 . no other odd resonances or eigenvalues exist if @xmath566 . the proof follows from the explicit formula @xmath569 by the implicit function theorem , we obtain that each of the zeros of @xmath570 is a real analytic function of @xmath481 $ ] with values in @xmath565 . the set of all odd resonances is therefore the union of a sequence of bounded real analytic curves . it is interesting to note that each of these resonance curves intersects the real axis , thereby ( by theorem [ thm.b3 ] ) giving rise to embedded eigenvalues . this happens at rational values of @xmath491 . more precisely , a direct computation shows that @xmath571 for @xmath24 if and only if @xmath572 \(ii ) for any @xmath547 we have @xmath580 and @xmath581 now suppose @xmath582 ; then @xmath583 and therefore , combining and , we obtain @xmath584 if @xmath585 or equivalently @xmath586 then @xmath587 , and the required result follows . it follows that for @xmath484 the even resonances are located in the rectangles @xmath588 just as in the odd case , the following statement shows that for each @xmath544 and @xmath484 , the rectangle @xmath589 contains exactly one resonance . [ cequals02 ] if @xmath566 there is an even resonance of the algebraic multiplicity one at @xmath567 for every even @xmath544 and an even eigenvalue of multiplicity one at @xmath568 for every non - zero odd @xmath544 . there are no other even resonances . the proof follows from the explicit formula @xmath590 just as in the odd case , we obtain that the resonances are given by branches of real analytic functions of @xmath484 with values in @xmath589 . however , in contrast with the odd case , the height of the rectangles @xmath589 is not uniformly bounded in @xmath491 . moreover , we have suppose that the conclusion of the theorem is false . then there exists a sequence @xmath594 such that @xmath595 is bounded . by passing to a subsequence we can assume that @xmath596 as @xmath597 . this would imply that @xmath598 by the joint continuity of the function @xmath599 . this is impossible by . formal calculations and numerical analysis suggest that the rate of divergence of @xmath600 as @xmath500 is logarithmic . thus , all even resonances move off to infinity and this provides partial explanation for the failure of the weyl law for @xmath483 . figure [ figure2 ] plots a typical even resonance curve as @xmath491 increases from @xmath4 to @xmath185 . the curve starts at @xmath602 when @xmath574 and passes through @xmath603 when @xmath604 and diverges to @xmath605 as @xmath498 . j. aguilar , j .- combes , _ a class of analytic perturbations for one - body schrdinger operators , _ commun . math . * 22 * ( 1971 ) , 269279 . d. borthwick , _ sharp upper bounds for perturbations of hyperbolic space , _ preprint , arxiv:0910.2439 . g. carron , _ d ' eterminant relatif et la fonction xi , _ american j. of math . * 124 * ( 2002 ) no . 2 , 307352 . t. christiansen , _ asymptotics for a resonance - counting function for potential scattering on cylinders , _ j. funct . * 216 * ( 2004 ) 172190 . e. b. davies , _ eigenvalues of an elliptic system , _ zeit . math . * 243 * ( 2003 ) , 719743 . e. b. davies and p. a. incani , _ spectral properties of matrices associated with directed graphs , _ proc . london math . soc . * 100 * , no . 1 ( 2010 ) , 5590 . p. exner , j. lipovsk , _ equivalence of resolvent and scattering resonances on quantum graphs , _ adventures in mathematical physics ( proceedings , cergy - pontoise 2006 ) , * 447 * , providence , r.i . , ( 2007 ) , 7381 . p. exner , j. lipovsk , _ resonances from perturbations of quantum graphs with rationally related edges , _ j. phys . a * 43 * ( 2010 ) 105301 . r. froese , _ asymptotic distribution of resonances in one dimension , _ * 137 * ( 1997 ) 251272 . f. gesztesy , m. mitrea , m. zinchenko , _ on dirichlet - to - neumann maps and some applications to modified fredholm determinants , _ preprint , arxiv:1002.0389v2 . a. incani , _ spectral theory of directed graphs , _ phd thesis , king s college london , 2009 . e. korotyaev , _ inverse resonance scattering on the real line , _ inverse problems * 21 * ( 2005 ) , 325341 . v. kostrykin and r. schrader , _ kirchhoff s rule for quantum wires , _ j. phys . a : math . * 32 * ( 1999 ) , 595630 . v. kostrykin and r. schrader , _ laplacians on metric graphs : eigenvalues , resolvents and semigroups . _ quantum graphs and their applications , 201225 , contemp . , * 415 * , amer . math . soc . , providence , ri , 2006 . v. kostrykin and r. schrader , _ heat kernels on metric graphs and a trace formula . _ adventures in mathematical physics , 175198 , contemp . math . , * 447 * , amer . math . soc . , providence , ri , 2007 . p. kuchment , _ quantum graphs : i. some basic structures , _ waves in random media * 14 * no . 1 ( 2004 ) , p. kuchment , _ quantum graphs : an introduction and a brief survey , _ pp.291314 in ` analysis on graphs and its applications ' , proc . symp . pure . kuchment , _ quantum graphs ii , _ j. phys . a : math . gen . * 38 * ( 2005 ) , 48874900 . r. e. langer , _ on the zeros of exponential sums and integrals , _ bull . math . soc . * 37 * ( 1931 ) , 213239 . c. j. moreno , _ the zeros of exponential polynomials ( i ) , _ comp . math . * 26 * ( 1973 ) 6978 . b .- s . ong , _ on the limiting absorption principle and spectra of quantum graphs , _ contemp . , * 415 * , amer . soc . , providence , ri , 2006 . l. parnovski , _ spectral asymptotics of laplace operators on surfaces with cusps , _ math . ann . * 303 * ( 1995 ) 281296 . m. reed , b. simon , _ methods of modern mathematical physics , _ vol . 4 , acad . press , 1978 . b. simon , _ resonances in n - body quantum systems with dilatation analytic potentials and the foundations of time - dependent perturbation theory _ , ann . of math . * 97 * ( 1973 ) , 247274 . j. sjstrand , m. zworski , _ complex scaling and the distribution of scattering poles , _ soc . * 4 * ( 1991 ) 729769 . p. stefanov , _ sharp upper bounds on the number of the scattering poles , _ j. funct . 231*(1 ) ( 2006 ) , 111142 . m. zworski , _ distribution of poles for scattering on the real line , _ j. funct . 73 * ( 1987 ) 277296 . |
we evaluate the coorbital corotation torque on a migrating protoplanet . the coorbital torque is assumed to come from orbit crossing fluid elements which exchange angular momentum with the planet when they execute a u - turn at the end of horseshoe streamlines . when the planet migrates inward , the fluid elements of the inner disk undergo one such exchange as they pass to the outer disk . the angular momentum they gain is removed from the planet , and this corresponds to a negative contribution to the corotation torque , which scales with the drift rate . in addition , the material trapped in the coorbital region drifts radially with the planet giving a positive contribution to the corotation torque , which also scales with the drift rate . these two contributions do not cancel out if the coorbital region is depleted , in which case there is a net corotation torque which scales with the drift rate and the mass deficit in the coorbital region , and which has same sign as the drift rate . this leads to a positive feedback on the migrating planet . in particular , if the coorbital mass deficit is larger than the planet mass , the migration rate undergoes a runaway which can vary the protoplanet semi - major axis by % over a few tens of orbits . this can happen only if the planet mass is sufficient to create a dip or gap in its surrounding region , and if the surrounding disk mass is larger than the planet mass . this typically corresponds to planet masses in the sub - saturnian to jovian mass range embedded in massive protoplanetary disks . runaway migration is a good candidate to account for the orbital characteristics of close orbiting giant planets , most of which have sub - jovian masses . these are known to cluster at short periods whereas planets of greater than two jovian masses are rare at short periods indicating a different type of migration process operated for the two classes of object . further , we show that in the runaway regime , migration can be directed outwards , which makes this regime potentially rich in a variety of important effects in shaping a planetary system during the last stages of its formation . | astro-ph0301171 | the study of the tidally induced migration of protoplanets embedded in protoplanetary disks has received renewed attention in the last few years following the discovery of extrasolar giant planets ( hereafter egps ) . it is in particular the best candidate to explain the short period egps ( the so - called hot jupiters ) which are likely to have begun to form further out in the disk and migrated radially inwards . when the planet mass is small ( i.e. when its hill radius is much smaller than the disk thickness ) , the migration rate can be evaluated using linear analysis , and is shown to be proportional to the planet mass , to the disk surface density , and inversely proportional to the square of the disk aspect ratio @xcite . the linear regime is often called the type i regime . it corresponds to a fast migration rate , although recent estimates @xcite show that the linear analytical estimate assuming a flat two dimensional disk has to be reduced by a factor @xmath1@xmath2 or more in a more realistic calculation which accounts for the disk vertical structure and a possible non - saturation of the corotation torque if the disk is viscous enough . migration in the type i regime is nevertheless still too fast , in the sense that the migration timescale it yields is shorter than the build - up timescale of a giant protocore ( see e. g. * ? ? ? we shall not address this issue here but rather assume that a giant planet embryo can form in the disk at a distance @xmath3 au , and with a mass @xmath4 , where @xmath5 @xmath6 is the critical mass above which rapid gas accretion begins . when this embryo mass is large enough , it enters another well studied migration regime , called type ii migration @xcite . in this regime , the protoplanet has a mass sufficient to open a gap in the disk , which is therefore split into an inner disk and an outer disk . the protoplanet then finds itself locked into the disk viscous evolution drifting inwards with it @xcite . as the protoplanet undergoes type ii migration towards the central object , it may accrete the surrounding nebula material . the accretion rate scales with the mass accretion rate onto the central object @xmath7 where @xmath8 is the disk surface density . here one assumes that the processes at work in the disk which contribute to the angular momentum exchange between neighboring rings can be adequately modeled by a phenomenological kinematic viscosity @xmath9 . on the other hand , the type ii migration timescale is of the order of the disk viscous timescale @xmath10 . the maximum mass that a giant protoplanet can accrete on its way to the central object should be therefore of the order of @xmath11 , that is of the order of the disk mass interior to the planet starting distance . noticeably this mass does not depend on the disk viscosity . if the planet does not migrate all the way to the central object before the disk is dispersed , then because more time is spent at larger radii , it is most likely to be left with a semi - major axis larger than the typical one for hot jupiters ( @xmath12@xmath13 au , see * ? ? ? this is consistent with the observed paucity of planets with masses exceeding two jovian masses at short periods @xcite . note too that planets undergoing type ii migration should tend to have higher masses at shorter periods . this is contrary to the observed trend . furthermore , as the planet mass grows , it becomes eventually larger than the surrounding disk mass , and its migration rate tends to drop , as the disk can not remove enough angular momentum from it . this has been investigated by @xcite . in this case , the amount of time necessary to bring the planet to a close orbit can be considerably larger than the disk viscous timescale , and can even exceed the disk lifetime . this migration slowing can be seen in the simulations of @xcite . it occurs _ soon _ after the planet has entered its type ii migration regime . from the above considerations , it is questionable whether the orbital characteristics of most close orbiters can be accounted for by type ii migration driven by the evolution of the disk . furthermore the vast majority of these have sub - jovian masses ( here one excepts and , which have large masses and which may have had a different origin ) . depending on the physical conditions of the protoplanetary disks in which they formed , they may not have fulfilled the gap opening criteria ( * ? ? ? * and references therein ) , with the consequence that they may have been involved in a migration regime intermediate between type i and type ii . this transitional regime has received little attention . @xcite and @xcite have worked out the feedback on the migration rate due to the nebula surface density profile perturbation under the action of the protoplanet lindblad torques . they introduced the concept of an inertial limit , that is the maximum mass of a protoplanet that can undergo steady state migration . it was suggested that masses above the inertial limit lead to a gap opening and to type ii migration . in their analysis , @xcite neglected the coorbital dynamics and the corotation torque it implies on the migrating planet . the purpose of this work is to give an evaluation of the corotation torque for a migrating planet , and to analyze its consequences on migration . we define the notation in section [ sec : notation ] , we present an appropriate expression for the corotation torque for a planet held on a fixed circular orbit in section [ sec : corotfix ] , we then derive the corotation torque for a migrating planet in section [ sec : derivation ] , and illustrate its properties using customized numerical simulations in section [ sec : numres ] . as we investigate the intermediate regime between type i and type ii migration , for which the disk response is affected by non - linear effects , the hill radius and the disk thickness have comparable orders of magnitude . the regime of interest thus involves mildly embedded protoplanets . we assume that it can be safely studied through two dimensional flat disk calculations provided a reasonable value is adopted for the gravitational potential smoothing length . in section [ sec : discussion ] we demonstrate the existence of runaway migration and relate the condition for its occurrence to the analytic discussion in section [ sec : derivation ] . finally in section [ sec : summary ] we summarize our results and we discuss their application to egps . we adopt a cylindrical coordinate system @xmath14 centered on the primary , the origin of which corotates with the planet which is located at azimuth @xmath15 . we consider a thin gaseous disk with mid - plane at @xmath16 , with surface density @xmath17 , orbiting a central point - like object of mass @xmath18 . the associated keplerian frequency is @xmath19 . in the unperturbed disk the orbital frequency is @xmath20 , which is usually slightly smaller than @xmath21 because the central acceleration is not only compensated for by the centrifugal acceleration , but also by a radial pressure gradient . the disk thickness is @xmath22 and the disk aspect ratio is @xmath23 . in this disk we consider an embedded planet with mass @xmath24 ( we note @xmath25 ) and semi - major axis @xmath26 . its orbital frequency is @xmath27 . we shall restrict ourselves to the case where its eccentricity @xmath28 can be neglected . we now write the angular frequency as @xmath29 , and we denote by @xmath30 the radial velocity of the fluid element at @xmath31 . the link between the corotation torque and the horseshoe orbit drag has been indicated by @xcite . a torque estimate for a planet on a fixed circular orbit embedded in a viscous disk has been given by @xcite who considered a steady flow as seen in the planet frame . we here derive an expression for the corotation torque in a disk with very small viscosity . then the specific vorticity is conserved along a streamline ( * ? ? ? * and references therein ) . we assume a steady flow in the planet frame . as we assume that the viscosity is low , we can write a bernoulli invariant along a streamline : @xmath32 where @xmath33 is the fluid specific enthalpy and @xmath34 , with @xmath35 being the gravitational potential . the bernoulli invariant is a useful label of the librating streamlines in the horseshoe region . the corotation torque expression , following @xcite , can be obtained by summing the contribution to the torque due to individual fluid elements over the horseshoe annular slab . the positive contribution coming from the outer fluid elements caught up by the planet , when they execute a u - turn in front of the latter , reads : @xmath36dr_+,\ ] ] where @xmath37 is the horseshoe zone half - width , @xmath38 is the specific angular momentum , @xmath39 is the corotation radius , and where we add an index @xmath40/ @xmath41 to any quantity to refer to its value on the outer/ inner part of its horseshoe streamline . the gradient of the bernoulli invariant is linked to the flow vorticity ( see e. g. * ? ? ? if we denote by @xmath42 the vertical component of the flow vorticity in the inertial frame , then we have : @xmath43 where we use the fact that @xmath44 , and where we evaluate this expression sufficiently far from the planet so that we can assume @xmath45 and so that we can neglect the dependency of @xmath46 upon @xmath47 . this allows one to transform eq . ( [ eq : tqp ] ) as : @xmath48 where we note @xmath49 the specific angular momentum drop of a fluid element as it switches from the outer horseshoe leg to the inner one . if @xmath50 is the inverse of the specific vorticity , then : @xmath51 where @xmath52 and @xmath53 are respectively the first and second oort s constants , evaluated at the planet orbit . here we assume that the planet mass is small enough that we can consider that a fluid element on an outer horseshoe leg at @xmath54 is mapped on the inner leg with radius @xmath55 , we have developed @xmath56 to first order in @xmath57 [ @xmath58 , and we have assumed @xmath59 . a similar treatment for the inner horseshoe leg yields the torque exerted on the protoplanet by the fluid elements which catch the planet up and are promoted to higher specific angular momentum orbits : @xmath60 it should be noted that the corotation torque , in the non - linear regime ( i. e. for @xmath37 finite ) should include all the fluid elements which corotate , in average , with the planet , that is to say all the fluid elements which librate in the corotating frame . this includes not only the horseshoe streamlines , but also the circumplanetary disk ( corresponding to the closed streamlines interior to the roche lobe ) . we find it more convenient to consider only the horseshoe drag exerted on the system \{planet + circumplanetary disk } , considered as a whole , and hereafter the planet should be understood as this system . we shall now use eq . ( [ eq : tqp_u ] ) and ( [ eq : tqm_u ] ) to evaluate the torque on a migrating object . we separate the orbital timescale @xmath61 ( which is also the horseshoe u - turn timescale from one leg to another ) from the horseshoe libration timescale @xmath62 , which is much longer . in particular , we neglect migration over the orbital timescale , while we consider it over the libration timescale . we say that migration is slow whenever the libration time of an outermost fluid element close to the separatrix between librating and circulating streamlines is short compared to the migration time across the horseshoe zone half width . as the former quantity is @xmath63 and the latter @xmath64 , the condition for slow migration reads : @xmath65 is the orbital time scale . we assume that eqs . ( [ eq : tqp_u ] ) and ( [ eq : tqm_u ] ) are still valid , provided care is taken about the evaluation of @xmath66 in particular , as the planet migrates , the `` impact parameter '' @xmath67 of a fluid element varies as its azimuth varies . we therefore have to consider the outer incident fluid elements at an azimuth @xmath68 . this azimuth needs to be small enough so that no significant radial drift occurs between the fluid element and the planet before the close encounter , and large enough so that the close encounter has not begun yet . similarly , the azimuth at which the incident inner fluid elements need to be considered has to be @xmath69 . we use the index @xmath70/ @xmath71 for the quantities relating to the close encounters originating from @xmath68 / @xmath69 ) . the positive part of the corotation torque is given by : @xmath72 while the negative part of the corotation torque is given by : @xmath73 note that we assume the value of @xmath37 to be the same as in the non - migrating case . although this needs to be reconsidered for a significant drift rate , this is surely true as long as migration is slow ( with the meaning defined earlier in this section ) . the numerical simulations presented in the next section will be used to check this assumption . the sum of eqs . ( [ eq : tqwp ] ) and ( [ eq : tqwm ] ) can be transformed so that values of @xmath74 and @xmath75 on the same side of the orbit are considered . care has to be taken about the sign of @xmath76 before making this transformation . we shall assume hereafter that @xmath77 . hence : @xmath78x^2dx,\ ] ] where we use the fact that @xmath79 for any @xmath57 in @xmath80 $ ] , since all of the fluid elements which execute a right u - turn are trapped in the coorbital region ( see also figs . [ fig : appendix ] and [ fig : appendix2 ] in appendix [ apa ] ) . we shall now temporarily assume a steady migration case ( @xmath81 ) . we call @xmath82 the mapping of a fluid element , between two close encounters with the planet . this notation is illustrated in fig . [ fig : interp ] . consider a fluid element initially located at the distance @xmath57 of the orbit just after a close encounter . its distance to the orbit just before the next close encounter is : @xmath83 where we assume that @xmath84 ( i. e. slow migration ) . within this approximation , the reciprocal map @xmath85 reads : @xmath86 because of the conservation of specific vorticity , one can write : @xmath87 , for any @xmath57 in @xmath88 $ ] ( [ eq : tqinter ] ) therefore reads : @xmath89 where @xmath90 , and where we have taken @xmath91 as an upper limit for both integrals , since the contribution of material close to the orbit is vanishingly small for slow migration is not mapped onto @xmath92 , and one should write @xmath93 as an upper limit of the l.h.s . integral of eq . ( [ eq : inter1 ] ) instead of @xmath94 , where @xmath93 is defined by @xmath95 . one finds @xmath96 . the contribution of material in the @xmath97 $ ] region is therefore @xmath98 , and scales therefore as @xmath99 , whereas the effect of migration on the corotation torque is in @xmath76 . the relative contribution of this innermost material is therefore vanishingly small for slow migration , and it is safe to write @xmath94 as an upper limit for the l.h.s . integral . ] . expanding the integrand to first order in @xmath100 , one can write the corotation torque given by eq . ( [ eq : tqinter ] ) as : @xmath101,\label{eq : sm1}\end{aligned}\ ] ] where the @xmath102 index stands for `` steady migration '' . ( [ eq : sm1 ] ) can be transformed into : @xmath103 . \label{eq : tqmod}\end{aligned}\ ] ] the first integral ( over @xmath92 , from @xmath104 to @xmath94 ) corresponds to the material which librates in the horseshoe region ( white trapezoidal area in fig . [ fig : interp ] ) . the corresponding torque expression is : @xmath105 where @xmath106 , the `` vorticity weighted coorbital mass '' , is defined as : @xmath107 this component @xmath108 of the torque arises because the librating fluid elements migrate radially with the planet and have to lose specific angular momentum at the same rate as this latter . for the case of an inward migration , the torque exerted on this region by the planet is negative . it exerts therefore a positive torque on the planet , and thus has a negative feed back on migration . a similar conclusion applies for the case of an outwards migration . the second integral in eq . ( [ eq : tqmod ] ) can be evaluated assuming @xmath109 ( i. e. slow migration ) , in which case it reduces to : @xmath110 as this torque corresponds to the integral over @xmath111 , it comes from the fluid elements of the hashed area of fig . [ fig : interp ] . these fluid elements are promoted to higher specific angular momentum trajectories after their ( unique ) close encounter with the planet . for the case of inwards migration , they therefore contribute negatively to the corotation torque and hence exert a positive feed back on migration . when one adds eqs . ( [ eq : gamma1 ] ) and ( [ eq : gamma2 ] ) , one gets the following torque expression : @xmath112 where we introduce the `` vorticity weighted coorbital mass deficit '' defined as : @xmath113b_p\nonumber\\ & = & \displaystyle 4\pi a\left[x_s\frac{\sigma_r(-x_s)}{b(-x_s)}- \int_{-x_s}^0\frac{\sigma_r(x)}{b(x)}dx\right]b_p . \label{eq : cmd}\end{aligned}\ ] ] if one neglects the radial variation of @xmath114 across the horseshoe region , the coorbital mass deficit appears as the mass difference between the mass of the horseshoe region , if the material in it had everywhere the surface density that it has at the inner separatrix , with its actual mass . as the coorbital region is generally depleted , this coorbital mass deficit is positive . one can then get an estimate of the torque for a non - steady migration using the following simple argument : if @xmath76 varies , the mass flux across the upstream separatrix varies , and this will have an impact on the torque exerted on the planet when the deficit or excess of in - flowing mass w.r.t . a steady state situation undergoes a close encounter with the planet , i. e. after a characteristic time equal to @xmath115 , where @xmath116 is half the libration time , i. e. the time a fluid element at a distance @xmath117 from the orbit needs to go from azimuth @xmath118 to azimuth @xmath119 . therefore @xmath120 is the time that a fluid element opposite the planet needs to drift to attain conjunction with it . one can therefore write : @xmath121\nonumber\\ & = & \gamma_{\rm sm}(\dot a)-\tau_{\rm lag}\ddot a\frac{d\gamma_{\rm sm}}{d\dot a}\nonumber\\ & = & 2b_p\,\delta m\ , a\dot a-\frac{\pi a^2b_p}{|a_p|x_s}\,\delta m\,\ddot a.\end{aligned}\ ] ] an exact method to evaluate the @xmath122 term in the slow migration limit , when one knows the specific vorticity profile across the horseshoe region , is provided in appendix [ apa ] . it can be noted that this torque expression cancels out for a planet held on a fixed circular orbit ( @xmath123 ) . this is expected as we have used @xmath124 , i. e. that the specific vorticity is conserved along a streamline . hence in the absence of migration , the torque is saturated . we could have used another dependency of @xmath125 on @xmath126 , involving a radial gradient of specific vorticity , which would have led to a constant term which we would interpret as the `` static '' part of the corotation torque . however , as our concern is to capture the planet drift effects on the torque , this would not have brought further insight regarding the runaway process we aim at characterizing . we note @xmath127 the other torque that is applied to the planet , which corresponds to non - librating ( i. e. , circulating ) material , which we assume to correspond to the differential lindblad torque . assuming that migration occurs with a negligible eccentricity , we have : @xmath128 this takes also the following form : @xmath129 which also takes the following form in the keplerian case : @xmath130 assuming that the @xmath76 variations , if any , occur on a time interval short enough to consider @xmath26 as a constant , one gets two different behaviors from eq . ( [ eq : adot1 ] ) : 1 . @xmath131 : the coorbital mass deficit is smaller than the planet mass , which is the case either for sufficiently low planet masses or for a large planet mass , when the planet mass is comparable to or larger than the surrounding disk . in that case the coorbital mass deficit can not become larger than the planet mass . then the corresponding homogeneous equation indicates that disturbances to @xmath26 are damped on the short timescale @xmath132 one can therefore discard transient behavior , retaining only the standard first order ode for migration : @xmath133 where the only difference with the usual expression is that we replace the planet mass alone @xmath24 by the `` planet effective mass '' @xmath134 . @xmath135 : the planet mass is large enough to open a significant dip in the disk , and this latter is substantial enough for the coorbital mass deficit to be larger than the planet mass . in that case the homogeneous ode associated to eq . ( [ eq : adot1 ] ) indicates that small perturbations to @xmath26 are exponentially growing on a timescale @xmath136 , which although depending on the exact value of @xmath137 , is of the order of a few tens of orbital periods ( assuming , as is reasonable for a mildly embedded object in a typical protoplanetary disk , that @xmath138 ) . the assumptions that we made to derive of eq . ( [ eq : tqexp ] ) , namely migration slow enough that the dip profile drifts instantaneously with the planet , and slow enough that the horseshoe zone crossing time is much larger than the libration time , rapidly break down , and one can only say at this point that this runaway regime is extremely fast , and that it occurs for @xmath139 . the actual behavior of a protoplanet in this regime has to be assessed through numerical simulations . since the ultimate sign of @xmath122 depends on the initial values of @xmath26 and @xmath76 , runaway can in principle occur outwards under specific initial conditions which need to be specified . we performed a series of dedicated numerical simulations to test the runaway regime and the validity of eq . ( [ eq : ode1 ] ) . the code that we used has already been described elsewhere ( see e. g. * ? ? ? * ) . as this code is an eulerian grid - based code , it must fulfill the courant condition on the timestep to ensure numerical stability . an improved algorithm resulting in a less demanding cfl condition , with the average azimuthal velocity at each radius subtracted out @xcite was used in order to increase the time step and speed up the code . the grid corotates with the guiding center of the planet osculating orbit . as a result , the planet motion with respect to the grid is slow , and mainly corresponds to a radial drift . our rotating frame is angularly accelerated . the corresponding acceleration , that is @xmath140 , is applied in much the same way as @xcite handles the coriolis acceleration in a rotating frame , so as to enforce angular momentum conservation . as these dedicated runs involve as accurate as possible a torque evaluation , we used non - reflecting boundary conditions so as to eliminate any reflected wave ( which can bring back to the planet the angular momentum previously removed from it ) and we used an initial profile with uniform specific vorticity ( i. e. with @xmath141 , which leads to a constant drift rate @xmath76 up to the center , if one only considers the differential lindblad torque ) . our mesh consists of @xmath142 sectors equally spaced in azimuth , divided radially in @xmath143 zones , the successive radii of zone boundaries being in geometric progression . our unit of mass is the central object mass @xmath18 , our unit of length is the initial planet semi - major axis @xmath144 and our unit of time is @xmath145 . in this system of units , the gravitational constant is therefore @xmath146 . the mesh outer boundary lies at @xmath147 and the inner boundary at @xmath148 . we use a uniform aspect ratio disk with @xmath149 . the resolution of our mesh is barely sufficient to accurately describe the differential lindblad torque acting on the planet . on the other hand , it is enough for a proper description of coorbital effects @xcite . our planet mass is @xmath150 ( which corresponds to a saturn mass planet if the central star has a solar mass ) . we used a number of disk surface density profiles @xmath151 with the following values @xmath152 , @xmath153 , @xmath154 , @xmath155 , @xmath156 , @xmath157 , @xmath158 , @xmath159 and @xmath160 . we used a uniform kinematic viscosity @xmath161 throughout the disk . the planet was initially held on a fixed circular orbit for @xmath162 orbits ( @xmath163 ) in order to give it a sufficient time to open a dip / gap around its orbit . this creates a depression of the specific vorticity profile which can lead to the effect we described in section 2 . it should be noted that the coorbital mass deficit that arises during the first @xmath162 orbits just scales with the disk surface density , that is to say : @xmath164 , where @xmath165 . ( [ eq : ode1 ] ) can therefore be written as : @xmath166 where the constants @xmath167 and @xmath168 both depend on @xmath24 and the disk parameters , but this dependency does not need to be considered here as we only vary the disk surface density in these runs . ( [ eq : d1 ] ) can be written as : @xmath169 where we introduce the critical surface density @xmath170 for migration runaway , with @xmath171 being a proportionality constant which is simply related to the differential lindblad torque , as one can see by letting @xmath172 . the novelty here is that the migration rate below the runaway limit grows faster than linearly with the disk surface density . our simulations aim at testing this super - linearity by checking whether the drift rate fulfills eq . ( [ eq : check ] ) or not . when it does , we determine the critical disk surface density for runaway and determine whether we indeed get a runaway for larger surface densities . we also check that the critical surface density so determined is consistent with coorbital mass deficit estimates . the protoplanet potential in the runs presented here is smoothed using a softening parameter @xmath173 , where @xmath174 is the disk thickness and @xmath175 . the results turn out to be sensitive to the value of the softening parameter because the horseshoe zone width crucially depends on it . lowering the softening parameter moves the separatrices away from the orbit . as a result , the coorbital mass deficit is increased , and the critical surface density for runaway is reduced . however , sensitivity to the softening parameter in a numerical simulation does not necessarily imply that the region within the roche lobe matters , but rather that coorbital processes matter , as their effectiveness depends strongly on the horseshoe zone width which is linked to it ( see * ? ? ? * for details ) . note that the discussion of the softening parameter by @xcite is valid only for the case when the specific vorticity has a linear dependence on the distance to corotation ( in which case the corotation torque scales as @xmath176 ) . here the specific vorticity profile has a depleted , more complex profile , and the analysis is no longer valid . we used a softening parameter comparable to the one derived by @xcite , which was found to give reasonable results for the case of mildly embedded planets . it should be kept in mind that if a smaller softening parameter is used , the critical surface density for runaway that we shall discuss later would be even lower than what we found . in that sense the extent of the runaway regime that we shall delineate later in this work can be considered as a conservative estimate . we show the temporal behavior of the planet semi - major axis for our nine runs at fig [ fig : a - vs - t ] . the origin of time is chosen at the planet release . the first @xmath162 orbits are therefore not shown . one can already notice that the radial drift rate is not proportional to the disk mass . the heaviest disk is indeed `` only '' @xmath177 times heavier than the lightest one we consider . for the lightest disk the planet migrates about @xmath178 % inwards over @xmath179 orbits , whereas for the heaviest disk the planet is already at @xmath180 after about @xmath181 orbits . [ fig : da - vs - t ] shows the migration rate as a function of time . we see that it is relatively constant for the runs @xmath182 to @xmath183 , that a slight global variation can be seen for the run @xmath184 , and that the other runs , @xmath185 to @xmath186 , display a strongly variable migration rate , which peaks at very large values . some oscillations in the migration rate for @xmath182 to @xmath185 can be seen . these can be identified with the planet crossing the boundaries of mesh zones ( i. e. the period for these oscillations is @xmath187 , where @xmath188 is the radial zone size , which explains why these oscillations are slower for the smaller migration rates ) . this also gives us an idea of the accuracy of the numerical scheme and of the torque dependency on the planet placement with respect to a mesh zone . this accuracy is satisfactory except for the run @xmath185 which exhibits large amplitude variations , but these are likely not relevant since the planet is then close to the grid inner boundary . we can check that the drift rate grows faster than linearly with the disk surface density for runs @xmath182 to @xmath184 . we estimate the average migration rate for these runs , over the whole time interval for runs @xmath182 to @xmath183 , and over the time interval @xmath177 to @xmath189 orbits for the run @xmath184 ( we discuss this choice later ) . [ fig : rel ] shows @xmath190 as a function of @xmath191 . the points are relatively well aligned , as was expected from eq . ( [ eq : check ] ) . a linear regression fit allows one to determine the critical surface density for runaway , and yields @xmath192 . the runs @xmath185 to @xmath186 , which correspond to disk surface densities larger than this runaway threshold , do indeed exhibit a very fast migration and a strongly time variable migration rate . one can also understand the time behavior of the drift rate in the first orbits of fig . [ fig : da - vs - t ] . as can be seen in eq . ( [ eq : estab ] ) , the timescale @xmath136 , over which the migration rate tends to its limiting value given by eq . ( [ eq : ode1 ] ) , increases when the disk surface density approaches its critical value . this is precisely the trend that we see in our runs . this is the reason we took @xmath177 orbits as a lower time value for estimating the average migration rate for run @xmath184 . the higher time value of @xmath189 orbits comes from the fact that by then the planet has already migrated a sizable fraction of its initial distance to the star such that its coorbital mass deficit may have been significantly altered . we finally check whether the critical surface density corresponds to a coorbital mass deficit comparable to the planet mass . we display at fig . [ fig : sl ] the surface density in the coorbital region ( in a @xmath193-plane ) and we superimpose a few streamlines . this allows us to get an estimate of the position of the separatrices , which we find at @xmath194 . once one knows the location of the separatrices , one can estimate the coorbital mass deficit . [ fig : cmd ] shows the inverse specific vorticity profile for run @xmath195 , from which one can estimate a coorbital mass deficit @xmath196 . this latter needs therefore to be @xmath197 times more massive to fulfill the runaway condition , from which we conclude that the runaway should occur for @xmath198 . this value is @xmath199 % larger than the value inferred from the linear regression fit . the agreement is not surprisingly rough , probably owing to the various assumptions made to derive eq . ( [ eq : cmd ] ) . in addition to the main runs involving freely migrating planets , we have performed a series of additional runs in which the migration rate @xmath76 is fixed in order to check the behavior described by eq . ( [ eq : tqexp ] ) . namely , in these runs , the planet was held on a fixed @xmath76 trajectory for 200 orbits . the run was started with semi major axis @xmath200 and ended with semi major axis @xmath201 , so that in all the runs the planet has semi major axis @xmath202 at @xmath203 orbits . the planet is given an instantaneous orbital frequency equal to @xmath204 . the runs are performed as before in the frame corotating with the guiding center . the resolution and other numerical values are unchanged with respect to what is described in the previous section . the values adopted for @xmath76 are @xmath205 , where @xmath206 is an integer ranging from @xmath207 to @xmath208 . the surface density in all these runs is @xmath209 , corresponding to @xmath210 . [ fig : gammaadot1 ] shows the torque as a function of time for each of the ten runs performed , and fig . [ fig : gvad ] shows the average value of the torque as a function of @xmath76 ( the average is taken over the time interval @xmath211 $ ] orbits , which eliminates initial transient behavior occurring over the first @xmath212 orbits , and which ensures that the average semi major axis of the planet over this time interval is the same for all the runs , and is @xmath213 . ) the torque modulation as the planet sweeps radially the mesh zones is again apparent in fig . [ fig : gammaadot1 ] . these plots confirm the linear dependence of the total torque upon the migration rate . one can check that the slope of @xmath214 as a function of @xmath76 is positive , i. e. the feed back is positive , and one can infer from the slope estimate the critical disk surface density for runaway . we note @xmath215 the specific torque acting on the planet . we can write , since @xmath81 : @xmath216 we also have the relationship : @xmath217 since @xmath218 scales with @xmath8 , and since the runaway starts for @xmath219 . if we call @xmath220 the slope of @xmath221 as a function of @xmath76 , then we have : @xmath222 the linear regression fit displayed at fig . [ fig : gvad ] yields @xmath223 , from which one infers @xmath224 , in good agreement ( within @xmath225 % ) with the estimate given by the linear regression fit of fig . [ fig : rel ] . this is however only in rough agreement with the result given by the coorbital mass deficit estimate . these additional runs also enable one to check whether the assumption that the horseshoe zone width does not depend upon the drift rate is valid or not . we evaluate the distance @xmath37 of the inner separatrix to the orbit at time @xmath203 orbits for all the runs . we find that @xmath226 and that this quantity exhibits no systematic trend with @xmath76 . the previous section illustrates the linear dependency of the coorbital corotation torque on the drift rate in the slow migration regime ( i. e. @xmath227 ) . in order to investigate the fast migration regime , and in particular in order to get an idea of the migration rates that can be achieved in a runaway episode , it is of interest to know the @xmath228 relationship for large values of @xmath76 . the problem in that regime is that if an external operator imposes a fixed large drift rate to the planet , this latter sweeps a sizable fraction of its initial orbital radius in a very short time , and no reliable value can be measured for the corresponding torque . a workaround can be found as follows . as the torque that we aim at measuring arises from a relative drift of the disk material with respect to the planet horseshoe region , one can mimic this drift by adding an external specific torque to the disk material in order to make the fluid elements drift radially , while the planet is held on a fixed circular orbit . that way a steady state situation can be achieved , which allows a precise measurement of the torque even for very large drift rates . namely , we performed a number of runs for which : 1 . the planet is held on a fixed circular orbit with radius @xmath202 , 2 . the disk material undergoes an additional , external specific torque , with expression : @xmath229 where @xmath230 is the disk material radial drift velocity at the planet orbit and @xmath231 the unperturbed surface density ; this expression ensures that the radial drift velocity in an axisymmetric situation , that is @xmath232 , corresponds to a steady state situation ( i. e. that @xmath233/\partial r=0 $ ] ) , 3 . a source of disk material with the adequate surface density is set at the grid outer / inner boundary for @xmath234 / @xmath235 . this ensures that no disk depletion occurs at large @xmath230 , which would modify the surface density profile and therefore would affect the torque value . this torque prescription leads to the radial drift of any structure in the surface density profile in much the same way as for a viscous drift , but contrary to this latter it does not lead to a radial spread of the profiles . we ran @xmath236 such configurations , with @xmath237 ( @xmath238 ) . as this corresponds to an inwards drift of the disk material , the upstream separatrix is the outer separatrix , and therefore the torque value should correspond to the ones measured at the previous section for positive values @xmath239 . the results are presented at figs . [ fig : large_adot ] and [ fig : large_adot2 ] . clearly there is a satisfactory agreement between these results and the results obtained at the previous section for the case of true slow migration , as can be seen on the left part of the plots . this validates this method as an alternate way of measuring the corotation torque dependence on the migration rate . this agreement can be understood using similar arguments as the ones used by @xcite to evaluate the corotation torque in a viscous disk . the librating fluid elements define a trapped region , the angular momentum of which is therefore constant in time . the external torque applied on this region is therefore exactly transmitted to the planet in a steady state situation . one can easily show that to lowest order in @xmath240 an expression similar to eq . ( [ eq : tqexp_one ] ) is obtained for the corotation torque . the agreement between the measured torques for a drifting planet and for a backward drifting disk , as well as the similarity of the flow topology in the @xmath241 plane in either case , suggests that the torque measurement for a drifting disk , even in the fast migration limit , gives a reasonable idea of the behavior of the corotation torque on a migrating planet . this new method also offers the advantage that the planet is fixed with respect to the grid , and therefore there is no torque modulation as observed in fig . [ fig : gammaadot1 ] when the planet sweeps the mesh zones , and thus it enables us to get a more precise estimate of the disk torque . the critical drift rate for fast migration is : @xmath242 below this rate , all of the disk fluid elements crossing the upstream separatrix undergo a horseshoe like close encounter with the planet , and contribute to the corotation torque , while above this rate , some of them miss the planet . in the fast regime , which corresponds to the right part of the plots of figs . [ fig : large_adot ] and [ fig : large_adot2 ] , the corotation torque reaches a maximum value and then slowly decays , while its characteristic order of magnitude is : @xmath243 in the series of runs presented in section [ sec : results_run_1 ] , @xmath218 scales with the disk surface density . the maximum drift rate should therefore roughly scale with the disk surface density during a runaway episode . this is in agreement with the results displayed at fig . [ fig : da - vs - t ] . since in the runaway regime the differential equation governing the time evolution of the planet semi - major axis is second order in time , it is formally possible to have an outward migration for an adequate choice of the initial @xmath26 and @xmath76 . for an outward migration , the upstream separatrix is the outer one . the higher the ( inverse of ) specific vorticity jump across that separatrix , the easier it is to get an outward migration . therefore , outwards migration should be easier to get for shallower surface density profiles ( corresponding to steeply increasing profiles of @xmath244 ) . the weakening of the differential lindblad torque for shallower surface density profiles @xcite plays in the same direction , since this torque tends to favor inward runaway rather than outward runaway . in order to illustrate this trend , we have performed a series of runs in which we hold the planet on a fixed @xmath245 orbit for @xmath212 orbital times , and then release it ( i. e. we allow it to freely migrate under the action of the disk torque ) . the planet mass , disk aspect ratio and viscosity , the grid resolution and the numerical algorithm were strictly the same as in section [ sec : results_run_1 ] , the disk surface density was @xmath246 , corresponding for @xmath247 to @xmath248 , i. e. approximately @xmath0 % above the runaway critical surface density . we tried four values for @xmath249 : @xmath250 , @xmath251 , @xmath252 , and @xmath94 , corresponding to an increasingly shallower surface density profile . the starting semi - major axis is @xmath253 , and the semi - major axis at the time of release ( @xmath254 orbital times ) is @xmath202 , which ensures that in the four runs the disk surface density at the planet orbit is the same at the release time , and only differs by its slope . we see in fig . [ fig : outwards ] that the planet undergoes an inward runaway for the two steepest surface density profiles , and an outwards runaway for the two shallowest . the maximum @xmath76 is of the same order of magnitude in the four cases , and the runaway starting time , corresponding to the short exponential regime in which migration can be considered as slow ( @xmath255 ) , is of the order of ten orbits , comparable to the outermost horseshoe libration time . this illustrates that a common mechanism is at work for the inward and outward runaways , and that these are tightly linked to the coorbital dynamics . in order to assess the importance of runaway migration in protoplanetary nebulae , we have tried to delineate the runaway migration domain borders in a ( planet mass , disk mass ) space , while keeping the disk aspect ratio and viscosity fixed . all our disks have @xmath256 , and a viscosity @xmath257 . we tried three values of the disk aspect ratio : @xmath258 , @xmath259 and @xmath12 , corresponding respectively to values for the @xmath249 parameter @xmath260 , @xmath261 and @xmath262 . we call disk mass the quantity @xmath263 , and disk reduced mass the quantity @xmath264 . the critical disk mass for runaway depends on the planet mass , as it depends on the dip shape around the orbit and on the position of the separatrices . in order to determine the disk critical mass , we measure the disk torque exerted on the planet , which is held on a fixed circular orbit , for the case where we exert no additional torque on the disk material ( we do not impose any additional disk radial drift other than the one arising from its viscous evolution ) , and for the case where we exert an additional torque on the disk material ( which corresponds to imposing an additional radial drift of the disk material with velocity @xmath230 ) . in the first case we measure a torque @xmath214 , and in the second case a torque @xmath265 . from eq . ( [ eq : tqexp ] ) and section [ sec : fast ] , these two torques can be written respectively as : @xmath266 where @xmath127 , the differential lindblad torque , is assumed to be independent of the disk radial drift velocity , and where the `` static '' part of the corotation torque vanishes since we consider initially a uniform specific vorticity disk . as the planet is held on a fixed circular orbit and a stationary state is reached , one can use eq . ( [ eq : deltamsigma ] ) , which yields : @xmath267 where @xmath215 and @xmath268 . the value @xmath230 must be chosen small enough so that it corresponds to the slow migration limit , and large enough that it allows an accurate measurement of @xmath269 . preliminary runs have shown that an accurate estimate of the critical surface density can be reached with a value of @xmath230 as small as @xmath270 , corresponding to a disk radial drift which amounts to less than one zone radial width over the whole simulation , which illustrates the fact that even a small resolution grid , with a reduced number of zones across the horseshoe region , captures remarkably well most of the features of the corotation torque , as noted by @xcite . the measurement of @xmath271 and @xmath272 can be performed in two different ways : * either we perform two different runs with constant values of the disk drift , @xmath94 and @xmath230 ( method 1 ) , * or we perform one run with a vanishing additional disk drift , and once a steady state is reached we switch the drift to @xmath230 . the new torque value can then be measured after a horseshoe libration time ( method 2 ) . the first method is better suited to small mass planets , for which the libration time is prohibitively long , while the second method is well suited for higher mass planets , since it allows almost a 50 % saving of cpu time compared to the first method . for intermediate , saturn mass planets we used both methods to check that they give comparable results . the results are presented at figs . [ fig : rwlim1 ] , [ fig : rwlim2 ] and [ fig : rwlim3 ] . these plots lead to a number of comments : * the thinner the disk , the larger the runaway domain . this corresponds to expectations : for a given planet mass and disk viscosity ( such that the planet mass is smaller than the viscous gap opening criterion ) , the thinner the disk , the deeper the dip opened around the orbit , and therefore the larger the coorbital mass deficit . * in the three cases , the mass most favorable to runaway ( corresponding to the minimum of the critical disk mass curve ) is @xmath273 @xmath274 , or typically a saturn mass . * runaway migration can be found in relatively massive protoplanetary disks ( a few times more massive than the minimum mass solar nebula , depending on the protogiant semi - major axis ) . * runaway migration should be common , in such disks , for giant protoplanets which reach a sizable fraction of a saturn mass . * the right part of the domain boundary shows a steep rise around @xmath275 @xmath274 in the three cases . * even in a disk with @xmath276 such as the mmsn , runaway migration is more difficult as one gets close to the star . the coorbital mass deficit scales indeed as @xmath277 . * the analysis performed breaks down for lower mass planets ( @xmath278 ) , as the torque can not reliably be estimated through a 2d calculation . the runaway limit for the low mass planets at the left of the vertical dashed line is likely higher than found in our analysis , as in their case only a fraction of the disk vertical extent is involved in the coorbital dynamics . the viscosity chosen for these runs is high enough that the dip viscous time @xmath279 is comparable to the libration time ( @xmath280 being the dip half - width ) . this time is also the dip opening time when the planet is `` switched on '' in the disk , or also the minimum time that the planet needs to sweep radially its own dip radial width in order for the surface density profile depression to follow the planet migration . it is therefore possible to talk unambiguously of _ the _ disk runaway critical mass for a given planet , independently of its `` preparation state '' . in other words , whether we release _ ab initio _ the planet in an unperturbed disk , or whether we hold it on a fixed circular orbit for a few hundreds of orbits in order to allow it to open a dip before being released makes no difference : a runaway is observed above the same disk mass , and the maximum drift rate measured is the same in both cases . if one tries to delineate the runaway domain for a disk with a much smaller viscosity , one finds that the estimate strongly depends on the disk `` preparation state '' , contrary to the case of figs . [ fig : rwlim1 ] to [ fig : rwlim3 ] . one can achieve indeed a significant coorbital mass deficit if one holds the planet on a fixed circular orbit for a sufficient amount of time . simulations performed for saturn mass planets in massive , low viscosity disks display a very erratic behavior , with an alternation of brief runaway episodes followed by moderately eccentric ( @xmath281 ) , halted migration episodes , the overall drift rate being a relatively small but sizable ( @xmath282 % ) fraction of the corresponding type i drift rate . it is of interest to evaluate the type i to runaway drift rate ratio at the critical disk mass @xmath283 for runaway , as a function of planet mass . at this boundary , eq . ( [ eq : tqfast ] ) leads to @xmath284 ( since @xmath285 ) , whereas the type i drift rate in the same disk , according to eq . ( 70 ) of @xcite , is given by : @xmath286 fig . [ fig : rw1ratio ] shows the ratio of the runaway to type i migration rate estimate , as a function of planet mass , for the three aspect ratios presented in figs . [ fig : rwlim1 ] to [ fig : rwlim3 ] . these ratios are estimated at the critical disk mass for runaway . since the maximum runaway drift rate scales roughly with the disk mass ( see sect . [ sec : fast ] ) , this ratio should not vary significantly if one considers disk masses higher than the critical one . one can see that this ratio is marginally larger than unity for slightly sub - saturn mass planets , and drops considerably for smaller and larger masses . the runaway drift rate estimate is however still considerably larger than the type ii viscous drift rate estimate for a jupiter mass planet , as it corresponds typically to the third of the type i drift rate of a saturn mass planet . the main source of the effect we have presented is the drift across the separatrices of the coorbital region of inner / outer disk material , in the cases of inward / outward planet drift respectively . we have assumed that the drift velocity of the planet with respect to the disk is @xmath76 , thereby neglecting the disk viscous drift rate @xmath287 . taking this additional drift into account would add a positive contribution to the corotation torque @xmath288 . for the case of a weak coorbital depletion , @xmath289 , so that this additional term scales as @xmath290 , the one - sided lindblad torque . this additional term , coupled to the one - sided lindblad torque , has been evaluated to order of magnitude by @xcite , and found to be small . since furthermore it does not participate in the feed back and therefore in the runaway , it is legitimate to have neglected it in the present analysis . the ability of the planet to maintain its coorbital mass deficit during a runaway episode deserves further investigation . before runaway , the coorbital region is partially cleared of disk material under the action of the lindblad torques . during the runaway drift , the lindblad torques are assisted in maintaining the coorbital mass deficit by the horseshoe dynamics , which traps the coorbital material . the streamlines are not exactly closed in the @xmath241 space , however , and the coorbital material can be lost . in that case the coorbital region is no longer depleted , and the planet switches to type i migration , which endows it with a comparable drift rate . no case has been found in which the runaway is maintained `` forever '' . a single runaway episode however can sizably affect the planet semi - major axis , by a factor @xmath1 or more . in this work we have neglected accretion onto the protoplanets , whereas the mass range for which the runaway mechanism is relevant seems to imply gas accretion . a runaway episode is likely to be accompanied by a simultaneous accretion of disk material . a simple way to take into account gas accretion is to consider that a fraction @xmath291 of the material which enters the coorbital zone through one separatrix and flows out along the other one , is kept by the planet . in the absence of accretion , this disk material contribution to the corotation torque amounts to @xmath292 . when accretion is considered , the contribution to the corotation torque reduces to : @xmath293 since the material that is accreted onto the planet loses / gains only half of the specific angular momentum that it would lose / gain otherwise . accretion therefore plays the role of a runaway moderator . it would be of interest to investigate in detail the interplay between runaway and accretion , taking consistently into account the corotation torque feed back reduction and the growth of the horseshoe region . we have evaluated the torque exerted on a protoplanet embedded in a gaseous disk produced by the fluid elements as they perform a horseshoe u - turn in the planet vicinity . we have interpreted this torque as the coorbital corotation torque . this torque exhibits a dependency upon the planet radial drift rate in the disk which is tied to the depletion of the coorbital region of the planet , that is to say to the existence of a dip or gap cleared around the orbit . the sign of the corotation torque is the same as that of the drift rate . hence it exerts a positive feed back on the migration process . this feed back leads to a migration runaway in massive disks when the coorbital mass deficit is larger than the planet mass . we have checked and illustrated the main properties of the corotation torque through customized numerical simulations . these showed the link between runaway and the coorbital mass deficit . they indicated that the migration rate for disks with sub - critical masses grows faster than linearly with disk mass , and they illustrated that the planet drift rate characteristic response time is the horseshoe libration time . the occurrence of inward runaway in disks with @xmath294 has been investigated , and found to be likely in thin ( @xmath295 % ) , massive disks with a mass several times that of the mmsn . typically it occurs for saturn - sized giant protoplanets , but it can involve planets up to one jupiter mass for sufficiently massive disks . the runaway drift rate is found to be comparable to the type i estimate given by the differential lindblad torque that one would obtain using linear theory even though that would be invalid for a planet held in a fixed circular orbit ( because the planet opens a significant dip around its orbit and the actual differential torque acting on it amounts to a small fraction of its linear estimate ) . it turns out that the corotation torque is at most equal to the differential lindblad torque in a @xmath294 disk . this prevents the possibility of outward runaway in these disks , and we indeed found no occurrence of outward runaways in such disks . additional runs with shallower disk profiles however have shown that outward runaway can occur , provided the planet is endowed with an adequate value for @xmath26 and @xmath76 ( this latter needs to be established at least over a libration time ) . the way we initiated outward runaway was artificial . @xcite have exhibited a two planet configuration engaged in an outward migration . their outer planet is of saturn mass , and is therefore a good candidate for outward runaway in a massive disk . an alternate possibility to initiate an outward motion is that the disk mass flow across the orbit is strongly variable in time . if a larger @xmath296 enters the coorbital zone through the outer separatrix , the corotation torque may be temporarily large ( and positive ) , providing the seed for an outward runaway . when the viscosity is large enough , that the viscous diffusion time across the horseshoe zone is faster than the time to migrate through it , relaxation and depletion of the coorbital region occurs in a sufficiently short amount of time that the coorbital mass deficit and the disk critical mass for runaway are well defined quantities . the situation turns out to be much more chaotic in a very low viscosity disk , although globally the average drift rate of the planet towards the central object amounts to a sizable fraction of the type i drift rate . inspection of figs . [ fig : rwlim1 ] to [ fig : rwlim3 ] shows that in massive disks , sub - jovian planets can still undergo a type i like drift , whereas on previous grounds they were expected to have a much slower ( type ii ) drift rate , comparable to the slow viscous disk drift rate . the runaway threshold is lower in thinner disks . if protoplanetary disks are flared ( i. e. @xmath297 ) , they could be extremely thin in their inner regions , which could assist runaway there . a runaway episode or a succession of them stops whenever the coorbital mass deficit that can be achieved is too small . if one assumes a disk surface density profile @xmath298 , then the coorbital mass deficit scales as @xmath299 . as the runaway condition reads @xmath300 , one would get , at the end of the runaway episodes , the relationship @xmath301 , where @xmath302 is the planet orbital period . this indicates a tendency for smaller masses to reach smaller periods . a recent analysis by @xcite indicates a paucity of planets with masses exceeding a jupiter mass at small periods @xmath303 a few days while sub jovian mass planets tend to cluster at these small periods . the observational data for the larger mass planets is consistent with a type ii migration scenario in which most time is spent at larger radii @xcite . however , the distribution of smaller mass objects would appear to require a relatively fast migration of the type discussed here followed by a slowing down or stopping near their current orbital locations . the fate of a giant protocore at the end of its runaway episode(s ) if it has not been brought close to the primary , would be a slow , type ii migration , together with a possible mass growth , along the lines already studied e. g. by @xcite followed possibly by stopping inside a magnetospheric cavity . an alternative scenario could be envisaged to account for the properties of egps . in this , the disks in which they formed would have been massive enough to sustain a succession of runaway episodes typically up to the central tenth of an astronomical unit where hot `` jupiters '' with @xmath304 @xmath274 , roughly corresponding to the most favorable mass for runaway , are found to accumulate ( or alternatively protocores that massive could be constituted in situ by the accumulation of lower mass bodies brought there by type i migration . ) some of these hot proto - saturns could then be involved in outward runaway . at the same time , gas accretion onto these cores would eventually endow them with a mass sufficient to prevent any further runaway migration episodes . these planets would then correspond to the massive ( @xmath305 @xmath274 ) extrasolar planets , which are found further out from their host stars than the hot jupiters . computational resources were available at the cgcv grenoble , and are gratefully acknowledged . we transform eq . ( [ eq : tqinter ] ) using the fact that a coorbital fluid element does not undergo any radial drift ( we assume a low viscosity ) between two successive close encounters and we write : @xmath306,\ ] ] where we use the fact that the distance @xmath307 of the fluid element to the central object , just after the @xmath70-close encounter at time @xmath308 , is the same as the distance @xmath309 , just before the @xmath71-close encounter at time @xmath310 . we now assume that the profile @xmath311 is independent of @xmath310 for any @xmath57 in @xmath80 $ ] , which corresponds to assuming that either migration is steady state , or that it is slow enough that the surface density profile responds much faster than @xmath312 . we can therefore write : @xmath313\frac{\partial w_r^-}{\partial x}\nonumber\\ & & + \frac 12[a(t)-a(t-\tau)]^2\frac{\partial^2 w_r^-}{\partial x^2}.\end{aligned}\ ] ] a taylor expansion in @xmath314 of this expression yields : @xmath315 the second term of the bracket , an upper limit of which is @xmath316 , is negligible compared to the first one , as long as the migration remains slow . ( [ eq : tqinter ] ) and ( [ eq : inter2 ] ) yield : @xmath317\right\}.\label{eq : tqf}\end{aligned}\ ] ] an order of magnitude estimate of the last term ( in @xmath122 ) of eq . ( [ eq : tqf ] ) can be given if one assumes e. g. a quadratic dependency of @xmath75 on @xmath57 : @xmath318 , in which case one has : @xmath319 . ( [ eq : tqf ] ) can then be rewritten as : @xmath320 balmforth , n. j. , korycansky , d. g. 2001 , , 326 , 833 foglizzo , t. & ruffert , m. 1997 , , 320 , 342 ivanov , p. b. , papaloizou , j. c. b. & polnarev , a. g. 1999 , , 307 , 79 kley , w. 1998 , , 338 , 37 lin , d. n. c. & papaloizou , j. c. b. 1986 , , 309 , 846 masset , f. s. 2002 , , 387 , 605 . 2001 , , 558 , 453 masset , f. & snellgrove , m. 2001 , , 320 , l55 masset , f. s. 2000a , , 141 , 165 . 2000b , in disks , planetesimals and planets , asp conference series , vol . 219 , 7580 miyoshi , k. , takeuchi , t. , tanaka , h. & ida , s. 1999 , , 516 , 451 nelson , r. p. , papaloizou , j. c. b. , masset , f. s. , & kley , w. 2000 , , 318 , 18 papaloizou , j. c. b. & terquem , c. 1999 , , 521 , 823 papaloizou , j. c. b. & lin , d. n. c. 1984 , , 285 , 818 tanaka , h. , takeuchi , t. & ward , w. r. 2002 , , 565 , 1257 trilling , d. e. , benz , w. , guillot , t. , lunine , j. i. , hubbard , w. b. & burrows , a. 1998 , , 500 , 428 trilling , d. e. , lunine , j. i. & benz , w. 2002 , astroph/0208184 ward , w. r. 1997 , icarus , 126 , 261 . 1992 , abstracts of the lunar and planetary science conference , 23 , 1491 . 1991 , abstracts of the lunar and planetary science conference , 22 , 1463 ward , w. r. & hourigan , k. 1989 , , 347 , 490 zucker , s. & mazeh , t. 2002 , , 568 , l113 |
we present the results of a high resolution ( ) 12 ksec _ chandra _ hrc - i observation of the starburst galaxy ic342 taken on 2 april 2006 . we identify 23 x - ray sources within the central region of ic342 and resolve the historical ultraluminous x - ray source ( ulx ) , x3 , near the nucleus into 2 sources , namely c12 and c13 . the brighter source c12 , with , is spatially extended ( ) . from the astrometric registration of the x - ray image , c12 is at r.a . = 03h:46m:48.43s , decl . = + 68d05m47.45s , and is closer to the nucleus than c13 . thus we conclude that source is not an ulx and must instead be associated with the nucleus . the fainter source c13 , with is consistent with a point source located at ( p.a. ) of c12 . we also analyzed astrometrically corrected optical _ hubble space telescope _ and radio very large array images ; a comparison with the x - ray image showed similarities in their morphologies . regions of star formation within the central region of ic342 are clearly visible in _ hst _ h image which contains 3 optical star clusters and our detected x - ray source c12 . we find that the observed x - ray luminosity of c12 is very close to the predicted x - ray emission from a starburst , suggesting that the nuclear x - ray emission of ic342 is dominated by a starburst . furthermore , we discuss the possibility of agn emission in the nucleus of ic342 , which we can not prove nor discard with our data . # 1to 0pt#1 | 0706.0770 | ic342 is a nearby ( 1.8mpc ; see @xcite for a review ) , almost face - on ( @xmath8 ; @xcite ) late - type sc / scd galaxy in the maffei group which is one of the closest groups to our galaxy with well developed spiral arms . infrared and radio observations indicated moderately strong nuclear starburst activities ( @xcite ; @xcite ) in ic342 , whereas x - ray observations with _ einstein _ @xcite , _ rosat _ @xcite , and _ asca _ @xcite indicated that the majority of the x - ray emission came from a few ultraluminous x - ray sources ( ulxs ) . ulxs are defined as off - nuclear x - ray sources whose luminosities are greater than @xmath9 , the eddington luminosity of a @xmath10 black hole . one of the ulx identified , x3 ( based on the designations from @xcite ) was located near the galactic center , and was reported to be coincident with the center of ic342 at the resolutions of the _ asca _ ( fwhm@xmath11 ) and _ rosat _ ( fwhm@xmath12 ) data . optical and near infrared observations of the inner region of ic342 revealed the presence of a nuclear star cluster @xcite . photometrically distinct , luminous , and compact stellar clusters were often found in the centers of late - type spiral galaxies , such as ngc6240 @xcite and ngc1808 @xcite . @xcite studied high spatial resolution ( fwhm@xmath13 millimeter interferometric observations of the @xmath14 emission of ic342 and concluded that gaseous matter were accumulating at the nucleus through streaming motions . this suggested that star cluster formation could be a repetitive process and this process could have provided the fuel for starburst activities . as a result , if x3 were indeed associated with the galactic center , its emission may not be from a single ulx , but could be associated with the nuclear star cluster . however , the early identification of x3 as an ulx was uncertain due to limited resolution of the early x - ray data . the higher light collecting power and lower background level of _ xmm - newton _ provided a clearer picture of the nuclear x - ray source x3 . a 10 ksec observation of ic342 was taken with the pn - ccd camera and two mos - ccd cameras using the medium filter on 2001 february 11 . two independent studies by @xcite ( hereafter k03 ) and @xcite ( hereafter b03 ) on the data set yielded similar results , with a total of about 35 sources detected . the slope of the x - ray luminosity function was determined to be @xmath15 , which was consistent with other starburst galaxies and galactic high mass x - ray binaries ( hmxbs ) . however , there were differences in the two studies about the nature and properties of the nuclear source . k03 found that the nuclear x - ray source was consistent with the instrumental psf of a point source located @xmath16 from the galactic center , suggesting a off - center scenario . k03 fit the nuclear source with a power - law + blackbody spectral model ( @xmath17 kev , @xmath18@xmath19 , photon index @xmath20 ) and found the disk temperature was about 5 to 10 times lower than expected . such cool disk system might be evidence of imbhs ( @xcite ; @xcite ) . however , we could not ignore the scenario of a stellar mass black hole with inner disk outflows during high accretion rate period @xcite , and that of magnetic corona atop a cool accretion disk in which x - rays were reprocessed to give soft photons @xcite . the black hole mass of the nuclear x - ray source was estimated to be between 220 @xmath21 and 15000 @xmath21 ( k03 ) . together with the unphysical spectral fit with the broken power - law model , k03 ruled out the possibility of an accreting stellar mass black hole with beamed relativistic jet emission . on the other hand , b03 suggested that the nuclear source was extended , mainly in the soft band , and its emission could be fit by a model of a point source plus a uniform disk with a radius of @xmath22 , which was consistent with a supernova heated hot gas . moreover , the nuclear source spectrum was fit by b03 with a mekal + power - law model ( @xmath23 kev , @xmath24 , @xmath25 , @xmath26@xmath19 ) , including contribution from the fe l - shell emission complexes at @xmath27 kev and @xmath28 kev . they found that a large fraction of the x - ray emission was likely to be thermal in origin . the relative contribution of the mekal component to the total absorbed flux was about 45% and 1% in the 0.52 kev and 210 kev band respectively , leading b03 to suggest that the hard component might be due to a number of circumnuclear x - ray binaries . the limited astrometric accuracy and spatial resolution of the previous x - ray observations restricted our ability to determine the exact nature of the x - ray properties at the center of ic342 . in this paper , we report a new high resolution ( fwhm @xmath29 ) _ chandra _ _ x - ray observatory _ observation of the starburst galaxy ic342 . the observation and the data reduction procedures of the x - ray data , and the associated optical data for comparison , are presented in section [ s : obsdatared ] . in section [ s : analysis ] , the analysis of the data , including source detection , photometry , and astrometric registration of the x - ray sources are described . in section [ s : nuclearsource ] , we present the results of the associated counterparts of the nuclear x - ray sources in optical and radio images and the properties of the nuclear sources . we discuss the nature of nuclear x - ray emission of ic342 in section [ s : nuclear ] and we present a summary of our findings in section [ s : summary ] . ic342 was observed on 2006 april 2 starting at 15:15:44 ( ut ) with the _ chandra _ high resolution camera ( hrc - i ) for a total exposure time of 12.16 ksec , with a photon energy range of 0.08 the nominal pointing of the instrument was near the center of the galaxy and the total region covered by the observation was @xmath1 which covered most of structure of the galaxy , including the spiral arms . the field of view of the _ chandra _ observation is presented in figure [ f : dss ] ( left ) , overlaid with the digital sky survey ( dss ) image of ic342 . the central @xmath30 of ic342 is magnified in figure [ f : dss ] ( right ) for reference . the x - ray data was analyzed using the chandra interactive analysis of observations software package ( ciao ) v.3.3 and chandra calibration database ( caldb ) v.3.2.1 software applied on the level-2 event fits file . the _ chandra _ data had gone through standard data processing ( sdp , or the pipeline ) with updated calibration procedures , including instrumental corrections and filtering good time intervals , applied . there were 4 other archival _ chandra _ ic342 data sets : three axaf ccd imaging spectrometer ( acis - s ) observations ( each 10 ksec ) in sub - array mode centered on one of the ulx ( x1 ) , and one 3 ksec hrc - i observation . for the 3 acis - s data , the detectors did not cover the nuclear region of ic342 as the sub - array mode was used to reduce pile - up . the 3 ksec hrc - i observation , on the other hand , had insufficient signal for source detection at the galactic center . therefore only the 12 ksec hrc - i image was analyzed in this work . ic342 was also observed 4 times between 2001 and 2005 with _ xmm - newton_. among these 4 observations , the data taken in year 2001 were studied by b03 and k03 . we have retrieved the 2001 epic data for the analysis of x - ray spectrum of the nuclear sources . the event files were reprocessed and filtered with the _ xmm - newton _ science analysis software ( sas v7.1.0 ) and we used the heasoft v6.2 and xspec v12.3.1 packages to perform spectral analysis . we also analyzed archival _ hubble space telescope _ ( _ hst _ ) broadband and emission - line images of ic342 for comparisons with the x - ray data . we searched the _ hst _ data archive for images with pointings that included the central part of the galaxy . the calibrated and data - quality images were retrieved , with preliminary ccd processing , including bias and dark subtraction , flat - fielding , and shutter - shading correction , automatically applied using calibration frames closest to the time of the observations . the fields were inspected to ensure that the galactic nucleus did not fall on the boundaries on any of the ccd chips . we found that most of the archival images were saturated at the central region of the galaxy . the only unsaturated images were the wide field planetary camera 2 ( wfpc2 ) images taken on 1996 january 7 in broadband _ ( f555w ) , _ i _ ( f814w ) , and in narrow - band h@xmath7 ( f656n ) , and they were selected for comparison with our _ chandra _ hrc - i data . the exposure time of each frame was 260 sec each for _ v _ and _ b _ band and 500 sec for h@xmath7 . mosaics were produced using the stsdas task wmosaic which removed known relative rotations and geometric distortions of the individual wpfc2 chips . astrometry of the calibrated _ hst _ images were corrected using the coordinates from the usno - b1.0 catalog @xcite for comparison with the positions of x - ray sources in the same reference frame . details of the astrometric registration of the data will be described in section [ s : anlyast ] . we performed source detection on the x - ray image using the wavdetect algorithm @xcite which employs a wavelet method to detect point sources . this algorithm was selected because it is able to effectively identify extended sources with properly chosen wavelet scales ; this is important for the analysis of our _ chandra _ ic342 image , which has an extended source near the galactic center . source detection results from the wavdetect algorithm depend on two parameters , the significance threshold ( _ sigthresh _ ) and the image scale . we set the wavelet radii , in pixels , to be ` 1.0 1.414 2.0 2.828 4.0 5.657 8.0 11.314 16.0 ' and the parameter _ sigthresh _ to be @xmath31 for all runs . this detection threshold corresponds to less than one false detection in the _ chandra _ hrc - i field due to statistical fluctuations in the background @xcite . the runtime of wavdetect depends heavily on the selected resolution of the image , and a resolution of @xmath32 was considered to be the practical size limit for the input data on most desktop computers . we tested running wavedetect on the central part of the image containing the galactic center with binning factors 1 , 2 , 4 , 8 , 16 , and 32 . an exposure map , created using ciao , was applied for each binning to normalize the images and to correct for instrumental artifacts caused by imperfections of the _ chandra _ mirror and detector , especially at the edges of the field . we found the list of sources detected depended on the binning factor applied . a binning factor of 4 was used in the final data reduction as a compromise between spatial resolution and faint source detection . the _ chandra _ field was divided into 4 equal quadrants with overlapping at the central regions before being recombined to the full - frame . the final image scale with a binning factor of 4 was @xmath33/pixel . to generate the final source list , we applied two additional selection criteria for the sources identified by the wavdetect algorithm : 1 . the source should be detected in at least 2 different binning images , with one of them being the one with a binning factor of 4 . the s / n of the source had to be greater than 3 in the binning factor 4 image . with these selection criteria we detected a total of 23 x - ray sources in our _ chandra _ hrc - i ic342 image . properties of the detected sources are summarized in table [ t : sourcelist ] . a total of 14 and 8 of our sources were also previously detected in the _ xmm - newton _ ( b03 ; k03 ) and the _ rosat _ observations @xcite respectively . the alternate _ xmm - newton _ and _ rosat _ source ids are listed in table [ t : sourcelist ] . the locations of all sources detected are shown overlaid on the dss image in figure [ f : dss ] . a majority of the identified sources are located in the spiral arms of the galaxy , as previously noted by b03 and k03 with _ xmm - newton _ observations . using the chandra deep field data @xcite , we estimated that there could be 910 background objects within our @xmath1 _ chandra _ frame , implying that @xmath34% of the sources currently detected could be background objects . with the high resolution _ chandra _ hrc - i image , we were able to resolve for the first time x - ray emission from the center of ic342 into two distinct components , namely c12 , the brighter and extended component near the center of the galaxy , and c13 , a much fainter component located @xmath5 from c12 at p.a . the two sources could be identified by the wavedetect algorithm when the binning factor were set to be either 2 or 4 , suggesting that they were genuine detections . images of the central @xmath35 of the _ chandra _ hrc - i image of ic342 are shown in figure [ f : central_xray ] , together with the positions of the two nuclear sources c12 and c13 , and the x - ray flux contours . details of the x - ray emission morphology will be discussed in section [ ss : xrayimage ] . we extracted the integrated counts of the 23 detected x - ray sources of ic342 using the ciao tool dmextract , and listed the results in table [ t : sourcelist ] . the source regions were defined in ds9 using circular apertures with different radii based on the relative brightness of the source compared to the local background and their off - axis angles . all the background regions were extracted using circular annuli centered on individual sources . the only exception was the nuclear source c12 , in which an elliptical aperture was used because of its extended nature . moreover , the background regions of the two nuclear sources c12 and c13 ( @xmath5 apart ) had to be modified to avoid overlapping . we used webpimms to convert the count rates for all detected x - ray sources into unabsorbed 0.0810 kev luminosity by assuming an absorbed power - law model with @xmath36@xmath19 and @xmath20 , adopted from the average parameters derived from spectral fitting of the _ xmm - newton _ data by k03 . these results are also presented in table [ t : sourcelist ] . this set of average parameters was similar to that derived from the same _ xmm - newton _ data set by b03 , with @xmath37@xmath19 and @xmath38 . the difference in 0.0810 kev luminosities obtained with these two sets of spectral parameters is about 10% , with slightly lower values for the parameters from k03 . for c12 , b03 and k03 derived the x - ray flux by fitting the source spectrum with a simple absorbed power - law model ( b03 : @xmath39@xmath19 and @xmath40 ; k03 : @xmath41@xmath19 and @xmath42 ) . for comparison , our adopted flux for c12 is smaller by about 40% from the b03 model and about 30% larger than the k03 simple power - law model by using the average the x - ray sources parameters . in this paper , we adopt the value from k03 , except for table 1 in which we use the average parameters from k03 as described above . there is a large uncertainty in the distance estimate to ic342 due to the large uncertainty in the amount of dust extinction toward the galactic center along that line - of - sight . distance estimates of ic342 range from 1.5 to 8 mpc @xcite . a distance of 1.8 mpc was assumed in this paper ( corresponding to a scale of @xmath43 87 pc ) , identical to that in boker et al . ( 1997 , 1999 ) and k03 , and similar to that ( 2 mpc ) assumed by @xcite . using an elliptical ( @xmath44 and @xmath45 for semiminor and semimajor axis ) integrating window to include all the diffused emission , we found the stronger central nuclear source c12 has @xmath46 net counts , corresponding to an unabsorbed x - ray luminosity of @xmath47 ; the weaker source c13 was one of the weakest source detected in our image , with a net count of @xmath48 ( s / n = 3.8 ) , and an unabsorbed luminosity @xmath49 . one of the main objectives of our study is to obtain accurate positions of x - ray sources near the center of ic342 . the _ chandra _ hrc - i images have nominal 90% and 99% confidence astrometric accuracy of @xmath50 and @xmath51 respectively , while the astrometric accuracies of the _ hst _ wfpc2 data is about @xmath52 @xcite . in order to identify optical counterparts of our _ chandra _ x - ray sources in the _ hst _ data , we first registered their positions to the same reference frame for comparison . we used the usno - b1.0 catalog @xcite to improve the astrometric accuracies for both the _ chandra _ and _ hst _ the advantage of using this is that proper motions are already included in the catalog , and hence can be corrected for . the nominal rms accuracy of the usno b1.0 catalog is @xmath53 . the astrometry of our _ chandra _ image can be improved when one or more x - ray sources are associated with optical or radio sources of known positions . as seen in figure [ f : dss ] , one of the brightest sources , c5 , is clearly associated with a bright stellar object within 1@xmath54 in the dss image . this source was also previously identified to be associated with a star in the _ rosat _ ( source 4 in @xcite ) and the _ xmm - newton _ observation ( x12 in k03 ) . the spectrum of x12 is soft and can be fitted with a blackbody model ( @xmath55 = 0.17 kev ) . b03 determined the x - ray to optical flux ratio @xmath56 of x12 to be @xmath57 - 3 , which was consistent with that expected from a foreground star of @xmath56 @xmath57 - 1 . b03 and k03 also identified it to be a foreground star . using the usno - b1.0 catalog , we found that the source c5 was offset by @xmath58 ( ra offset by @xmath59 and dec offset by @xmath60 ) from this foreground star , which was within the expected @xmath61 pointing accuracy of the _ chandra _ data . while there are other x - ray sources which could be associated with point sources in the dss image , their positional offsets are either @xmath62 or their s / n ratios were too low ( @xmath63 ) to make firm registrations with their optical counterparts . thus we corrected the pointing of the _ chandra _ image using the ciao command wcsupdate by linearly shifting it until c5 and the star coincided . this astrometric reference was applied to all the sources and their shifted positions were tabulated in table [ t : sourcelist ] . the astrometry of the _ hst _ images were corrected using the iraf task ccmap . with many bright stars within the wfpc2 frame , we could visually identify stars from the catalog in the _ hst _ data products and compare them with the usno - b1.0 catalog to improve the astrometry . a total of 7 stars were used for astrometric registration for the _ v _ and _ i _ band data , and 3 stars for the h@xmath7 data . the rms positional errors returned by ccmap were @xmath64 for _ v _ , @xmath65 for _ i _ , and @xmath66 for h@xmath7 . these registrations are similar to that of our _ chandra _ hrc - i images , hence allowing meaningful spatial comparison of the sources in different wavelengths . the high spatial resolution of the _ chandra _ hrc - i image ( @xmath67 ) provides an unprecedented detailed look into the x - ray emitting structure of the ic342 galaxy center . detailed structures of the x - ray emission from the ic342 center are illustrated in figure [ f : central_xray ] with the central @xmath35 of the _ chandra _ image . the two nuclear sources , c12 and c13 , are separated from other x - ray sources and stand out from the low diffuse x - ray background . the brighter source , c12 , was observed to have at least two components . the first component is the core emission of radius @xmath68 at the center , roughly consistent with that bounded by the x - ray contour line of 1.4 counts . the second component is the diffuse emission out to a radius of @xmath69 , with an asymmetric extension along the north - south direction . the spatial extent , derived from the x - ray image at 99.5% linear scale , is @xmath70 in the north - south and @xmath71 in the east - west directions ( 127 pc @xmath72 82 pc assuming a distance 1.8 mpc ) , resulting in a major - to - minor axis ratio of @xmath73 . the observed north - south elongation is similar to that observed in the near infrared wavelength by @xcite , who detected a diffused emission with a major - axis of 110 pc and a high surface brightness source at the center . the authors interpreted the results as evidence for a small - scale nuclear stellar bar and a central concentration of stars . we determined the x - ray luminosity of the core and the circumnuclear diffuse regions of c12 by extracting respectively the source counts of the inner @xmath74 region with a circular aperture , and the source counts of an annulus of @xmath75 . the resulting x - ray luminosities ( assuming the same spectral and distance parameters as listed in section [ s : anlypho ] ) are @xmath76 and @xmath77 , implying roughly equal flux contributions in x - ray from these two regions . to analyze the structure of emission from c12 , its radial flux profile was extracted and normalized at the center . the results are shown in figure [ f : rp ] ( top - left ) . as a comparison , the normalized radial flux profile of the _ chandra _ psf at the location of c12 derived from the _ chandra _ psf simulator , the _ chandra _ ray tracer ( chart ) , is also shown ( _ dash _ ) . the emission from the core component of c12 at the central @xmath78 is consistent with that from a point source . however , the diffused emission component at radius @xmath79 does not decrease as steeply as that of the psf . we also computed the radial flux profile of a point source plus a uniform @xmath80 circular disk at the location of c12 computed from the simulation task marx . the results ( _ dot - dash _ ) , also shown in figure [ f : rp ] ( top - left ) , are generally consistent with that observed for c12 , with a hint of slowly - decreasing surface brightness in the range @xmath81 for c12 . a kolmogorov - smirnov test ( k - s test ) was performed to measure the difference between radial profile of c12 and the models . the k - s test statistics @xmath82 , the maximum vertical deviation between the two curves , is 0.78 ( confidence probability@xmath83 ) for c12 and a point source , and 0.33 ( confidence probability = 6.0% ) for c12 and a point source plus a @xmath69 disk . the surface brightness profiles for selected bright x - ray sources c3 , c6 , and c21 , are also plotted in figure [ f : rp ] ( top - right ) , [ f : rp ] ( bottom - left ) , and [ f : rp ] ( bottom - right ) , respectively for comparison . the k - s statistic @xmath82 of the reference sources c3 , c6 , and c21 are in a much lower range of 0.170.33 ( corresponding to confidence probabilities 2.2%76.0% ) . these results reinforce the conclusion that the emission from c12 is not consistent with that from a point source ; the x - ray emission of c12 might be due to multiple sources or diffuse emission or a combination of both . however , we can not conclude whether the emission is consistent with the various components proposed by @xcite due to the limited resolution of the x - ray data . the newly detected x - ray source c13 is one of the faintest source in our _ chandra _ observation , with @xmath84 . we failed to extract the radial profile of c13 , and can not determine whether it is a point source or a diffused source because of insufficient count statistics . we also could not exclude the possibility that the emission was actually part of the nebulosity of the extended emission component of c12 . we compared the _ chandra _ image of the galaxy center of ic342 with the optical observations . the registered positions of the x - ray nuclear sources c12 and c13 are overlaid on the _ hst _ _ v _ band and h@xmath85 images of the central @xmath86 region of the ic342 in figures [ f : vband ] ( top ) and [ f : halpha ] ( top ) respectively . the spatial extent of the c12 core emission is also shown by a @xmath74 circle ( _ dash _ ) in the figures . the _ hst _ _ i _ band image was also analyzed but was found to be similar to the _ v _ data and thus not shown . in the _ hst _ images , there are many star forming regions at the center of ic342 with obscured emissions and bright knots . both _ hst _ images show that the optical emission from the central @xmath87 of ic342 is concentrated in three star clusters aligned in the north - south direction . the star clusters have also been observed previously in the near - ir co band by @xcite . we named the three clusters here as the nuclear star cluster ( nsc ; middle ) , sc1 ( top ) , and sc2 ( bottom ) . the centroid of c12 lies close to that of the optical maximum of the nsc , with an offset of @xmath33 measured in all the @xmath88 , @xmath89 , and h@xmath7 images . the observed separations were only slightly larger than the uncertainty of the x - ray bore sight correction of c12 ( @xmath90 ) in our _ chandra _ image . using the near - ir co band observations , @xcite deduced that the nsc is a relatively young cluster with an age of @xmath91 yrs , and a mass of @xmath92 . these results favored an instantaneous burst model rather than a constant star formation rate model for the nsc , meaning the dominant stellar population are of the same age as the nsc . also , nsc could have formed as a result of gas flowing to the nucleus initiated by torques from a nuclear stellar bar . the h@xmath7 image is consistent with this scenario in that it shows gas spiraling around the nsc ( see figure [ f : vband ] and [ f : halpha ] ) . the x - ray contours are overlaid on the broadband _ v _ and h@xmath7 images of ic342 in figures [ f : vband ] ( bottom ) and [ f : halpha ] ( bottom ) respectively for a comparison of the diffuse x - ray emission with the detailed structure of the optical galaxy center . these contour lines encircle the central @xmath93 region of ic342 and its structure is found to generally agree with those observed in the optical wavelengths , despite the lower angular resolution of the x - ray data . moreover , the x - ray diffused emission is elongated along the north - south axis , which is similar to the alignment of the three optical nuclear star clusters . these coincidence suggest that the observed x - ray nuclear emission is associated with the ongoing starburst and the three optical star clusters . however , the newly detected x - ray nuclear source c13 does not coincide with any bright features in the h@xmath7 image , suggesting that it is not associated with the star formation activities at the galaxy center . for comparisons with the x - ray and optical data , we also obtained very large array ( vla ) 2 cm ( 15 ghz ) and 6 cm ( 5 ghz ) radio images of ic342 . the observations and the data reduction procedures have already been described in tsai et al . these images have spatial resolutions of @xmath94 at 6 cm and @xmath95 at 2 cm , respectively , with a @xmath96 uncertainty in the absolute position of detected sources . the positions of all the radio sources detected in the 2 cm and 6 cm are marked in figures [ f : vband ] ( top ) and [ f : halpha ] ( top ) . the optical images of ic342 presented in @xcite were not astrometrically registered for comparison with the radio images . however , our current analysis presents improved astrometry on the x - ray and optical data , allowing accurate identifications of radio counterparts for the x - ray and optical sources . our revised astrometry explains the slight differences in the radio source positions of our figures [ f : vband ] and [ f : halpha ] compared with the figure 2 of @xcite . the brightest radio source , source a using the naming scheme of @xcite , is near the optical star cluster sc1 and it was identified as a supernova remnant . one of the radio source , source j , lies very close to the center of the optical emission from nsc at a distance of @xmath97 , and a distance @xmath61 to the center of the nuclear x - ray source c12 . this source is very weak , with only a @xmath98 detection in 2 cm and was actually not detected in the 6 cm data . @xcite explained source j as a region situated between sc1 and nsc , requiring the excitation from a star cluster powered by at least 70 o7 stars . however , with our corrected astrometry , source j was found to be much more likely to be associated with the center of the nsc . we will discuss the possibility of the radio emission from a radio - quiet agn in section [ ss : agn ] . it should also be noted that none of the radio sources identified as regions by @xcite , except source j , could be associated with any star clusters in the optical images . this could be due to the high internal extinction @xmath99 in the central regions of ic342 , consistent with the presence of high molecular gas concentrations , which could fuel more star formation in the region . the radio contours from the 6 cm vla image are overlaid on the _ hst _ image and _ chandra _ contours in figures [ f : vband ] ( bottom ) and [ f : halpha ] ( bottom ) . the morphology of the radio emission in the central region is very different from both the x - ray and optical images . c12 is encompassed by several radio sources aligned from northeast to southwest that have been identified as supernova remnants and regions @xcite . the majority of the bright radio sources and the radio diffused radio emission are located west of c12 , suggesting enhanced star forming activity in that region . on the other hand , the faint x - ray source c13 is about @xmath100 from a compact region ( source l in tsai et al . we noticed that the position of the source l listed in tsai et al . ( 2006 ) was incorrect and the revised position should be r.a . = 03h:46m:49.13s , decl . = + 68d05m46.4s ( tsai 2007 , private communication ) , which was used here . with the pointing accuracy of the _ chandra _ data at @xmath29 , we concluded that c13 is unlikely to be associated with any radio emissions . to investigate the long term variability of the nuclear source c12 , we compared the observed x - ray luminosities at the center of ic342 over a period of 13 years using data taken by _ chandra _ ( this paper ; 2006 april 2 ) , _ xmm - newton _ ( b03 , k03 ; 2001 february 11 ) , and _ rosat _ ( @xcite ; 1991 february 13 ) . we could not directly compare luminosities from the literature because of the different assumptions made in those works . for example , the luminosity values in the earlier _ asca _ and _ rosat _ publications were computed assuming a distance to ic342 of 4.5 mpc . we recalculated the luminosities from all the observations by assuming the same energy range , spectral model , and assumed distance of ic342 . thus instead of quoting from the literature , we extracted the fluxes of c12 from the archival data of _ xmm - newton _ and _ rosat _ , together with our _ chandra _ observation , and analyzed them with the procedures described in section [ s : anlypho ] . the fluxes from the different observatories were converted from source counts using the spectral model derived from _ xmm - newton _ since it had the highest spectral resolution . we adopted the simpler spectral model assumed for c12 ( named x21 ) in k03 , i.e. , a power - law model with @xmath101@xmath19 and @xmath102 . furthermore , we scaled the luminosities and their errors to a distance of 1.8 mpc and to the same energy range as hrc - i ( 0.110 kev ) . the scaled luminosities and errors of the nuclear source c12 from the different x - ray missions are listed table [ t : lumtable ] . the results from the two brightest x - ray sources in our _ chandra _ image , c3 and c6 , are also listed for reference . the spectral models used were : 1 . c3 : power - law model with @xmath103@xmath19 and @xmath104 2 . c6 : disk blackbody model with @xmath105@xmath19 and @xmath106 we note that there was an _ asca _ observations taken on september 1993 . however , the poor spatial resolution of _ asca _ ( fwhm @xmath11 ) and the large extraction radius ( @xmath107 ) used in analyzing the nuclear source @xcite implies that the _ asca _ results might suffer from confusion problems . in our _ chandra _ data , a @xmath107 circular region centering at c12 would have included six other sources ( c9 , c10 , c13 , c14 , c15 , and c18 ) . using the present _ chandra _ photometric information , we estimated that 46% of the measured _ asca _ counts of c12 could be due to confusion , implying a 0.110 kev luminosity of @xmath108 . however , this estimate did not include uncertainties due to long term spectral and flux variability information of all these x - ray sources . as a result , we did not include the _ asca _ data point in the present variability study . on the other hand , though the spatial resolution of _ rosat _ and _ xmm - newton _ were lower than _ chandra _ , the confusion problem in the two observations were not as significant as that in _ asca_. the spatial extent of c12 were similar in both _ xmm - newton _ and _ rosat _ , about @xmath109 , and a circular region of this size would include both c12 and c13 . nonetheless , the source counts of c13 is only 6% of c12 , and thus its contribution is negligible unless it significant varies in flux . thus , we decided to ignore the confusion effect in the _ xmm - newton _ and the _ rosat _ data . this analysis is the first to study the luminosity of c12 over a long period of time . in the energy range 0.110 kev , the observed luminosity of the ic342 nuclear x - ray source c12 varies roughly by a factor of 2 over the 3 observations spanning a period of over 15 years . the observed variations of the luminosity of c12 is actually the smallest amongst the three x - ray sources investigated . it is worth noting that the flux uncertainties tabulated in table [ t : lumtable ] only include photon statistics . uncertainties in the relative calibration of various x - ray detectors have not been included . however , previous studies indicate that variations of normalized flux in different x - ray instruments are small , at a level of @xmath110 @xcite , and thus are not expected to account for all the flux variability observed . another uncertainty in the current study is the assumption of a single spectral model of c12 over the entire period as there was no spectral information from the _ chandra _ and _ rosat _ data . we also studied the short - term variability of c12 . we extracted the source and background lightcurves of c12 using dmextract and high level background intervals were filtered . the short - term lightcurves of c12 were studied . however , we were not able to detect any short - term variability on timescale of the observation because of the observed low count level . the high resolution _ chandra _ hrc - i observation resolves for the first time the x - ray emission at the center of ic342 , revealing the structure of the nucleus of this starburst galaxy . as seen from figure [ f : vband ] and [ f : halpha ] , both the x - ray and the optical emission are confined to a region of radius @xmath111 in the center of ic342 . the most intense x - ray emission ( inner @xmath74 ) is coincident with the nsc seen in both _ v _ band and h@xmath7 images . there are many regions of star formation within the central region of ic342 which were clearly visible in h@xmath7 . it is this region that contains the 3 optical star clusters and includes our detected _ chandra _ x - ray source c12 . the extension seen in the x - ray diffuse emission of c12 is most probably due to starburst activities induced by star clusters and massive stars . this region also contains radio sources which aligns from northeast to southwest . as discussed in @xcite , these radio sources could be interpreted as regions which were small clusters that required excitation by individual o star clusters , each with over 5000 o stars ( @xcite ; @xcite ) . possible contributors to the x - ray emission from a nuclear starburst include supernovae , o stars , high mass x - ray binaries ( hmxbs ) , ulxs , and an obscured agn . using our _ chandra _ x - ray data , we attempted to explain in the following the stellar nature in the nsc . many previous infrared observations of ic342 @xcite have proposed a starburst scenario for the nuclear region . the model of the central starburst includes a nuclear star cluster surrounded by a mini spiral arms - ring system on the scale of @xmath112 ( e.g. , fig . 10 in @xcite ) , as supported by observations of molecular line emissions . the x - ray image generally agrees with the spatial scales of this model , with the size of the c12 emission @xmath113 is roughly the size of the proposed central ring ( @xmath114 ) , while the mini spiral arms that extend in the north - south direction to @xmath115 would be beyond the spatial extent of c12 ( see figure 1 of @xcite for a comparison of the optical and @xmath116co ( 1 - 0 ) infrared emission ) . this model suggests that most of the star formation occurs in the central ring region rather than along the spiral arms . our _ chandra _ observation is consistent with this scenario since the x - ray emission is concentrated in the central 100 pc . this star formation structure is smaller in size compared to those in other starburst galaxies , which have radii of order 1 kpc @xcite . we could study the starburst properties of ic342 using the results from starburst galaxies and agn survey @xcite with the far - infrared to soft x - ray luminosity ratio . the far - infrared flux measured by the _ infrared astronomical satellite _ ( @xmath117 ) at 60 @xmath118 m was @xmath119 jy in the inner r=@xmath120 region of ic342 @xcite . @xcite found that the luminosity of the compact regions in the nucleus was of the order of @xmath121 to the total ir luminosity of ic342 . therefore we estimated the far - infrared flux in the galaxy center of ic342 to be @xmath122 jy . we derived the soft ( @xmath123 kev ) x - ray flux of c12 with webpimms using the spectral model and parameters listed in section [ s : anlypho ] and obtained @xmath124 . this led to a flux ratio @xmath125 with a range of -3.1 to -3.4 . this is in agreement with the value for star formation galaxies , -3.33 ( @xmath126 ) , and slightly smaller than the average value of seyfert 2 galaxies , -2.54 ( @xmath127 ) @xcite . @xcite derived a model to predict x - ray emission from starbursts using the brackett @xmath128 emission . the model suggests that the x - ray emission arises from the accretion of material in population i binary systems which are formed together with the ob stars . the total x - ray emission in the starburst is calculated to be @xmath129 where @xmath130 is the observed brackett @xmath128 emission , @xmath131 is the average x - ray luminosity of binary systems and was assumed to be @xmath132 , and @xmath133 is the number of ob stars per massive binary systems and was assumed to be 500 @xcite . the brackett @xmath128 emission from the inner 100 pc of ic342 was measured to be @xmath134 @xcite , thus implying a predicted x - ray luminosity from the starburst of @xmath135 . our observed 0.35 kev x - ray luminosity of c12 is about 20% of this predicted luminosity . this might suggest that the x - ray emission from the starburst component in c12 was not dominated by population i binary systems as suggested in the @xcite model . it might instead be due to thermal emission such as coronal from young stars , supernovae / supernova remnants , and the hot interstellar medium heated by supernovae and stellar winds . this might suggest that the soft x - ray emission in c12 was due to part of starburst component and some other thermal emission . we will discuss the possible thermal origin of the diffuse emission from c12 in section [ ss : other ] . another possible origin of the core x - ray emission from the galactic center is an agn . using the _ xmm - newton _ data , b03 proposed that the hard x - ray from c12 could be coming from a hidden agn , although there was no direct evidence to confirm its existence in ic342 . the starburst - agn connection has been a topic that generated great interests but the exact relationship remains uncertain . the triggering mechanism for both phenomena could be the interacting or the merging of gas - rich galaxies , which generates fast compression of the available gas in the inner galactic regions , causing both the onset of a major starburst and the fueling of a central black hole , hence raising the agn . examples of this class of objects include ngc6240 @xcite , ngc4303 @xcite , and ngc1808 @xcite . one particularly interesting object for comparison is ngc1808 , where a similar core+diffuse x - ray emission morphology was observed . the extended x - ray emission of the ngc1808 center ( @xmath136850 pc ) is dominated by soft radiation ( @xmath137 1 kev ) , and follows the same orientation as the h@xmath7 and optical - uv emission , suggesting intense star formation in the region . in contrast to ic342 , two x - ray point sources ( with a separation of @xmath138 pc ) were resolved in the core of ngc1808 . spectral information from the _ chandra _ acis data showed that they were associated respectively with hot gas emission and hard x - ray emission . the nuclear location from the 2mass data of ngc1808 was closer to the softer point source ( source s1 in @xcite ) . furthermore , both the x - ray spectra of the centers of ic342 and ngc1808 can be modeled with absorbed power - law plus mekal models , and the fe k@xmath7 emission is missing in both spectra . @xcite compared the ratios of the detected lines such as and in ngc1808 with those observed in m82 and concluded a starburst - agn coexistence in ngc1808 . with astrometrically corrected x - ray and radio data , we were able to study the radio loudness of the possible agn at the center of ic342 . as discussed in section [ ss : radio ] , radio observations of ic342 reveal a group of weak sources within 5@xmath54 radius of the center , with an integrated @xmath139 . this is slightly lower than the the range of @xmath140 observed from the 48 low - luminosity agns ( llagns ) sampled by @xcite . a weak radio source , source j , is near the x - ray center source c12 , with @xmath141 mjy , and @xmath142 mjy ( 3-@xmath143 limit ) , while a stronger source , source a , with @xmath144 mjy , and @xmath145 mjy , is @xmath146 away from c12 @xcite . therefore if we concentrate on the central @xmath147 radius region of ic342 , corresponding to the size of the core component of c12 , then the integrated radio flux in the region would be @xmath148 . together with the hard ( 2 10 kev ) x - ray flux of the c12 core determined from webpimms ( @xmath149 ) , we could determine the radio loudness parameter @xmath150 , as defined by @xcite , to be @xmath151 . this value is comparable to the average for seyfert galaxies ( @xmath152 ) as determined from a sample of 51 agn candidates @xcite . based on the distribution and flux of the radio sources in the nucleus , we could classify that ic342 as a radio - quiet type agn , if there existed one . radio - quiet agn included liners , seyfert ( 1 and 2 ) galaxies , and qsos @xcite . assuming the x - ray luminosity of c12 were due entirely to an agn , then the hard x - ray luminosity observed @xmath153 would be just below the lower range of other liners observed with _ _ asca _ , and _ chandra _ ( in the range @xmath154 ) , but would be low when compared to seyfert 1 galaxies or quasars ( typically @xmath155 ) . another property to consider is the x - ray flux variability . as shown in table [ t : lumtable ] , we observed the ic342 center x - ray flux to have varied by a factor of @xmath156 ( or flux variability amplitude of @xmath157 ) over 15 years in the three observations , which is comparable to the mean long term flux variability amplitude of 68% for type 1 agn and 48% for type 2 agn @xcite . on the other hand , we also notice that the power law index of the x - ray central source determined from the _ xmm - newton _ spectral fit , @xmath102 ( k03 ) , is slightly steeper than the typical value of radio - quiet agn and liners at @xmath158 @xcite . moreover , the absorption column density , @xmath159 ( b03 , k03 ) , is much lower than typical agn values of @xmath160 , together with no intrinsic absorption in the x - ray spectral fit . there was also no sign of the 6.4 kev fe k@xmath7 line emission , typically an indicator of agn , in the _ xmm - newton _ epic spectrum ( b03 , k03 ) . however , this line is also absent in a number of starburst galaxies identified as having hidden agn @xcite and thus we can not exclude the possibility of the presence of an agn in ic342 . the positional coincidence of the x - ray center with the optical star clusters in ic342 could suggest a scenario of coexistence of the nuclear star cluster and an agn , as discussed by other authors @xcite . we attempt to place a lower limit to the black hole mass . assuming that the black hole is accreting at sub - eddington rate , we could place a lower limit of the black hole mass at the center of ic342 at @xmath161 given the bolometric luminosity of the galaxy @xmath162 as determined by @xcite . on the other hand , it had been reported that nuclear star clusters are more massive than the coexisting black holes and their mass ratio @xmath163/@xmath164 @xcite . this implies @xmath165 , the nuclear star cluster mass as determined by ir data by @xcite . another estimate of the black hole mass comes from the @xmath166 relation @xcite , which had been shown to apply even in the massive star cluster regime @xcite . using the velocity dispersion measurements for the nuclear star cluster ( @xmath167 km s@xmath168 , @xcite ) , and the @xmath166 relation of @xcite , we derived @xmath169 , which is consistent with the limits described above . based on these calculations , the range of black hole mass is suggestive of an intermediate mass black hole ( imbh ) . however , we should not rule out the alternative possibility of an inactive supermassive black hole ( mass @xmath170 ) similar to known seyfert nuclei that the nuclear emission is due to advection - dominated accretion flows ( adaf ; @xcite ) . further spectral analysis could help to address whether this is the case . the spatial extent of the x - ray emission from the nucleus of ic342 could alternatively indicate the presence of populations of stellar objects uniformly distributed throughout the core region . these objects could be x - ray binaries , ob stars , or supernova remnants . for example , as pointed out by @xcite , populations of massive x - ray binaries evolved from high mass stars could be formed in starburst events , and such populations would be consistent with models associated with starbursts . it had been suggested that photoionization from these stars might have accounted for the the enhanced @xmath171 millimeter emission in the central ring of the mini spiral model of ic342 @xcite . we considered the possibility of multiple unresolved point sources , consisting of a mixture of stars and xrbs in the nuclear star cluster , contributing to the observed x - ray structure at the center of ic342 . a burst of star formation usually resulted in the production of large number of ob stars @xcite . with the typical x - ray luminosity for o stars @xmath172 @xcite , and the number of o stars in the nucleus of ic342 at @xmath1364000 estimated from the brackett @xmath128 emission @xcite , we derived the x - ray luminosity from these stars at @xmath173 . this is less than 0.1% of the total observed x - ray luminosity of c12 . the more luminous hmxbs , usually associated with young stellar population like o stars , might have also contributed to the observed x - ray emission . this could be inferred from the ratio of total x - ray luminosity from hmxb ( with @xmath174 ) to the number of o stars , i.e. @xmath175 @xcite . for the nucleus of ic342 , @xmath131 was estimated to be about @xmath176 ( 10%90% of x - ray luminosity of c12 ) , and the corresponding number of x - ray binaries was about 10 as inferred from @xmath130 . if these o stars and xrbs were uniformly distributed in a region of radius 50 pc ( the central @xmath74 region of c12 ) , the mean separation ( @xmath29 for hmxbs and @xmath177 for o stars ) would be smaller than the spatial resolution of _ chandra _ hrc - i . as a result , a model with 4000 o stars and @xmath178 hmxbs could account for both the observed morphology and flux of core component of the x - ray emission of c12 . we also considered whether the diffuse component of c12 could also be explained by the model above . as shown in figure [ f : halpha ] , the nucleus contains many regions of star formation , which could be the result of an outflow induced by the stellar winds of the massive stars or x - ray binaries present in the star clusters , as observed in m82 @xcite . the exact nature of this diffuse emission remains uncertain given the limited amount of x - ray data available . further observations with possible spectral analysis of this diffuse emission could help to link the starburst activities and the x - ray emission of the stellar objects in the nuclear star cluster . this paper aimed to study the nature of the x - ray emission in the nucleus of ic342 using high resolution ( @xmath179 ) _ chandra _ hrc - i observation . the historical ulx at the center of ic342 , x3 , was resolved into 2 sources , namely c12 and c13 , for the first time since the x - ray telescopes prior to _ chandra _ did not have enough spatial resolution . the brighter source c12 , with over 90% of the x - ray nuclear flux , was found to be located very close to the optical galactic center . the source c12 was also revealed as an extended source with a complex x - ray morphology , with a core component of diameter @xmath74 that is roughly consistent with a point source of @xmath180 , and a diffuse component that extends out to a diameter of @xmath181 , or @xmath182 pc . thus , we concluded that c12 is not an ulx . the much weaker source c13 , with @xmath183 , is consistent with a point source situated @xmath5 at p.a . = @xmath184 from c12 . registration of the x - ray image with optical ( _ hst _ ) and radio ( vla ) data allowed us to find the counterparts of x - ray sources . the registered position of c12 is r.a . = 03h:46m:48.43s , decl . = + 68d05m47.45s , which is located @xmath33 from the optical nsc @xcite and @xmath185 from a weak radio source ( source j in @xcite ) , and within the boresight correction @xmath90 of the _ chandra _ data . intense star formation was observed in the _ hst _ h@xmath7 image of the inner 100 pc of ic342 , where the 3 optical star clusters ( sc1 , sc2 , and nsc ) and our _ chandra _ x - ray source c12 are found . moreover , the diffuse x - ray emission extends preferentially along the north - south direction , which is the same as the alignment of the star clusters in the optical . this suggests that the nuclear x - ray emission is associated with star formation . we discussed possible contributors to the x - ray emission in the nuclear region of ic342 , with a main focus on the extended source c12 . a starburst model for ic342 based on molecular line observations @xcite that included a mini spiral arms - ring system was found to be consistent with the spatial extent of c12 . furthermore , the predicted x - ray emission from starbursts as derived from the brackett @xmath128 emission @xcite is close to the actual observed x - ray luminosity of c12 , indicating that x - ray from the nucleus is dominated by starburst activities . on the other hand , the hard emission from previous _ xmm - newton _ observation suggested that there could be an agn in ic342 ( b03 , k03 ) . the luminosity of c12 , @xmath186 , is consistent with that of a liner at ( typical luminosity @xmath187 ) . by comparing our current observation with earlier x - ray missions , we found a possible long term variability factor of 1.8 in the ic342 nuclear x - ray emission over 15 years . this amount of variation would be consistent with that from an agn . weak radio emission of @xmath188 in the vicinity of c12 is consistent with a radio - quiet agn . the observed properties ( luminosity , absorption column density ) are similar to those of type 2 low - luminosity liners or seyfert 2 galaxies . the black hole mass as inferred from the @xmath166 relation for the agn of @xmath189 , is consistent with an imbh , but we should not ignore the possibility of an inactive supermassive black hole . in conclusion , our current data did not confirm the existence of an agn in the core of ic342 , but it could neither be discarded . the lack of any spectral information of our _ chandra _ hrc - i data did not allow for detailed studies of the surface brightness distribution of the nuclear source c12 in different x - ray energy bands . in particular , separate spectral information on the core and diffuse components of the x - ray galactic center emission of ic342 would allow us the opportunity to put a tighter constraint on the co - existence of a low luminosity agn and a starburst in ic342 . with the x - ray emission from the core and diffused components at roughly the same level , further high spectral resolution observations with the _ chandra _ acis could shed new light on how the nucleus of ic342 fit into the starburst - agn family . we thank l. sjouwerman and c.w . tsai for generously providing the radio maps and helpful discussions . this work is based on observations obtained with _ chandra _ , support for this work was provided by the national aeronautics and space administration through chandra award number g06 - 7110x issued by the chandra x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of the national aeronautics space administration under contract nas8 - 03060 . mak acknowledges support from hku under the grant of postgraduate studentship . pun acknowledges support of a rgc grant from the government of the hong kong sar . kong acknowledges support from the national science council , taiwan , through a grant nsc96 - 2112-m-007 - 037-my3 . ccccccccl c1&3:45:10.62&68:02:31.95&1.1&0.7&@xmath190&5.60&@xmath191 & + c2&3:45:40.52&68:03:10.65&0.7&0.5&@xmath192&4.93&@xmath193 & _ rosat_#1 , xmm x2 + c3&3:45:55.62&68:04:55.89&0.2&0.4&@xmath194&27.17&@xmath195 & ic342 x1 , _ rosat_#3 , xmm x7 + c4&3:45:59.65&68:05:38.62&0.7&0.5&@xmath196&3.81&@xmath197 & xmm x9 + c5&3:46:06.58&68:07:05.70&0.5&0.4&@xmath198&7.14&@xmath199 & _ rosat_#4 , xmm x12 , @xmath200=12.8 @xmath133=11.82 + c6&3:46:15.79&68:11:12.98&0.3&0.4&@xmath201&18.76&@xmath202&ic342 x2 , _ rosat_#5 , xmm x13 + c7&3:46:22.17&68:05:05.84&0.4&0.4&@xmath203&3.03&@xmath204 & xmm x14 + c8&3:46:27.36&68:04:10.83&0.5&0.4&@xmath205&3.79&@xmath206 & + c9&3:46:43.66&68:06:11.58&0.2&0.4&@xmath207&7.60&@xmath208 & _ rosat_#6 , xmm x19 + c10&3:46:44.95&68:05:55.00&0.2&0.4&@xmath209&2.46&@xmath210 & + c11&3:46:45.56&68:09:46.43&0.9&0.4&@xmath211&3.09&@xmath212 & _ rosat_#7 , xmm x20 + c12&3:46:48.43&68:05:47.45&0.2&0.4&@xmath213&14.66&@xmath214 & nucleus , ic342 x3 , _ rosat_#8 , xmm x21 + c13&3:46:49.50&68:05:45.10&0.3&0.4&@xmath215&3.77&@xmath216 & + c14&3:46:52.82&68:05:04.96&0.2&0.4&@xmath217&2.40&@xmath218 & xmm x24 + c15&3:46:57.49&68:06:19.51&0.2&0.4&@xmath219&9.78&@xmath220 & _ rosat_#9 , xmm x26 + c16&3:47:01.86&68:02:37.02&0.6&0.5&@xmath221&3.73&@xmath222 & + c17&3:47:04.15&68:09:05.82&0.5&0.5&@xmath223&2.89&@xmath224 & + c18&3:47:04.89&68:05:18.18&0.2&0.4&@xmath225&3.02&@xmath226 & + c19&3:47:10.90&68:02:53.00&0.5&0.5&@xmath227&2.86&@xmath228 & + c20&3:47:18.38&67:51:14.59&1.4&0.6&@xmath229&8.88&@xmath230 & _ b_=11.7 _ r_=10.0 + c21&3:47:18.73&68:11:29.79&0.8&0.5&@xmath231&7.36&@xmath232 & xmm x30 + c22&3:47:22.91&68:08:59.81&0.4&0.4&@xmath233&7.49&@xmath234 & xmm x31 + c23&3:48:06.93&68:04:54.32&0.8&0.7&@xmath235&3.54&@xmath236 & + [ t : sourcelist ] region of ic342 displayed in square root scale to bring out the features . * top * : locations of the 3 optical star clusters are labeled by black circles . the centroids of the x - ray center sources c12 and c13 ( blue cross ) , and a @xmath74 diameter region centered around of c12 ( blue dashed circle ) , are also shown . radio sources detected in the vla 2 cm and 6 cm maps are marked and labeled ( red cross ) according to the naming convention by @xcite . * bottom * : overlay of _ chandra _ x - ray ( blue ; same as figure [ f : central_xray ] ) and radio 6 cm contours ( red ) on the optical data . contour levels of radio 6 cm flux are 0.37 , 0.93 , 1.49 , 2.05 , 2.61 mjy beam@xmath168 . , title="fig:",width=453 ] + region of ic342 displayed in square root scale to bring out the features . * top * : locations of the 3 optical star clusters are labeled by black circles . the centroids of the x - ray center sources c12 and c13 ( blue cross ) , and a @xmath74 diameter region centered around of c12 ( blue dashed circle ) , are also shown . radio sources detected in the vla 2 cm and 6 cm maps are marked and labeled ( red cross ) according to the naming convention by @xcite . * bottom * : overlay of _ chandra _ x - ray ( blue ; same as figure [ f : central_xray ] ) and radio 6 cm contours ( red ) on the optical data . contour levels of radio 6 cm flux are 0.37 , 0.93 , 1.49 , 2.05 , 2.61 mjy beam@xmath168 . , title="fig:",width=453 ] + |
we consider the valence - bond - solid ground state of the-deformed higher - spin aklt model (-vbs state ) with real . we investigate the eigenvalues and eigenvectors of a matrix ( matrix ) , which is constructed from the matrix product representation of the-vbs state . we compute the longitudinal and transverse spin - spin correlation functions , and determine the correlation amplitudes and correlation lengths . | 1009.4018 | in one - dimensional quantum systems , a completely different behavior for the integer spin chains from the half - integer spin chains was predicted by haldane @xcite . the antiferromagnetic isotropic spin-1 model introduced by affleck , kennedy , lieb and tasaki ( aklt model ) @xcite , whose ground state can be exactly calculated , has been a useful toy model to validate haldane s prediction of the massive behavior for integer spin chains . moreover , it lead to a deeper understanding for integer spin chains such as the discovery of the special type of long - range order @xcite . the aklt model has been generalized to higher - spin models , anisotropic models , etc @xcite . the hamiltonians are essentially linear combinations of projection operators with nonnegative coefficients , and their ground states are called valence - bond - solid ( vbs ) state . recently , the vbs state was investigated in perspective of its relation to quantum information and experimental implementation by means of optical lattices , see refs . @xcite for example . there are largely three types of representations for the ground state which are equivalent to each other : the schwinger boson representation , the spin coherent representation and the matrix product representation . for isotropic higher - spin models , the spin - spin correlation functions @xcite and the entanglement entropy @xcite have been calculated by utilizing the spin coherent representation and the properties of legendre polynomials . for the @xmath0-deformed spin-1 model , spin - spin correlation functions were evaluated @xcite from the matrix product representation . in this paper , we consider the ground state of a @xmath0-deformed higher - integer - spin model which was constructed recently in ref . @xcite ( @xmath0-vbs state ) . from its matrix product representation , we analyze one and two point functions of the @xmath0-vbs ground state . we notice that a matrix , which is constructed from the matrix product representation , plays a fundamental role in computing correlation functions , especially spin - spin two point correlation functions . investigating the structure of the matrix in detail , we obtain its eigenvalues and eigenvectors . utilizing the results , we determine the correlation amplitudes and correlation lengths of the longitudinal and transverse spin - spin correlation functions . this paper is organized as follows . in the next section , we briefly review the quantum group @xmath2 , and investigate the finite dimensional highest weight representation in terms of schwinger bosons . in section [ sec : qvbs ] , we precisely define the higher - spin generalization of the @xmath0-deformed aklt model on an @xmath3-site chain , and rigorously derive its @xmath0-vbs ground state in a matrix product form . the squared norm of the state will be written in terms of the trace of the @xmath3-th power of a matrix @xmath1 , which plays an important role in this paper . in section [ sec : spec - of - g ] , we obtain the eigenvalues and eigenvectors of @xmath1 . utilizing them , we compute one and two point functions in section [ sec : correlation ] . especially , we determine the correlation amplitudes and correlation lengths of the longitudinal and transverse spin - spin correlation functions . section [ sec : conclusion ] is devoted to the conclusion of this paper . we introduce several notations , fixing a real number @xmath0 throughout this paper . let us define the @xmath0-integer , @xmath0-factorial and @xmath0-binomial coefficient for @xmath4 as @xmath5=\frac{q^n - q^{-n}}{q - q^{-1 } } , \quad [ n]!= \begin{cases } \displaystyle \prod_{i=1}^{n } [ i ] & n\in { \mathbb n},\\ 1 & n=0 , \end{cases } \\ \left [ \begin{array}{c } n \\ k \end{array } \right ] = \begin{cases } \displaystyle \frac { [ n ] ! } { [ k ] ! [ n - k ] ! } & k=0,\dots , n,\\ 0 & \rm otherwise , \end{cases } \end{split}\end{aligned}\ ] ] respectively . the quantum group @xmath6 @xcite is defined by generators @xmath7 and @xmath8 with relations @xmath9 = \frac{q^h - q^{-h}}{q - q^{-1}},\quad \left[h , x^{\pm}\right ] = \pm 2 x^{\pm } .\end{aligned}\ ] ] the comultiplication is given by @xmath10 @xmath6 has the schwinger boson representation , where the generators are realized as @xmath11 with @xmath0-bosons @xmath12 and @xmath13 satisfying @xmath14 = -a , \quad [ n_a , a^\dagger ] = a^\dagger , \quad [ n_b , b ] = -b , \quad [ n_b , b^\dagger ] = b^\dagger.\end{aligned}\ ] ] we denote the space where @xmath15-dimensional highest weight representation of @xmath2 is realized by @xmath16 . the basis of @xmath16 is given by @xmath17 ! [ j - m ] ! } } the weyl representation which we describe below , is an equivalent representation to the schwinger boson representation , and is efficient for practical calculation . let us denote the @xmath0-bosons @xmath12 and @xmath13 acting on the @xmath18-th site as @xmath19 and @xmath20 . the weyl representation is to represent @xmath21 , @xmath22 , @xmath19 and @xmath20 on the space of polynomials @xmath23 $ ] as @xmath24 where @xmath25 the generators of @xmath2 are now represented as @xmath26 the tensor product of two irreducible representations has the clebsch - gordan decomposition @xmath27 } & { \left [ \begin{array}{ccc } s_1 & s_2 & j \\ m_1 & m_2 & m \end{array } \right ] } = \delta_{m_1+m_2,m}(-1)^{s_1-m_1 } q^{m_1(m_1+m_2 + 1)+ \ { s_2(s_2 + 1)-s_1(s_1 + 1)-j(j+1 ) \ } /2 } \nn \\ & \times \sqrt { \frac{[j+m ] ! [ j - m ] ! [ s_1-m_1 ] ! [ s_2-m_2 ] ! [ s_1+s_2-j ] ! [ 2j+1 ] } { [ s_1+m_1 ] ! [ s_2+m_2 ] ! [ s_1-s_2+j ] ! [ s_2-s_1+j ] ! [ s_1+s_2+j+1 ] ! } } \\ & \times \sum_{z=\mathrm{max}(0,-s_1-m_1,j - s_2-m_1)}^ { \mathrm{min}(j - m , s_1-m_1,s_2+j - m_1 ) } \frac { ( - q^{m+j+1 } ) ^z [ s_1+m_1+z ] ! [ s_2+j - m_1-z]!}{[z ] ! [ j - m - z ] ! [ s_1-m_1-z ] ! [ s_2-j+m_1+z ] ! } , \nn\end{aligned}\ ] ] is the @xmath0-analog of the clebsch - gordan coefficient @xcite . ( the factor @xmath28 is missing in ref . this coefficient is compatible with the inverse of the decomposition @xmath29 } |s;m_1\rangle \otimes|s;m_2\rangle .\end{aligned}\ ] ] for later purpose , we will also investigate the clebsch - gordan decomposition of @xmath2 in terms of the schwinger boson or the weyl representation . utilizing @xmath30 one can show that the highest weight vector @xmath31 acting on the @xmath18-th and @xmath32-th sites is given by @xmath33 moreover , we can show the following : [ x^-^nv_j ] @xmath34 ! \sum_{\mu=0}^n q^{-2\mu s } { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^\mu \left(x_\beta y_\alpha \right)^{n-\mu } \\ & \times \prod_{\nu=1}^{2s - j } \left ( x_\alpha y_\beta -q^{2(\nu - s-1)}x_\beta y_\alpha \right ) . \end{split } \label{vectorexpression}\end{aligned}\ ] ] a proof of this proposition is given in appendix [ sec : proof ] . let @xmath35 . noting @xmath36 } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } = 0,\end{aligned}\ ] ] for @xmath37 , one can see that @xmath38 the model we treat in this paper is an anisotropic integer spin-@xmath39 hamiltonian on an @xmath3-site chain with the periodic boundary condition @xmath40 where @xmath41 , and @xmath42 , which acts on the @xmath43-th and @xmath44-th sites , is the @xmath6 projection operator from @xmath45 to @xmath46 as @xmath47 } { \left [ \begin{array}{ccc } s & s & j \\ m_1 ' & m_2 ' & m_1'+ m_2 ' \end{array } \right]}\\ & \quad\quad\times \delta_{m_1+m_2 , m_1'+ m_2 ' } \otimes \end{split}\end{aligned}\ ] ] the nonnegativity @xmath48 implies that all the eigenvalues of @xmath49 are nonnegative . ( of course , @xmath50 is the hermitian conjugate of @xmath51 . ) moreover , we will see that the energy of the ground state @xmath52 is zero : @xmath53 since we set @xmath54 , we find that is equivalent to @xmath55 noting the nonnegativity . from proposition [ x^-^nv_j ] , one observes that any vector in @xmath56 of the @xmath43-th and @xmath44-th sites has the form @xmath57 where @xmath58 does not depend on @xmath59 or @xmath60 . thus , the condition imposes the restriction that @xmath52 has the form @xmath61 with some polynomials @xmath62 such that this form is consistent with for @xmath63 . the unique choice of @xmath62 with such consistency is a constant ( which can be set to be 1 ) , and we achieve the unique ground state @xmath64 in the schwinger boson representation , we have @xmath65 which is a generalization of the @xmath66 case @xcite . note that each site have the correct spin value : @xmath67 ( @xmath68 ) where @xmath69 . our ground state is a @xmath0-deformation of the valence - bond - solid ( vbs ) state , which we call @xmath0-vbs state , see figure [ q - vbs ] . -vbs state . each line is a @xmath0-deformed valence bond , and the circle @xmath70 represents the @xmath0-symmetrization of spin-1/2 particles @xmath71 at each site.,height=94 ] the schwinger boson representation of the ground state can be transformed into the following equivalent form called the matrix product representation @xcite , which generalizes the @xmath66 @xcite or @xmath72 @xcite case . noting , we have @xmath73,\end{aligned}\ ] ] where @xmath74 is an @xmath75 vector - valued matrix acting on the @xmath43-th site whose element is given by @xmath76 \left [ \begin{array}{c } s \\ i ' \end{array } \right ] [ s - i+ i ' ] ! [ s+i- i ' ] ! } \ |s ; i'-i\rangle_k \\ & = : h_{i i ' } |s ; i'-i \rangle_k , \quad\quad ( 0\le i , i'\le s ) . \end{split}\end{aligned}\ ] ] the symbol @xmath77 for two @xmath75 vector - valued matrices @xmath78 is defined by @xmath79 which is associative . for example , for @xmath80 , @xmath81 and the product in the form is calculated as @xmath82 with @xmath83 then the matrix product ground state is @xmath84 we define @xmath85 by replacing each ket vector in the matrix @xmath74 by its corresponding bra vector : @xmath86 for example , for @xmath80 , @xmath87 now we introduce `` @xmath1 matrix '' , which will play an important role in our study . let us set an @xmath88 dimensional vector space @xmath89 and its dual orthogonal space @xmath90 as @xmath91 here , @xmath92 ( @xmath93 ) is an orthonormal ( dual orthonormal ) basis . we define an @xmath94 matrix @xmath1 acting on the space @xmath89 as @xmath95 or equivalently as @xmath96 we also introduce @xmath97 for an operator @xmath98 acting on the one - site vector space @xmath99 as @xmath100 each element of the matrix @xmath1 can be expressed explicitly as @xmath101 where @xmath102 \left [ \begin{array}{c } s \\ b \end{array } \right ] \left [ \begin{array}{c } s \\ c \end{array } \right ] \left [ \begin{array}{c } s \\ d \end{array } \right ] [ s - a+c ] ! [ s+a - c ] ! [ s - b+d ] ! [ s+b - d ] ! } \ . \end{split}\end{aligned}\ ] ] each element of @xmath97 for @xmath103 and @xmath104 , which act on @xmath105 as @xmath106 can be also expressed as @xmath107 the squared norm of the ground state is calculated as @xmath108 \operatorname{tr}\left [ g_1 \star\cdots\star g_l \right ] \\ = & \operatorname{tr}\left[\left ( g_1^\dagger \star \cdots \star g_l^\dagger \right ) \otimes \left ( g_1 \star \cdots \star g_l \right ) \right ] \\ = & \operatorname{tr}\left[\left ( g_1^\dagger \otimes g_1 \right ) \star \cdots \star \left ( g_l^\dagger \otimes g_l \right ) \right ] \\ = & \operatorname{tr}g^l . \label{psipsi = trgl } \end{split}\end{aligned}\ ] ] note that the elements of @xmath109 are no longer vectors , and thus we can replace the symbol @xmath77 by the usual product in the third line of . the one point function @xmath110 of an operator @xmath98 can be written in terms of @xmath1 and @xmath97 as @xmath111 \operatorname{tr}\left[a_1g_1 \star g_2 \star \cdots \star g_l \right ] } { \langle\psi|\psi\rangle } = \frac{\operatorname{tr}g_a g^{l-1 } } { \operatorname{tr}g^l } , \label{onepoint } \end{split}\end{aligned}\ ] ] where @xmath112 is defined by @xmath113 . in the same way , the two point function of @xmath98 and @xmath114 can also be written in terms of @xmath115 and @xmath116 as @xmath117 investigating the eigenvalues and eigenvectors of the matrix @xmath1 will be crucial for the analysis of correlation functions . in the next section , we study the @xmath1 matrix in detail . @xcite , we conjectured that the spectrum of @xmath1 is given by @xmath118}!\right)^2 { \left [ \begin{array}{c } 2s+1 \\ s-\ell \end{array } \right ] } , \quad(\ell=0,1,\dots , s ) , \label{eigenvalues}\end{aligned}\ ] ] where the degree of the degeneracy of each @xmath119 is @xmath120 . one can easily find that @xmath121 in this section , we prove the conjecture by giving an exact form for the eigenvector corresponding to each eigenvalue . + first one observes that the @xmath1 matrix has the following block diagonal structure : @xmath122 the size of each block @xmath123 is @xmath124 . each element of @xmath123 is @xmath125}!{\left [ s+a - c \right ] } ! \\ & \times \sqrt { { \left [ \begin{array}{c } s \\ a \end{array } \right ] } { \left [ \begin{array}{c } s \\ a+j \end{array } \right ] } { \left [ \begin{array}{c } s \\ c \end{array } \right ] } { \left [ \begin{array}{c } s \\ c+j \end{array } \right ] } } \ . \end{split}\end{aligned}\ ] ] we construct intertwiners among the @xmath126 blocks @xmath127 . this helps us to construct eigenvectors of each block from another block with a smaller size . ( the same idea was used in ref . @xcite to study the spectrum of a multi - species exclusion process ) . let us define a family of linear operators @xmath128 as @xmath129}{\left [ s - a - j+1 \right]}}{{\left [ j \right]}{\left [ s - j+1 \right ] } } } & c = a , \\ \displaystyle -q^{1-a - j}\sqrt { \frac{{\left [ a \right ] } { \left [ s - a+1 \right]}}{{\left [ j \right]}{\left [ s - j+1 \right ] } } } & c = a-1 , \\ 0 & \rm otherwise \end{cases } \\ \nonumber\end{aligned}\ ] ] for @xmath130 , and @xmath131}{\left [ s - a+j+1 \right]}}{{\left [ -j \right]}{\left [ s+j+1 \right ] } } } & c = a , \\ \displaystyle -q^{1-a+j}\sqrt { \frac{{\left [ a \right ] } { \left [ s - a+1 \right]}}{{\left [ -j \right]}{\left [ s+j+1 \right ] } } } & c = a-1 , \\ 0 & \rm otherwise . \end{cases}\\ \nonumber\end{aligned}\ ] ] for @xmath132 . by direct calculation , one finds the matrix @xmath133 enjoys the intertwining relation @xmath134 [ intertwining ] with the use of proposition [ intertwining ] , one can show the following : [ eigen - structure ] each block @xmath123 has a simple ( nondegenerated ) spectrum @xmath135 and the corresponding eigenvectors are given by @xmath136 } ! { \left [ i+\ell \right ] } ! { \left [ s - i \right ] } ! } { [ s ] ! [ \ell ] ! [ s -i-\ell ] ! [ i ] ! } } |i , i+\ell{{\rangle\!\rangle } } & j\ge 0 , \\ \displaystyle \sum_{0\le i\le s-\ell}q^{(\ell+1)i } \sqrt{\frac{{\left [ s-\ell \right ] } ! { \left [ i+\ell \right ] } ! { \left [ s - i \right ] } ! } { [ s ] ! [ \ell ] ! [ s -i-\ell ] ! [ i ] ! } } |i+\ell , i{{\rangle\!\rangle } } & j<0 , \end{cases}\end{aligned}\ ] ] for @xmath137 , and @xmath138 for @xmath139 . [ gmatrixstructure ] @xmath140{270}{$\mapsto$ } } } \\ w_{s-1 } & |\lambda_s{{\rangle\!\rangle}}_{s-1 } & |\lambda_{s-1}{{\rangle\!\rangle}}_{s-1 } \\ \ \downarrow\scriptstyle{i}_{s-1 } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } \\ w_{s-2 } & |\lambda_s{{\rangle\!\rangle}}_{s-2 } & |\lambda_{s-1}{{\rangle\!\rangle}}_{s-2 } & |\lambda_{s-2}{{\rangle\!\rangle}}_{s-2 } \vspace{2 mm } \\ \ \vdots & \quad\vdots & \quad \vdots & \quad\vdots \quad\quad\quad\quad\quad\ \ddots \vspace{2 mm } \\ w_2 & |\lambda_s{{\rangle\!\rangle}}_2 & |\lambda_{s-1}{{\rangle\!\rangle}}_2 & |\lambda_{s-2}{{\rangle\!\rangle}}_2 \quad \ \dots & \,|\lambda_2{{\rangle\!\rangle}}_2 \\ \ \downarrow\scriptstyle{i}_2 & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } \\ w_1 & |\lambda_s{{\rangle\!\rangle}}_1 & |\lambda_{s-1}{{\rangle\!\rangle}}_1 & |\lambda_{s-2}{{\rangle\!\rangle}}_1 \quad \ \dots & \,|\lambda_2{{\rangle\!\rangle}}_1&|\lambda_1{{\rangle\!\rangle}}_1 \\ \ \downarrow\scriptstyle{i}_1 & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \quad{\text{\rotatebox[origin = c]{270}{$\mapsto$ } } } & \\ w_0 & |\lambda_s{{\rangle\!\rangle}}_0 & |\lambda_{s-1}{{\rangle\!\rangle}}_0 & |\lambda_{s-2}{{\rangle\!\rangle}}_0 \quad \ \dots & \ , |\lambda_2{{\rangle\!\rangle}}_0 & |\lambda_1{{\rangle\!\rangle}}_0 & |\lambda_0{{\rangle\!\rangle}}_0 \\ \ \uparrow_{i_{-1 } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \\ w_{-1 } & |\lambda_s{{\rangle\!\rangle}}_{-1 } & |\lambda_{s-1}{{\rangle\!\rangle}}_{-1 } & |\lambda_{s-2}{{\rangle\!\rangle}}_{-1 } \quad \dots & \ , |\lambda_2{{\rangle\!\rangle}}_{-1}&|\lambda_1{{\rangle\!\rangle}}_{-1 } \\ \ \uparrow_{i_{-2 } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } \\ w_{-2 } & |\lambda_s{{\rangle\!\rangle}}_{-2 } & |\lambda_{s-1}{{\rangle\!\rangle}}_{-2 } & |\lambda_{s-2}{{\rangle\!\rangle}}_{-2 } \quad \dots & \ , |\lambda_2{{\rangle\!\rangle}}_{-2 } \vspace{2 mm } \\ \ \vdots & \quad\vdots & \quad \vdots & \quad\vdots \quad\quad\quad\quad\quad\ { \text{\rotatebox[x=1mm , y=2.5mm]{270}{$\ddots$}}}\vspace{2 mm } \\ w_{-s+2 } & |\lambda_s{{\rangle\!\rangle}}_{-s+2 } & |\lambda_{s-1}{{\rangle\!\rangle}}_{-s+2 } & |\lambda_{s-2}{{\rangle\!\rangle}}_{-s+2 } \\ \ \uparrow_{i_{-s+1 } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } \\ w_{-s+1 } & |\lambda_s{{\rangle\!\rangle}}_{-s+1 } & |\lambda_{s-1}{{\rangle\!\rangle}}_{-s+1 } \\ \ \uparrow_{i_{-s } } & \quad { \text{\rotatebox[origin = c]{90}{$\mapsto$ } } } \\ w_{-s } & |\lambda_s{{\rangle\!\rangle}}_{-s } \\ \end{array}\end{aligned}\ ] ] figure [ structure ] is helpful to understand how the eigenvectors are constructed . we prove this theorem below for only @xmath141 since one can show it for @xmath142 in the same way . _ proof of theorem [ eigen - structure ] . _ first , by direct calculation given below , we find that @xmath143 has an eigenvalue @xmath144 and its eigenvector is @xmath145 defined by . each element of @xmath146 is calculated as @xmath147}!{\left [ s+a - c \right ] } ! \\ & \quad\times \sqrt { { \left [ \begin{array}{c } s \\ a \end{array } \right ] } { \left [ \begin{array}{c } s \\ a+j \end{array } \right ] } { \left [ \begin{array}{c } s \\ c \end{array } \right ] } { \left [ \begin{array}{c } s \\ c+j \end{array } \right ] } } q^{(j+1)c}\displaystyle \sqrt{\frac{{\left [ s - j \right ] } ! { \left [ c+j \right ] } ! { \left [ s - c \right ] } ! } { { \left [ s \right ] } ! { \left [ j \right ] } ! { \left [ s - c - j \right ] } ! { \left [ c \right ] } ! } } \\ & = ( -1)^j q^{(a+j - s-1)(s+1)-(j+1 ) } \sqrt { { \left [ \begin{array}{c } s \\ a \end{array } \right ] } { \left [ \begin{array}{c } s \\ a+j \end{array } \right ] } \displaystyle \frac{{\left [ s - j \right]}!}{{\left [ s \right ] } ! { \left [ j \right ] } ! } } \\ & \quad\times { \left [ s \right ] } ! \sum_{0\le c\le s - j } q^{(c+1)(s+j+2 ) } \frac{{\left [ s - a+c \right]}!{\left [ s+a - c \right]}!}{{\left [ s - c - j \right ] } ! [ c ] ! } . \end{split}\end{aligned}\ ] ] using the formula @xmath148 } { \left [ \begin{array}{c } \beta+k \\ k \end{array } \right ] } q^{k(\alpha+\beta+2 ) } = { \left [ \begin{array}{c } \alpha+\beta+n+1 \\ n \end{array } \right ] } q^{n(1+\beta)},\end{aligned}\ ] ] we obtain @xmath149 } { \left [ \begin{array}{c } s \\ a+j \end{array } \right ] } \displaystyle \frac{{\left [ s - j \right]}!}{{\left [ s \right ] } ! { \left [ j \right ] } ! } } \\ & \quad\times { \left [ s \right]}!\ , q^{-sj+a(j - s)+s^2 + 2s+2 } { \left [ \begin{array}{c } 2s+1 \\ s - j \end{array } \right ] } { \left [ s - a \right]}!{\left [ a+j \right ] } ! \\ & = ( -1)^j\left({\left [ s \right]}!\right)^2 { \left [ \begin{array}{c } 2s+1 \\ s - j \end{array } \right ] } q^{(j+1)a } \sqrt{\displaystyle \frac{{\left [ s - j \right]}!{\left [ a+j \right]}!{\left [ s - a \right]}!}{{\left [ s \right]}!{\left [ j \right]}!{\left [ s - a - j \right]}!{\left [ a \right ] } ! } } \\ & = \lambda_j { { \langle\!\langle}}a , a+j |\lambda_j{{\rangle\!\rangle}}_j . \end{split}\end{aligned}\ ] ] note that the first element of @xmath145 is 1 by the definition : @xmath150 . next , we show by induction that @xmath143 has eigenvalues @xmath119 ( @xmath151 ) and their corresponding eigenvectors are given by @xmath152 defined by . suppose the theorem is true for @xmath153 , that is to say that the block @xmath154 has the eigenvalues @xmath119 and their corresponding eigenvectors @xmath155 @xmath156 with @xmath157 for @xmath158 . additionally , suppose that the first element of each @xmath159 is 1 . using the intertwining relation , one finds @xmath160 . we also find that the first element of @xmath161 is 1 , and thus @xmath162 is nonzero . furthermore , thanks to @xmath163 , the vectors @xmath162 ( @xmath164 ) are distinct ( in other words , @xmath165 is injective ) . we have already constructed the remaining eigenvector of @xmath143 explicitly , which is @xmath166 with its eigenvalue @xmath144 distinct from @xmath167 . the conjecture for the eigenvalues of the @xmath1 matrix that we exhibited in the beginning of this section follows as a simple corollary of theorem [ gmatrixstructure ] . moreover , we constructed their eigenvectors which are important for computing spin - spin correlation functions . [ prop - squared - norm ] the squared norm of @xmath152 is @xmath168 ! [ \ell-|j| ] ! [ s-\ell ] ! [ |j| ] ! } { [ s ] ! [ \ell+|j| ] ! [ s-|j| ] ! [ 2\ell+1 ] } , \label{squared - norm } \end{aligned}\ ] ] where we denote the transpose of @xmath169 by @xmath170 . we prove this proposition only for @xmath141 . _ proof of proposition [ prop - squared - norm ] . _ one can easily show that the product of intertwiners ( which is also an intertwiner ) has the following form by induction : @xmath171 } \sqrt { \frac { [ j ] ! [ s-\ell ] ! [ a ] ! [ s - c ] ! [ c+\ell ] ! [ s-(a+j ) ] ! } { [ \ell ] ! [ s - j ] ! [ c ] ! [ s - a ] ! [ a+j ] ! [ s -(c + \ell ) ] ! } } . \end{split}\end{aligned}\ ] ] then , @xmath172 is calculated as @xmath173!\right)^2 [ j ] ! } { [ s ] ! \left([\ell]!\right)^2 [ s - j ] ! } \sum_{0 \le a\le s - j \atop 0\le i , i'\le s } ( -1)^{i+i ' } q^ { ( i+i')(\ell+j+1)-2a\ell } \\ & \quad\quad \times { \left [ \begin{array}{c } \ell - j \\ a - i \end{array } \right ] } { \left [ \begin{array}{c } \ell - j \\ a - i ' \end{array } \right ] } \frac{[s - i ] ! [ i+\ell ] ! [ s - i ' ] ! [ i'+\ell ] ! [ a ] ! [ s-(a+j ) ] ! } { [ i ] ! [ s-(i+\ell ) ] ! [ i ' ] ! [ s-(i'+\ell ) ] ! [ s - a]![a+j ] ! } . \end{split}\end{aligned}\ ] ] the triple sum has the closed form @xmath174!\right)^2 [ \ell - j]![s+\ell+1 ] ! } { [ s-\ell ] ! [ j+\ell ] ! [ 2\ell+1]},\end{aligned}\ ] ] which finishes the proof . in the last section , we investigated the eigenvalues and eigenvectors of the @xmath1 matrix . by utilizing theorem [ gmatrixstructure ] and noting , the one point function @xmath175 can be represented as @xmath176 in the thermodynamic limit @xmath177 . as an application , we can calculate the probability of finding @xmath178 value as @xmath179![s - m]!}{[2s+1 ] ! } \sum_{i=0}^{s } q^{(s+2)(2i - m - s ) } \left [ \begin{array}{c } s \\ i - m \end{array } \right ] \left [ \begin{array}{c } s \\ i \end{array } \right ] . \end{split}\end{aligned}\ ] ] the two point function can be also represented as @xmath180 in the thermodynamic limit . inserting , , , , , and into , one finds the large - distance ( @xmath181 ) behaviors of the spin - spin correlation functions @xmath182 and @xmath183 are @xmath184 [ s+2 ] } { q^{2s-2 } [ s ] ( [ 2s+1 ] ! ) ^2 } ( { } _ 0 { { \langle\!\langle}}\lambda_1|g_{s^z}| \lambda_0 { { \rangle\!\rangle}}_0)^2 \left(-\frac{[s ] } { [ s+2 ] } \right)^r , \\ \langle s_1^+ s_r^- \rangle&=- \frac{[2 ] [ 3 ] [ s+2 ] } { q^{3s-2 } ( [ 2s+1 ] ! [ s ] ) ^2 } ( { } _ { -1 } { { \langle\!\langle}}\lambda_1|g_{s^-}| \lambda_0 { { \rangle\!\rangle}}_0)^2 \left(-\frac{[s ] } { [ s+2 ] } \right)^r,\end{aligned}\ ] ] where @xmath185 ! [ s+ i'-i ] ! \left [ \begin{array}{c } s \\ i \end{array } \right ] \left [ \begin{array}{c } s \\ i ' \end{array } \right ] , \end{split } \\ \begin{split } & \!\!\!\!\!\ ! { } _ { -1 } { { \langle\!\langle}}\lambda_1|g_{s^-}| \lambda_0 { { \rangle\!\rangle}}_0 = { } _ 0 { { \langle\!\langle}}\lambda_0|g_{s^+}| \lambda_1 { { \rangle\!\rangle}}_{-1 } \\ & = -q^{-s^2-s/2 + 1/2 } \sum_{i=0}^{s } \sum _ { i'=0}^{s-1 } q^{(s+2)i+(s+3 ) i ' } \sqrt{\left[\begin{array}{c } s \\ i'+1 \end{array } \right ] \left [ \begin{array}{c } s \\ i ' \end{array } \right ] } \\ & \quad \times \sqrt{(s+i- i ' ) [ s+i- i ' ] ( s - i+ i'+1 ) [ s - i+ i'+1 ] } \\ & \quad \times \sqrt { [ i'+1][s - i ' ] [ s]^{-1 } } [ s+ i'-i ] ! [ s+i- i'-1 ] ! \left [ \begin{array}{c } s \\i \end{array}\right ] . \end{split}\end{aligned}\ ] ] note that the terms with @xmath186 and @xmath187 in dominate the large - distance behaviors of @xmath188 and @xmath189 , respectively , since @xmath190 both @xmath182 and @xmath191 exhibit exponential decay with correlation length @xmath192 } { [ s ] } \right)^{-1},\end{aligned}\ ] ] generalizing the results for @xmath66 @xcite or @xmath72 @xcite case . in this paper , we investigated one and two point functions of the @xmath0-vbs ground state of an integer spin model ( the @xmath0-deformed higher - spin aklt model ) . the formulation of correlation functions by use of the matrix product representation of the ground state shows that the structure of a matrix , which we call @xmath1 matrix , plays an important role . we obtained the eigenvalues and eigenvectors of the @xmath1 matrix with the help of constructing intertwiners connecting different blocks of @xmath1 . then we calculated the spin - spin correlation functions by use of the eigenvalues and eigenvectors of the @xmath1 matrix , and determined the correlation amplitudes and correlation lengths of the longitudinal and transverse spin - spin correlation functions . it is interesting to investigate other types of correlation functions . for example , the entanglement entropy , which is defined in terms of the reduced density matrix , is a typical quantification of the entanglement of quantum systems . it is intriguing to calculate the entanglement entropy for the @xmath0-deformed model and observe the change from the isotropic point @xcite ( see also refs . @xcite for other vbs states ) . the authors thank atsuo kuniba and kazumitsu sakai for useful discussion . ca also thanks kirone mallick and andreas schadschneider for the very kind hospitality during his stay in europe . this work is supported by grant - in - aid for young scientists ( b ) 22740106 and global coe program `` education and research hub for math - for - industry . '' we prove proposition [ x^-^nv_j ] by induction . suppose that proposition [ x^-^nv_j ] holds for @xmath193 . we calculate the four terms of the action of @xmath194 on @xmath193 , separately . @xmath195 ! ( x_\alpha x_\beta)^{j-(n+1 ) } \\ = & \nn \sum_{\mu=0}^n q^{-2\mu s + j - n } { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^\mu \left(x_\beta y_\alpha \right)^{n+1-\mu } \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1)}x_\beta y_\alpha \right ) , \end{aligned}\ ] ] @xmath196!(x_\alpha x_\beta)^{j-(n+1 ) } \\ = & \nn \sum_{\mu=0}^n q^{-2\mu s -2\mu - 4s + j + n } { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^\mu \left(x_\beta y_\alpha \right)^{n+1-\mu } \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s)}x_\beta y_\alpha \right),\end{aligned}\ ] ] @xmath197!(x_\alpha x_\beta)^{j-(n+1 ) } \\ = & \nn \sum_{\mu=0}^n q^{-2\mu s - 2\mu -2s + j + n } { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu+1 } \left(x_\beta y_\alpha \right)^{n-\mu } \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s)}x_\beta y_\alpha \right),\end{aligned}\ ] ] @xmath198!(x_\alpha x_\beta)^{j-(n+1 ) } \\ = & \nn \sum_{\mu=0}^n q^{-2\mu s -2s - j + n } { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu+1 } \left(x_\beta y_\alpha \right)^{n-\mu } \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1)}x_\beta y_\alpha \right).\end{aligned}\ ] ] @xmath199 gives @xmath200 } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu } \left(x_\beta y_\alpha \right)^{n-\mu } \\ & \quad \times \left ( x_\alpha y_\beta -q^{-2s } x_\beta y_\alpha \right ) \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s ) } x_\beta y_\alpha \right ) \end{split } \nonumber \\ & \begin{split } = & \sum_{\mu=0}^n q^{-2\mu s - 2\mu -2s + j + n } { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu } \left(x_\beta y_\alpha \right)^{n-\mu } \\ & \quad \times \left ( x_\alpha y_\beta -q^{2(s - j ) } x_\beta y_\alpha \right ) \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1 ) } x_\beta y_\alpha \right ) \end{split } \nonumber \\ & \begin{split } = & \sum_{\mu=0}^n q^{-2\mu s - 2\mu -2s + j + n } { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu+1 } \left(x_\beta y_\alpha \right)^{n-\mu } \\ & \quad \times \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1 ) } x_\beta y_\alpha \right ) \\ & - \sum_{\mu=0}^n q^{-2\mu s - 2\mu - j + n } { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu } \left(x_\beta y_\alpha \right)^{n+1-\mu } \\ & \quad \times \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1 ) } x_\beta y_\alpha \right ) . \label{third - second } \end{split}\end{aligned}\ ] ] dividing the first term of @xmath199 by @xmath201 , we obtain @xmath202 { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu+1 } \left(x_\beta y_\alpha \right)^{n-\mu } \\ & \quad \times \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1 ) } x_\beta y_\alpha \right ) \end{split } \nonumber \\ \begin{split } = & \sum_{\mu=1}^{n+1 } q^{-2\mu s - \mu + n + 1 } [ j-\mu+1 ] { \left [ \begin{array}{c } j \\ \mu-1 \end{array } \right ] } { \left [ \begin{array}{c } j \\ n+1-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu } \left(x_\beta y_\alpha \right)^{n+1-\mu } \\ & \quad \times \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1 ) } x_\beta y_\alpha \right ) , \label{mochotto1 } \end{split}\end{aligned}\ ] ] where we replaced @xmath203 . dividing @xmath199 the second term of by @xmath201 , we obtain @xmath204 { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu } \left(x_\beta y_\alpha \right)^{n+1-\mu } \\ & \quad \times \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1 ) } x_\beta y_\alpha \right ) . \label{mochotto2 } \end{split}\end{aligned}\ ] ] since @xmath205}=0 $ ] , we can extend @xmath206 in and to @xmath207 . finally we have @xmath208 { \left [ \begin{array}{c } j \\ \mu \end{array } \right ] } { \left [ \begin{array}{c } j \\ n+1-\mu \end{array } \right ] } \left(x_\alpha y_\beta \right)^{\mu } \left(x_\beta y_\alpha \right)^{n+1-\mu } \\ & \quad \times \prod_{\nu=1}^{2s - j } \left(x_\alpha y_\beta -q^{2(\nu - s-1 ) } x_\beta y_\alpha \right ) , \end{split } \label{final}\end{aligned}\ ] ] and since @xmath209 ! ( x_\alpha x_\beta)^{j-(n+1 ) } & = \frac{\eqref{first}+\eqref{third}-\eqref{second}-\eqref{fourth}}{q - q^{-1 } } \\ & = \frac{\eqref{first}+\eqref{third - second}-\eqref{fourth}}{q - q^{-1 } } \\ & = \eqref{mochotto1 } + \eqref{mochotto2 } \\ & = \eqref{final } , \end{split}\end{aligned}\ ] ] proposition [ x^-^nv_j ] is true for @xmath210 . haldane , phys . lett . a 93 ( 1983 ) 464 . haldane , phys . 50 ( 1983 ) 1153 . i. affleck , t. kennedy , e.h . lieb and h. tasaki , comm . phys . 115 ( 1988 ) 477 . m. de nijs and k. rommelse , phys . b 40 ( 1989 ) 4709 . h. tasaki , phys . 66 ( 1991 ) 798 . d. p. arovas , a. auerbach and f.d.m . haldane , phys . rev . lett . 60 ( 1988 ) 531 . a. klmper , a. schadschneider and j. zittartz , j. phys . a 24 ( 1991 ) a. klmper , a. schadschneider and j. zittartz , z. phys . b 87 ( 1992 ) 281 . a. klmper , a. schadschneider and j. zittartz , europhys . ( 1993 ) 293 . k. totsuka and m. suzuki , j. phys . a 27 ( 1994 ) 6443 . m.t . batchelor and c.m . yung , int . b 8 ( 1994 ) 3645 . m. greiter and s. rachel , phys . rev . b 75 ( 2007 ) 184441 . d. schuricht and s. rachel , phys . b 78 ( 2008 ) 014430 . tu , g - m . zhang and t. xiang , phys . b 78 ( 2008 ) 094404 . tu , g - m . zhang , t. xiang , z - x . liu and t - k . ng , phys . b 80 ( 2009 ) 014401 . v. karimipour and l. memarzadeh , phys . rev . b 77 ( 2008 ) 094416 . d. p. arovas , k. hasebe , x - l . qi and s - c . zhang , phys . b 79 ( 2009 ) 224404 . f. verstraete , m.a . martin - delgado , j.i . cirac , phys . 92 ( 2004 ) 087201 j.j . garca - ripoll , m.a . martin - delgado , j. i. cirac , phys . 93 , 250405 ( 2004 ) w.d . freitag and e. mller - hartmann , z. phys . b 83 ( 1991 ) 381 . y. xu , h. katsura , t. hirano and v.e . korepin , j. stat . ( 2008 ) 347 . korepin and y. xu , int . j. mod . b 24 ( 2010 ) 1361 . a. n. kirillov and n. yu . reshetikhin , adv . ser . in math . vol . 7 ( 1988 ) 285 k. motegi , phys . a 374 ( 2010 ) 3112 . v. drinfeld , sov . math - dokl . 32 ( 1985 ) 254 . m. jimbo , lett . 10 ( 1985 ) 63 . k. totsuka and m. suzuki , j. phys . matter 7 ( 1995 ) 1639 . c. arita , a. kuniba , k. sakai and t. sawabe , j. phys . a 42 ( 2009 ) 345002 . h. fang , v.e.korepin and v. roychowdhury , phys . 93 ( 2004 ) 227203 . h. katsura , t. hirano and y. hatsugai , phys . b 76 ( 2007 ) 012401 . h. katsura , t. hirano and v.e . korepin , j. phys . a 41 ( 2008 ) 135304 . h. katsura , n. kawashima , a. kirillov , v.e . korepin and s. tanaka , j. phys . a 43 ( 2010 ) 255303 . |
we study the loop model on the honeycomb lattice with real value by means of a cluster algorithm . the formulation of the algorithm is based on the equivalence of the loop model and the low - temperature graphical representation of a-color ashkin - teller model on the triangular lattice . the latter model with integer can be simulated by means of an embedding swendsen - wang - type cluster method . by taking into account the symmetry among loops of different colors , we develop another version of the swendsen - wang - type method . this version allows the number of colors to take any real value . as an application , we investigate the , and loop model at criticality . the determined values of various critical exponents are in excellent agreement with theoretical predictions . in particular , from quantities associated with half of the loops , we determine some critical exponents that corresponds to those for the tricritical potts model but have not been observed yet . dynamic scaling behavior of the algorithm is also analyzed . the numerical data strongly suggest that our cluster algorithm _ hardly _ suffers from critical slowing down . # 1([#1 ] ) # 1to 0pt#1 # 1proof of # 1 . # 1#2 # 1 # 1 s h p [ section ] [ theorem]definition [ theorem]proposition [ theorem]lemma [ theorem]corollary [ theorem]conjecture [ theorem]question | cond-mat0605165 | in monte carlo studies of statistical systems undergoing phase transitions , critical slowing down is one of the prominent problems . consider a monte carlo algorithm with dynamic exponent @xmath6 . in order to generate a given number of effectively independent samples , one has to spend computing effort @xmath7 , where @xmath8 is the linear sytem size and @xmath9 accounts for the volume of the sytem of interest . for the local metropolis algorithm for the potts model , the dynamic exponent is around @xmath10 . thus , in two dimensions , the required computing effort grows like @xmath11 . therefore , a central task of computational statistical physics is to develop algorithms such that @xmath12 vanishes or is significantly suppressed . for a discussion of monte carlo methods , see ref . @xcite . for the potts model , a significant breakthrough was the invention of the swendsen - wang cluster method @xcite and its single - cluster version the wolff cluster method @xcite . for the two- and three - dimensional ising model , the dynamic exponent of the sw algorithm is about @xmath13 and @xmath14 @xcite , respectively . in comparison with the metropolis simulations , the critical slowing down is significantly suppressed . in addition to the potts model , another important class of models in statistical physics is the @xmath0 model . the @xmath0 model is defined in terms of @xmath2-component spins on a lattice , with an isotropic pair coupling of the form @xmath15 , where @xmath16 and @xmath17 are a pair of neighboring lattice sites and @xmath18 is a function . a particularly interesting case is the honeycomb @xmath0 model , where function @xmath18 is @xmath19 , with @xmath20 a measure of the inverse temperature . it turns out that the @xmath0 model has a nice graph representation @xcite ; the graph consists of a number of nointeracting and nonoverlapping loops on the honeycomb lattice . however , in contrast to the potts model , an efficient cluster algorithm is still lacking for the @xmath0 model . when simulating the @xmath0 model , one has to apply a metropolis - like local algorithm , except for some special cases such as @xmath21 or @xmath22 . even worse is that local updates of loop configurations require some global connectivity information . therefore , the computing effort grows like @xmath23 as @xmath8 increases , where the exponent @xmath24 accounts for the effective critical slowing down due to the global - connectivity - checking procedure . the value of @xmath24 is close to @xmath4 in two dimensions , unless some complicated data structure is applied . apart from the above computational considerations , developing efficient algorithms for the @xmath0 loop model is also highly desirable from physical point of view . this is because , while much exact information about the critical properties of the @xmath0 loop model has been accumulated in the past decades , many open questions still exist ; for some recent publications , see e.g. , refs . @xcite . as a generalization , dilution can be introduced into the @xmath0 loop model and the potts model . for a sufficient number of diluted sites , a different universality class can arise at the so - called tricritical point . in contrast to the tricritical potts model , exact results for the tricritical @xmath0 model are scarce . therefore , a high - precision numerical study of the @xmath0 model is of great importance . in this work , we solved the long - standing problem of developing an efficient algorithm for the @xmath0 loop model for @xmath1 . we apply the newly developed cluster algorithm , an embedding swendsen - wang - type cluster method , to the honeycomb @xmath0 loop model without dilution , for which many exact predictions are available . the numerical data confirm the exact predictions by coulomb gas theory and by conformal field theory . further , the dynamical data imply that , for @xmath25 , the embedding cluster algorithm _ hardly _ suffers from critical slowing down . let us consider a plane graph @xmath26 and its dual graph @xmath27 . let @xmath28 and @xmath29 denote the total numbers of vertices and of edges of @xmath30 , respectively , and @xmath31 and @xmath32 for graph @xmath33 . the dual relation between @xmath30 and @xmath33 guarantees @xmath34 ; there is one - to - one correspondence between edges of @xmath30 and @xmath33 . on the edges of @xmath30 , bonds are placed such that they form a number of closed paths , or _ loops_. in addition , it is required that these loops can not share a common bond . the weight of such a loop configuration is given by @xmath35 , and the partition sum is @xmath36 where the sum is over all satisfied loop configurations . symbol @xmath37 denotes the sum of the lengths of all loops , and @xmath38 is the number of loops . in principle , parameters @xmath39 and @xmath2 can be any real or complex numbers . in this work , we shall only consider real numbers @xmath40 and @xmath41 , so that we have a probabilistic interpretation . first , let us further restrict our attention to the case where @xmath1 is an integer , and define a @xmath2-color ashkin - teller ( at ) model @xcite on the dual graph @xmath33 was considered . nevertheless , the ashkin - teller model with @xmath42 has also received considerable research attentions ; see e.g. , refs . @xcite ] . on every vertex @xmath43 of @xmath33 , one simultaneously places @xmath2 independent ising spins @xmath44 ; say they have color @xmath45 . on every edge @xmath46 , spins in the same color interact via couplings @xmath47 , and any two pairs of spins @xmath44 and @xmath48 that are in different colors interact via couplings @xmath49 . the latter involves four - spin interactions . the hamiltonian reads @xmath50 the partition sum is then given by @xmath51 where the sum is over all spin configurations @xmath44 with @xmath52 , as represented by a single symbol @xmath53 . apparently , in the low - temperature region @xmath54 , systems defined by eq . ( [ def_h_at ] ) have @xmath55 ground states . the @xmath56 case reduces to the isotropic version of the standard ashkin - teller model @xcite . in this work , we shall concentrate on model ( [ def_h_at ] ) in the infinite - coupling limit : @xmath57 , @xmath58 , but @xmath59 fixed at a finite value . in order to see the physical implication of this limit , let us consider the spin configurations on the ends of a given edge @xmath60 of graph @xmath33 . from eq . ( [ def_h_at ] ) , if there are @xmath61 unequal ising variables in the same color on edge @xmath60 , the boltzmann weight reads @xmath62 \nonumber \\ k=1 \ ; : & \ ; & \exp \ { j_2 ( n-2 ) + \frac{1}{2 } j_4 [ n(n-1)-2(n-1 ) ] \ } \nonumber \\ & \ ; & \vdots \nonumber \\ k\hspace{0.5 cm } \ ; : & \ ; & \exp \ { j_2 ( n-2 k ) + \frac{1}{2 } j_4 [ n(n-1)-2 k ( n-1 ) + k ( k-1 ) ] \ } \ ; . \label{def_weight}\end{aligned}\ ] ] normalizing these weights by that of case @xmath63 , one has table [ table_weights_at ] . lc configuration & normalized weight + + @xmath63 & @xmath64 + @xmath65 & @xmath66 } \equiv e^{-2j } $ ] + @xmath67 & @xmath68 + in words , for any given edge @xmath69 , only two types of spin configurations are allowed on its ends : spins in the same color @xmath70 and @xmath71 are all equal , or there is _ at most _ one pair of unequal spins in the same color . the relative weight of the latter over the former is @xmath72 . from now on , we shall refer to the at model whose relative weights are given in table [ table_weights_at ] as the infinite - coupling at model ( iat ) . it turns out that , at least for integer @xmath2 , the loop model defined by eq . ( [ def_z_loop ] ) can be exactly mapped onto the @xmath2-color iat model . for this purpose , let us consider the low - temperature expansion of the iat model , which can be performed in a similar way to the low - temperature expansion of onsager s ising model ( see e.g. , ref . @xcite ) . for each color of ising spin variable @xmath44 , we represent those unequal neighboring spins by lines of color @xmath73 on the corresponding edge of the graph @xmath30 ( recall that the iat model itself is defined on the dual graph @xmath33 ) . in words , if two adjacent spins of color @xmath73 are unequal , draw a line of color @xmath73 on the edge of @xmath30 ; otherwise , do nothing . do this for all pairs of nearest - neighbor spins . apart from effects caused by boundaries , one obtains an eulerian subgraph @xmath74 ; i.e. , for every vertex , the number of lines touching it must be _ even_. the lines are therefore joined together to form polygons ( loops ) ; these polygons have color @xmath73 . conversely , these polygons divide the plane into spin - up and spin - down domains for ising variable @xmath44 . for any such set of loops , there are just two corresponding spin configurations : they are related to each other by flipping all spins @xmath44 . do this for all colors of ising variables @xmath53 ( recall that we have used symbol @xmath53 to represent all spin variables @xmath44 for @xmath75 ) . one obtains a configuration containing @xmath2-color loops . any such set of loops corresponds to @xmath55 spin configurations . for the iat model , the zero weight of configuration @xmath76 for any @xmath77 guarantees that there is at most one pair of unequal spins in same color on each edge of @xmath78 . therefore , although loops on graph @xmath30 can intersect , they can never share a common edge . an example of the low - temperature graph for the @xmath56 iat model on the triangular lattice is shown in fig . [ fig_at_loop ] . ( 0.5,-1.4)(6,7 ) ( -1,1.73205)5(0,0)(10,0 ) ( 2,0)6(0,0)(-4,6.9282 ) ( -1,1.73205)4 ( 2,0)5 ( 0,0)(1,1.73205 ) ( -1,1.73205)5 ( 2,0)6(0,0)0.24 ( -1,1.73205)4 ( 2,0)5 ( 0,1.73205)(0,1.1547)(1,0.57735)(1.5,0.866025 ) ( -0.5,0.866025)(0,1.1547 ) ( 1.0,0.57735)(1,0 ) ( 3,1.73205)0.24 - 9090 ( 4,3.46415)0.24 - 9090 ( 6,3.46415)0.24 - 9090 ( 7,1.73205)0.24 - 9090 ( 5,5.19615)0.24 - 9090 ( -1,5.19615)0.2490270 ( 0,3.46415)0.2490270 ( 3,0.57735)(4,1.1547)(4,2.3094)(5,2.88675)(6,2.3094)(6,1.1547)(7,0.57735)(8,1.1547)(8,2.3094)(7,2.88675)(7,4.04145 ) ( 6,4.6188)(6,5.7735)(5,6.35085)(4,5.7735)(4,4.6188)(3,4.04145)(3,2.88675 ) ( 2,2.3094)(2,1.1547)(3,0.57735 ) ( -1,4.04145)(-1,2.88675)(0,2.3094 ) ( 1,2.88675)(1,4.04145)(0,4.6188)(0,5.7735 ) ( -1,6.35085)(-2,5.7735)(-2,4.6188)(-1,4.04145 ) ( -4,-0.8)0.24 ( -2.0,-0.8 ) : @xmath79 ( 1.0,-0.8)0.24 ( 1.0,-0.8)0.2490270 ( 3.0,-0.8 ) : @xmath80 ( 6.0,-0.8)0.24 ( 6.0,-0.8)0.24 - 9090 ( 8.0,-0.8 ) : @xmath81 let the energy of a ground state be zero . table [ table_weights_at ] tells that , in comparison with a ground state , a pair of unequal spins must receive an energy penalty @xmath72 , denoted as @xmath39 . the partition sum of the iat model can then be written as @xmath82 where the second sum is over all possible color arrangements for a given loop configuration . the factor @xmath55 accounts for the number of ground states . note that each loop can randomly take one of the @xmath2 colors with equal probability , and the second sum can then be easily evaluated . this leads to @xmath83 where @xmath37 and @xmath38 again represent the sum of the lengths of loops and the number of loops , respectively . apart from a trivial constant , the partition sum ( [ def_z_at_limit ] ) is equal to eq . ( [ def_z_loop ] ) . in other words , an exact one - to-@xmath55 mapping has been established between the loop model ( [ def_z_loop ] ) and the @xmath2-color iat model defined by table [ table_weights_at ] . the weight @xmath39 of the line segment in the loop configuration is related to @xmath84 by @xmath85 . note that , if the colors of loops are ignored , the loop representation of the iat model is also a low - temperature graph of the ising variable @xmath86 , the product of all @xmath2 colors of variables @xmath44 . this observation is vital in the development of cluster algorithms for the @xmath0 loop model , as shown later . given the well - known swendsen - wang or wolff cluster algorithm for onsager s ising model , an analogous cluster algorithm is readily available for the at model described by eq . ( [ def_h_at ] ) . the so - called direct or embedding algorithms can be formulated . a detailed stuty of the swendsen - wang - type algorithms , including its dynamical behavior , was carried out by salas and sokal @xcite in the context of the @xmath56 at model on the square lattice ( not in the infinite - coupling limit ) . in the present work , we shall only consider the embedding version of the swendsen - wang - type algorithm . for the at model ( [ def_h_at ] ) , for a given color of ising variable @xmath44 , the effective hamiltonian , conditioned on the other spin configurations @xmath48 for @xmath87 , can be written as @xmath88 where the effective nearest - neighbor coupling reads @xmath89 when all pairs of spins @xmath90 and @xmath91 for @xmath92 are equal , the effective interaction @xmath93 is finite . otherwise , one has the coupling @xmath94 . thus , the effective coupling is no longer translation - invariant . nevertheless , the effective coupling @xmath95 is always ferromagnetic . in this case , when the sw cluster algorithm is applied to update the @xmath44 spins , no frustration phenomena occur as in cluster simulation of the random ising model . therefore , it is plausible that the efficiency of such a cluster algorithm is not influenced too much by the inhomogeneous effective couplings . a swendsen - wang step for updating a given ising variable @xmath44 can then be written as * * step 1 : place occupied bonds . * given a spin configuration @xmath53 , for each edge @xmath69 of @xmath33 , place an occupied bond with probability @xmath96 @xmath97 * * step 2 : construct clusters . * two vertices that are connected through a chain of occupied bonds are said to be in the same cluster . on the basis of occupied bonds , the whole set of vertices @xmath98 of graph @xmath33 is decomposed into a number of clusters @xmath99 . * * step 3 : update spins @xmath44 . * for each cluster of @xmath99 , randomly assign the value of all spins on its vertices to be @xmath100 or @xmath101 with equal probability . this completes a swendsen - wang cluster step for spin variable @xmath44 . since all other spin variables @xmath48 with @xmath92 are kept fixed , this cluster method is called an embedding cluster algorithm . we shall refer to this method as the embedding sw algorithm for the iat model . a valid embedding algorithm for the @xmath2-color at model must involve the update of all colors of ising spin variables @xmath53 . this can be done sequentially or randomly . owing to the exact mapping between the loop and the iat model , for integer @xmath1 , the embedding method described in sec . ii is already a valid cluster algorithm for the @xmath0 loop model on graph @xmath30 . in particular , for @xmath22 , this cluster method just reduces to the standard swendsen - wang cluster algorithm for the ising model on graph @xmath33 . the procedure in sec . ii was already applied to simulate the @xmath56 iat model on the honeycomb , square , and triangular lattices @xcite . in comparison with the metropolis method , the critical slowing down is considerably suppressed , but still observable . at the critical point of the triangular iat model , which corresponds both to the critical baxter - wu model and to the @xmath102 loop model on the honeycomb lattice , it was found @xcite that the dynamical exponent @xmath12 is about @xmath103 . in this section , we shall show that the cluster algorithm in sec . ii can be further improved such that it can simulate the loop model with noninteger @xmath1 and that critical slowing down _ hardly _ exists . let us consider the ising spin @xmath104 . as mentioned earlier , any loop configuration of the @xmath0 loop model on @xmath30 is also a low - temperature graph of the @xmath105-spin configurations on @xmath33 . but information about the colors of loops can not be encoded in the @xmath105-spin configurations . in terms of loop configurations , the simulation of @xmath44-spin variable in the embedding sw procedure in sec . ii is equivalently to update the loops of color @xmath73 while keeping all other loops of color @xmath92 unchanged . our first task is then to rewrite the embedding sw procedure in sec . ii by using spin variable @xmath105 and loop colors as the dynamical variables , instead of using @xmath44 for @xmath106 . this leads to the following four steps : * * step 1 : place occupied bonds . * given a spin configuration @xmath105 on @xmath33 and the color information for loops on @xmath30 , for each edge @xmath69 , place an occupied bond with probability @xmath96 @xmath107 * * step 2 : construct clusters * on the basis of occupied bonds . note that the @xmath105 spins on the vertices of a given cluster can have different signs . * * step 3 : update spins @xmath105 . * for each cluster , flip all spins on its vertices with probability @xmath108 . * * step 4 : update loops . * consider the low - temperature graph of the @xmath105 spin configuration , let the colors of the existing loops remain unchanged , and assign the color of all newly generated loops to be @xmath73 . do steps 1 - 4 for each of the @xmath2 colors of loops . step 4 is redundant for the procedure in sec . ii , since the colors of loops are already encoded in the spin configuration @xmath44 with @xmath109 . for any satisfied loop configuration of the @xmath0 loop model , the color of each loop can randomly take any value of @xmath2 colors with equal probability . therefore , instead of keeping all loop colors @xmath92 unchanged , one can randomly reassign the color of each loop in step 4 , irrespective of the existing color information . on this basis , one can reformulate the embedding sw procedure as * * step 1 : construct loops and assign loop colors . * given a spin configuration @xmath105 on @xmath33 , construct the loop configuration on @xmath30 . for each loop , assign its color to be @xmath64 with probability @xmath110 and to be @xmath111 with probability @xmath112 . we say loops of color 1 and 0 to be ` active ' and ` inactive ' , respectively . * * step 2 : update the @xmath105-spin configuration . * this is done by the swendsen - wang - type procedure , in which the bond - occupation probability is @xmath113 a cluster simulation of the @xmath0 loop model can then be achieved by repeating the above two steps . note that @xmath2 is _ no longer _ required to be an integer ; it is sufficient to demand that @xmath1 , since @xmath110 should not be bigger than 1 . in summary , in comparison with the embedding sw procedure described in sec . ii , the last version of the algorithm has two prominent features . first , @xmath114 is no longer required to be an integer . second , after a completion of the update of ising spin variable , the loop colors are randomly reassigned , irrespective of the existing colors . this is possible because all the @xmath2 colors of spin variables @xmath44 ( @xmath106 ) are symmetric to each other . in other words , the symmetry among the @xmath2-color spin variables is used in the last version of the embedding cluster algorithm . if this symmetry is broken in some way , e.g. , one can let couplings @xmath47 depend on spin colors , the last version is then not applicable . but a slightly modified version of the procedure in sec . ii still works . because of usage of the symmetry among the @xmath2-color spin variables , one would expect that modes of the critical slowing down , associated with this symmetry , are significantly suppressed in the last version of the embedding sw algorithm . thus , one expects that it is more efficient than the procedure in sec . this will be numerically confirmed later . * remark : * for a plane graph @xmath30 of degrees @xmath115 , the loop assignment based on the @xmath105 spin configuration is not unique , since loops are allowed to share a common vertex . nevertheless , this should not affect the validity of the present embedding sw cluster algorithm . we conclude this section by pointing out that our embedding sw cluster algorithm for the @xmath0 loop model is of a similar spirit to the chayes - machta algorithm @xcite for the @xmath116-state potts model with real @xmath117 . the honeycomb lattice is of degree 3 , and thus the loops of the @xmath0 loop model never intersect . in this case , it has been known that the loop model can be exactly mapped on an @xmath0 spin model with partition sum @xcite @xmath118 where @xmath119 is the edge set of the honeycomb lattice . here , @xmath120 represents a @xmath2-dimensional unit vector namely @xmath121 spin model is defined such that @xmath122 and the factor @xmath39 is replaced by @xmath123 . this model has been under intensive investigation in the past two decades , and much exact information is available . for @xmath124 , the model undergoes a second - order phase transition ; for @xmath56 , the transition is of infinite order , the so - called kasteleyn - thouless ( kt ) transition ; for @xmath125 , it displays a lattice - gas - type transition . the renormalization exponents for critical @xmath0 loop systems for @xmath126 is a function of @xmath2 . for @xmath125 , it is expected that the transition is in the same universality class of baxter s hard - hexagon lattice gas model on the triangular lattice ( an exact mapping between the @xmath127 loop model and the hard - hexagon lattice gas can be established ) . the critical frontier for @xmath126 is exactly located at @xcite @xmath128 it was observed @xcite that a _ critical _ @xmath0 model corresponds with a _ @xmath129-state potts model . from exact calculations , conformal field theory @xcite , and coulomb gas theory @xcite , exact values of renormalization exponents can be obtained for the tricritical potts model as @xmath130 where the coulomb - gas couplings @xmath131 is related to @xmath116 as @xmath132 the symbols @xmath133 , @xmath134 , and @xmath135 represent exponents of the leading , subleading , and next - subleading thermal scaling fields , respectively , and @xmath136 and @xmath137 are for the magnetic scaling fields . therefore , exact values of critical exponents for the @xmath0 model described by eq . ( [ def_on ] ) as a function of @xmath2 are also known . we applied the embedding sw cluster algorithm in sec . iii to simulate the @xmath0 loop model on the honeycomb lattice ; the ising variable @xmath105 is defined on the vertices of the dual lattice i.e . , the triangular lattice . the sampled quantities can be classified into three types ; they are associated with loop configurations , ising spin variables , and distributions of clusters @xmath99 formed in the sw step . the first type includes the total lengths of loops and of the number of loops , as normalized by the volume of the system @xmath138 and denoted by @xmath139 and @xmath140 , respectively . here , the volume @xmath138 is the total number of vertices on the corresponding triangular lattice . the second - moment fluctuations of @xmath139 are also sampled as @xmath141 , where symbol @xmath142 represents the statistical average . taking into account the mapping between the loop model and the iat model , we also measured both the total lengths and the number of ` active ' loops , denoted by @xmath143 and @xmath144 ; the fluctuations of @xmath143 , denoted as @xmath145 , were also measured . the subscript @xmath146 means ` active ' . the second type of quantities concern the ising variables associated with loop configurations . a natural quantity to sample is the magnetic susceptibility @xmath147 , where @xmath73 is the magnetization density for spin variable @xmath105 . after each embedding sw step , we also reconstructed ising spins @xmath148 from these active loops . the associated susceptibility @xmath149 was then sampled . in addition , we consider the size distribution of domains enclosed by those active loops . these domains can be obtained by placing occupied bonds between all equal nearest - neighbor @xmath148 ising spins . they are named the ising clusters in the standard ising model . we measured the second moment of these ising clusters @xmath150 where @xmath151 is the number of @xmath148 spins in the @xmath16th domain . the last type of quantities contains information for the embedding sw simulations . they include the number of occupied bonds per spin @xmath152 , the number of clusters per spin @xmath153 , and the second and fourth moment of cluster sizes @xmath154 where @xmath155 or @xmath156 , and @xmath157 is the size of the @xmath16th cluster formed in the embedding sw simulations . in the conventional sw simulation of the ising model ( or more generally of the potts model ) , quantities @xmath158 and @xmath159 are exactly related to the second and the fourth moment of the magnetization density . the bond - number density is also exactly related to the energy density in the ising model . on the basis of the above sampled quantities , we sampled ratios @xmath160 at criticality , these amplitudes ratios are dimensionless and universal . they are known to be very useful in locating critical points in monte carlo studies of statistical systems undergoing phase transitions . for the sw simulation of the ising model , quantity @xmath161 reduces to @xmath162 defined on the basis of magnetic susceptibility . for a test of the algorithm , we performed simulations for @xmath163 and @xmath164 . the system sizes took @xmath165 values in range @xmath166 , where @xmath8 is the linear system size of the triangular lattice . periodic boundary conditions were applied . measurements were taken after every swendsen - wang step . for @xmath163 , eq . ( [ critical_point ] ) yields the critical point at @xmath167 . simulations were performed in range @xmath168 . about @xmath169 samples were taken for each size @xmath8 . parts of the @xmath162 data are shown in fig . [ fig_qm1_5 ] . the intersections of the @xmath162 data for different system sizes rapidly converge to the expected value @xmath170 . this implies the validity of the embedding cluster algorithm for the @xmath0 loop model with noninteger @xmath2 . 10.0 cm according the least - squares criterion , we fitted the @xmath162 data by @xmath171 where @xmath172 is an integer . the terms with amplitudes @xmath173 and @xmath174 describe finite - size corrections . the term with exponent @xmath175 is supposed to arise from the leading irrelevant scaling field . thus , the value of @xmath175 is given by @xmath135 in eq . ( [ critical_exponent ] ) . for @xmath163 , eq . ( [ critical_exponent ] ) yields @xmath176 . other finite - size corrections can arise from various sources , such as from the subleading irrelevant scaling field , from the analytical part of the free energy @xmath177 , and from the second derivative of @xmath177 with respect to the leading irrelevant field . it is not _ a prior _ clear how many or which such correction terms can be observed in the numerical data . thus , one should make various fits by taking into account all possible sources of finite - size corrections . final estimates of interested quantities are then obtained by comparing the results of various fits . we let exponent @xmath178 and @xmath175 to be fitted by the numerical data , and fixed @xmath179 at @xmath180 for simplicity . all the @xmath162 data in range @xmath166 can be successfully described by eq . ( [ fit_q1 ] ) . the results are @xmath181 , @xmath182 , @xmath183 , and @xmath184 . the value of @xmath178 is consistent with the exact value @xmath185 in eq . ( [ critical_exponent ] ) , and that of @xmath175 agrees with @xmath186 . the estimated critical point is also in excellent agreement with the exact prediction @xmath187 . 10.0 cm next , we plotted parts of the @xmath161 data in fig . [ fig_qw1_5 ] . the intersections of the @xmath161 data reflect the percolation threshold of clusters formed in the embedding sw procedure . in cluster simulations of statistical systems , the efficiency can be reflected by the average size of the formed clusters . the efficiency will be limited if the average size is either too small or too big : little change of configurations is made for the former , and large amount of effort has to be carried out for the latter ( probably it also accompanies by little change ) . ideally , the percolation threshold of clusters formed in monte carlo simulations should coincide or be very close to the phase transition of statistical systems . this is indeed the case in the present embedding sw algorithm for the @xmath0 loop model , as reflected by fig . [ fig_qw1_5 ] . the fits of the @xmath188 data by eq . ( [ fit_q1 ] ) yields @xmath189 , @xmath190 , and @xmath191 . the estimated percolation threshold is consistent with the thermal critical point @xmath192 . 10.0 cm we also simulated the @xmath164 loop model . equation ( [ critical_point ] ) predicts the critical point at @xmath193 . simulations were performed in range @xmath194 , and system sizes took @xmath165 values in range @xmath195 . parts of the @xmath162 data are shown in fig . [ fig_q1_9 ] . as expected , the intersections of the @xmath162 data for different sizes converge rapidly to @xmath196 . the fits of the @xmath162 data by eq . ( [ fit_q1 ] ) yields @xmath197 , @xmath198 , @xmath199 , and @xmath200 . the value of @xmath178 is more or less consistent with that of @xmath201 . but that of @xmath175 does not agree with that of @xmath202 . in this case , we expect that the dominant finite - size corrections are not described by exponent @xmath135 . a plausible scenario for the dominant finite - size corrections might be the following . as mentioned earlier , the mapping between the @xmath0 model on graph @xmath30 and the iat model on the dual graph @xmath33 is only exact up to boundary effects . even though these boundary effects are expected to vanish in the thermodynamic limit @xmath203 , they can not be neglected for finite @xmath8 . according to a simple argument , these boundary effects vanish as a function of @xmath204 . indeed , exponent @xmath101 is consistent with the numerical values of @xmath175 for @xmath163 and @xmath164 . for @xmath205 , and @xmath4 , we performed extensive simulations right at the exactly predicted critical points given by eq . ( [ critical_point ] ) . periodic boundary conditions were applied , and the system sizes took @xmath206 values in range @xmath207 . samples were taken after every embedding sw step . the number of samples is @xmath208 for @xmath209 , and @xmath210 for @xmath211 ( wenan and henk : monte carlo data are not complete yet ) . for later convenience , we list in table . [ table_exponent_on ] the exact values of critical exponents for @xmath212,and @xmath4 , as predicted by eq . ( [ critical_exponent ] ) ; also included are the exact values of critical points given by eq . ( [ critical_point ] ) . r|lllll|r @xmath2 & @xmath133 & @xmath134 & @xmath135 & @xmath136 & @xmath137 & @xmath213 + + @xmath64 & @xmath214 & @xmath64 & @xmath215 & @xmath216 & @xmath217 & @xmath218 + @xmath219 & @xmath220 & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 + @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 + @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 + @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath214 & @xmath244 & @xmath245 + the critical @xmath56 loop model on the honeycomb lattice is a very special case . it can be mapped both onto the standard @xmath56 iat model on the triangular lattice and the baxter - wu model with three - spin interactions on the triangular lattice . the latter has been exactly solved by various methods , and is believed to be in the same universality class as the @xmath156-state potts model . further , it is known that the @xmath56 loop model undergoes a kt transition , which is an infinite - order phase transition . * energy - like quantities . * the energy - like quantities include the density of the length of loops @xmath139 and of the loop number @xmath140 . the exact information for the critical properties of the @xmath102 model tells one that , at criticality , the finite - size behavior of @xmath139 and @xmath140 is governed by the subleading thermal exponent @xmath134 in eq . ( [ critical_exponent ] ) . from the analysis of monte carlo data , as shown later , it turns out that the finite - size scaling behavior of bond - occupation density @xmath152 , a quantity associated with the embedding sw step , is also described by @xmath134 . instead of being listed in tables , the @xmath139 and @xmath140 data are shown in fig . [ fig_e2_00a ] ; the vertical axis shows the data for @xmath246 , @xmath247 , and @xmath248 , where constants @xmath249 , @xmath250 , and @xmath251 were obtained from the numerical fits . 10.0 cm we made a least - squares fit of the monte carlo data for energy - density - like quantities to the general ansatz @xmath252 the symbols @xmath119 and @xmath253 should be replaced by specific quantities in the fits . for instance , depending on which quantity is going to be analyzed , @xmath119 can be @xmath139 , @xmath140 , and @xmath152 , and @xmath253 can be @xmath249 , @xmath250 , and @xmath251 . in eq . ( [ fit_energy ] ) , the term with @xmath254 describes boundary effects arising from the mapping between the ising - spin configuration and the loop representation . the term with @xmath173 arises from the least irrelevant scaling field . however , the value of @xmath175 is not very clear . according to conformal field theory , if an operator with exponent @xmath255 exists , a sequence of exponents @xmath256 , @xmath257 , @xmath258 , can in principle also exist . therefore , @xmath175 can be @xmath135 or @xmath259 etc . in the analysis of our monte carlo data , we have tried to take into account all possible sources of finite - size corrections . the fits yield @xmath260 and @xmath261 for the @xmath262 data , @xmath263 and @xmath264 for the @xmath265 data , and @xmath266 and @xmath267 for the @xmath268 data . the values of @xmath178 are all in good agreement with the prediction @xmath269 . within the sampled quantities , the density of the lengths and of the number of active loops , @xmath143 and @xmath270 , are also energy - like quantities . since the color of each loop can randomly take one of @xmath2 colors , the statistical mean value of @xmath271 is just @xmath272 for any size @xmath8 ; the same applies to quantity @xmath140 . therefore , we do not need to analyze them . * specific - heat - like quantities . * quantities @xmath273 and @xmath145 account for the fluctuations of @xmath139 and @xmath143 , respectively . the @xmath273 and @xmath145 data are respectively shown in figs . [ fig_c2_00a ] and [ fig_ci2_00a ] . these suggest that the finite - size behavior of @xmath273 is governed by exponent @xmath274 , while that of @xmath145 is governed by @xmath275 . 10.0 cm 10.0 cm an observation is that the behavior of the magnetic susceptibility @xmath276 , for the @xmath277 for @xmath106 , is also described by exponent @xmath275 . actually , for the present case @xmath56 , because of the symmetry between two ising variables @xmath148 and @xmath278 for the iat model , it can be shown that , in the thermodynamic limit @xmath203 , @xmath279 and @xmath280 are equivalent to each other ( wenan henk : this statement has not been carefully checked ) . the data for specific - heat - like quantities were fitted by @xmath281 where symbols @xmath99 and @xmath282 should be replaced by specific quantities in the fits . the fits of the @xmath283 data yield @xmath284 and @xmath285 , and those of @xmath145 give @xmath286 and @xmath287 . * susceptibility - like quantities . * quantity @xmath276 for the @xmath105 ising variable has been discussed earlier . the fits of the @xmath288 data also yield @xmath289 . other susceptibility - like quantities include @xmath290 for spin @xmath148 , @xmath291 for the ising clusters of spin variable @xmath148 , and @xmath158 for the clusters formed in the embedding sw step . the data of these quantities are shown in figs . [ fig_mi2_00a ] and [ fig_si2_00a ] . 10.0 cm 10.0 cm the data for susceptibility - like quantities were fitted by @xmath292 where symbols @xmath293 and @xmath294 should be replaced by specific quantities in the fits . exponent @xmath295 is left to be fitted by numerical data . the fits of the @xmath296 and @xmath297 data by eq . ( [ fit_chi ] ) yield @xmath298 and @xmath299 , respectively . both results are in good agreement with the exact value of @xmath300 , as given in table [ table_exponent_on ] . this implies that the percolation threshold of the clusters formed in the embedding sw step coincides with the thermal critical point @xmath301 . the fit of @xmath302 yield @xmath303 . there is no prediction of the exact value for this exponent . * binder ratios . * in the monte carlo simulations , we also samped dimensionless amplitude ratios @xmath162 and @xmath304 , defined by eq . ( [ def_ratio ] ) . the @xmath162 and @xmath304 data are shown in fig . [ fig_q2_00a ] . 10.0 cm the data for amplitude ratios were fitted by @xmath305 where exponent @xmath175 is left to be determined by numerical data . the fits of the @xmath306 data yield @xmath307 and @xmath308 , and those of @xmath304 give @xmath309 and @xmath310 . both estimates of @xmath175 agree with integer @xmath101 , which can be explained as @xmath259 . the analysis of the monte carlo data for the honeycomb @xmath0 model with @xmath311 , and @xmath233 follows an analogous procedure as that for @xmath56 . thus , the details of the fitting procedure are skipped here . the fitting results are listed in tables [ table_fit_energy ] and [ table_fit_chi ] . rr|lllll @xmath2 & & @xmath139 & @xmath140 & @xmath312 & @xmath313 & @xmath314 + + @xmath219 & @xmath315 & @xmath316 & @xmath317 & @xmath318 & @xmath319 & @xmath320 + & @xmath315(exact ) & @xmath321 & @xmath322 & @xmath322 & @xmath322 & + & @xmath323 & @xmath324 & @xmath325 & @xmath326 & @xmath327 & @xmath328 + @xmath329 & @xmath315 & @xmath330 & @xmath331 & @xmath332 & @xmath331 & @xmath333 + & @xmath315(exact ) & @xmath334 & @xmath335 & @xmath335 & @xmath335 & + & @xmath323 & @xmath336 & @xmath337 & @xmath338 & @xmath339 & @xmath340 + @xmath233 & @xmath315 & @xmath341 & @xmath342 & @xmath343 & @xmath344 & @xmath345 + & @xmath315(exact ) & @xmath346 & @xmath347 & @xmath347 & @xmath347 & + & @xmath323 & @xmath348 & @xmath349 & @xmath350 & @xmath351 & @xmath352 + @xmath240 & @xmath315 & @xmath353 & @xmath354 & @xmath355 & @xmath356 & @xmath357 + & @xmath315(exact ) & @xmath358 & @xmath359 & @xmath360 & @xmath361 & @xmath362 + & @xmath323 & @xmath363 & @xmath364 & @xmath365 & @xmath366 & @xmath367 + .[table_fit_chi ] fitting results for susceptibility - associated quantities . symbol @xmath368 represents the universal values of those amplitude ratios at criticality . we also give the predicted values for exponent @xmath255 to be fitted . the exponent @xmath255 for amplitude ratios is for the dominant finite - size corrections . the question mark means that we are not very sure about the prediction of @xmath175 . [ cols=">,>,<,<,<,<,<,<",options="header " , ] for @xmath56 , quantities associated with those active loops or the spin variable @xmath148 have clear physical meaning , since they appear naturally in the @xmath56 iat model described by eq . ( [ def_h_at ] ) . in this case , the scaling behavior of specific - heat - like quantity @xmath145 and susceptibility @xmath149 can be predicted from the exact results for the baxter - wu model or for the triangular iat model . the numerical results fit well with the predictions . nevertheless , no exact results seem to exist for the finite - size behavior of the geometric quantity @xmath302 , the second - moment of the ising clusters for spin @xmath148 . for noninteger @xmath2 , however , the symmetry between the active and the inactive loops breaks . to our knowledge , there are thus far no prediction for exponents governing the scaling behavior of quantities @xmath145 and @xmath149 , as listed in tables [ table_fit_energy ] and [ table_fit_chi ] . on the other hand , the following scenario seems plausible . the estimated exponent @xmath255 for @xmath145 in table [ table_fit_energy ] is actually the mixed effect of @xmath133 and @xmath134 , while that for @xmath149 is due to the effective mixture of @xmath136 and @xmath137 . we did try to make fits for the @xmath145 and @xmath149 by taking into account such a mixture effect , but the results seem worse than those by eqs . ( [ fit_specific_heat ] ) and ( [ fit_chi ] ) . an observation is that , independent of the @xmath2 value , the exponents for the scaling behavior of @xmath149 and @xmath158 are alway equivalent . this strongly suggests that the percolation threshold of clusters formed in the embedding sw step coincides with the thermal critical point for any @xmath1 . a rigorous proof is lacking . during the monte carlo simulations , the value of every sampled quantity was saved on hard disk after every sweep . statistical analyses were then performed on these data . both static and dynamic information of sampled quantities can then be obtained . let symbol @xmath177 be a given quantity , the unnormalized autocorrelation function is then defined as @xmath369 the normalized autocorrelation function is @xmath370 typically , @xmath371 decays exponentially ( @xmath372 ) for large @xmath373 , and the exponential autocorrelation time is defined as @xmath374 in addition , we define the integrated autocorrelation time @xmath375 here , the factor @xmath108 is purely a matter of convention ; it follows the definition in ref . the integrated autocorrelation time controls the statistical error in monte carlo measurements of @xmath376 . more precisely , the sample mean has variance @xmath377 where @xmath378 is the total length of monte carlo simulations . parts of the numerical data for the normalized autocorrelation function are shown in fig [ fig_pchi2_00a ] for susceptibility @xmath276 , in fig . [ fig_ps1a2_00a ] the density of loop lengths @xmath139 , and in fig . [ fig_pmi2_00a ] for susceptibility @xmath149 for spin variable @xmath148 . the good collapse of the @xmath379 data strongly suggests that the critical slowing down is _ absent _ in the embedding sw simulation of the @xmath102 loop model on the honeycomb lattice . for quantity @xmath149 , after a single monte carlo step , the autocorrelation function drops to a value smaller @xmath13 . this means that two subsequent samples are almost effectively independent . except for very small @xmath373 , the data lines in figs.[fig_pchi2_00a ] , [ fig_ps1a2_00a ] , and [ fig_pmi2_00a ] are rather straight ; the scattering phenomena at the right - hand side are due to statistical noise . therefore , the autocorrelation function is almost a purely exponential function of time : @xmath380 . 10.0 cm 10.0 cm 10.0 cm the integrated autocorrelation time @xmath381 defined in eq . ( [ def_tau_int ] ) was also measured for all sampled observables . apart from the constant factor @xmath108 , quantity @xmath381 is just the area underline the @xmath379 data line as a function of time @xmath373 . parts of the @xmath381 data are shown in fig . [ fig_tau2_00a ] . 10.0 cm we fitted the @xmath381 data for the sampled quantities by @xmath382 where the dynamic exponent @xmath12 is left free to be determined by the numerical data . we assume that correction exponent @xmath101 , which appears in static quantities , also exists in the dynamical data . nevertheless , it turns out that , for @xmath383 , all the @xmath381 data for quantities in fig . [ fig_tau2_00a ] can be fitted by eq . ( [ fit_tau ] ) with @xmath384 . the results are shown in table [ table_fit_tau ] . these show that , for all quantities shown in table [ table_fit_tau ] , the integrated autocorrelation time @xmath381 converge rapidly to constants as @xmath203 ; the values of these constants are all rather small . quantity @xmath276 has the largest autocorrelation time @xmath381 , and @xmath139 has the next largest @xmath381 . for @xmath311 and @xmath233 , the general behavior of @xmath379 and @xmath385 for quantities listed in table [ table_fit_tau ] is similar as that for @xmath56 . namely , the autocorrelation function @xmath371 is almost a purely exponential function of time @xmath373 . for quantity @xmath149 , after a single monte carlo step , the value of @xmath379 drops to a significantly small value , which is about @xmath386 for @xmath387 . the quantities that have the largest two values of correlation time @xmath381 are @xmath276 and @xmath139 ; @xmath139 has the largest value for @xmath387 and @xmath276 for the others . therefore , we only show the @xmath139 data for @xmath387 in fig . [ fig_ps1a1_25 ] and the @xmath276 data for @xmath388 and @xmath233 in figs . [ fig_pma1_50 ] and [ fig_pma1_75 ] , respectively . 10.0 cm 10.0 cm 10.0 cm the @xmath381 data for the quantities in table [ table_fit_tau ] are also shown in figs . [ fig_tau1_25 ] , [ fig_tau1_50 ] and [ fig_tau1_75 ] for @xmath311 , and @xmath233 , respectively . 10.0 cm 10.0 cm 10.0 cm from fig . [ fig_tau1_25 ] for @xmath387 , it is not completely clear whether the critical slowing down is absent or not for quantities @xmath139 and @xmath276 . nevertheless , the approximately straight line of quantity @xmath139 suggests that , even the critical slowing down exists , the integrated correlation time @xmath385 may diverge in a logarithmic format or in a power law of @xmath8 with a very small exponent @xmath12 . we fitted the @xmath381 data by eq . ( [ fit_tau ] ) . for the @xmath389 data for @xmath387 , it turns out they can still be described by eq . ( [ fit_tau ] ) when a correction term with @xmath390 is also included . for the other quantities and for the other values @xmath2 , the @xmath381 data for @xmath383 all can be well described by eq . ( [ fit_tau ] ) with @xmath391 . the results are shown in table [ table_fit_tau ] ; the value of @xmath174 is @xmath392 for @xmath389 with @xmath387 . rr|llll quantity & & @xmath387 & @xmath388 & @xmath393 & @xmath394 + + @xmath139 & @xmath395 & @xmath396 & @xmath397 & @xmath398 & @xmath399 + & @xmath400 & @xmath401 & @xmath402 & @xmath403 & @xmath404 + @xmath405 & @xmath395 & @xmath406 & @xmath407 & @xmath408 & @xmath409 + & @xmath400 & @xmath402 & @xmath410 & @xmath411 & @xmath412 + @xmath149 & @xmath395 & @xmath413 & @xmath414 & @xmath415 & @xmath416 + & @xmath400 & @xmath417 & @xmath418 & @xmath419 & @xmath420 + @xmath312 & @xmath395 & @xmath421 & @xmath422 & @xmath423 & @xmath424 + & @xmath400 & @xmath425 & @xmath426 & @xmath427 & @xmath428 + @xmath429 & @xmath395 & @xmath430 & @xmath431 & @xmath432 & @xmath433 + & @xmath400 & @xmath434 & @xmath435 & @xmath436 & @xmath437 + taking into account the approximate linearity of the @xmath389 data for @xmath387 in fig . [ fig_tau1_25 ] , we also fitted the @xmath389 data by @xmath438 indeed , all the data can be well described by eq . ( [ fit_tau_25 ] ) ; we obtain @xmath439 , @xmath440 , and @xmath441 . by making use of the equivalence of the loop configurations of the @xmath0 loop model and the low - temperature graph of the ashkin - teller model in the infinite - coupling limit ( iat ) , we formulated an embedding swendsen - wang - type algorithm for the @xmath0 loop model with real value @xmath1 . for @xmath22 , this algorithm reduces to the conventional swendsen - wang method for the ising model . with some modifications , an embedding wolff - type method ( single - cluster version ) is readily available . we then applied our cluster algorithm to the @xmath0 loop model on the honeycomb lattice . the numerical data reveal the finite - size scaling behavior of several quantities . the associated exponents are confirmed to be those exact values predicted by the coulomb gas theory and by conformal field theory . the dynamical data strongly imply that the embedding cluster algorithm suffers little from critical slowing down . this is somewhat impressive in the sense that the dynamic exponent @xmath12 of the swendsen - wang type algorithm must satisfy the li - sokal bound @xcite : @xmath442 , unless it can be proved that the amplitude of terms with exponent @xmath443 vanishes . the value of @xmath444 can be easily calculated from eq . ( [ critical_exponent ] ) , which yields @xmath445 , @xmath446 , @xmath447 , and @xmath64 for for @xmath387 , @xmath226 , @xmath233 , and @xmath4 , respectively . the absence of a dynamic exponent @xmath448 in the embedding sw simulations of the @xmath0 loop model implies that the amplitude for terms with exponent @xmath449 indeed vanishes . we argue that this is because our embedding sw cluster algorithm has made use of the symmetry among the @xmath2 colors of ising spins . similar scenarios exist elsewhere . for instance , it can be proved that , in metropolis simulations of the ising model , the dynamic exponent @xmath12 must satisfy @xmath450 . however , the dynamic exponent @xmath12 of the swendsen - wang algorithm is much smaller than @xmath451 . this is because the symmetry between the up- and down - pointing ising spins is fully taken into account in the swendsen - wang algorithm . to have a better understanding of our above argument , let us consider another version of swendsen - wang - type cluster method for the iat or the @xmath0 model with integer @xmath2 , as described in sec . this algorithm directly simulates each color of the spin variables @xmath44 in the iat model for @xmath452 . in the language of loop configurations , the algorithm only updates loops in the same color ; it does not interchange or reassign colors of loops . in other words , the symmetry among loops in different colors is not taken into account . such a cluster algorithm has been applied to the critical @xmath56 iat model on the triangular lattice , which is equivalent to the critical @xmath102 model on the honeycomb lattice or the baxter - wu model . it was found that the dynamic exponent @xmath12 for the integrated correlation time is @xmath453 ; indeed , @xmath12 satisfies the li - sokal bound : @xmath454 . instead of the individual spin variables @xmath44 , the final version of the embedding sw algorithm simulates the product variable @xmath455 . after each update of spin variable @xmath105 , the colors of loops are reassigned randomly , irrespective of the existing colors of loops . in this sense , it is natural that dynamic exponent @xmath456 vanishes . the @xmath22 loop model is just the ising model . from universality , one has @xmath458 for the @xmath459 model on any planar graph . on the honeycomb lattice , eqs . ( [ critical_exponent ] ) and ( [ relation_gq ] ) tell that @xmath460 is a monotonically decreasing function of @xmath5 . therefore , one has @xmath461 for @xmath25 . in this sense , our cluster algorithm still satisfies the li - sokal bound . since the value of @xmath460 decreases as a function of @xmath2 , it is also expected that , as @xmath2 increases , the value of @xmath381 decrease . this is consistent with @xmath395 in table [ table_fit_tau ] . it is clear from sec . iii that the embedding cluster algorithm described in this work can not be applied to the @xmath462 case . nevertheless , since all the loop configurations are the low - temperature graphs of the @xmath105 spin variable , and vice versa . it seems that , for @xmath463 , a reweighting modification of the swendsen - wang simulation of the ising model can still be useful . such a procedure can be described as * * step 1 . * for a spin configuration @xmath464 , in which the number of loops is @xmath38 , generate a new spin configuration @xmath465 by using the swendsen - wang algorithm . * * step 2 . * derive the loop information for the new configuration @xmath465 , and calculate the loop number @xmath466 . accept the new configuration with probability @xmath467 , namely , set @xmath468 . otherwise , keep the old spin configuration @xmath105 . repeating of these two steps forms a valid ` cluster ' algorithm for @xmath469 . for the @xmath0 loop model on the honeycomb lattice , it turns out that , for @xmath470 and small system sizes @xmath471 , this algorithm works pretty well . 100 k. binder and d. stauffer , _ applications of the monte carlo method in statistical physics _ ( springer - verlad , berlin , 1984 ) . swendsen and j .- s . wang , phys . lett . * 58 * , 86 ( 1987 ) . u. wolff , phys . lett . * 62 * , 361 ( 1989 ) . g. ossola and a.d . sokal , nucl . b * 691 * , 259 ( 2004 ) . stanley , phys . * 20 * , 589 ( 1968 ) . e. domany , d. mukamel , b. nienhuis , and a. schwimmer , nucl . b * 190 * , 279 ( 1981 ) . b. nienhuis , phys . * 49 * , 1062 ( 1982 ) . b. nienhuis , j. stat . phys . * 34 * , 731 ( 1984 ) . l. chayes , l.p . pryadko , and k. shtengel , nucl . b * 570 * , 590 ( 2000 ) . w. guo , h.w.j . blte , and f.y . wu , phys . lett . * 85 * , 3874 ( 2000 ) . jacobsen , n. read , and h. saleur , phys . lett . * 90 * , 090601 ( 2003 ) . w. janke and a.m.j . schakel , phys . lett . * 95 * , 135792 ( 2005 ) . w. guo , b. nienhuis , and h.w.j . blte , phys . lett . * 96 * , 045704 ( 2006 ) . j. ashkin and e. teller , phys . rev . * 64 * , 178 ( 1943 ) . c. fan , phys . a * 39 * , 136 ( 1972 ) . grest and m. widom , phys . b * 24 * , 6508 ( 1981 ) . e. fradkin , phys . lett . * 53 * , 1967 ( 1984 ) . baxter , _ exactly solved models in statistical mechanics _ ( academic press , london new york , 1982 ) . g. musial , l. debski , and g. kamieniarz , phys . rev . b * 66 * , 012407 ( 2002 ) . j. salas , and a.d . sokal , j. stat . phys . * 87 * , 1 ( 1997 ) . y. deng , j. salas , t. garoni , a. sportielle , and a.d . sokal , unpublished . l. chayes and j. machta , physica a * 254 * , 477 ( 1998 ) . cardy , _ phase transitions and critical phenomena _ , edited by c. domb and j.l . ( academic press , london , 1987 ) , vol . 11 , p. 55 , and references therein . sokal , _ monte carlo methods in statistical mechanics : foundations and new algorithms _ , lecture notes at the cargse summer school , 1996 . x .- li and a.d . sokal , phys . lett . * 63 * , 827 ( 1989 ) . |
we report on an attempt to accurately wavelength calibrate four nights of data taken with the keck hires spectrograph on qso phl957 , for the purpose of determining whether the fine structure constant was different in the past . using new software and techniques , we measured the redshifts of various ni ii , fe ii , si ii , etc . lines in a damped ly system at . roughly half the data were taken through the keck iodine cell which contains thousands of well calibrated iodine lines . using these iodine exposures to calibrate the normal th - ar keck data pipeline output we found absolute wavelength offsets of 500 m / s to 1000 m / s with drifts of more than 500 m / s over a single night , and drifts of nearly 2000 m / s over several nights . these offsets correspond to an absolute redshift of uncertainty of about ( ) , with daily drifts of around ( ) , and multiday drifts of nearly ( ) . the causes of the wavelength offsets are not known , but since claimed shifts in the fine structure constant would result in velocity shifts of less than 100 m / s , this level of systematic uncertainty may make it difficult to use keck hires data to constrain the change in the fine structure constant . using our calibrated data , we applied both our own fitting software and standard fitting software to measure , but discovered that we could obtain results ranging from significant detection of either sign , to strong null limits , depending upon which sets of lines and which fitting method was used . we thus speculate that the discrepant results on reported in the literature may be due to random fluctuations coming from under - estimated systematic errors in wavelength calibration and fitting procedure . .2truein | 0904.4725 | the fine structure constant today , @xmath9 , is usually thought of as a fundamental , unchanging , constant of nature , but recently both experimental and theoretical papers have challenged this assumption , ( e.g. , see the review by garcia - berro , isern , and kubyshin ( 2007 ) ) . motivated especially by the possible experimental detection ( see below ) of a change in @xmath0 , we applied for and took four nights of keck data , half of it through the keck iodine cell , on qso phl957 . our goal was to get an extremely well - calibrated , high signal / noise absorption spectrum on a distant damped ly@xmath0 ( dla ) system ( @xmath1 ) in order to measure the value of the fine structure constant more than 10 billion years ago and compare it to the value today . the basic method to determine @xmath10 , is to measure differences in redshifts between different atomic transitions of elements in the same physical system , and use the fact that for small changes in @xmath0 the energy level of a given atomic transition can be approximated as @xmath11 where @xmath12 is the frequency of the transition on earth , @xmath13 is the frequency in the high redshift cloud , and the @xmath14-values measure the dependence of @xmath13 on @xmath0 and have been calculated for many common transitions . ( see for example , murphy et al . 2001a , dzuba , et al . 2002 , and porsev , et al . 2007 for more detailed discussion and table [ tab : lineinfo ] for the values of @xmath14 for various transitions . ) the values of @xmath14 depend upon the electron orbital configurations of the initial and final quantum states , and therefore different transitions have different values of @xmath14 . if all transitions had the same @xmath14 , then all wavelength shifts would be the same and one could absorb any change in @xmath0 into the determination of the redshift of the physical system . however , since different transitions have different values of @xmath14 , the relative transition wavelengths will differ from what they are in the lab if @xmath15 . for example , murphy , et al . ( 2001a ; 2001b ; 2003 ; 2004 ) , used a many - multiplet method on keck hires data of 143 absorption systems in the redshift range @xmath16 to find a significant reduction of @xmath0 in the past , @xmath17 , while chand , et al . ( 2004 ) and srianand et al ( 2004 ) , used the same method for a subset of transitions on vlt / uves data on 23 absorbers to find @xmath18 . this latter analysis was criticized by murphy , et al . ( 2008 ) who reanalyzed the chand et al . data to get @xmath19 , consistent with their previous result . in the meantime other groups returned results , for example , levshakov , et al . ( 2006 ) used the vlt / uves spectrograph to study fe ii lines in one system at @xmath20 and found a null result , @xmath21 , and in another system at @xmath22 to find @xmath23 ( levshakov , et al . murphy , et al . ( 2006 ) have also criticized these results , claiming that the data do not allow limits as strong as those reported . subsequently , the levshakov , et al ( 2006 ) results were weakened to @xmath24 for the @xmath20 system and @xmath25 for the @xmath22 system ( molaro , et al . 2008 ) , still a null result inconsistent with the detections . given the above inconsistent results , we were particularly interested in the fe ii @xmath261608 and fe ii @xmath261611 transitions since these have @xmath14 values that are both large and more importantly of opposite sign . thus , if @xmath0 was different in the past , the relative positions of these two lines should be significantly shifted from their laboratory values . for our dla at @xmath1 , a relative shift between fe ii @xmath261608 and fe ii @xmath261611 of about @xmath27 m / s is expected if the murphy et al . value @xmath28 is correct . ( fe ii @xmath261608 shifts by @xmath29 m / s , while fe ii @xmath261611 shifts by @xmath30 m / s in the rest frame . ) thus our goal was to centroid these lines to better than 50 m / s , so as to determine @xmath8 in a single ion in a single absorption system . our method , which is close to that used by levshakov , et al 2006 , contrasts with that of murphy et al . ( 2001a ; 2001b ; 2003 ; 2004 ) and chand et al . ( 2004 ) where the signal / noise was not high enough to detect the @xmath8 signal in any single pair of lines ; they did a statistical averaging over many transitions in many absorption systems , and thus might be subject to systematic errors in selection , calibration , or averaging procedures . since we expected to have superbly well - calibrated spectra , and these two fe ii lines appear in the same echelle order we hoped we could give convincing evidence for or against a change in the fine structure constant . for our work , in addition to the fe ii lines , there are also several ni ii , si ii , al ii , and al iii lines that fall in the wavelength range covered by the iodine cell and that we can use . in what follows , besides the fe ii @xmath261608/@xmath261611 pair , various other sets of lines are used . potentially we could fit all 16 lines that have calculated @xmath14 values and that appear in our spectra : fe ii @xmath261608/@xmath261611/@xmath262344 , ni ii @xmath261709/@xmath261741/@xmath261751 , si ii @xmath261526/@xmath261808 , al iii @xmath261854/@xmath261862 , al ii @xmath261670 , cr ii @xmath26 2062/@xmath262056/@xmath262066 , and zn ii @xmath262026/@xmath262062 . if we restrict ourselves to lines that occur at wavelengths for which we have iodine spectra we would use only the 9 lines : fe ii @xmath261608/@xmath26 1611 , niii @xmath261709/@xmath261741/@xmath261751 , si ii @xmath261808 , al iii @xmath261854/@xmath261862 , and al ii @xmath261670 . if we worry that saturated lines ( those with minimum flux less than 10% ) may not be accurately fit we can restrict ourselves to the 7 lines that meet the above criteria and are unsaturated : feii @xmath261611 , niii @xmath261709/@xmath261741/@xmath261751 , siii @xmath261808 , and al iii @xmath261854/@xmath261862 . finally , we note that al iii has a systematically higher ionization potential than the other ions and is a sub - dominant ionization state of al . it thus could exist in a physically different location . thus , we most reliably consider the five calibratable , unsaturated , singly ionized transitions : feii @xmath261611 , niii @xmath261709/@xmath261741/@xmath261751 , and siii @xmath261808 . phl957 is a bright ( b=16.6 ) qso at @xmath31 , with a damped ly alpha system at @xmath1 ( beaver , et al 1972 ) . we obtained 5.5 hours of data on november 1 , 2002 , 4 hours on oct 3 , 2004 , 4 hours on oct 4 , 2004 , and 5 hours on oct 5 , 2004 . five of the 11 exposures taken in 2002 had the iodine cell in place , while 6 of the 13 exposures taken in 2004 were taken through the iodine cell . table [ tab : journal ] shows how the iodine cell exposures were interspersed with the non - iodine cell exposures . table [ tab : journal ] also shows the times that the relevant th / ar calibration arcs were taken , as well the temperatures inside the hires enclosure , and whether or not the echelle gratings were moved between the the qso exposures and the relevant th / ar exposures . the data from 2002 were acquired through the c1 decker ( fwhm @xmath32 ) , with the kv380 blocking filter in place . the data from 2004 were also observed through the c1 decker but with the kv418 blocking filter , and the ccd mosaic was binned by two in the spatial dimension . the 2002 spectra were extracted and combined using makee ( barlow 2002 ) . the 2004 spectra were extracted and combined using the xidlxavier / idl ] package hireduxxavier / hiredux / hires_doc.html ] ( bernstein , et al . in preparation ) . the 2002 and 2004 data can not be easily combined since a new mosaic of ccd s was installed in the spectrograph in between these runs . we note that the 2002 data have a substantially worse signal / noise per pixel ( 42 for the non - iodine cell exposures ) compared to the combined 2004 non - iodine cell data which have a s / n of about 70 per pixel . the keck iodine cell has been used extensively in searches for extra solar planets using the doppler technique ( butler , et al . the cell is placed in the beam and superposes several narrow absorption lines per angstrom on the qso spectrum between 4950 and 5900 . as pointed out by murphy , et al . ( 2001b ) , the different optical paths of the th - ar lamp and the qso spectrum are a possible source of systematic error in the wavelength calibration . using the iodine cell , wavelength calibration errors can be dramatically reduced since issues such as atmospheric dispersion , guiding errors at the slit , and all changes to the optics of the spectrograph are shared by both the iodine lines as well as the qso spectra . in addition , the th - ar lamp spectra are done at different times than the qso spectra . figure [ fig : iodine ] shows a sample of the iodine cell spectra taken with the fourier transform spectrometer ( fts ) at kpno with a resolution of around 170,000 and a signal / noise of 700 per pixel ( butler , et al . , 1996 ; marcy , 2008 ) . this high resolution and s / n means that we do not expect the iodine spectrum to be a significant source of wavelength calibration error . figure [ fig : fe67 ] shows a portion of echelle order 67 taken both with and without the iodine cell . in the bottom panel of figure [ fig : fe67 ] one can see the iodine lines as well as the fe ii @xmath261608 and fe ii @xmath261611 lines of the dla towards phl957 . the top panel of figure [ fig : fe67 ] does not contain the iodine lines . in order to do the wavelength calibration we started from the xidl th - ar calibration . xidl takes th - ar wavelengths that have been transformed from vacuum values to air values so that lines can be identified . using these identified lines a polynomial is fit over the full 2d spectrum . the wavelength scale found is then transformed back to vacuum values using the inverse edlen formula . after finding the th - ar wavelength scale we use our iodine measurements with two similar but independent methods to recalibrate . in both methods we convolved the high signal / noise iodine spectrum measured at knpo ( butler , et al . 1996 ) with a gaussian and then minimized the @xmath33 of the difference between this convolved spectrum and the phl957 spectrum shifted by an amount @xmath34 . in the first method a 3-parameter fit was performed in each wavelength bin of the phl957 spectra , with bin sizes from 1 to 10 angstroms being used . the fit returned the wavelength shift , @xmath34 , the sigma of the convolution gaussian , as well as a multiplicative continuum offset and the formal errors of these quantities . using this method , regions of the phl957 spectra that had strong lines were not included in the fit since these could distort the results ; a linear interpolation of bins on either side of the line region was used to find a correction at the line center . using the second method , strong lines in the phl957 spectra were fit and subtracted from the data before differencing with the convolved iodine spectra . this allowed direct determination of a wavelength correction even at line centers ( except for saturated regions which are removed as above ) . also , in the second method rather than fitting for both @xmath34 and the convolution sigma in each wavelength bin , a rolling 5 to 10 angstrom bin iterative method was used that found the wavelength correction for each data point and also the one best convolution sigma for the entire echelle order . we also tried setting the sigma of our gaussian convolution kernal to the resolution expected from the physical setup of the telescope , @xmath35 . this sigma was consistent with but slightly smaller than the sigma we found by fitting , but the wavelength offsets found either way agreed to within errors . in this method the errors in the wavelength calibration were found from the values of @xmath34 that caused the @xmath33 to increase by 1 . the methods of finding the continuum of the fts iodine spectra were also somewhat different for the two methods . in the first method , we used the highest flux value in a variable size wavelength bin as the continuum value . in the second method , we averaged the three largest flux values ( and their corresponding wavelengths ) in each one angstrom bin , and then set the continuum by fitting a spline through these averaged points . determining the continua of the phl957 spectra taken with the iodine cell is difficult because of the large number of unresolved overlapping iodine lines . the first method used a 4th order polynomial fit to the set flux values within 2-sigma of the highest flux found over a certain wavelength range . the second method found the phl957 continuum using a standard continuum fitting program , thus probably underestimating it . because of this underestimate , we repeated our analysis using several different possible continua and discovered that our final wavelength calibration results were robust for all plausible phl957 continua . a comparison between the results of the two methods showed agreement in @xmath34 within errors , and also good agreement in the reported errors on @xmath34 . we also tried a method that used a cross - correlation between the fts and phl957 spectra , which also worked , but did not seem as accurate as either of the other two methods . we will use the results of our second method throughout the rest of the paper . our final recalibration shift data is smoothed with a 5-angstrom box filter to reduce fitting noise . figure [ fig : convolve ] compares a small portion of echelle order 67 of phl957 with the convolved and shifted fts iodine cell spectrum . note that the th - ar calibrated spectra used as input to our recalibration pipeline should not include the standard keck hires heliocentric correction that accounts for the doppler shift from the changing motion of the earth around the sun . we recalibrate in the lab frame , and then make our own , more accurate , solar system barycentric motion correction using the code developed by marcy and butler ( 2008 ) . this barycentric correction should be good to better than 1 m / s . we found that the accuracy of the wavelength recalibration as measured by our calculated error depended strongly on the size of the bin used for the convolution . thus we report values from a 10 bin analysis as a compromise between small uncertainty and constancy of the wavelength calibration over the width of a single bin . figure [ fig : vshiftall ] shows an example ( exposure 3 - 1 ) of the resultant wavelength recalibation shift over the entire wavelength range for which we have both qso and fts iodine spectra . this figure is a major result of this work . we see several interesting effects . first the size of the shift is significant : typically from 500 m / s to 1000 m / s corresponding to a substantial fraction of a ccd pixel ( roughly 1300 m / s / pixel or 0.023 / pixel at @xmath36 . ) contrast these systematic velocity errors with an expected shift of 136 m / s between fe ii @xmath261608 and fe ii @xmath261611 due to the @xmath8 claim described in the introduction . we note previous workers have found similar keck hires shifts of 1000 m/s or larger ( e.g. page 734 of the night airglow line paper by osterbrock et al . 2000 ) , and figure 4 of suzuki et al . 2003 ) but this does not seem to be widely appreciated . second , there is a clear pattern seen in each echelle order , with the shift increasing from the edge of each order and reaching a maximum near the middle . for the purpose of measuring @xmath8 it is not the overall shift discussed above that is important , but the relative shift between the transitions being compared . depending on the echelle order figure [ fig : vshiftall ] shows relative shifts of 300 m / s to 800 m / s within the same order . these shifts could be dangerous since , depending upon the lines being being compared , they could result in a systematic relative velocity shift between lines , thereby mocking a changing @xmath0 . next , we are interested in how this wavelength recalibration shift varies with time . thus , figure [ fig : vshift67 ] shows a more detailed look at echelle order 67 which contains the fe ii @xmath261608/@xmath261611 lines . in the figure , different lines are the recalibration shifts for each of the five exposures that used the iodine cell and are labeled by their i d s , e.g. 3 - 0 is the first exposure of phl957 taken on oct-3 - 2004 . see table [ tab : journal ] for more details . we note that each exposure has a similar , but not identical shape as a function of wavelength , and that there is a different wavelength offset for each exposure . while the variation over this order for each exposure is typically 500 m / s or less , the shift between different exposures can be as much as 2000 m / s . this means that the th - ar wavelength calibration that gives us the wavelength solution as a function of pixel , is not stable and drifts with time . it is of interest to note that recalibration offsets between nights are in general larger than the drifts during each night . one sees that on oct 3 and 4 , when the iodine cell exposures where taken one after the other , the shift during the night was substantially smaller than the inter - day shifts , while on oct 5 , when there was three hours between iodine cell exposures , there was a larger shift . figure [ fig : vshift67 ] is probably more important than the previous figure , since it shows that the th - ar calibration is not stable over time . if the wavelength shifts shown in figure [ fig : vshiftall ] were stable in time , they could be removed and would have little effect on the measurement of @xmath8 . but large systematic wavelength shifts during the night mean that measurements at the desired level of precision may be difficult with th - ar calibration alone . it is important to note that even though the calibration errors reported here are much larger than the final velocity precision needed to determine @xmath8 , it is possible that these calibration errors average out and do not ruin the final @xmath8 determination . in the many multiplet method , many different lines are compared across many different absorbers at many different redshifts . if the signs of the calibration errors are random , the calibration errors may average away . a complete discussion of this possibility is beyond the scope of this work , but will be presented elsewhere . we made some preliminary attempts at understanding the causes of these unexpected wavelength calibration shifts . while complete understanding of the causes and exploration of methods to correct for the shifts are beyond the scope of this paper , we report some ideas and preliminary results here . we hope to use additional data and analysis to find a more complete understanding , which will then be published elsewhere . first , it is interesting that during all three nights the shifts became more negative at about 500 m / s per hour . thus we plotted the shifts vs time , and also vs. the various temperatures , etc . that hires reports . with only six iodine exposures it is difficult to discern a pattern and impossible to that prove a pattern exists , and in fact we did not see any fully convincing trends . the best partial trend is a decrease in calibration shift with , where is the difference in hires enclosure temperature between when the th / ar calibration exposure was taken and when the qso exposure was taken . we plot this trend in figure [ fig : deltatin ] , where the 6 qso exposures are shown as asterisks and individually labeled . the data is given in table [ tab : journal ] . we see that the trend is badly broken by the 4th exposure on oct 5 ( 5 - 3 or i d 2097 ) , which originally led us to discount temperature as the main culprit . however , we have some exposures of a star ( hd209833 ) taken the same nights through the iodine cell , and we plot the calibration shifts vs for them as small boxes . the complete analysis of these star exposures is beyond the scope of this paper , but we see that they fall on the same trend line , showing that exposure 5 - 3 seems to be an exception . next , we check whether or not the echelle gratings were moved between the time of the th / ar arc exposure and the qso exposure . this is shown in table [ tab : journal ] . we see that only on oct 3 ( exposures 3 - 1 and 3 - 2 ) were the gratings not moved between th / ar calibration and the qso exposures . in figure [ fig : deltatin ] we see that the dispersion of exposures 3 - 1 and 3 - 2 from the trend line is smaller than for exposures 4 - 1 and 4 - 2 . exposure 5 - 0 is near the trend line , but of course 5 - 3 is way off , probably for some other reason . thus we see a weak hint that moving the gratings between th / ar calibration and science exposure can cause a calibration error . to understand the problem with exposure 5 - 3 we note that a different th / ar arc was used for 5 - 3 than for the other iodine exposure ( 5 - 0 ) taken that night . to check whether something went wrong with the arc exposure ( id=2107 ) used to calibrate 5 - 3 , we redid the iodine recalibration using arc exposure 2026 , taken much earlier that same night . this point is also plotted in figure [ fig : deltatin ] as a labeled triangle . the point shifts substantially , but keeps the same ( large ) distance from the trend line . thus we see that while temperature surely plays an important role and a correction perhaps can made for this , there are other parameters that seem to be important , not all of which are understood at this point . moving the gratings between th / ar and science exposures may increase the calibration error . in what follows we will not attempt to use the temperature trend to make any correction , though in future work this might be possible . we note that other contributions to the calibration shifts are possible . n. suzuki ( private communication ) suggested that the cause could be that the hires spectrograph is mounted at a small angle with respect to the optical axis , so that the light path rotates as the telescope moves , resulting in variable vignetting in comparison to light from the th - ar lamp which is fixed to the spectrograph . more generally , the optical path of the th / ar calibration light differs from the optical path of the qso light resulting in different wavelength to pixel mapping . p. molaro ( private communication ) suggested that the cause could be the changing position of the qso image centroid in the slit . we hope to check this with additional exposures at a future date . m. murphy ( private communication ) suggested that the cause could be temperature and pressure differences between the th / ar calibration exposures and the data exposures , as well as the resetting of the echelle gratings . the analysis presented above was , in large part , motivated by the comments of m. murphy . it would be useful to study and understand these wavelength calibration shifts , perhaps by analyzing other data , since it might then be possible to model and remove them . this will be pursued in more detail elsewhere . overall , the wavelength calibration errors reported here may seem to be a remarkable result , but we note that osterbrock et al , ( 2000 ) recalibrated the keck hires th - ar tube wavelengths using several night sky airglow lines and found similar ( @xmath37 ) calibration errors . in addition , both the magnitude and time variation of these shifts were detected using ly alpha forest lines by suzuki , et al . ( 2003 ) . it is interesting to ask whether sky lines themselves could be used to calibrate the qso spectra without use of the iodine cell . in fact , the current version of makee data reduction pipeline includes an option to calibrate using sky lines . we used our iodine method to check on the wavelength accuracy of the makee sky line calibration , but still found substantial wavelength shifts both within and across echelle orders , and as a function of time . our original hope was that if there would be a time - stable wavelength calibration found from the iodine cell exposures then that then could be applied to all the exposures . when the iodine cell is in place we get well - calibrated wavelengths , but substantially less s / n . this is because iodine lines cover basically the entire spectrum , decreasing the number of photons , adding spurious components to line fitting programs , and also making the continuum difficult to determine . so , it is the exposures without the iodine cell that carry the highest signal / noise and that we wish to use to test for a changing alpha . since again , as figure [ fig : vshift67 ] shows , the wavelength calibrations vary substantially over the course of a night , it is not clear how to use the calibration in our fits . we tried several methods before settling on the following : 1 ) for exposures taken with the iodine cell we add the wavelength recalibration correction to each wavelength output by the standard hires pipeline . 2 ) for exposures taken without the iodine cell , we interpolate/ extrapolate the wavelength correction using the two iodine exposures nearest in time , and assuming that the correction changes linearly with time . 3 ) we then add the barycentric correction to every data point . at this point we have wavelength corrected spectra that can be fit . we can either add the spectra for each order together ( rebinning since the spectra are no longer on a common wavelength scale ) or just combine and sort the data files for each order together giving more flux measurements to be fit . we can check the effect of our recalibration by seeing how well sharp features in the spectra line up . for example , figure [ fig : checkvshift ] shows a close - up of the left edge of the saturated fe ii @xmath261608 line before and after iodine recalibration . the panel labeled ` th / ar ' ( a ) shows an overlay of the six iodine exposures calibrated with the standard hires pipeline , while the panel labeled ` iodine ' ( a ) show the same after our recalibration . the features line up significantly better after recalibration ; spreads of more than 2000 m / s become significantly less than 1000 m / s . the other two panels show overlays of the seven exposures taken without the iodine cell . the panel labeled ` th / ar ' ( b ) is again the result from the standard hires calibration , while the panel labeled ` iodine interp ' ( b ) shows the same seven exposures , but recalibrated using the interpolation scheme described above . for the non - iodine exposures , there is some improvement in alignment of the line edge , but not nearly as much as for the iodine exposures . here spreads of 1000 m / s or more seem to remain even after recalibration . since for signal / noise and fitting reasons , we must use mainly the non - iodine exposures , it is disappointing that our efforts to recalibrate may not pay off . our inability to model the calibration shifts between iodine exposures seems to be the main culprit . in what follows , we mostly just combine the recalibrated , continuum subtracted spectra for all non - iodine exposures into one large file , and then sort by wavelength and fit . in some cases we rebin the data combining several data points and adding the errors in quadrature , and in some cases we just use the co - added spectra from the standard hires pipeline . we treat the data from 2002 separately since it was taken on a different ccd chip . we also treat the data with and without the iodine cell separately since the iodine lines add significant effective noise to the phl957 spectra . we fit the spectra using a code we developed based upon the cern library minuit minimization program and the humdev voigt profile calculation routine ( wells , 1999 ) . we compared our results with those of vpfit ( carswell et al . 2008 ) and dude ( kirkman et al . 2003 ) and found agreement for individual lines . using our own code allows us easily to do joint fits with the redshifts of transitions varying independently and also allows us to add additional parameters as needed . figures [ fig : linefita ] and [ fig : linefitb ] show fit results for several transitions for the combined 2004 data . the fit parameters are the redshift , the line width ( b - value ) , and the column density . we note that several of the lines are saturated , making wavelength centroiding more difficult and less accurate . the s / n of each 2004 non - iodine spectra is approximately 25 per pixel , giving a total s / n for the 7 non - iodine exposures of about 70 per pixel . the individual iodine exposures have a s / n of around 19 per pixel , for a co - added total of about 47 per pixel , though the iodine lines cause the effective s / n to be lower than this . the co - added 2002 data have s / n approximately 42 per pixel . our total signal / noise is thus quite high for a high redshift object , and we want to first estimate the minimum possible error on @xmath8 that could be obtained with these spectra . to do this , we use a fisher matrix type method suggested by murphy , webb , and flambaum ( 2006 ) , and bouchy , pepe , and queloz ( 2001 ) . for continuum normalized flux spectrum @xmath38 with 1-sigma error array @xmath39 , the minimum possible uncertainty in velocity contributed by pixel @xmath40 is : @xmath41 thus more precise velocity measurements come from pixels with large flux gradients and small errors . the minimal possible uncertainty in the velocity of a portion of spectrum is thus : @xmath42 ^ 2\right]^{-1/2 } , \label{eqn : eqsigmav}\ ] ] where the sum is over pixels . finally , the minimum uncertainty in @xmath8 can be found by performing a least - squares fit to a version of equation [ eqn : eqdelalpha ] : @xmath43 where @xmath44 numbers the lines that are being compared , @xmath45 is a constant offset ( degenerate with the system redshift ) , and the minimum error in @xmath8 is just the fit uncertainty in the slope of this linear equation . murphy , webb , and flambaum ( 2006 ) performed this analysis for their data and for the data of chand , et al . ( 2004 ) , and levshakov , et al . ( 2006 ) , finding that while their own errors were ( barely ) within the allowable minimum , the reported uncertainties of the others were smaller than the minimum possible . the corrected version of the levshakov results ( molaro , et al . 2008 ) does seem to be in agreement with the minimum error limit . we would like to perform such an estimate , but first note that previous estimates of minimum errors did not include uncertainties in the @xmath14 values . table [ tab : lineinfo ] shows that these uncertainties can be significant , and inclusion of these uncertainties will increase both the fisher matrix minimum errors and also the error on @xmath8 . one can include these uncertainties , @xmath46 , in the fit and obtain a more realistic minimum error estimate . in this case , rather than a simple linear least - squares fit one must use a method that allows errors in both the ordinate and the abscissa . since the uncertainties in @xmath14 are theoretical estimates and not statistical errors , this method is not technically completely consistent , but it should give a reasonable idea of the size of the effect . as we find below , the smallness of the claimed values for @xmath8 imply that the uncertainties in @xmath14-values are not very important . table [ tab : lineinfo ] shows the minimum values of @xmath47 for the portion of the combined 2004 spectra containing each of the transitions that have a calculated q - value . these results were calculated using co - added spectra from the standard xidl hires pipeline that did not include the iodine cell wavelength recalibration ; this simplifies the calculation but has little effect on these minimum error results . the values range from 25 m / s for the strong fe ii @xmath261608 line to 153 m / s for the weak lines such as fe ii @xmath261611 . it is interesting to note that the maximum velocity precisions obtainable with our data seem to be sufficient to measure the shifts predicted by a changing value of @xmath8 at the level claimed by murphy et al . however , we also see that the weakness of the fe ii @xmath261611 line means that the precision obtained from using only the fe ii @xmath261608/@xmath261611 pair may not be good enough for this purpose . one can combine all these minimum @xmath47 values using equation [ eqn : eqsigmav ] to get an overall minimum velocity error , but since @xmath8 is determined by differences in redshifts this is not appropriate . if one just wanted to determine how accurately the redshift of the entire system could be determined , then one could combine the minimum errors in table [ tab : lineinfo ] to find @xmath48 m / s . if we included only the lines for which we have iodine cell calibration the result is @xmath49 m / s . the 2002 data alone would give 38 m / s precision , which if combined with the 2004 data would give an ultimate velocity precision of 12 m / s . of course , as discussed below there are several important systematic errors that greatly increase these uncertainties . next we perform the least squares fit to find minimum possible errors on @xmath8 from this data and display the results in table [ tab : minsigdelalpha ] . these uncertainties , e.g. @xmath50 , are reasonably competitive with errors quoted by murphy , et al . ( 2001a ; 2003 ; 2004 ) , with chand , et al . ( 2004 ) , and with levshakov , et al ( 2006 ) . we again note that the latter two groups seem to have produced measurements of @xmath8 with errors smaller than their minimum possible errors , something murphy , webb , and flambaum ( 2006 ) have criticized , and which levshakov et al . subsequently corrected ( molaro , et al . 2008 ) . we note that minimum possible error coming from analysis of the fe ii @xmath261808/@xmath261611 pair is @xmath51 , substantially less precise than could be found using several lines . as mentioned this is mostly because the fe ii @xmath261611 line is so weak ( due to its rather low oscillator strength ) and thus its redshift can not be measured very precisely . therefore , our original idea of using just these two lines is probably not that useful . in addition , examination of the lines shows that it is only the saturated regions of the fe ii @xmath261608 line profile that are detectable in the fe ii @xmath261611 line profile . thus a joint @xmath33 fit can not accurately recover their relative velocity offset . it seems it is better to compare strong lines with other strong lines , and weak lines with other weak lines . we do this in the following section . we can also test the importance of the uncertainties in @xmath14 listed in table [ tab : lineinfo ] , by doing a linear fit that allows errors in both the abscissa and ordinate . in this case , the resulting minimum error depends upon the fit value of @xmath8 , that is , the slope of the line given in equation [ eqn : eqvj ] . this is to be expected since if @xmath52 , it does not matter what the values of @xmath53 are , while if there is a large slope then uncertainty in @xmath53 will propagate to uncertainty in @xmath54 , and therefore uncertainty in @xmath8 . a simple way to estimate the increase in @xmath8 uncertainty is to just do this propagation of errors , that is , change : @xmath55 where @xmath56 . the results of this error propagation are also shown in table [ tab : minsigdelalpha ] for the murphy , et al . ( 2003 ) value @xmath57 . we note that because the slope @xmath8 is small , the increase in uncertainty is also quite small , no larger than around 14% depending on the set of lines used . thus previous workers who ignored this effect are not making a large underestimation of their @xmath8 uncertainties and limits . because of the smallness of this effect , we will not consider the uncertainty in @xmath14 in what follows . it is important to realize that the minimum possible errors discussed above are never reached in a real observation for several reasons as follows : \1 . as extensively discussed in murphy , et al . ( 2008 , section 2.2.2 ) , comparison between transitions with different degrees of saturation will give velocity uncertainties derived from voigt profile fitting substantially greater than the fisher matrix minimum . this is because velocity precision coming from the sharp edges of the saturated profiles can not be directly compared with the central regions of unsaturated profiles . especially in multicomponent systems , this will result in degeneracy between the component fit parameters , and make the redshift determination of any one component less accurate . this is evident when the covariance matrix for one of our 5 component fits is examined . we see strong correlation between the redshifts of the various components and between these redshifts and various other parameters such as the b - values . we can investigate this effect directly in our data by comparing the formal parameter fit errors with the minimum errors calculated above . for example , a 5 component fit of the saturated fe ii @xmath261608 line using the co - added data gives a minimum possible redshift error of 33.4 m / s , while the fitting program returns an error of 184 m / s on the strongest component . for si ii @xmath261808 , an unsaturated line , the effect is smaller , but still pronounced : 32.6 m / s minimum error vs. 55 m / s fit result . we note that the minimum errors found from using formulas [ eqn : eqsigmavi ] and [ eqn : eqsigmav ] on the individual exposures are quite consistent with the minimum errors found from the co - added data : for 3-component fit to fe ii 1@xmath261608 we find 32.6 m / s for co - added data , vs. 27.4 m / s from combining all 7 non - iodine cell exposures . for si ii @xmath261808 the comparison is 50.4 m / s from the co - added compared to 49.9 m / s for individual exposures . note that errors returned from the fit software are actually also overly optimistic because : \2 . as discussed above there are other sources of error such as wavelength calibration , \3 . the errors on the flux are not really gaussian which means that larger than gaussian deviations occur , \4 . it is very difficult to determine the number of components needed to fit each transition , especially for saturated , or nearly saturated lines . the formal fit error returned on a component redshift does not include the possibility of another component , which may be apparent in some transitions and not in others . the @xmath33 surface can have several local minima , and small measurement errors can switch one between needing another component and not needing another component . since it is the difference between component redshifts that is used to determine @xmath8 , this can be a source of systematic error . for example , we found examples where a region of a spectrum was formally slighter better fit by two nearby components than by a single component , but the redshift of the components in the two cases shifted by around 1500 m / s , a value large compared to what is expected from @xmath8 . murphy et al . ( 2008 ) investigated this effect in detail ( see their figure 7 ) and found that while this effect does exist if the the number of fit components is less than the true number of components , this systematic error is greatly reduced if the number of fit components is equal to or greater than the true number of components . we tried several methods of deriving a value of @xmath8 from our voigt profile fits . the simplest method is to fit each of our n lines independently , and then do a least squares linear fit of equation [ eqn : eqvj ] . the value of @xmath58 can be found from @xmath59 where @xmath60 is the redshift obtained by averaging over all the transitions detected in a given velocity component . the result of this simple direct method for component 1 ( the component with the largest column density ) of the 9 lines for which we have both q - values and an iodine spectrum is shown as the solid line in figure [ fig : delalphanew9 ] . here we plot @xmath61 vs. the wavelength corrected velocity @xmath62 for each line , as well as the best linear fit for the 9 lines ( solid line ) and the best linear fit for the 5 unsaturated , lines that are singly ionized ( dashed line ) . for component 1 of the 9 lines , the fit slope is @xmath63 , with @xmath64 , formally a very strong detection , inconsistent with other measurements . for component 1 and the 5 lines , the fit slope is @xmath65 , with @xmath66 , formally a nice null result consistent with , but much weaker than results in the literature , and a factor of 3 worse than the minimum possible error listed in table [ tab : minsigdelalpha ] . however , we do not believe either of these results . if a changing fine structure were the explanation for the variation in values of @xmath62 in the 9 line fit , then the points in figure [ fig : delalphanew9 ] would lie on a straight line within errors . we see from the figure and from the terrible @xmath67 that this is not the case . when the goodness of fit to a model is very bad , it means that either the model is incorrect or that there are systematic errors not included in the data uncertainties . in such cases the errors derived from the fit can not be trusted , and a better way to estimate the uncertainty is to increase the errors on the data points until @xmath68 . doing this for the 9 line fit gives @xmath69 ( @xmath70 ) , a null result with a more realistic error estimate , but far from the sensitivity we expected from our data . another way to see that the first result above is meaningless is to do the fit including the errors in @xmath14 . this gives an entirely different answer which also is a terrible fit . however , when the errors are scaled as above , this fit then gives nearly the same result as above but with larger errors . the problem with the 9 line fit seems to be the saturated lines fe ii @xmath261608 and al ii @xmath261670 , as well as the aluminum lines al iii @xmath261854 and @xmath261862 . since al iii is in a different ionization state than the other ions , it might exist at a different physical location , and so maybe should be left out of the fits . we want to discuss in more detail the saturated lines that are far from the others , but before doing that we note that this pattern repeats also in the @xmath53 vs. @xmath62 fits of the 2nd strongest component , the results of which are given in table [ tab : delalpha ] . the fit redshifts of the first components of the fe ii @xmath261608 and al ii @xmath261670 lines differ from the dashed 5 line fit by about @xmath71 . some insight into possible reasons for this can be gained by noting that if we force this component of fe ii @xmath261608 to be the same as the average of the unsaturated transitions , we also get a completely satisfactory fit to the voigt profile : @xmath72 for the free fit , and @xmath73 ( @xmath74 ) for the 3 component fit with one redshift forced to the average value . the reported formal fit errors on these redshifts are @xmath75 , substantially smaller than the difference between the free fit and the forced fit . thus we see another possible source of error in determining @xmath8 . the formal fit errors seem to underestimate the true range of acceptable redshift values , and therefore overestimate the precision with which @xmath8 can be determined by this method . we note that when we force the redshift of the first component of fe ii @xmath261608 to agree with the other lines , the 2nd and 3rd component also come into alignment . thus there is a degeneracy where the redshift and b - value of one component can play off the redshift and b - values of other components . another way this can be seen , is from the fact that the value derived for @xmath8 depends greatly on the error reported for the fe ii @xmath261608 redshift . figure [ fig : delalphanew9 ] shows that the linear fit for @xmath8 is driven by the tiny error on the fe ii @xmath261608 redshift . increasing that error by a factor of ten ( to roughly equal the errors on the other redshifts ) changes the value to @xmath76 . examination of the al ii @xmath261670 line shows a similar problem ; again the line is saturated and in this case there are certainly more than 3 detectable components . we are only allowing 3 total components so that we can compare component to component across other transitions where only 3 components are detectable , but the fitting program can find different nearly equivalent ways to fit the line to 3 components since saturation means less shape information is available . while the effect is largest for the saturated lines , we find the same effect for the al iii @xmath261864 and al ii @xmath261862 , unsaturated lines , where there are only 3 detectable components . here again while the best fit redshift is far from the others , we can get a completely acceptable fit to the voigt profile ( @xmath77 ) even while forcing the first component to the average . again , when one component is forced to agree , the other components then also agree . thus , we note a systematic error that can give trouble in this type of fitting . when a system has two ( or more ) components nearby in redshift space , there can be a near degeneracy where the two redshifts can play off each other , moving significantly at the level of precision we are after , but yet still giving very good voigt profile fits . note that a linear fit to the 5 line set that includes the q errors gives nearly the same answer as shown by the dashed line in figure [ fig : delalphanew9 ] . this is to be expected since the slope of the line , @xmath8 , is small . we can repeat the above analysis for the 2nd and 3rd strongest components . the results are given in table [ tab : delalpha ] for various sets of lines . in these cases we mostly find null results such as @xmath78 with @xmath79 for all nine lines . this is because the lower column densities mean larger formal fit errors . there are additional components detectable in some of the lines , but they are much weaker and do not contribute much . to get a final answer using this method one could average the above results , which actually gives a result with larger formal error than from the first component alone , again showing that systematic errors are dominating this procedure . the above method , while useful for showing errors arising from fitting degeneracies , is not really correct because for it to work one must match up the various components fit in different lines in order to properly compare their redshifts . as noted , these fitting degeneracies allow the redshifts of two nearby components to be traded against each other and against the b - values , resulting in correlations that cause the uncertainty in the component redshifts to be larger than the formal fit errors . more properly , one should do a large joint fit for all system components simultaneously , that is tie together the different velocity components . this is standard practice and is used by murphy et al . , chand , et al . levshakov , et al . , thus , for example , the fit to the 9 lines above by the previous method used a total of 81 parameters : redshift , column density and b - value for each of 3 components of each of 9 lines . a global joint fit might might have 9 parameters for the redshift , column density , and b - value of each of 3 components , plus some column density offsets to allow for different elemental abundances ( 1 parameter per component per additional element or 12 including fe ii , ni ii , si ii , and al ii and al iii ) . an additional b - value offset is probably necessary for each component of the al iii lines since these are in a different ionization state ( 3 more parameters ) . then adding @xmath8 as a final parameter , there would be a total of 25 parameters . the results of such a global joint fit method are shown in table [ tab : delalpha ] for various sets of lines . table [ tab : delalpha ] shows that results we find for @xmath8 vary over a wide range of values , from significant detection @xmath80 for the fe ii @xmath261608/fe ii @xmath81 pair , to to a null result @xmath82 for the nine - line fit . looking over all the results in table [ tab : delalpha ] we see that both the method used and the lines selected can make a significant difference in the final result . if our errors were under control , this should not be the case . note that in most cases , the errors for the global fit method method hover around @xmath83 ; roughly half the errors of around @xmath84 found in the simple method described in the previous section . in table [ tab : fitcomponents ] we display the voigt profile components that resulted from one of our joint global fits . one might worry that the puzzling results we find come from errors in our fitting program , so as a final check , we redid our calculation of @xmath8 using the new version of vpfit ( carswell , et al . 2008 ) which includes the possibility of doing a global joint fit for @xmath8 . these results are also shown in table [ tab : delalpha ] . for the 6 line set ( fe ii @xmath261611/@xmath261608 , ni ii@xmath261709/@xmath261741/@xmath261751 , si ii@xmath261808 ) we studied above we find @xmath85 ( @xmath86 ) , while for the set of all 16 lines we find find @xmath87 ( @xmath88 ) . both these results are within the range found by our different fitting methods . we tried and failed to give a definitive answer to the question of whether the fine structure constant was different at early times in high redshift ly alpha systems . in order to investigate this problem in detail we used data taken through the keck iodine cell and wrote our fitting software from scratch . using the iodine cell for wavelength calibration we found a serious source of systematic error that did not allow calibration of the keck hires spectrograph to the precision needed . we also found degeneracies in the fitting procedures that added to the calibration systematic errors . due to all the systematic errors we were able to derive various results , running from very significant detections , to strong null limits . does this imply that a meaningful measurement of or limit on @xmath8 is impossible using keck hires ? it is not clear . perhaps more careful attention to voigt profile fitting , as advocated by m. murphy ( private communication 2008 ) will solve the fit degeneracy problem . perhaps a careful selection of absorbers would also help , or trying to focus on systems with a single component . perhaps the wavelength calibration errors we discovered can be corrected , or perhaps they will average away if a large sample of absorbers is considered . in this paper we raise these questions , but do not answer them . we also used a fisher matrix technique to investigate the minimum possible errors in @xmath8 that our spectra s signal / noise would allow . we found our fit results did not exceed these limits . for example , for the set of 9 lines used in most of our analyses , the minimum possible error on @xmath8 , as given in table [ tab : minsigdelalpha ] , is @xmath89 , a result consistent with our fit results . since the calculation of minimum possible errors is not difficult , we suggest that workers always calculate them and never report results with uncertainties smaller than the data theoretically can allow . while we have not yet looked carefully at data or analysis done by other workers in the field , we worry that some of the systematic errors and overestimation of precision we found here may also be present in other analyses . thus one possible explanation for the discrepant findings on @xmath8 discussed in the introduction , is that several workers in the field are overestimating the precision of their measurements and the discrepancies reported in the literature are due to random fluctuations occurring within the larger , under - reported , systematic errors . at this point it is not clear how to make further progress in this subject using keck hires , but other techniques such as frequency combs ( steinmetz , et al . 2008 ) may become available and be of use in resolving the question . in addition , proposed new instruments ( e.g. , codex for e - elt or espresso for the vlt ) are being designed for doppler measurement stability and will hopefully be be free of these problems . we thank david kirkman , bob carswell , marc rafelski , joel heinrich , john johnson , and michael j. simmonds for helpful discussions . we especially thank michael murphy for many insightful comments , suggestions and questions , including several that led to corrections of errors in early versions of this paper . finally , we thank nao suzuki for discussion of an early version of this paper where we discovered that he had already understood and published several of the points made here . k.g . and j.b.w . were supported in part by the doe under grant de - fg03 - 97er40546 . is partially supported by an nsf career grant ( ast-0548180 ) , and j.x.p . and a.m.w . are supported in part by nsf grant ( ast-07 - 09235 ) . the w. m. keck observatory is operated as a scientific partnership among the california institute of technology , the university of california and the national aeronautics and space administration . the observatory was made possible by the generous financial support of the w. m. keck foundation . the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . barlow , t. 2002 , makee keck observatory hires data reduction software ( pasadena : caltech ) , http://spider.ipac.caltech.edu/staff/tab/makee /index.html beaver , e.a . , et al . , 1972 , apj , 178 , 95 bouchy , f. , pepe , f. , and queloz , d. , 2001 , a&a , 374 , 733 butler , r.p . , et al . , 1996 , pasp , 108 , 500 carswell , r.f . , webb , j.k . , cooke , a.j . , irwin , m.j . , 2008 , vpfit version 9.5 web page , http://www.ast.cam.ac.uk/ rfc / vpfit.html chand , h. , srianand , r. , petitjean , p. , aracil , 2004 , a&a , 417 , 853 dzuba , v.a . & flambaum , v. v. , 2008 , astro - ph/0805.0461v2 dzuba , v.a . , flambaum , v. v. , kozlov , m. g. , and marchenko , m. , 2002 , phys . rev . a , 66 , 022501 garcia - berro , e. , isern , j. , & kubyshin , y.a . , 2007 , astron . , 14 , 113 kirkman , d. , et al . , 2003 , apj suppl , 149 , 1 levshakov , s.a . , et al . , 2007 , a&a , 466 , 1077 levshakov , s.a . , et al . , 2006 , a&a , 449 , 879 marcy , g. w. , 2008 , unpublished . marcy , g. w. , & butler , r.p . , 2008 , unpublished . molaro , p. , reimers , d. , agafonova , i.i . , & levshakov , s.a . , 2008 , j. st , 163 , 173 murphy , m. t. , et al . , 2001a , mnras , 327 , 1208 murphy , m. t. , webb , j. k. , flaumbaum , v. v. , churchill , c. w. , prochaska , j. x. , 2001b , mnras , 327 , 1236 murphy , m. t. , webb , j. k. , and flambaum , v. v. , 2003 , mnras , 345 , 609 murphy , m. t. , webb , j. k. , and flambaum , v. v. , 2006 , arxiv : astro - ph/0611080 murphy , m. t. , webb , j. k. , and flambaum , v. v. , 2008 , mnras , 384 , 1053 osterbrock , d. e. , et al . , 2000 , pasp , 112 , 733 porsev , s.g . , , 2007a , phys . a 76 , 052507 srianand , r. , chand , h. , petitjean , & p. , aracil , b. , 2004 , phys . , 92 , 121302 steinmetz , t. , et al . 2008 , science , 321 , 1335 . suzuki , n. , et al . 2003 , pasp , 115 , 1050 . wells , r.j . , 1999 , j.quant . transfer , 62 , 29 cccccccccc 1 nov 02 & 33 & out & 5:26 & 1800 & 3.75 & & & & + 1 nov 02 & 34 & out & 5:58 & 1800 & 3.76 & & & & + 1 nov 02 & 35 & in & 6:30 & 1800 & 3.79 & & & & + 1 nov 02 & 36 & in & 7:01 & 1800 & 3.76 & & & & + 1 nov 02 & 47 & out & 7:58 & 1800 & 3.61 & & & & + 1 nov 02 & 48 & out & 8:30 & 1800 & 3.68 & & & & + 1 nov 02 & 49 & in & 9:03 & 1800 & 3.68 & & & & + 1 nov 02 & 50 & in & 9:34 & 1800 & 3.78 & & & & + 1 nov 02 & 60 & out & 10:34 & 1800 & 3.71 & & & & + 1 nov 02 & 61 & out & 11:06 & 1800 & 3.75 & & & & + 1 nov 02 & 62 & in & 11:39 & 1800 & 3.71 & & & & + 3 oct 04 & 67/3 - 0 & out & 9:18 & 3600 & 4.26 & 66 & 09:15 & no & 0.042 + 3 oct 04 & 68/3 - 1 & in & 10:31 & 3600 & 4.21 & 66 & 09:15 & no & 0.097 + 3 oct 04 & 69/3 - 2 & in & 11:33 & 3600 & 4.07 & 66 & 09:15 & no & 0.236 + 3 oct 04 & 70/3 - 3 & out & 12:35 & 3600 & 4.00 & 66 & 09:15 & no & .0306 + 4 oct 04 & 1096/4 - 0 & out & 9:25 & 3600 & 3.00 & 1144 & 15:35 & yes & -0.153 + 4 oct 04 & 1098/4 - 1 & in & 10:56 & 3600 & 2.97 & 1144 & 16:35 & yes & -0.125 + 4 oct 04 & 1099/4 - 2 & in & 11:57 & 3600 & 3.01 & 1144 & 16:35 & yes & -0.167 + 4 oct 04 & 1100/4 - 3 & out & 12:58 & 3600 & 2.93 & 1144 & 16:35 & yes & -0.083 + 5 oct 04 & 2094/5 - 0 & in & 8:25 & 3600 & 2.87 & 2026 & 3:01 & yes & 0.125 + 5 oct 04 & 2095/5 - 1 & out & 9:27 & 3600 & 2.86 & 2107 & 15:46 & yes & 0.430 + 5 oct 04 & 2096/5 - 2 & out & 10:29 & 3600 & 2.91 & 2107 & 15:46 & yes & 0.375 + 5 oct 04 & 2097/5 - 3 & in & 11:30 & 3600 & 3.62 & 2107@xmath90 & 15:46 & yes & -0.333 + 5 oct 04 & 2098/5 - 4 & out & 12:32 & 2700 & 3.55 & 2107 & 15:46 & yes & -0.264 + cccccc fe ii @xmath261608.45 & 67 & @xmath91 & 25.0 & yes + fe ii @xmath261611.20 & 67 & @xmath92 & 153 & yes + al ii @xmath261670.79 & 65 & @xmath93 & 34.0 & yes + ni ii @xmath261709.60 & 63 & @xmath94 & 83.1 & yes + ni ii @xmath261741.55 & 62 & @xmath95 & 48.7 & yes + ni ii @xmath261751.92 & 62 & @xmath96 & 70.8 & yes + si ii @xmath261808.01 & 60 & @xmath97 & 36.4 & yes + al iii @xmath261854.72 & 58 & @xmath98 & 76.0 & yes + al iii @xmath261862.79 & 58 & @xmath99 & 125 & yes + si ii @xmath261526.71 & 71 & @xmath100 & 28.8 & no + zn ii @xmath262026.14 & 53 & @xmath101 & 129 & no + zn ii @xmath262062.66 & 52 & @xmath102 & 229 & no + cr ii @xmath262056.26 & 52 & @xmath103 & 89.9 & no + cr ii @xmath262062.24 & 52 & @xmath104 & 102 & no + cr ii @xmath262066.16 & 52 & @xmath105 & 143 & no + fe ii @xmath262344.21 & 46 & @xmath106 & 41.7 & no + all 16 lines ( 2002 and 2004 data ) & @xmath107 & @xmath108 & + all 16 lines & @xmath109 & @xmath110 & @xmath111 + 9 calibratable lines & @xmath112 & @xmath113 & @xmath89 + 7 calibratable , unsaturated lines & @xmath114 & @xmath115 & @xmath116 + fe ii @xmath261608/@xmath261611 pair & @xmath117 & @xmath118 & @xmath119 + ccrrc simple ( comp 1 ) & the ( 5 ) & @xmath120 & 1.09 & 0.35 + simple ( comp 1 ) & ( 5)+fe ii @xmath261608 & @xmath121 & 323 & @xmath122 + simple ( comp 1 ) & ( 5)+al iii @xmath261854/@xmath261862 & @xmath123 & 22 & @xmath124 + simple ( comp 1 ) & ( 5)+fe ii @xmath261608 + al iii @xmath261854/@xmath261862 & @xmath125 & 248 & @xmath126 + simple ( comp 1 ) & the ( 9 ) & @xmath127 & 264 & 0 + simple ( comp 2 ) & the ( 5 ) & @xmath128 & 0.66 & 0.57 + simple ( comp 2 ) & ( 5)+fe ii @xmath261608 & @xmath129 & 75 & @xmath130 + simple ( comp 2 ) & the ( 9 ) & @xmath131 & 92 & @xmath132 + simple ( comp 3 ) & ( 5)+fe ii @xmath261608 & @xmath133 & 3.8 & 0.004 + global joint fit & fe ii @xmath261608 and fe ii @xmath261611 pair & @xmath134 & 0.96 & - + global joint fit & the ( 5 ) & @xmath135 & 0.80 & - + global joint fit & ( 5 ) + fe ii @xmath261608 & @xmath136 & 0.92 & - + global joint fit & ( 5 ) + al iii @xmath261854/@xmath261862 & @xmath137 & 0.83 & - + global joint fit & the ( 9 ) & @xmath138 & 1.01 & - + vpfit global ( w / o iodine ) & ( 5 ) + fe ii @xmath261608 & @xmath139 & 3.1 & @xmath140 + vpfit global ( w / o iodine ) & all 16 & @xmath141 & 6.0 & @xmath142 + |
we prove knudsen s law for a gas of particles bouncing freely in a two dimensional pipeline with serrated walls consisting of irrational triangles . dynamics are randomly perturbed and the corresponding random map studied under a skew - type deterministic representation which is shown to be ergodic and exact . * keywords : * knudsen s law , random billiards , random maps , irrational polygons , ergodic billiards . | 1006.4831 | in his nowadays classical studies on the kinetic theory of gases , the danish physicist m. knudsen experimentally observed that , no matter how an inert gas was injected into a pipeline , the direction in which a molecule rebounds from the pipeline s solid wall is asymptotically independent of its initial trajectory . that is , the fraction of particles leaving the surface in a given direction is proportional to @xmath0 , where @xmath1 is the angle that such a particle s trajectory defines , measured with respect to the normal to the surface ( @xcite ) . this behaviour is referred to as the * _ knudsen s ( cosine ) law _ * ever since . in the experiment , the gas is injected at a very low pressure so that interactions between particles are negligible . the physical justification of knudsen s law is the following . first of all , one assumes that particles bounce at the pipeline s wall _ elastically_. this means that the energy of a particle is preserved in a collision which , in turn , implies the _ refection law _ : the angle that the incident direction forms with the normal to a surface , @xmath2 , equals the angle @xmath3 of the reflected direction , @xmath4 . it has been proved that this law is valid in a first approximation if we do not take into account thermal effects . since we also assume that particles do not interact with one another , we must conclude that the microscopic irregularities on the pipeline s surface are responsible for the destruction of any particular pattern in the original gas distribution . indeed , even if we assume that the bounces are perfectly elastic , microscopic holes in the boundary of the pipeline and imperfections in relief are dimensionally comparable to the molecules of the gas and have therefore a disruptive ( i.e. , unpredictable ) effect on particle collisions : after many bounces , we are in the so - called _ knudsen s regime _ in which the reflected direction is independent of the incident one . however , this argument does not explain by itself why the reflected angles are distributed according to the _ cosine law_. the theory of _ billiards _ helps clarify this point ( @xcite , @xcite ) . the irregularities of the pipeline surface can be reasonably modelled as cavities or microscopic cells with a dispersive geometry such that , once a particle has entered one of them , it comes out of it with a rather arbitrary direction , even if all the collisions inside that cavity are elastic . if the pipeline is made of a uniform material , we can model one such cell as a billiard table and the pipeline wall as an infinite row of such billiard tables that a particle moving freely enters and exits . with this description , knudsen s cosine law can be then seen as a consequence of the fact that , for sufficiently dispersive billiard geometries , the liouville measure is the unique measure preserved by the billiard flow ( see @xcite for more details ) . unfortunately , this need not be the case for polygonal ( i.e. , non - dispersive ) tables such as fig . [ fig pipeline ] . so polygonal geometries require a slightly different approach . since the characteristic size of this cell is infinitely small compared with the diameter of the pipeline , the exact position along the open side of the cell ( dotted line in fig . [ fig pipeline ] ) at which the particle enters a cell coming from the previous one is considered to be randomly distributed with uniform probability . following @xcite , we will refer to these billiards as _ random billiards_. more concretely , the dynamics of a particle bouncing inside the pipeline are given by the _ first return map _ to the open side of a the billiard cell , which defines a markov process . thus , the dynamics are characterised by a transition operator @xmath5 such that , for any reflected direction @xmath6 $ ] , @xmath7 is the probability that a molecule takes a direction in @xmath8 $ ] after the next rebound . these concepts will be reviewed in sections [ section preliminaries ] and [ section random billiards ] . in this formalism , knudsen s law can be mathematically written as follows . let @xmath9 be the initial angle distribution with which the gas is injected into the pipeline . that is , if @xmath8 $ ] , @xmath10 represents the proportion of particles that will hit the pipeline s wall for the first time with an incident angle @xmath11 . denote by @xmath12 the distribution after @xmath13 collisions . in this context , knudsen s cosine law can be expressed in two different ways . the * _ strong knudsen s law _ * claims that , for almost any initial distribution @xmath9 , after many collisions @xmath12 converges to the cosine distribution , @xmath14 . \label{eq strong law}\ ] ] on the other hand , the * _ weak knudsen s law _ * states that if , after any collision , we count the number of particles reflected with a given angle @xmath1 and then divide by the total number of particles , this quantity is proportional to @xmath15 . explicitly,@xmath16 obviously ( [ eq strong law ] ) implies ( [ eq weak law ] ) , so the strong knudsen s law implies the weak one . deterministic dynamical systems can be sometimes understood as idealizations of _ real _ systems , the latter usually subjected to negligible perturbations in a first approximation . the use of a dynamical approach to study a gas is one such example . in this paper , we will study the properties of the random billiard obtained as a result of modelling the rough surface of the pipeline by a zig - zag geometry as shown in fig . [ fig pipeline ] . our billiard cell is an isosceles triangle with open side @xmath17 defined by an angle @xmath18 . this model was introduced in @xcite but has not been studied in depth yet . for example , in @xcite nothing is said about the ergodic properties of this billiard table when @xmath19 ( it is shown that it is not ergodic under the liouville measure if @xmath20 though ) . thus it is not at all clear if the dynamics resulting from a random billiard as in fig . [ fig billiard table ] will in fact follow knudsen s law ( weak or strong ) . in this paper , we will prove that it is actually the case . more concretely , the main contributions of our paper are the following : 1 . little is known about the ergodicity of the flow of deterministic billiards given by irrational polygon tables ( @xcite ) . remarkably , we show that , for a concrete example , randomising the billiard in a sensible way ( i.e. , choosing the point at which the particle enters the billiard randomly but uniformly distributed on one side of the table ) implies its ergodicity and exactness with respect to the liouville measure . 2 . unlike other attempts to prove ( weak ) knudsen s law , where one usually assumes very irregular geometries at a microscopic level responsible for the dispersive effects , we deal with an extremely simple pattern . this has additional advantages as a simpler model capturing the main features of a system can be simulated and developed more easily . 3 . in the literature , ergodicity of random billiards relies on the ergodicity of the first return map ( see @xcite ) . we use here a different approach and take advantage of a skew - type representation of random maps recently introduced in @xcite . the techniques used to prove the exactness of our model can be certainly adapted to study other ( random ) billiard tables and random maps in general . finally , we prove that the strong knudsen s law holds for our model . as we mentioned before , even if one assumes dispersive billiard geometries , one will only prove that the first return map is ergodic , i.e. , the weak law . the paper is structured as follows : in section [ section model ] we introduce the random billiard behind our model . we recall some basic definitions and properties of billiards in section [ section preliminaries ] and the concept of random billiard in section [ section random billiards ] . in section [ section random maps ] , we review the skew - type representation of random maps introduced in @xcite and show the relationship between this representation and the dynamical evolution of absolutely continuous measures . we prove that this skew - type representation is exact in section [ section exactness of s ] , which uses some auxiliary results presented separately in section [ section ergodicity of ss ] . finally , in section [ section knudsen s law ] , we illustrate our results with numerical simulations . suppose we have a gas of non - interacting particles moving freely inside a pipeline . our model consists of a two dimensional infinite pipeline whose rough walls are modelled as a sequence of cells built as a juxtaposition of a _ fundamental cell _ ( fig . [ fig billiard table ] ) . in this section , we are going to fix the geometry of such cell and introduce the main hypothesis behind the dynamics of the particles . the content of this section is extracted from @xcite . both the geometry and dynamics of our model are summarised in figure [ fig pipeline ] . we assume that the walls of the pipeline are not smooth but describe a serrated regular pattern built from a fixed isosceles triangle with an open side . we characterise that triangle by one of its angles @xmath21 . the dynamics of a particle moving freely inside the pipeline are as follows : the particle enters one of the cells with some angle @xmath2 and bounces on its sides elastically , that is , according to the reflection law . the particle eventually comes out of the cell with a given direction @xmath22 , then crosses the pipeline , and reaches another cell on the opposite wall , and the process recurs . since the diameter of the pipeline is several orders of magnitude bigger than the characteristic length @xmath17 of the cell , it is plausible to think that , every time the particle enters a new cell through the open side , it does so at a point uniformly distributed on @xmath17 . in the literature , one encounters examples of particle dynamics on bounded domains where the random perturbation is introduced in the observations of @xmath22 every time the particles bounces off the wall ( see for example @xcite ) , which implies that collisions are no longer elastic . this perturbation is explained as a consequence of the roughness of the wall . observe that , in our model , randomness is introduced in a completely different way . we assume elastic collisions but , instead of considering dispersive geometries responsible for the random bounces , the noise is introduced in a sensible way without modifying the ( rather elementary ) geometry of the problem . the trajectory that a particle describes after it comes out of a billiard cell until it enters another one on the opposite wall is completely irrelevant for dynamical purposes . in practice , we can better understand our dynamical system as the closed billiard table obtained by fixing together two triangular cells so that they share the same open side as in figure [ fig billiard table ] . when the particle crosses that open side , it retains its direction but it is assigned a different position on @xmath17 according to a uniform law . at a uniformly distributed random point.,width=377,height=181 ] the * _ incoming angle _ * @xmath2 ( resp . * _ outgoing angle _ * ) is then the angle in @xmath23 $ ] with which the particle enters ( resp . leaves ) the cell measured with respect to the normal to the open side @xmath17 . one often shifts @xmath23 $ ] by @xmath24 and measure @xmath1 in @xmath25 $ ] , i.e. with respect to horizontal . in @xcite , for any incoming angle @xmath26 $ ] , feres gives the different possible directions with which a particle may come out of the fundamental cell ( fig . [ fig billiard table ] ) . they are four , given by maps @xmath27\rightarrow\mathbb{r}$ ] defined as@xmath28 and any @xmath29 applies with certain probability @xmath30 . the probabilities associated to the maps ( [ eq tau maps ] ) are as follows:@xmath31{ll}1 & \theta\in\lbrack0,\alpha)\\ u_{\alpha}\left ( \theta\right ) & \theta\in\lbrack\alpha,\pi-3\alpha)\\ 2\cos(2\alpha)u_{2\alpha}(\theta ) & \theta\in\lbrack\pi-3\alpha,\pi-2\alpha)\\ 0 & \theta\in\lbrack\pi-2\alpha,\pi ] \end{array } \right . \smallskip\nonumber\\ p_{2}\left ( \theta\right ) & = \left\ { \begin{array } [ c]{ll}0 & \theta\in\lbrack0,\pi-3\alpha)\\ u_{\alpha}\left ( \theta\right ) -2\cos(2\alpha)u_{2\alpha}(\theta ) & \theta\in\lbrack\pi-3\alpha,\pi-2\alpha)\\ u_{\alpha}\left ( \theta\right ) & \theta\in\lbrack\pi-2\alpha,\pi-\alpha)\\ 0 & \theta\in\lbrack\pi-\alpha,\pi ] \end{array } \right . \smallskip\nonumber\\ p_{3}\left ( \theta\right ) & = \left\ { \begin{array } [ c]{ll}0 & \theta\in\lbrack0,2\alpha)\\ 2\cos(2\alpha)u_{2\alpha}(-\theta ) & \theta\in\lbrack2\alpha,3\alpha)\\ u_{\alpha}\left ( -\theta\right ) & \theta\in\lbrack3\alpha,\pi-\alpha)\\ 1 & \theta\in\lbrack\pi-\alpha,\pi ] \end{array } \right . \smallskip\nonumber\\ p_{4}\left ( \theta\right ) & = \left\ { \begin{array } [ c]{ll}0 & \theta\in\lbrack0,\alpha)\\ u_{\alpha}\left ( -\theta\right ) & \theta\in\lbrack\alpha,2\alpha)\\ u_{\alpha}\left ( -\theta\right ) -2\cos(2\alpha)u_{2\alpha}(-\theta ) & \theta\in\lbrack2\alpha,3\alpha)\\ 0 & \theta\in\lbrack3\alpha,\pi ] , \end{array } \right . \tag{m2}\label{eq probabilities model}\ ] ] where@xmath32 and @xmath18 is assumed to be smaller than @xmath33 . as we will see later , this family of maps and probabilities is all we need to prove knudsen s law . this section aims at recalling the main concepts of the theory of ( deterministic ) billiards and one of its key results , namely , that the _ billiard map _ preserves the liouville measure . the content of this section was extracted from @xcite , which the reader is referred to for an exhaustive exposition on billiards . a * _ billiard table _ * @xmath34 is the closure of a bounded open domain @xmath35 whose boundary @xmath36 is a finite union of smooth compact curves ( @xmath37 , @xmath38 ) . @xmath39 are called the * _ walls _ * of the billiard table and are assumed to intersect each other only at their endpoints or * _ corners_*. for any @xmath40 , @xmath41 is defined by a @xmath37 map @xmath42\subset \mathbb{r}\rightarrow\mathbb{r}^{2}$ ] such that the second derivative @xmath43 either vanishes or is identically zero . this condition prevents the wall @xmath41 from having inflection points or line segments . a wall such that @xmath44 is called * _ focusing _ * or * _ dispersing _ * if @xmath45 points inwards or outwards @xmath34 , respectively . otherwise @xmath46 and the wall is called * _ flat_*. in this framework , we consider the dynamics of a point particle moving freely within @xmath34 . let @xmath47 denote the position of a free particle moving within a billiard table . we suppose that the particle bounces at @xmath48 * _ elastically_*. that is , if it collides with the wall @xmath49 at a point @xmath50 that is not a corner , the incident angle @xmath2 that the velocity @xmath51 forms with the normal vector @xmath13 to @xmath49 at @xmath50 equals the angle of reflection @xmath3 between @xmath13 and @xmath51 after the collision , i.e. @xmath4 . this implies that @xmath51 has constant norm , which we will assume equal to @xmath52 for the sake of simplicity . under these assumptions , the * _ billiard flow _ * @xmath53 is the flow that this particle defines on the * _ phase space _ * @xmath54 . since the particle follows a straight trajectory between two consecutive collisions , the dynamical information of the system is contained in the geometry of the boundary @xmath48 and how the particle bounces off it . therefore , instead of studying the billiard flow , first one introduces the cross - section @xmath55 $ ] as the set of all postcollisional velocity vectors , where @xmath56 $ ] measures the angle between @xmath51 and the normal vector @xmath13 pointing to the interior of @xmath34 , and then considers the first return map @xmath57 that the billiard flow induces on @xmath58 . that is , given a point of @xmath58 , @xmath57 gives the next position at where the particle collides and its velocity after that collision . the map @xmath57 is often called the * _ billiard map_*. it is a well known result in the theory of billiards that the billiard map preserves the * _ liouville measure _ * @xmath59 on @xmath58 , where @xmath60 is the lebesgue measure on @xmath48 and @xmath61 is the measure on @xmath23 $ ] given by @xmath62 , @xmath63\right ) $ ] ( ( * ? ? ? * ; * ? ? ? * lemma 2.35 ) ) . however , except for a few general results , little is known about the ergodicity of billiard maps with respect to the liouville measure for explicit examples of boundary geometries . for example , dispersive billiards are ergodic ( ( * ? ? ? * ; * ? ? ? * theorem 6.20 ) ) while regular polygons are not . finally , as we pointed out in section [ section model ] , it is customary in the literature to translate @xmath23 $ ] and measure @xmath1 in @xmath64 $ ] , in which case @xmath61 is given by @xmath65 , @xmath66\right ) $ ] . we will follow this convention throughout the paper . in this section we are going to get back to the concept of _ random billiard _ and its different representations . let @xmath34 be a billiard table with boundary @xmath67 . remove one of the walls of @xmath48 , for example @xmath68 , so that we obtain a billiard table with an _ open side_. without loss of generality , we can assume that the open side is just a line segment @xmath17 as in figure [ fig billiard table ] . this identification can be carried out by the map @xmath69\subset\mathbb{r}\rightarrow\mathbb{r}^{2}$ ] defining @xmath68 . on the other hand , suppose a particle enters the billiard from the open side at @xmath70 with an angle @xmath71 $ ] with respect to @xmath68 . denote by @xmath72 $ ] the angle associated to the velocity of that particle when it returns to the open side for the first time . the * _ first return map _ * ( to the open side ) is then the map @xmath73\rightarrow \overline{pq}\times\lbrack0,\pi]$ ] induced from the billiard map @xmath57 such that @xmath74 , i.e. , the particle leaves the billiard at @xmath75 with an angle @xmath76 . the liouville measure @xmath59 on @xmath77 $ ] induces a measure on @xmath78 $ ] by restriction in a natural way . we will continue referring to this measure as the liouville measure and denoting it by @xmath59 , where @xmath60 now stands for the lebesgue measure on @xmath17 . it can be proved that the first return map @xmath79 also preserves the liouville measure ( ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * exercise i.5.1 ) ) . furthermore , [ lemma ergodicity first return]if the billiard map @xmath80\rightarrow\partial d\times\lbrack0,\pi]$ ] is ergodic with respect to @xmath59 , so is the first return map @xmath81\rightarrow\overline{pq}\times\lbrack0,\pi]$ ] . let @xmath82)$ ] be an invariant set , i.e. , @xmath83 . this set can be naturally regarded as a measurable set of @xmath77 $ ] since @xmath84 is identified with the wall @xmath68 . since @xmath57 is ergodic , @xmath85 has full @xmath86}$]-measure , which means that@xmath87}(({\textstyle\bigcup\nolimits_{n\geq1 } } f^{-n}\left ( a\right ) ) \bigcap\left ( \overline{pq}\times\lbrack 0,\pi]\right ) ) = 1.\ ] ] if @xmath88\right ) $ ] , then @xmath89 for some @xmath90 and , from the very definition of the first return map that , @xmath91 for some @xmath92 . therefore,@xmath93\right ) = \bigcup \nolimits_{m\geq1}t^{-m}\left ( a\right ) = a,\ ] ] where the last equality follows from the invariance of @xmath94 . consequently , @xmath95}(a)=1 $ ] and @xmath79 is ergodic . let @xmath96 be the probability space built upon @xmath17 with the normalised lebesgue measure @xmath60 on the borel @xmath97-algebra @xmath98 . the * _ random billiard _ * associated to @xmath34 is then the time - discrete random dynamical system @xmath99\rightarrow \lbrack0,\pi]~|~x\in\overline{pq}\right\ } $ ] , where @xmath17 is regarded as a probability space . roughly speaking , the random billiard thus built models the outgoing state a free particle that entered the billiard through the open side at a point uniformly distributed along @xmath17 . when regarded as a probability space , @xmath17 will be denoted by @xmath100 . the random billiard @xmath99\rightarrow\lbrack0,\pi ] ~|~x\in\omega\right\ } $ ] defines a transition probability kernel @xmath101\times\mathcal{b}\left ( [ 0,\pi]\right ) \rightarrow\lbrack0,1]$ ] given by@xmath102\text { and } a\in\mathcal{b}\left ( [ 0,\pi]\right ) . \label{eq probability kernel}\ ] ] we say that @xmath9 is * _ invariant _ * with respect to @xmath103\rightarrow\lbrack0,\pi]~|~x\in\omega\right\ } $ ] if @xmath104}k(\theta , a)d\nu(\theta)\text { for any } a\in \mathcal{b}\left ( [ 0,\pi]\right)\ ] ] and that the random billiard is * _ ergodic _ * with respect to @xmath9 if @xmath79 is ergodic with respect to @xmath105 . for example , @xmath106 , @xmath107\right ) $ ] , is an invariant measure because @xmath59 is invariant with respect to the first return map @xmath79 ( ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * proposition 2.1 ) ) . moreover , since dispersive billiards are ergodic with respect to the liouville measure , the corresponding random billiard is also ergodic with respect to @xmath61 ( lemma [ lemma ergodicity first return ] ) . in this situation , birkhoff s ergodic theorem implies that @xmath108 where @xmath109\rightarrow\lbrack0,\pi]$ ] is the projection onto the second factor . that is , the average number of particles reflected with some angle @xmath11 converges to @xmath110 so ( weak ) knudsen s law holds for dispersive billiards . in general , for non - dispersive billiards , little is know about the ergodicity of the first return map . this is indeed the case of the random billiard introduced in section [ section model ] ( fig . [ fig billiard table ] ) . recall that , as far as we know , determining whether a general irrational triangular billiard is ergodic is still an open problem ( ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * section 7 ) ) and only a few particular examples were proved to be ergodic ( @xcite ) . therefore , at this point , it is not clear at all if knudsen s law holds for the pipeline model introduced in section [ section model ] . to prove it , the first return map must be replaced with a more convenient representation . given a random billiard , the probability kernel ( [ eq probability kernel ] ) contains all the dynamical information of the system and determines its properties . this implies that , as far as the dynamics is concerned , the underlying probability space @xmath100 ( which in section [ section random billiards ] was equal to @xmath96 ) plays a secondary role . in particular , other probability spaces @xmath111 giving rise to the same probability kernel are available . indeed , if @xmath112 $ ] is a transition probability kernel on a measurable space @xmath113 , kolmogorov s existence theorem on markov processes guarantees that there exists a probability @xmath114 defined on the borel @xmath97-algebra of the topological space @xmath115 , @xmath116 for all @xmath117 , such that the chain @xmath118 is markovian and has transition probability kernel @xmath119 ( ( * ? ? ? * ; * ? ? ? * theorem 2.11 ) ) , where @xmath120 is the projection onto the @xmath121th factor . in this statement , @xmath122 has to be a @xmath97-compact hausdorff space and @xmath123 the borel @xmath97-algebra . unfortunately , @xmath124 is not very manageable to work with explicitly . for instance , the representation @xmath73\rightarrow\overline { pq}\times\lbrack0,\pi]$ ] given by the first return map introduced in section [ section random billiards ] only involves finite dimensional spaces and hence seems more convenient . we are now going to consider another skew - type representation for random maps recently introduced in @xcite . as we will show in section [ section exactness of s ] , this representation will be crucial to prove the asymptotic properties of the random map defined by our random billiard fig . [ fig billiard table ] . let @xmath125 be a general measure space and let @xmath126 be the unit interval regarded as a probability space , where @xmath60 stands for the lebesgue measure . later on , we will apply the results of this section to @xmath127 $ ] . for any @xmath128 , let @xmath129 and @xmath130 $ ] be measurable mappings such that @xmath131 is a measurable partition of the unity , i.e. , @xmath132 for any @xmath133 ( compare to ( [ eq probabilities model ] ) and ( [ eq tau maps ] ) in section [ section model ] ) . we define the * _ random dynamical system _ * @xmath134 such that @xmath135 with probability @xmath136 . the transition probability kernel of @xmath137 is given by@xmath138 this probability kernel defines the evolution of an initial distribution ( probability measure ) @xmath9 on @xmath113 under the random map @xmath137 iteratively as@xmath139}k(\theta , a)d\nu^{(n)}(\theta),~~n\geq1,\label{eq 15}\ ] ] where @xmath140 . a measure @xmath9 is called * _ _ * invariant if @xmath141 . following @xcite , consider now @xmath142 and set @xmath143 such that @xmath144 is a finite measurable partition of @xmath100 . we define the * _ skew - type representation _ * of the random map @xmath137 as the map @xmath145 such that@xmath146 where@xmath147 the map @xmath148 is well defined because if @xmath149 then @xmath150 necessarily . moreover , @xmath151 thus defined is @xmath152-measurable ( ( * ? ? ? * lemma 3.1 ) ) and it is a skew - type representation of the random dynamical system @xmath137 . as we will see , the map @xmath151 is extremely useful to study the properties of @xmath137 . for example , @xmath9 is an invariant measure of @xmath137 on @xmath153 if and only if @xmath105 is an invariant measure of @xmath151 on @xmath154 ( ( * ? ? ? * lemma 3.2 ) ) . moreover , since we know that the sine @xmath65 is invariant by the transition probability kernel ( [ eq 16 ] ) of our billiard table , we conclude that @xmath61 is invariant by @xmath137 built from ( [ eq probabilities model ] ) and ( [ eq tau maps ] ) , and therefore @xmath59 is invariant by the corresponding map @xmath151 . in an abuse of terminology , we will continue calling @xmath59 the * _ liouville measure when referring to _ * @xmath151 . in this context , we will say that the random map @xmath137 with initial distribution @xmath9 is * _ ergodic _ * ( resp . * _ mixing _ * , * _ exact _ * ) if @xmath151 is ergodic ( resp . mixing , exact ) with respect to @xmath105 . in section [ section exactness of s ] , we are going to prove that the @xmath151 associated to ( [ eq probabilities model ] ) and ( [ eq tau maps ] ) is exact . [ h ] s.jpg unlike the first return map @xmath79 ( section [ section random billiards ] ) , the map @xmath151 has no dynamical interpretation and is a purely auxiliary tool to represent and study @xmath137 . therefore , we need to establish the relationship between @xmath151 and the transition probability kernel ( [ eq 16 ] ) , which carries all the dynamical information . this is the content of theorem [ theorem 1 ] . before stating that relationship , we need to introduce some notation and an auxiliary lemma , whose proof is included separately in the appendix for the sake of a clearer exposition . [ definition slices]we define iteratively the sets @xmath155 , where @xmath156 for any @xmath157 , as@xmath158 [ lemma slices go to zero]let @xmath159 the fibre through @xmath133 . if @xmath160 , then@xmath161 [ theorem 1]let @xmath9 be a measure on @xmath113 and @xmath140 . then,@xmath162 : = \int_{[0,1)\times x}\mathbf{1}_{a}\left ( \pi_{2}\circ s^{n}\right ) d\left ( \lambda\otimes\nu\right ) = \nu^{(n)}\left ( a\right ) .\ ] ] first of all , it is not difficult to check that@xmath163 this can be proved iteratively rewriting @xmath119 in terms of ( auxiliary ) dirac deltas , @xmath164 where @xmath165 , @xmath166 . on the other hand , since @xmath167 is the disjoint union of the sets @xmath168 , @xmath156,@xmath169 where , in the second line , we have used that @xmath170 on the set @xmath168 . if now we apply fubini s theorem,@xmath171 but by lemma [ lemma slices go to zero ] , for any fixed @xmath133,@xmath172 therefore,@xmath173 [ corollary koopman]let @xmath174 , @xmath175 , be the * koopman operator * associated to @xmath151 , i.e. , @xmath176 . let @xmath177 be the projection onto the second factor . if @xmath9 is a measure on @xmath153 , then@xmath178}u_{s}^{n}\left ( 1_{\pi_{2}^{-1}(a)}\right ) d\left ( \lambda\otimes\nu\right ) , ~a\in \mathcal{b},~n\in\mathbb{n}.\ ] ] let @xmath151 be the skew - type representation of the random billiard introduced in section [ section model ] and let @xmath179 be a probability measure absolutely continuous with respect to the liouville measure @xmath61 with radon - nikodym derivative @xmath180,\mu\right ) $ ] . one can easily check that @xmath181 , where @xmath182 and @xmath183 $ ] . if @xmath151 is mixing ( or exact ) then , from corollary [ corollary koopman ] , @xmath184 ( ( * ? ? ? * proposition 4.4.1 ( b ) ) ) , which implies the strong knudsen s law . we will prove that @xmath151 is exact in section [ section exactness of s ] and give more details about the strong law in section [ section knudsen s law ] . the proof uses the fact that the pull - back @xmath185 of a characteristic function @xmath186 , @xmath187 \right ) $ ] , can not be invariant by @xmath188 ( section [ section ergodicity of ss ] ) . it is worth observing that the strong law , unlike the standard approach to random billiards available in the literature from the first return map ( see [ eq 18 ] ) , is a consequence of the asymptotic properties of the skew - type representation @xmath151 . let @xmath151 be the skew - type representation of the random map ( [ eq tau maps ] ) and ( [ eq probabilities model ] ) . in this section , we are going to show that @xmath190\rightarrow\lbrack0,1)\times \lbrack0,\pi]$ ] can not leave invariant any characteristic function @xmath191 , where @xmath192 \right ) $ ] has probability @xmath193 and @xmath177 denotes the projection onto the second factor . this result will be used in the next section to prove that @xmath151 is exact . first , we will show that our model exhibits a very useful symmetry . [ prop symmetry]let @xmath151 be the skew - type representation of the random map ( [ eq tau maps ] ) and ( [ eq probabilities model ] ) . let @xmath194\rightarrow\mathbb{r}$ ] be a function such that @xmath195 is @xmath189-invariant , i.e.,@xmath196 and let @xmath197\rightarrow\lbrack0,\pi]$ ] be defined by @xmath198 . then @xmath199 is also @xmath189-invariant . it is not difficult to realise from ( [ eq probabilities model ] ) that@xmath200 just using that the the function @xmath201 introduced in ( [ eq definition function u ] ) has period @xmath202 . furthermore , straightforward computations shows that@xmath203 given an index @xmath204 , we define its conjugate index @xmath205 in the following manner:@xmath206 on the other hand , ( [ eq 13 ] ) is equivalent to@xmath207 we have to check that this property holds for @xmath208 . so let @xmath209 $ ] such that @xmath210 . since @xmath211 and @xmath212 for any @xmath213 , we have that @xmath214 implies@xmath215{r}p_{i}\left ( x\right ) = p_{\widetilde{i}}\circ\phi\left ( x\right ) > 0\\ p_{j}\left ( \tau_{i}\left ( x\right ) \right ) = p_{\widetilde{j}}\circ \phi\circ\tau_{i}\left ( x\right ) = p_{\widetilde{j}}\circ\tau_{_{\widetilde { j}}}\circ\phi\left ( x\right ) > 0 \end{array } \right\ } \longrightarrow(p_{\widetilde{i}}\cdot(p_{\widetilde{j}}\circ \tau_{_{\widetilde{j}}}))\circ\phi\left ( x\right ) > 0.\ ] ] therefore , using the invariance of @xmath216 expressed in ( [ eq 34])@xmath217 so @xmath208 also satisfies ( [ eq 34 ] ) and is @xmath189-invariant . [ prop g invariant s2 and phi]let @xmath218,\mu\right ) $ ] and suppose that @xmath195 is invariant by @xmath189 and @xmath219 . if @xmath18 in ( [ eq probabilities model ] ) and ( [ eq tau maps ] ) is irrational , then @xmath216 is constant . let @xmath220\right/ _ { 0\sim\pi}$ ] be the quotient space obtained by identifying @xmath221 . this space is homeomorphic to the unit circle @xmath222 . the function @xmath223\rightarrow\mathbb{r}$ ] induces a function @xmath224:\left . [ 0,\pi]\right/ _ { 0\sim\pi}-\{[0]\}\rightarrow \mathbb{r}$ ] on the quotient that is well defined except at @xmath225 $ ] , the equivalence class of @xmath226 , because maybe @xmath227 . we are going to prove that @xmath224 $ ] is invariant by an irrational rotation . since constants are the only integrable functions invariant by irrational rotations , @xmath224 $ ] must be constant and @xmath216 too . recall that @xmath228 reads as@xmath229 we are going split the interval @xmath230 $ ] into subintervals and , for each of them , we will explore the invariance properties of @xmath216 implied by ( [ eq 34 ] ) . ( i ) : : if @xmath231 , then @xmath232 so@xmath233 ( ii ) : : if @xmath234 , then @xmath235 . writing @xmath236 with @xmath237,@xmath238 now , since @xmath216 is @xmath219-invariant,@xmath239 therefore,@xmath240 \left ( \left [ x\right ] \right ) = \left [ g\right ] \left ( \left [ x+6\alpha\right ] \right ) \text { if } x\in\lbrack\pi -4\alpha,\pi-3\alpha).\ ] ] ( iii ) : : if @xmath241 then @xmath242 . writing @xmath243 with @xmath237,@xmath244 using @xmath245,@xmath246 that is,@xmath240 \left ( \left [ x\right ] \right ) = \left [ g\right ] \left ( \left [ x+6\alpha\right ] \right ) \text { if } x\in\lbrack\pi -3\alpha,\pi-2\alpha).\ ] ] ( iv ) : : suppose @xmath247 , @xmath248 with @xmath237 . since @xmath249 on @xmath250 , by ( [ eq 34]),@xmath251 since @xmath216 is invariant by @xmath219,@xmath252 summarising , @xmath253 which implies@xmath254\left ( [ x]\right ) = [ g]\left ( [ x+2\alpha]\right ) \text { if } x\in\lbrack\pi-2\alpha,\pi-\alpha).\ ] ] now @xmath237 and we can use the @xmath189-invariance expressed in the first item so that@xmath240 \left ( \left [ x\right ] \right ) = \left [ g\right ] \left ( \left [ x+2\alpha\right ] \right ) = \left [ g\right ] \left ( \left [ x+6\alpha\right ] \right ) .\ ] ] ( v ) : : finally , let @xmath255 , @xmath256 with @xmath237 . since @xmath257 on @xmath258@xmath259 by ( [ eq 34 ] ) . using the @xmath219-invariance of @xmath216,@xmath260 that is , @xmath261 , which again implies@xmath254\left ( [ x]\right ) = [ g]\left ( [ x+2\alpha]\right ) \text { if } x\in\lbrack\pi-\alpha,\pi).\ ] ] now @xmath262 and @xmath263 as by assumption @xmath264 ; we are in the situation of ( i ) so@xmath240 \left ( \left [ x\right ] \right ) = \left [ g\right ] \left ( \left [ x+2\alpha\right ] \right ) = \left [ g\right ] \left ( \left [ x+6\alpha\right ] \right ) .\ ] ] in conclusion , if @xmath216 is invariant by @xmath189 and @xmath219 , then@xmath240 \left ( \left [ x\right ] \right ) = \left\ { \begin{array } [ c]{l}\left [ g\right ] \left ( \left [ x+4\alpha\right ] \right ) \text { \ if } x\in(0,\pi-4\alpha)\\ \left [ g\right ] \left ( \left [ x+6\alpha\right ] \right ) \text { \ if } x\in\lbrack\pi-4\alpha,\pi ) . \end{array } \right . \label{eq 31}\ ] ] let @xmath265 such that @xmath266 . observe that @xmath267 because @xmath264 . applying the invariance expressed in ( [ eq 31 ] ) iteratively , we have@xmath240 \left ( \left [ x\right ] \right ) = \left [ g\right ] \left ( \left [ x+\left ( 4k+6\right ) \alpha\right ] \right ) \text { for any } x\in\left ( 0,\pi\right ) . \label{eq 39}\ ] ] in other words , @xmath268 $ ] is invariant by the rotation of angle @xmath269 . if @xmath270 is irrational , @xmath216 must be constant . [ propositon characteristic function invariant]let @xmath271\right ) $ ] and suppose that @xmath272 is @xmath189-invariant . if @xmath18 in ( [ eq probabilities model ] ) and ( [ eq tau maps ] ) is irrational , then @xmath273 equals either @xmath52 or @xmath226 . suppose that @xmath193 . we want to prove that @xmath272 can not be @xmath189-invariant . since constants are trivially @xmath189-invariant , we can subtract @xmath273 from @xmath186 so that @xmath274 is still @xmath189-invariant and has expectation @xmath226 . let @xmath275 and define @xmath276 , which is clearly @xmath219-invariant because @xmath277 . since @xmath195 is @xmath151-invariant , so are @xmath278 ( proposition [ prop symmetry ] ) and @xmath279 . by proposition [ prop g invariant s2 and phi ] , @xmath280 is constant and equal to its expectation @xmath281 $ ] . but @xmath282 = \operatorname*{e}\nolimits_{\mu}\left [ g\right ] = 0 $ ] because @xmath283 and @xmath284 are invariant by @xmath219 . therefore , @xmath285 = 0 $ ] and @xmath280 , which is constant , must be equal to @xmath226 . that is,@xmath286 since@xmath287{l}1-\mu\left ( c\right ) > 0\text { if } x\in c\\ -\mu\left ( c\right ) < 0\text { if } x\notin c , \end{array } \right.\ ] ] ( remember that we assumed @xmath193 ) , we conclude from ( [ eq 25 ] ) that @xmath288 and@xmath289{l}g\left ( x\right ) = 1/2\longleftrightarrow x\in c,\\ g(x)=-1/2\longleftrightarrow x\in c^{c}. \end{array}\ ] ] moreover , looking carefully at the proof of proposition [ prop g invariant s2 and phi ] , we have@xmath240 \left ( \left [ x\right ] \right ) = \left\ { \begin{array } [ c]{l}\left [ g\right ] \left ( \left [ x+4\alpha\right ] \right ) \text { \ if } x\in(0,\pi-4\alpha)\\ -\left [ g\right ] \left ( \left [ x+6\alpha\right ] \right ) \text { \ if } x\in\lbrack\pi-4\alpha,\pi ) , \end{array } \right.\ ] ] where instead of @xmath245 we have now used @xmath290 . therefore , ( see ( [ eq 39]))@xmath240 \left ( \left [ x\right ] \right ) = -\left [ g\right ] \left ( \left [ x+\left ( 4k+6\right ) \alpha\right ] \right ) \text { for any } x\in(0,\pi ) , \label{eq 40}\ ] ] where @xmath267 is such that @xmath291 . the minus sign in ( [ eq 40 ] ) tells us that the rotation @xmath292 of angle @xmath293 sends @xmath294 to @xmath295 and vice versa . let @xmath296 and @xmath297 the fourier expansions of @xmath186 and @xmath298 respectively . since @xmath299}$ ] , we have that@xmath300 on the other hand , we already argued that @xmath301 . imposing that the fourier coefficients of @xmath186 and @xmath298 are unique , we deduce that@xmath302 as @xmath303 is irrational , @xmath304 for any @xmath305 , which implies that @xmath306 for any @xmath305 so @xmath186 is constant a.s .. but this is clearly a contradiction because ( [ eq 25 ] ) implied @xmath307 . in this section , we are going to prove that @xmath151 is exact . this will imply knudsen s strong law for the random billiard introduced in section [ section model ] . let @xmath308 be a measurable transformation of a probability space @xmath309 . from the measurability of @xmath79 , we have the chain of @xmath97-algebras@xmath310 the map @xmath79 is called * _ exact _ * if the @xmath97-algebra @xmath311 only contains sets of measure either @xmath226 or @xmath52 . it is not difficult to prove that sets in @xmath312 are characterised by the property @xmath313 this characterisation , in turn , implies that a map @xmath79 is exact if and only if@xmath314 for any @xmath315 such that @xmath316 ( ( * ? ? ? * section 2.2 ) ) . equation ( [ eq 32 ] ) reads at the level of @xmath317 spaces as@xmath318 where @xmath319 and @xmath320 is the * _ koopman operator _ * defined on @xmath317 . if a map is exact , then@xmath321 which means that @xmath322 $ ] in @xmath317 as @xmath323 ( see ( * ? ? ? * section 2.5 ) ) . if @xmath9 is a probability absolutely continuous with respect to @xmath61 with radon - nikodym derivative @xmath324 , then@xmath325 \underset{n\rightarrow \infty}{\longrightarrow}\mu\left ( a\right ) \operatorname*{e}\nolimits_{\mu } \left [ f\right ] = \mu(a),\ ] ] that is , the sequence of measures @xmath326 converges ( weakly ) to @xmath61 for any absolutely continuous measure @xmath179 ( ( * ? ? ? * proposition 4.4.1 ( b ) ) , ( * ? ? ? * section 2.6 ) ) . we want to show that the skew - type representation @xmath327\rightarrow\lbrack0,1)\times\lbrack0,\pi]$ ] of the model introduced in section [ section model ] is exact with respect to the liouville measure to conclude that , for any absolutely continuous measure @xmath179 on @xmath328\right ) $ ] and any @xmath329)$ ] , @xmath330}\mathbf{1}_{a}\left ( \pi_{2}\circ s^{n}\right ) d\nu=\int_{[0,\pi]}u_{s}^{n}\left ( \mathbf{1}_{\pi_{2}^{-1}(a)}\right ) d\lambda\otimes\nu\underset{n\rightarrow\infty}{\longrightarrow } \mu\left ( a\right ) , \label{eq 33}\ ] ] i.e. , that the strong knudsen s law holds . observe from ( [ eq 33 ] ) that we only need to consider the action of the koopman operator @xmath331 on square integrable functions that are the pull - back by @xmath332 of functions in @xmath333,\mathcal{b}\left ( [ 0,\pi]\right ) , \mu\right ) $ ] . we are going to denote @xmath334,\mathcal{b}\left ( [ 0,\pi]\right ) , \mu\right ) $ ] simply by @xmath335\right ) $ ] for the sake of a clearer notation . consequently , we do not need to check that @xmath151 is exact for the borel @xmath336-algebra of @xmath337 $ ] but it is enough to consider a smaller one , namely , the smallest @xmath97-algebra that makes both the functions in @xmath338\right ) \right ) $ ] and @xmath151 measurable . for any @xmath339 , let @xmath340\right ) \right\ } \right ) , \ ] ] where @xmath168 are as in definition [ definition slices ] . for any @xmath341\right ) $ ] , the sets @xmath342 and @xmath343 will be called * generators * ( of @xmath344 ) . the sequence of @xmath97-algebras @xmath345 define a filtration,@xmath346 indeed , for any @xmath341\right ) $ ] and any @xmath347 , a generator @xmath348 can be written as@xmath349 which implies @xmath350 . we define @xmath351 as the limit of this filtration . @xmath352 @xmath353\rightarrow\lbrack0,1)\times\lbrack0,\pi]$ ] is @xmath351-measurable . for any @xmath354\right ) $ ] , @xmath355 is also @xmath351-measurable . the second statement of the proposition is obvious . to prove that @xmath151 is @xmath351-measurable it is enough to show that @xmath356 for the generators of @xmath351 . so let @xmath357 such that @xmath358 where @xmath341\right ) $ ] and @xmath347 . then , by definition of the sets @xmath168@xmath359 which clearly belongs to @xmath351 . the following two lemmas aim at getting a better insight of the structure of the @xmath97-algebra @xmath351 . as an immediate consequence of them , it will be enough to show that @xmath151 is exact by looking at its action on the generators of @xmath351 . [ prop finite union generators]any set @xmath360 can be expressed as a finite union of disjoint generators . first to all , observe that , for a fixed @xmath339 , the sets @xmath168 are disjoint and form a partition of @xmath361 $ ] . the complement of a generator @xmath362 is@xmath363 \backslash u\right ) , \ ] ] where @xmath364 denotes the disjoint union , so it can be expressed as a finite union of generators . on the other hand , if we take a countable union of generators @xmath365 , @xmath366,@xmath367 where @xmath368 is in @xmath369 . but@xmath370 where the indices @xmath371 can only take finite number of possibilities . therefore , a countable union of generators reduces to a finite union . the union @xmath372 of a filtration @xmath373 is an algebra of sets . the @xmath97-algebra generated by an algebra of sets @xmath374 is characterised by containing all the sets in @xmath374 and the limit of all monotone sequence of sets . that is , @xmath375 coincides with the monotone class generated by @xmath374 ( ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * section 1.3 , theorem 1 ) ) . this observation is the key to proving the following lemma : [ lemma a contains generator]if @xmath376 has positive probability , then @xmath94 contains a generator of positive probability . let @xmath377 . if @xmath378 , then @xmath360 for some @xmath339 . by proposition [ prop finite union generators ] , @xmath94 can be expressed as a finite union of generators and at least one of them must have positive probability . if @xmath379 , then there exists a monotone sequence of sets @xmath380 such that @xmath381 . we are going to deal with the cases @xmath382 increasing or decreasing separately . * if @xmath382 is increasing , then @xmath383 . since @xmath384 , at least one @xmath385 must have positive measure , @xmath386 . but @xmath387 for some @xmath388 , so it contains a generator of positive probability . * let @xmath389 be a decreasing sequence such that @xmath390 . suppose that @xmath391 ( otherwise there is nothing to prove ) . we have@xmath392 so that @xmath94 can be expressed as the complement of a countable union of disjoint sets . rename @xmath393 and @xmath394 , and express any @xmath395 as a disjoint union of generators ( lemma [ prop finite union generators]),@xmath396 in this decomposition , we only consider sets of strictly positive probability . then,@xmath397 where @xmath398 . we claim that there exists a generator @xmath399 of strictly positive probability that does not intersect @xmath400 , i.e. , @xmath401 . + let @xmath402 be a sequence appearing in the decomposition ( [ eq 38 ] ) . if @xmath403 , then@xmath404 \right\backslash \pi_{2}^{-1}\left ( \bigcup\nolimits_{m\in i_{i_{1}^{n} ... i_{r}^{n}}}u_{m}\right ) \right ) \bigcap j_{i_{1}^{n} ... i_{r}^{n}}\ ] ] is a generator of strictly positive measure that does not intersect @xmath400 . if @xmath405 for any finite sequence @xmath402 in ( [ eq 38 ] ) , then @xmath406 a.s . and since we assumed that @xmath407 , there must exist some set @xmath408 with positive probability that does not appear in the decomposition of @xmath400 . that is , @xmath409\right ) \bigcap j_{i_{1} ... i_{k}}\subseteq a$ ] . for any @xmath357 , @xmath410 in other words , @xmath353\rightarrow\lbrack0,1)\times \lbrack0,\pi]$ ] is an exact endomorphism of @xmath411,\mathcal{f},\lambda\otimes\mu\right ) $ ] . let @xmath357 be such that @xmath412 . by lemma [ lemma a contains generator ] , @xmath94 contains a generator @xmath413 of positive measure for some @xmath341\right ) $ ] , @xmath414 . it is an immediate consequence of the definitions that , after @xmath13 iterations , the set @xmath415 has all its fibers of length @xmath52 , that is,@xmath416 let @xmath417 and let @xmath159 be the fiber at @xmath209 $ ] . looking carefully at ( [ eq tau maps ] ) and ( [ eq probabilities model ] ) , one can see that@xmath418 where @xmath419 denotes the identity on @xmath25 $ ] , and@xmath420 these remarks imply that , if @xmath421 , @xmath422 since all the fibers of @xmath423 have length @xmath52 , @xmath424 and , iteratively,@xmath425 define @xmath426 . obviously @xmath427 , the fibers of @xmath294 have all length @xmath52 , and @xmath428 . we want to show that @xmath429 . observe that @xmath430 because @xmath431 . from @xmath427 we have@xmath432 but @xmath294 and @xmath433 have the same measure ( @xmath189 is @xmath59-preserving ) , so @xmath186 is a @xmath189-invariant function a.s .. since @xmath434 a.s . ( the fibers of @xmath294 have all length @xmath52 a.s . ) , @xmath435 has full measure by proposition [ propositon characteristic function invariant ] . therefore , @xmath436 and @xmath437 this proves that @xmath189 is exact . in general , we have@xmath438 so that , using again that @xmath151 is @xmath59-preserving,@xmath439 which proves that @xmath440 is increasing and converges to @xmath52 , i.e. , @xmath151 is also exact . to sum up , the skew - type representation @xmath441\rightarrow\lbrack0,1)\times\lbrack0,\pi]$ ] associated to the random map ( [ eq tau maps ] ) and ( [ eq probabilities model ] ) is exact which implies , by ( [ eq 33 ] ) , that @xmath442 as @xmath323 for any initial distribution @xmath9 on @xmath25 $ ] absolutely continuous with respect to the law @xmath443 , @xmath444\right ) $ ] . since @xmath445 on @xmath446 , @xmath61 and the lebesgue measure @xmath60 are absolutely continuous with respect to each other on @xmath447 . in other words , _ strong knudsen s law _ @xmath442 _ holds for any initial distribution absolutely continuous with respect to _ @xmath448__. _ _ in this section , we are going to show that numerical simulations are in accordance with theoretical results . with this aim , we take an arbitrary initial distribution on @xmath230 $ ] such as in the following picture and we make it evolve according to our random system ( [ eq tau maps ] ) and ( [ eq probabilities model ] ) . initial.jpg + initial normalised distribution of particles . [ fig initial distribution ] since we can only simulate a finite number particles in a computer , in this experiment we take a total amount of 30000 balls and divide @xmath449 $ ] in subintervals of the same length ( 45 in our experiment ) . that is , we approximate the initial distribution by a step function . in each subinterval , we put the proportion of balls according to the probability density function above . the initial angle associated to any of those balls is the middle point of the interval where they fall . the simulation then goes as follows . at the @xmath121-_th _ step , we take the @xmath450-_th _ particle with angle @xmath451 and a random number @xmath452 uniformly distributed between @xmath226 and @xmath52 , one for each particle . if @xmath453 , then @xmath454 , where @xmath455 are as in ( [ eq tau maps ] ) , and so on . after a long number of iterations , the distribution stabilises as follows : final.jpg + final normalised distribution of particles . in red , @xmath456 . [ fig final distribution ] as we can see , the outline of the final distribution tends to the graph of @xmath457 . the small inaccuracy is explained by the fact that only a finite ( i.e. , a discrete ) number of initial angles is considered . the smaller the subintervals in which @xmath230 $ ] is divided are , the better the final distribution approximates @xmath457 . we have repeated this experiment over several initial distributions obtaining always similar results , which experimentally confirms the validity of knudsen s law for our model . [ proof of lemma [ lemma slices go to zero]]to start with , will prove that @xmath458 from the very definition of @xmath151 and @xmath155 , we have@xmath459 on the other hand , let @xmath460 . then @xmath461 for some @xmath462 . let @xmath463 . it is obvious that @xmath464 as @xmath465 because @xmath466 is not empty . moreover , @xmath467 . therefore @xmath468 and@xmath469 identifying both fibers @xmath466 and @xmath470 as subsets of @xmath471 , the map@xmath289{rcl}s_{x}:i_{x}\bigcap j_{i_{1}\cdots i_{n}}\subset\lbrack0,1 ) & \longrightarrow & i_{\tau_{i_{n}}(x)}\bigcap j_{i_{1}\cdots i_{n-1}}\\ y & \longmapsto & \varphi_{i_{n}}(y,\tau_{i_{n}}(x))=\frac{1}{p_{i_{n}}(x)}\left ( y-{\textstyle\sum\nolimits_{j=1}^{i_{n}-1 } } p_{j}(x)\right ) , \end{array } \label{eq 37}\ ] ] is then a diffeomorphism . after having identified @xmath472 as a subset of @xmath471 , we can compute its lebesgue measure @xmath473 . consequently,@xmath474 applying iteratively ( [ eq 19 ] ) , we obtain@xmath475 n. chernov and r. markarian . _ introduction to the ergodic theory of chaotic billiards . _ monografas del instituto de matemtica y ciencias afines * 19*. instituto de matemtica y ciencias afines ; pontificia universidad catlica del per , lima , 2001 . |
we consider the problem of minimizing a function , which is the sum of a linear function and a composition of a strongly convex function with a linear transformation , over a compact polyhedral set . jaggi and lacoste - julien showed that the conditional gradient method with away steps employed on the aforementioned problem without the additional linear term has linear rate of convergence , depending on the so - called pyramidal width of the feasible set . we revisit this result and provide a variant of the algorithm and an analysis that is based on simple duality arguments , as well as corresponding error bounds . this new analysis ( a ) enables the incorporation of the additional linear term , ( b ) does not require a linear - oracle that outputs an extreme point of the linear mapping of the feasible set and ( c ) depends on a new constant , termed the vertex - facet distance constant " , which is explicitly expressed in terms of the problem s parameters and the geometry of the feasible set . this constant replaces the pyramidal width , which is difficult to evaluate . | 1504.05002 | consider the minimization problem @xmath0 where @xmath1 is a compact polyhedral set , @xmath2 and @xmath3 is strongly convex and continuously differentiable over @xmath4 . note that for a general matrix @xmath5 , the function @xmath6 is not necessarily strongly convex . when the problem at hand is large - scale , first order methods , which have relatively low computational cost per iteration , are usually utilized . these methods include , for example , the class of projected ( proximal ) gradient methods . a drawback of these methods is that under general convexity assumptions , they posses only a sublinear rate of convergence @xcite , while linear rate of convergence can be established only under additional conditions such as strong convexity of the objective function @xcite . luo and tseng @xcite showed that the strong convexity assumption can be relaxed and replaced by an assumption on the existence of a local error bound , and under this assumption , certain classes algorithms , which they referred to as feasible descent methods " , converge in an asymptotic linear time . the model ( [ eq : problem ] ) with assumptions on strong convexity of @xmath7 , compactness and polyhedrality of @xmath8 was shown in @xcite to satisfy the error bound . in @xcite wang and lin extended the work @xcite and showed that there exists a _ global _ error bound for problem ( [ eq : problem ] ) with the additional assumption of compactness of @xmath8 ; and derived the exact linear rate for this case . we note that the family of feasible descent methods " include the block alternating minimization algorithm ( under the assumption of block strong convexity ) , as well as gradient projection methods , and therefore are usually at least as complex as evaluating the orthogonal projection operator onto the feasible set @xmath8 at each iteration . an alternative to algorithms which are based on projection ( or proximal ) operators are _ linear - oracle_-based algorithms such as the conditional gradient ( cg ) method . the cg algorithm was presented by frank and wolfe in 1956 @xcite , for minimizing a convex function over a compact polyhedral set . at each iteration , the algorithm requires a solution to the problem of minimizing a linear objective function over the feasible set . it is assumed that this solution is obtained by a call to a linear - oracle , i.e. , a black box which , given a linear function , returns an optimal solution of this linear function over the feasible set ( see an exact definition in section [ sec : cg ] ) . in some instances , and specifically for certain types of polyhedral sets , obtaining such a linear - oracle can be done more efficiently than computing the orthogonal projection onto the feasible set ( see examples in @xcite ) , and therefore the cg algorithm has an advantage over projection - based algorithms . the original paper of frank and wolfe also contained a proof of an @xmath9 rate of convergence of the function values to the optimal value . levitin and polyak showed in @xcite that this @xmath9 rate can also be extended to the case where the feasible set is a general compact convex set . cannon and culum proved in @xcite that this rate is in fact _ tight_. however , if in addition to strong convexity of the objective function , the optimal solution is in the interior of the feasible set , then linear rate of convergence of the cg method can be established @xcite . epelman and freund @xcite , as well as beck and teboulle @xcite showed a linear rate of convergence of the conditional gradient with a special stepsize choice in the context of finding a point in the intersection of an affine space and a closed and convex set under a slater - type assumption . another setting in which linear rate of convergence can be derived is when the feasible set is uniformly ( strongly ) convex and the norm of the gradient of the objective function is bounded away from zero @xcite . + another approach for deriving a linear rate of convergence is to modify the algorithm . for example , hazan and garber used _ local _ linear - oracles in @xcite in order to show linear rate of convergence of a localized " version of the conditional gradient method . a different modification , which is viable when the feasible set is a compact polyhedral , is to use a variation of the conditional gradient method that incorporates away steps . this version of the conditional gradient method , which we refer to as _ away steps conditional gradient _ ( ascg ) , was initially suggested by wolfe in @xcite and then studied by guelat and marcotte @xcite , where a linear rate of convergence was established under the assumption that the objective function is strongly convex , as well as an assumption on the location of the optimal solution . in @xcite jaggi and lacoste - julien were able to extend this result for the more general model ( [ eq : problem ] ) for the case where @xmath10 , without restrictions on the location of the solution . we note that the ascg requires that the linear - oracle will produce an optimal solution of the associated problem which is an extreme point . we will call such an oracle a _ vertex linear - oracle _ ( see the discussion in section [ sec : vertexlinearoracles ] ) . * contribution . * in this work , our starting point and main motivation are the results of jaggi and lacoste - julien @xcite . our contribution is threefold : * we extend the results given in @xcite and show that the ascg algorithm converges linearly for the general case of problem , that is , for any value of @xmath5 and @xmath11 . + the additional linear term @xmath12 enables us to consider much more general models . for example , consider the @xmath13-regularized least squares problem @xmath14 where @xmath15 is a compact polyhedral , @xmath16 and @xmath17 . since @xmath18 is compact , we can find a constant @xmath19 for which @xmath20 for any @xmath21 . we can now rewrite the model as @xmath22 } { \left\|{{{\bf b}}{{\bf x}}-{{\bf c}}}\right\|}^2+\lambda y,\ ] ] which obviously fits the general model ( [ eq : problem ] ) * the analysis in @xcite assumes the existence of a _ vertex _ linear - oracle on the set @xmath23 , rather than an oracle for the set @xmath8 . this fact is not significant for the pure " cg algorithm , since it only requires a linear - oracle and not a _ vertex _ linear - oracle . this means that for the cg algorithm , a linear - oracle on @xmath23 can be easily obtained by applying @xmath5 on the output of the linear - oracle on @xmath8 . on the other hand , this argument fails for the ascg algorithm that specifically requires the oracle to return an extreme point of the feasible set , and finding such a vertex linear - oracle on @xmath23 might be a complex task , see section [ sec : vertexlinearoracles ] for more details . our analysis only requires a vertex linear - oracle on the original set @xmath8 . * we present an analysis based on simple duality arguments , which are completely different than the geometric arguments in @xcite . consequently , we obtain a computable constant for the rate of convergence , which is explicitly expressed as a function of the problem s parameters and the geometry of the feasible set . this constant , which we call the vertex - facet distance constant " , replaces the so - called _ pyramidal width _ constant from @xcite , which reflects the geometry of the feasible set and is obtained as the optimal value of a very complex mixed integer saddle point optimization problem whose exact value is unknown even for simple polyhedral sets . * paper layout . * the paper is organized as follows . section [ sec : preliminaries ] presents some preliminary results and definitions needed for the analysis . in particular , it provides a brief introduction to the classical cg algorithm and linear oracles . section [ sec : awaystepconditionalgradient ] presents the ascg algorithm and the convergence analysis , and is divided into four subsections . in section [ sec : vertexlinearoracles ] the concept of vertex linear - oracle , needed for the implementation of ascg , is presented , and the difficulties of obtaining a vertex linear - oracle on a linear transformation of the feasible set are discussed . in section [ sec : ascgmethod ] we present the ascg method with different possible stepsize choices . in section [ sec : rateconvergenceanalysis ] , we provide the rate of convergence analysis of the ascg for problem , and present the new _ vertex - facet distance _ constant used in the analysis . finally , in section [ sec : findingomegaforpolyhedrons ] , we demonstrate how to compute this new constant for a few examples of simple polyhedral sets . * notations . * we denote the cardinality of set @xmath24 by @xmath25 . the difference , union and intersection of two given sets @xmath24 and @xmath26 are denoted by @xmath27 , @xmath28 and @xmath29 respectively . subscript indices represent elements of a vector , while superscript indices represent iterates of the vector , i.e. , @xmath30 is the @xmath31th element of vector @xmath32 , @xmath33 is a vector at iteration @xmath34 , and @xmath35 is the @xmath31th element of @xmath33 . the vector @xmath36 is the @xmath31th vector of the standard basis of @xmath37 , @xmath38 is the all - zeros vector , and @xmath39 is the vector of all ones . given two vectors @xmath40 , their dot product is denoted by @xmath41 . given a matrix @xmath42 and vector @xmath43 , @xmath44 denotes the spectral norm of @xmath45 , and @xmath46 denotes the @xmath47 norm of @xmath32 , unless stated otherwise . @xmath48 , @xmath49 and @xmath50 represent the transpose , rank and image of @xmath45 respectively . we denote the @xmath31th row of a given matrix @xmath45 by @xmath51 , and given a set @xmath52 , @xmath53 is the submatrix of @xmath45 such that @xmath54 for any @xmath55 . if @xmath45 is a symmetric matrix , then @xmath56 is its minimal eigenvalue . if a matrix @xmath45 is also invertible , we denote its inverse by @xmath57 . given matrices @xmath58 and @xmath59 , the matrix @xmath60\in { \mathbb{r}}^{n\times { ( m+k)}}$ ] is their horizontal concatenation . given a point @xmath32 and a closed convex set @xmath8 , the distance between @xmath32 and @xmath8 is denoted by @xmath61 . the standard unit simplex in @xmath37 is denoted by @xmath62 and its relative interior by @xmath63 . given a set @xmath64 , its convex hull is denoted by @xmath65 . given a convex set @xmath66 , the set of all its extreme points is denoted by @xmath67 . we start by presenting two technical lemmas . the first lemma is the well known _ descent lemma _ which is fundamental in convergence rate analysis of first order methods . the second lemma is _ hoffman s lemma _ which is used in various error bound analyses over polyhedral sets . [ lemma : descentlemma ] let @xmath68 be a continuously differentiable function with lipschitz continuous gradient with constant @xmath69 . then for any @xmath70 we have @xmath71 [ lemma : hoffman ] let @xmath8 be a polyhedron defined by @xmath72 , for some @xmath42 and @xmath73 , and let @xmath74 where @xmath75 and @xmath76 . assume that @xmath77 . then , there exists a constant @xmath78 , depending only on @xmath45 and @xmath79 , such that any @xmath80 satisfies @xmath81 a complete and simple proof of this lemma is given in ( * ? ? ? 299 - 301 ) . defining @xmath82 as the set of all matrices constructed by taking linearly independent rows from the matrix @xmath83^t$ ] , we can write @xmath78 as @xmath84 we will refer to @xmath78 as the _ hoffman constant _ associated with matrix @xmath83^t$ ] . throughout the article we make the following assumption regarding problem . [ ass : fx = gex ] 1 . [ ass : f lipschitz ] @xmath6 is continuously differentiable and has a lipschitz continuous gradient with constant @xmath69 . [ ass : g strongly convex ] @xmath7 is strongly convex with parameter @xmath85 . [ ass : xpolyhedron ] @xmath8 is a nonempty compact polyhedral set given by @xmath86 for some @xmath42 , @xmath87 . we denote the optimal solution set of problem by @xmath88 . the diameter of the compact set @xmath8 is denoted by @xmath89 , and the diameter of the set @xmath23 ( the diameter of the image of @xmath8 under the linear mapping associated with matrix @xmath5 ) by @xmath90 . the two diameters satisfy the following relation : @xmath91 we define @xmath92 to be the maximal norm of the gradient of @xmath7 over @xmath23 . problem possesses some properties , which we present in the following lemmas . [ lemma : optimalissubspace ] let @xmath88 be the optimal set of problem . then , there exists a constant vector @xmath93 and a scalar @xmath94 such that any optimal solution @xmath95 satisfies @xmath96 and @xmath97 . although the proof of the lemma in the given reference is for polyhedral sets , the extension for any convex set is trivial . [ lemma : ofupperbound ] let @xmath98 be the optimal value of problem . then , for any @xmath80 @xmath99 where @xmath100 . let @xmath101 be some optimal solution of problem , so that @xmath102 . then for any @xmath80 , it follows from the convexity of @xmath6 that @xmath103 where the last two inequalities are due to the cauchy - schwartz inequality and the definition of @xmath104,@xmath89 and @xmath90 . the following lemma provides an _ error bound _ , i.e. , a bound on the distance of any feasible solution to the optimal set . this error bound will later be used as an alternative to a strong convexity assumption on @xmath6 , which is usually needed in order to prove a linear rate of convergence . this is a different bound than the one given in @xcite , since it relies heavily on the compactness of the set @xmath8 , thus enabling to circumvent the use of the so - called gradient mapping . [ lemma : errbound ] for any @xmath80 , @xmath105 where @xmath106 , and @xmath78 is the hoffman constant associated with matrix @xmath107^t$ ] . lemma [ lemma : optimalissubspace ] implies that the optimal solution set @xmath88 can be defined as @xmath108 where @xmath109 for some @xmath110 and @xmath111 . for any @xmath80 , applying lemma [ lemma : hoffman ] with @xmath112^t$ ] , we have that @xmath113 where @xmath78 is the hoffman constant associated with matrix @xmath107^t$ ] . now , let @xmath80 and @xmath114 . utilizing the @xmath85-strong convexity of @xmath7 , it follows that @xmath115 by the first order optimality conditions for problem , we have ( recalling that @xmath80 and @xmath95 ) @xmath116 therefore , @xmath117 \frac{\sigma_g}{2}{\left\|{{{\bf e}}{{\bf x}}-{{\bf t}}^*}\right\|}^2&\leq { \left\langle{\nabla f({{\bf x}}^*)},{{{\bf x}}-{{\bf x}}^*}\right\rangle}+\frac{\sigma_g}{2}{\left\|{{{\bf e}}{{\bf x}}-{{\bf e}}{{\bf x}}^*}\right\|}^2\\ & = { \left\langle{\nabla g({{\bf e}}{{\bf x}}^*)},{{{\bf e}}{{\bf x}}-{{\bf e}}{{\bf x}}^*}\right\rangle}+{\left\langle{{{\bf b}}},{{{\bf x}}-{{\bf x}}^*}\right\rangle}+\frac{\sigma_g}{2}{\left\|{{{\bf e}}{{\bf x}}-{{\bf e}}{{\bf x}}^*}\right\|}^2 \end{aligned}\ ] ] now , using we can continue to obtain @xmath118 we are left with the task of upper bounding @xmath119 . by the definitions of @xmath94 and @xmath6 we have that @xmath120 { \left\langle{{{\bf b}}},{{{\bf x}}}\right\rangle}-s^*&={\left\langle{{{\bf b}}},{{{\bf x}}-{{\bf x}}^*}\right\rangle}\\ & = { \left\langle{\nabla f({{\bf x}}^*)},{{{\bf x}}-{{\bf x}}^*}\right\rangle}-{\left\langle{\nabla g({{\bf e}}{{\bf x}}^*)},{{{\bf e}}{{\bf x}}-{{\bf e}}{{\bf x}}^*}\right\rangle}\\ & = { \left\langle{\nabla f({{\bf x}}^*)},{{{\bf x}}-{{\bf x}}^*}\right\rangle}- { \left\langle{\nabla g({{\bf t}}^*)},{{{\bf e}}{{\bf x}}-{{\bf t}}^*}\right\rangle}. \end{aligned}\ ] ] therefore , using , as well as the cauchy - schwartz inequality , we can conclude the following : @xmath121 on the other hand , exploiting , the convexity of @xmath6 and the cauchy - schwartz inequality , we also have that @xmath122 { \left\langle{{{\bf b}}},{{{\bf x}}}\right\rangle}-s^*&= { \left\langle{\nabla f({{\bf x}}^*)},{{{\bf x}}-{{\bf x}}^*}\right\rangle}- { \left\langle{\nabla g({{\bf t}}^*)},{{{\bf e}}{{\bf x}}-{{\bf t}}^*}\right\rangle}\\ & \leq f({{\bf x}})-f^*-{\left\langle{\nabla g({{\bf t}}^*)},{{{\bf e}}{{\bf x}}-{{\bf t}}^*}\right\rangle}\\ & \leq f({{\bf x}})-f^*+{\left\|{\nabla g({{\bf t}}^*)}\right\|}{\left\|{{{\bf e}}{{\bf x}}-{{\bf t}}^*}\right\|}. \end{aligned}\ ] ] combining , , and the fact that @xmath123 , we obtain that @xmath124 moreover , the definitions of @xmath104 and @xmath90 imply @xmath125 , @xmath126 , and since @xmath80 , it follows from lemma [ lemma : ofupperbound ] that @xmath127 . utilizing these bounds , as well as to bound results in @xmath128 ( { \left\langle{{{\bf b}}},{{{\bf x}}}\right\rangle}-s^*)^2&\leq \left(f({{\bf x}})-f^*+g{\left\|{{{\bf e}}{{\bf x}}-{{\bf t}}^*}\right\|}\right)^2\\ & = ( f({{\bf x}})-f^*)^2 + 2g{\left\|{{{\bf e}}{{\bf x}}-{{\bf t}}^*}\right\|}(f({{\bf x}})-f^*)+g^2{\left\|{{{\bf e}}{{\bf x}}-{{\bf t}}^*}\right\|}^2\\ & \leq ( f({{\bf x}})-f^*)c+2gd_{{\bf e}}(f({{\bf x}})-f^*)+g^2\frac{2}{\sigma_g}(f({{\bf x}})-f^*)\\ & = ( f({{\bf x}})-f^*)\left(c+2gd_{{\bf e}}+\frac{2g^2}{\sigma_g}\right)\\ & = ( f({{\bf x}})-f^*)\left({\left\|{{{\bf b}}}\right\|}d+3gd_{{\bf e}}+\frac{2g^2}{\sigma_g}\right ) . \end{aligned}\ ] ] plugging and back into , we obtain the desired result : @xmath129 in order to present the cg algorithm , we first define the concept of linear oracles . given a set @xmath8 , an operator @xmath130 is called a * linear oracle * for @xmath8 , if for each @xmath131 it returns a vector @xmath132 such that @xmath133 for any @xmath80 , i.e. , @xmath134 is a minimizer of the linear function @xmath135 over @xmath8 . linear oracles are black - box type functions , where the actual algorithm used in order to obtain the minimizer is unknown . for many feasible sets , such as @xmath136 balls and specific polyhedral sets , the oracle can be represented by a closed form solution or can be computed by an efficient method . the cg algorithm and its variants are linear - oracle based algorithms . the original cg algorithm , presented in @xcite also known as the frank - wolfe algorithm is as follows . [ alg : cg ] + input : a linear oracle @xmath137 + initialize : @xmath138 for @xmath139 1 . compute @xmath140 . 2 . choose a stepsize @xmath141 . 3 . update @xmath142 . the algorithm is guaranteed to have an @xmath143 rate of convergence for stepsize determined according to exact line search @xcite , adaptive stepsize @xcite and predetermined stepsize @xcite . this upper bound on the rate of convergence is tight @xcite and therefore variants , such as the ascg were developed . the ascg algorithm was proposed by frank - wolfe in @xcite . a linear convergence rate was proven for problems consisting of minimizing strongly convex objective functions over polyhedral feasible sets in @xcite under some restrictions on the location of the optimal solution , and in @xcite without such restrictions . jaggi and lacoste - julien @xcite showed that the latter result is also applicable for the specific case of problem where @xmath10 ( or more generally @xmath144 ) , provided that an appropriate linear - oracle is available for the set @xmath23 . in this section , we extend this result for the general case of problem , i.e. , for any @xmath5 and @xmath11 . furthermore , we explore the potential issues with obtaining a linear - oracle for the set @xmath23 , and suggest an alternative analysis , which only assumes existence of an appropriate linear - oracle on the original set @xmath8 . moreover , our analysis differs from the one presented in @xcite by the fact that it is based on duality rather than geometric arguments . this approach enables to derive a computable constant for the rate of convergence , which is explicitly expressed as a function of the problem s parameters and the geometry of the feasible set . we separate the discussion of the ascg into four sections . in section [ sec : vertexlinearoracles ] we define the concept of _ vertex linear oracles _ , which is needed for the ascg method , and the issues of obtaining such an oracle for linear transformations of simple sets . section [ sec : ascgmethod ] contains a full description of the ascg method itself , including the concept of vertex representation , and representation reduction . in section [ sec : rateconvergenceanalysis ] we present the rate of convergence analysis of the ascg for problem , as well as introduce the new computable convergence constant @xmath145 . finally , in section [ sec : findingomegaforpolyhedrons ] we demonstrate how to compute @xmath145 for three types of simple sets . the ascg algorithm requires a linear oracle which is a _ vertex linear oracle _ , a concept that we now define explicitly . given a polyhedral set @xmath8 with vertex set @xmath146 , a linear oracle @xmath147 is called a * vertex linear oracle * for @xmath8 , if for each @xmath131 it returns a vertex @xmath148 such that @xmath133 for any @xmath80 . notice that , according to the fundamental theorem of linear programming ( * ? ? ? * theorem 2.7 ) , the problem of optimizing any linear objective function over the compact set @xmath8 always has an optimal solution which is a vertex . therefore , the vertex linear oracle @xmath149 is well defined . we also note that in this paper the term vertex " is synonymous with the term extreme point " in @xcite , jaggi and lacoste - julien proved that the ascg algorithm is affine invariant . this means that given the problem @xmath150 where @xmath7 is a strongly convex function and @xmath5 is some matrix , applying the ascg algorithm on the equivalent problem @xmath151 where @xmath152 , yields a linear rate of convergence , which depends only on the strong convexity parameter of @xmath7 and the geometry of the set @xmath153 ( regardless of what @xmath5 generated it ) . however , assuming that @xmath5 is not of a full column rank , i.e. , @xmath6 is not strongly convex , retrieving an optimal solution @xmath154 from the optimal solution @xmath155 requires solving a linear feasibility problem . this feasibility problem is equivalent to solving the following constrained least squares problem : @xmath156 which , for a general @xmath5 , may be more computationally expensive than simply applying the linear oracle on set @xmath8 . moreover , in order to apply the algorithm to problem , a vertex linear oracle must be available for the set @xmath152 . assuming there exists a vertex linear oracle @xmath149 for @xmath8 , constructing such an oracle @xmath157 for @xmath23 may incur an additional computational cost per iteration . a naive approach to construct a general linear oracle @xmath158 , given @xmath149 , is by the formula @xmath159 however , the output @xmath160 of this linear oracle is not guaranteed to be a vertex of @xmath23 , and therefore , in order to obtain a vertex linear oracle @xmath161 , a vertex @xmath134 of @xmath23 with the same objective function value as @xmath162 must still be found . as an example , take @xmath8 to be the unit box in three dimensions , @xmath163 ^ 3\subseteq { \mathbb{r}}^3 $ ] , and let @xmath5 be given by @xmath164 we denote the vertex set @xmath146 of the set @xmath8 by the letters a - h as follows : @xmath165 and the linear mappings of these vertices by the matrix @xmath5 by a-h : @xmath166 the vertex set of @xmath23 is @xmath167 . [ fig : x and ex ] and @xmath23 ] the sets @xmath8 and @xmath23 are presented in figure [ fig : x and ex ] . notice that finding a vertex linear oracle for @xmath8 is trivial , while finding one for @xmath23 is not . in particular , a vertex linear oracle for @xmath8 may be given by any operator @xmath169 satisfying @xmath170 given the vector @xmath171 , we want to find @xmath172 using the naive approach , described in , we obtain a vertex of @xmath8 by applying the vertex linear oracle @xmath149 described in with parameter @xmath173 , which may return either one of the vertices b , c , g or h. if vertex c is returned , then its mapping c does not yield a vertex in @xmath23 . therefore , the oracle @xmath157 must now search for a vertex with the same objective function value , or alternatively , discover that c lies on the face defined by b and h , and consequently return one of these vertices . obviously , this is true for any @xmath174 such that @xmath175 returns one of the vertices c , d , e or g. this 3d example illustrates that , even for a simple @xmath8 , understanding the geometry of the set @xmath23 , let alone constructing a vertex linear oracle over it , is not trivial and becomes more complicated as the dimension of the problem increases . we aim to show that given a vertex linear oracle for @xmath8 , the ascg algorithm converges in a linear rate for problem . since in our analysis we do not assume the existence of a vertex linear oracle for @xmath23 , but rather a vertex linear oracle for @xmath8 , the computational cost per iteration is independent of the matrix @xmath5 , and depends only on the geometry of @xmath8 . we will now present the ascg algorithm . in the following we denote the vertex set of @xmath8 as @xmath176 . moreover , as part of the ascg algorithm , at each iteration @xmath34 the iterate @xmath33 is represented as a convex combination of points in @xmath146 . specifically , @xmath33 is assumed to have the representation @xmath177 where @xmath178 . let @xmath179 , then @xmath180 and @xmath181 provide a compact representation of @xmath33 , and @xmath33 lies in the relative interior of the set @xmath182 . throughout the algorithm we update @xmath180 and @xmath183 via the vertex representation updating ( vru ) scheme . the ascg method has two types of updates : a _ forward step _ , used in the classical cg algorithm , where a vertex is added to the representation , and an _ away step _ , unique to this algorithm , in which the coefficient of one of the vertices used in the representation is reduced or even nullified . specifically , the away step uses the direction @xmath184 where @xmath185 and step size @xmath186 so that @xmath187 and so @xmath188 . moreover , if @xmath189 , then @xmath190 is nullified , and consequently , the vertex @xmath191 is removed from the representation . this vertex removal is referred to as a _ drop step_. the full description of the ascg algorithm and the vru scheme is given as follows . [ alg : ascg ] + input : a vertex linear oracle @xmath192 + initialize : @xmath193 where @xmath194 , @xmath195 for any @xmath196 and @xmath197 for @xmath139 1 . compute @xmath198 . 2 . compute @xmath199 . 3 . if @xmath200 , then set @xmath201 and @xmath202 . + otherwise , set @xmath203 and @xmath204 4 . choose a stepsize @xmath141 . update @xmath205 . employ the vru procedure with input @xmath206 and obtain an updated representation @xmath207 . the stepsize in the ascg algorithm can be chosen according to one of the following stepsize selection rules , where @xmath208 and @xmath209 are as defined in the algorithm . @xmath210 [ remark : monotone ] it is simple to show that under the above two choice of stepsize strategies , the sequence of function values @xmath211 is nonincreasing . since the convergence rate analyses for both of these stepsize options is similar , we chose to conduct a unified analysis for both cases . following is exact definition of the vru procedure . [ alg : updatingschemeforukmuk ] + * input : * @xmath33 - current point . + @xmath212 - vertex representation of @xmath33 , + @xmath213 - current direction and stepsize , + @xmath214 - candidate vertices . + * output : * updated vertex representation @xmath207 of @xmath215 . + if @xmath216 ( away step ) then 1 . update @xmath217 for any @xmath218 . 2 . update @xmath219 . 3 . if @xmath220 ( drop step ) , then update @xmath221 , otherwise @xmath222 . else ( @xmath223 - forward step ) 1 . update @xmath224 for any @xmath225 . 2 . update @xmath226 . 3 . if @xmath227 , then update @xmath228 , otherwise update @xmath229 . update @xmath230 with @xmath231 being a representation reduction procedure with constant @xmath232 . the vru scheme uses a representation reduction procedure @xmath231 with constant @xmath232 , which is a procedure that takes a representation @xmath233 of a point @xmath32 and replaces it by a representation @xmath234 of @xmath32 such that @xmath235 and @xmath236 . we consider two possible options for the representation reduction procedure : 1 . @xmath231 is the trivial procedure , meaning it does not change the representation , in which case its constant is @xmath237 . the procedure @xmath231 is some implementation of the carathodory theorem ( * ? ? ? * section 17 ) , in which case its constant is @xmath238 . using this option will accelerate the algorithm when the number of vertices is not polynomial in the problem s dimension . a full description of the incremental representation reduction ( irr ) scheme , which applies the carathodory theorem efficiently in this context , is presented in appendix [ appx : caratheodory ] . we will now prove the linear rate of convergence for the ascg algorithm for problem . in the following we use @xmath239 to denote the _ index set of the active constraints at @xmath32 _ , @xmath240 similarly , for a given set @xmath241 , the set of active constraints for all the points in @xmath241 is defined as @xmath242 we present the following technical lemma , which is similar to a result presented by jaggi and lacoste - julien @xcite . in @xcite the proof is based on geometrical considerations , and utilizes the so - called pyramidal width constant " , which is the optimal value of a complicated optimization problem , whose value is unknown even for simple sets such as the unit simplex . in contrast , the proof below relies on simple linear programming duality arguments , and in addition , the derived constant @xmath145 , which replaces the pyramidal width constant , is computable for a many choices of sets @xmath8 . [ lemma : pyrwidth2 ] given @xmath243 and @xmath244 . if there exists a @xmath245 such that @xmath246 and @xmath247 , then @xmath248 where @xmath249 for @xmath250 by the fundamental theorem of linear programming @xcite , we can maximize the function @xmath135 on @xmath8 instead of on @xmath146 and get the same optimal value . similarly , we can minimize the function @xmath251 on @xmath252 instead of on @xmath241 , and obtain the same optimal value . therefore , @xmath253 \max_{{{\bf p}}\in v , { { \bf u}}\in u}{\left\langle{{{\bf c}}},{{{\bf p}}-{{\bf u}}}\right\rangle}&=\max_{{{\bf p}}\in v}{\left\langle{{{\bf c}}},{{{\bf p}}}\right\rangle}-\min_{{{\bf u}}\in u}{\left\langle{{{\bf c}}},{{{\bf u}}}\right\rangle}\\ & = \max_{{{\bf x}}\in x } { \left\langle{{{\bf c}}},{{{\bf x}}}\right\rangle}-\min_{{{\bf y}}\in \operatorname{conv}(u)}{\left\langle{{{\bf c}}},{{{\bf y}}}\right\rangle}\\ & = \max_{{{\bf x}}:{{\bf a}}{{\bf x}}\leq{{\bf a } } } { \left\langle{{{\bf c}}},{{{\bf x}}}\right\rangle}+\max_{{{\bf y}}\in \operatorname{conv}(u)}{\left\{{-{\left\langle{{{\bf c}}},{{{\bf y}}}\right\rangle}}\right\}}. \end{aligned}\ ] ] since @xmath8 is nonempty and bounded , the problem in @xmath32 is feasible and bounded above . therefore , by strong duality for linear programming , @xmath254 plugging back into we obtain : @xmath255 \max_{{{\bf p}}\in v , { { \bf u}}\in u}{\left\langle{{{\bf c}}},{{{\bf p}}-{{\bf u}}}\right\rangle}&=\min_{{{\boldsymbol \eta}}\in{\mathbb{r}}^m_+ : { { \bf a}}^t{{\boldsymbol \eta}}={{\bf c } } } { \left\langle{{{\bf a}}},{{{\boldsymbol \eta}}}\right\rangle}+\max_{{{\bf y}}\in \operatorname{conv}(u)}{\left\{{-{\left\langle{{{\bf c}}},{{{\bf y}}}\right\rangle}}\right\}}\\ & = \min_{{{\boldsymbol \eta}}\in{\mathbb{r}}^m_+ : { { \bf a}}^t{{\boldsymbol \eta}}={{\bf c}}}\max_{{{\bf y}}\in \operatorname{conv}(u ) } { \left\langle{{{\bf a}}-{{\bf a}}{{\bf y}}},{{{\boldsymbol \eta}}}\right\rangle}. \end{aligned}\ ] ] since @xmath256 is in @xmath252 , we have that @xmath257 for any value of @xmath258 , and therefore , @xmath259 using strong duality on the rhs of , we obtain that @xmath260 denote @xmath261 and @xmath262 . from the definition of @xmath263 , it follows that @xmath264 for all @xmath265 , and that for any @xmath266 there exists at least one vertex @xmath265 such that @xmath267 , and hence , @xmath268 which in particular implies that @xmath269 since @xmath270 , we can conclude from ( [ 539 ] ) and ( [ 541 ] ) that @xmath271 therefore , replacing the rhs of the set of inequalities @xmath272 in by the bounds given in , we obtain that @xmath273 combining , , and it follows that @xmath274 where @xmath275 we will now show that it is not possible for @xmath276 to satisfy @xmath277 . suppose by contradiction @xmath276 satisfies does satisfy @xmath277 . then @xmath278 is a feasible solution of problem for any @xmath279 , and since @xmath247 we obtain that @xmath280 as @xmath281 , and thus @xmath282 . however , since @xmath146 contains a finite number of points , the lhs of is bounded from above , and so @xmath283 in contradiction . therefore , there exists @xmath266 such that @xmath284 . since @xmath285 , the vector @xmath286 is well defined . moreover , @xmath287 satisfies @xmath288 and @xmath289 { { \bf a}}_i\overline{{{\bf x}}}={{\bf a}}_i{{\bf z}}\frac{\omega_x}{|u|{\left\|{{{\bf z}}}\right\|}}&\leq { \left\|{{{\bf a}}_i}\right\|}{\left\|{{{\bf z}}}\right\|}\frac{\zeta}{|u|{\left\|{{{\bf z}}}\right\| } \varphi } \leq \frac{\zeta}{|u|},\quad \forall i\in \overline{j } , \end{aligned}\ ] ] where the first inequality follows from the cauchy - schwartz inequality and the second inequality follows from the fact that if @xmath290 , then @xmath291 and so @xmath292 . consequently , and imply that @xmath287 is a feasible solution for problem . therefore , @xmath293 , which by yields @xmath294 the constant @xmath145 represents a normalized minimal distance between the hyperplanes that contain facets of @xmath8 and the vertices of @xmath8 which do not lie on those hyperplanes . we will refer to @xmath145 as _ the vertex - facet distance of @xmath8_. examples for the derivation of @xmath145 for some simple polyhedral sets can be found in section [ sec : findingomegaforpolyhedrons ] . the following lemma is a technical result stating that the active constraints at a given point are the same as the active constraints of the set of vertices in its compact representation . [ lemma : ix_iu ] let @xmath80 and the set @xmath243 satisfy @xmath295 , where @xmath296 . then @xmath297 . it is trivially true that @xmath298 since @xmath32 is a convex combination of points in the affine space defined by @xmath299 . we will prove that @xmath300 . any @xmath301 satisfies @xmath302 . assume to the contrary , that there exists @xmath303 such that some @xmath304 satisfies @xmath305 . since @xmath306 and @xmath307 , it follows that @xmath308 in contradiction to the assumption that @xmath303 . [ cor : boundoninnerprod ] for any @xmath309 which can be represented as @xmath295 for some @xmath296 and @xmath243 , it holds that , @xmath310 for any @xmath309 define @xmath311 . it follows from lemma [ lemma : ix_iu ] that @xmath312 . for any @xmath95 , the vector @xmath313 satisfies @xmath314 and , from the convexity of @xmath6 , as well as the optimality of @xmath101 , @xmath315 . therefore , invoking lemma [ lemma : pyrwidth2 ] achieves the desired result . we now present the main theorem of this section , which establishes the linear rate of convergence of ascg for problem . this theorem is an extension of ( * ? ? ? * thorem 7 ) , and the proof follows the same general arguments , while incorporating the use of the error bound from lemma [ lemma : errbound ] and the new constant @xmath145 . [ thm : convergenceofascg ] let @xmath316 be the sequence generated by the ascg algorithm for solving problem using a representation reduction to procedure @xmath231 with constant @xmath232 , and let @xmath98 be the optimal value of the problem . then for any @xmath317 @xmath318 where @xmath319 @xmath106 with @xmath78 being the hoffman constant associated with matrix @xmath107^t$ ] , @xmath100 , and @xmath145 is the vertex - facet distance of @xmath8 given in ( [ defomega ] ) . for each @xmath34 we will denote the stepsize generated by exact line search as @xmath320 and the adaptive stepsize as @xmath321 . then @xmath322 from lemma [ lemma : descentlemma ] ( the descent lemma ) , we have that @xmath323 assuming that @xmath324 , then for any @xmath95 we have that @xmath325 { \langle{\nabla f({{\bf x}}^k)},{{{\bf d}}^k}\rangle}&=\min{\left\ { { { \langle{\nabla f({{\bf x}}^k)},{{{\bf p}}^k-{{\bf x}}^k}\rangle } , { \langle{\nabla f({{\bf x}}^k)},{{{\bf x}}^k-{{\bf u}}^k}\rangle}}\right\}}\\ & \leq { \langle{\nabla f({{\bf x}}^k)},{{{\bf p}}^k-{{\bf x}}^k}\rangle}\\ & \leq { \langle{\nabla f({{\bf x}}^k)},{{{\bf x}}^*-{{\bf x}}^k}\rangle}\\ & \leq f^*-f({{\bf x}}^k ) , \end{aligned}\ ] ] where the first equality is derived from the algorithm s specific choice of @xmath208 , the third line follows from the fact that @xmath326 , and the fourth line follows from the convexity of @xmath6 . in particular , @xmath327 , and by it follows that @xmath321 is equal to @xmath328 we now separate the analysis to three cases : ( a ) @xmath223 and @xmath329 , ( b ) @xmath216 and @xmath329 , and ( c ) @xmath330 . in cases ( a ) and ( b ) , it follows from that @xmath331 using inequalities , and , we obtain @xmath332 subtracting @xmath98 from both sides of the inequality and using , we have that @xmath333 f({{\bf x}}^{k+1})-f^ * & \leq f({{\bf x}}^k)-f^*+\frac{\overline{\gamma}^k}{2}{\langle{\nabla f({{\bf x}}^k)},{{{\bf d}}^k}\rangle}\\ & \leq ( f({{\bf x}}^k)-f^*)\left(1-\frac{\overline{\gamma}^k}{2}\right ) . \end{aligned}\ ] ] in case ( a ) , @xmath334 , and hence @xmath335 in case ( b ) , we have no positive lower bound on @xmath209 , and therefore we can only conclude , by the nonnegativity of @xmath209 , that @xmath336 however , case ( b ) is a drop step , meaning in particular that @xmath337 , since before applying the representation reduction procedure @xmath231 , we eliminate one of the vertices in the representation of @xmath338 . denoting the number of drop steps until iteration @xmath34 as @xmath339 , and the number of forward steps until iteration @xmath34 as @xmath340 , it follows from the algorithm s definition that @xmath341 ( at each iteration we add a vertex , remove a vertex , or neither ) and @xmath342 ( the number of removed vertices can not exceed the number of added vertices ) , and therefore @xmath343 . we arrive to case ( c ) . in this case , implies @xmath344 which combined with and results in @xmath345 from the algorithm s specific choice of @xmath208 , we obtain that @xmath346 applying the bound in and the inequality @xmath347 to , it follows that @xmath348 by the definitions of @xmath191 and @xmath349 , and since applying representation reduction procedure @xmath231 ensures that that @xmath350 , corollary [ cor : boundoninnerprod ] implies that for any @xmath95 , @xmath351 lemma [ lemma : errbound ] implies that there exists @xmath95 such that @xmath352 , which combined with convexity of @xmath6 , bounds from below as follows : @xmath353 { \langle{\nabla f({{\bf x}}^k)},{{{\bf u}}^k-{{\bf p}}^k}\rangle}^2 & \geq \left(\frac{\omega_x}{n}\right)^2\frac{{\left\langle{\nabla f({{\bf x}}^k)},{{{\bf x}}^k-{{\bf x}}^*}\right\rangle}^2}{{\left\|{{{\bf x}}^k-{{\bf x}}^*}\right\|}^2}\\ & \geq \left(\frac{\omega_x}{n}\right)^2\frac{(f({{\bf x}}^k)-f({{\bf x}}^*))^2}{{\left\|{{{\bf x}}^k-{{\bf x}}^*}\right\|}^2}\\ & \geq \left(\frac{\omega_x}{n}\right)^2\frac{(f({{\bf x}}^k)-f^*)^2}{\kappa(f({{\bf x}}^k)-f^*)}\\ & = \frac{(\omega_x)^2}{n^2\kappa}(f({{\bf x}}^k)-f^ * ) , \end{aligned}\ ] ] which along with yields @xmath354 f({{\bf x}}^{k+1})-f^ * & \leq f({{\bf x}}^k)-f^*-\frac{{\langle{\nabla f({{\bf x}}^k)},{{{\bf u}}^k-{{\bf p}}^k}\rangle}^2}{8\rho d^2}\\ & \leq ( f({{\bf x}}^k)-f^*)\left(1-\frac{(\omega_x)^2}{8\rho\kappa d^2n^2}\right ) \end{aligned}\ ] ] therefore , if either of the cases ( a ) or ( c ) occurs , then by ( [ 666 ] ) and , it follows that @xmath355 where @xmath356 is defined in ( [ defalpha ] ) . we can therefore conclude from cases ( a)-(c ) that until iteration @xmath34 we have at least @xmath357 iterations for which ( [ 718 ] ) holds , and therefore @xmath358 applying lemma [ lemma : ofupperbound ] for @xmath359 we obtain @xmath360 , and the desired result ( [ 613 ] ) follows . in this section , we demonstrate how to compute the vertex - facet distance constant @xmath145 for a few simple polyhedral sets . we consider three sets : the unit simplex , the @xmath361 ball and the @xmath362 ball . we first describe each of the sets as a system of linear inequalities of the form @xmath363 . then , given the parameters @xmath45 and @xmath364 , as well as the vertex set @xmath146 , @xmath145 can be computed by its definition , given by ( [ defomega ] ) . _ _ the unit simplex @xmath365 can be represented by @xmath366 { { \bf a}}=\begin{bmatrix}-{{\bf i}}_{n \times n}\\{{\bf 1}}_n^t\\-{{\bf 1}}_n^t\end{bmatrix}\in { \mathbb{r}}^{(n+2)\times n},\ ; { { \bf a}}=\begin{bmatrix}{{\bf 0}}_n\\1\\1\end{bmatrix}\in{\mathbb{r}}^{(n+2)}. \end{aligned}\ ] ] the set of extreme points is given by @xmath367 . notice that since there are only @xmath368 extreme points which are all affinely independent , using a rank reduction procedure which implements the carathodory theorem is the same as applying the trivial procedure that does not change the representation . in order to calculate @xmath145 , we first note that @xmath369 , and therefore @xmath370 and @xmath371 which means that @xmath372 . _ _ the @xmath361 ball is given by the set @xmath373 therefore @xmath374 and each row of the matrix @xmath375 is a vector in @xmath376 . the set of extreme points is given by @xmath377 , and therefore has cardinality of @xmath378 . finally , we have that @xmath379 and @xmath380 which means that @xmath381 . _ _ the @xmath362 ball is represented by @xmath366 { { \bf a}}=\begin{bmatrix}{{\bf i}}\\-{{\bf i}}\end{bmatrix}\in { \mathbb{r}}^{2n\times n},\ ; { { \bf a}}=\begin{bmatrix}{{\bf 1}}\\{{\bf 1}}\end{bmatrix}\in{\mathbb{r}}^{2n}. \end{aligned}\ ] ] the set of extreme points is given by @xmath382 , which in particular implies that @xmath383 . therefore , for large - scale problems , using the representation reduction procedure , which is based on carathodory theorem , is crucial in order to obtain a practical implementation . from the definition of @xmath45 and @xmath146 , it follows that @xmath384 and @xmath385 which implies that @xmath386 . 10 a. beck and m. teboulle . a conditional gradient method with linear rate of convergence for solving convex linear systems . , 59(2):235247 , 2004 . a. beck and m. teboulle . gradient - based algorithms with applications to signal recovery problems . in d. palomar and y. eldar , editors , _ convex optimization in signal processing and communications _ , pages 139162 . cambridge university press , 2009 . d. p. bertsekas . . athena scientific , belmont , ma , 2nd edition , 1999 . d. bertsimas and j. n. tsitsiklis . , volume 6 . athena scientific belmont , ma , 1997 . m. d. canon and c. d. cullum . a tight upper bound on the rate of convergence of frank - wolfe algorithm . , 6(4):509516 , 1968 . j. dunn and s. harshbarger . conditional gradient algorithms with open loop step size rules . , 62(2):432 444 , 1978 . m. epelman and r. m. freund . condition number complexity of an elementary algorithm for computing a reliable solution of a conic linear system . , 88(3):451485 , 2000 . m. frank and p. wolfe . an algorithm for quadratic programming . , 3(1 - 2):95110 , 1956 . d. garber and e. hazan . a linearly convergent conditional gradient algorithm with applications to online and stochastic optimization . d. goldfarb and m. j. todd . chapter ii : linear programming . in g. nemhauser , a. r. kan , and m. todd , editors , _ optimization _ , volume 1 of _ handbooks in operations research and management science _ , pages 73 170 . elsevier , 1989 . j. guelat and p. marcotte . some comments on wolfe s ` away step ' . , 35(1):110119 , 1986 . o. gler . . graduate texts in mathematics . springer , new york , ny , usa , 2010 . a. j. hoffman . on approximate solutions of systems of linear inequalities . , 49(4):263265 , 1952 . s. lacoste - julien and m. jaggi . an affine invariant linear convergence analysis for frank - wolfe algorithms . , 2014 . e. levitin and b. t. polyak . constrained minimization methods . , 6(5):787823 , 1966 . y. nesterov . , volume 87 . springer , 2004 . z. quan luo and p. tseng . error bounds and convergence analysis of feasible descent methods : a general approach . , 46 - 47(1):157178 , 1993 . r. t. rockafellar . . princeton university press , 2nd edition , 1970 . wang and c .- j . iteration complexity of feasible descent methods for convex optimization . , 15:15231548 , 2014 . p. wolfe . , chapter chapter 1:convergence theory in nonlinear programming . north - holland publishing company , 1970 . in this section we will show a way to efficiently and incrementally implement the constructive proof of carathodory theorem , as part of the vru scheme , at each iteration of the ascg algorithm . we note that this reduction procedure does not have to be employed , and instead the trivial procedure , which does not change the representation can be used . in that case , the upper bound on the number of extreme points in the representation is just the number of extreme points of the feasible set @xmath8 . + the implementation described in this section will allow maintaining a vertex representation set @xmath180 , with cardinality of at most @xmath387 , at a computational cost of @xmath388 operations per iteration . for this purpose , we assume that at the beginning of iteration @xmath34 , @xmath338 has a representation with vertex set @xmath389 , such that the vectors in the set are affinely independent . moreover , we assume that at the beginning of iteration @xmath34 , we have at our disposal two matrices @xmath390 and @xmath391 . we define @xmath392 to be the matrix whose @xmath31th column is the vector @xmath393 for @xmath394 , where @xmath395 is called the reference vertex . the matrix @xmath396 is a product of elementary matrices , which ensures that the matrix @xmath397 is in row echelon form . the implementation does not require to save the matrix @xmath398 , and so at each iteration , only the matrices @xmath396 and @xmath399 are updated . let @xmath400 be the vertex set and let @xmath401 be the coefficients vector at the end of iteration @xmath34 , before applying the rank reduction procedure . updating the matrices @xmath402 and @xmath403 , as well as @xmath400 and @xmath401 , is done according to the following _ incremental representation reduction _ scheme , which is partially based on the proof of carathodory theorem presented in ( * ? ? ? * section 17 ) . [ alg : caratheodory ] + * input * : representation @xmath207 of point @xmath404 , set @xmath405 of affinely + independent vectors , and matrices @xmath406 and @xmath407 . + * output * : updated representation @xmath207 of @xmath404 , and matrices @xmath408 + and @xmath409 . 1 . set @xmath410 . 2 . update @xmath411 . 3 . if @xmath412 , then set the matrix @xmath402 to be empty and @xmath413 . 4 . else , if @xmath414 , then set @xmath415 . [ step : irr_dropstep ] else , if @xmath416 ( drop step ) , then 1 . find @xmath417 such that @xmath418 . [ step : irr_changereferencevertex ] if @xmath419 ( the reference vertex was removed ) , then remove the first column of @xmath399 and change reference vertex to @xmath420 , using the update formula @xmath421 where @xmath422 . ( a non - reference vertex was removed ) , remove column @xmath423 from @xmath402 . 6 . else , if @xmath424 ( forward step ) , then 1 . find @xmath425 . 2 . compute @xmath426 . 3 . update the matrix @xmath427 $ ] . [ step : irr_findrowrank]compute @xmath428 - the row rank of @xmath429 . 5 . if @xmath430 , then 1 . [ step : irr_solsyseq ] find a solution @xmath431 of the following system + @xmath432 2 . set the vector @xmath433 to be + @xmath434 3 . compute @xmath435 and @xmath436 and set + @xmath437 4 . update @xmath438 for all @xmath439 . [ step : irr_findnullset ] compute @xmath440 . [ step : irr_removevertices ] for each @xmath441 remove column @xmath442 matrix @xmath402 . [ step : irr_updaterepresentation ] update @xmath443 . 7 . [ step : irr_rowelimination]if @xmath429 is not in row echelon form , then construct a matrix @xmath444 , as a composition of elementary matrices , such that @xmath445 is row echelon form , and update @xmath446 and @xmath447 . notice that in order to compute the row rank of the matrix @xmath402 in step [ step : irr_findrowrank ] , we may simply convert the matrix to row echelon form , and then count the number of nonzero rows . this is done similarly to step [ step : irr_rowelimination ] , and requires ranking of at most one column . we will need to rerank the matrix in step [ step : irr_rowelimination ] only if @xmath430 , and subsequently at least one column is removed in step [ step : irr_removevertices ] . the irr scheme may reduce the size of the input @xmath400 only in the case of a forward step , since otherwise the vertices in @xmath400 are all affinely independent . nonetheless , the irr scheme _ must _ be applied at each iteration in order to maintain the matrices @xmath448 and @xmath396 . the efficiency of the scheme relies on the fact that only a small number of vertices are either added to or removed from the representation . the potentially computationally expensive steps are : step [ step : irr_changereferencevertex ] - replacing the reference vertex , step [ step : irr_findrowrank ] - finding the row rank of @xmath402 , step [ step : irr_solsyseq ] - solving the system of linear equalities , step [ step : irr_removevertices ] - removing columns corresponding with the vertices eliminated from the representation , and step [ step : irr_rowelimination ] - the ranking of the resulting matrix @xmath429 . step [ step : irr_changereferencevertex ] can be implemented without explicitly using matrix multiplication and therefore has a computational cost of @xmath388 . since @xmath448 was in row echelon form , step [ step : irr_findrowrank ] requires a row elimination procedure , similar to step [ step : irr_rowelimination ] , to be conducted only on the last column of @xmath429 , which involves at most @xmath449 operations and an additional @xmath388 operation for updating @xmath403 . moreover , since @xmath448 was full column rank , the irr scheme guarantees that in step @xmath450 the vector @xmath431 has a unique solution , and since @xmath429 is in row echelon form , it can be found in @xmath388 operations . moreover , in step [ step : irr_removevertices ] , the specific choice of @xmath451 ensures that the reference vertex @xmath395 is not eliminated from the representation , and so there is no need to change the reference vertex at this stage . furthermore , it is reasonable to assume that the set @xmath24 satisfies @xmath452 , since otherwise the vector @xmath404 , produced by a forward step , can be represented by significantly less vertices than @xmath33 , which , although possible , is numerically unlikely . therefore , assuming that indeed @xmath452 , the matrix @xmath444 , calculated in step [ step : irr_rowelimination ] , applies a row elimination procedure to at most @xmath453 rows ( one for each column removed from @xmath402 ) or one column ( if a column was added to @xmath402 ) . conducting such an elimination on either row or column takes at most @xmath388 operations , which may include row switching and at most @xmath368 row addition and multiplication . therefore , the total computational cost of the irr scheme amounts to @xmath388 . |
we consider a single - level quantum dot tunnel - coupled to one normal and one superconducting lead . we employ a diagrammatic real - time approach to calculate the finite - frequency current noise for subgap transport . the noise spectrum gives direct access to the internal dynamics of the dot . in particular the noise spectrum shows sharp dips at the frequency of the coherent oscillations of cooper pairs between dot and superconductor . this feature is most pronounced when the superconducting correlation is maximal . furthermore , in the quantum - noise regime , , the noise spectrum exhibits steps at frequencies equal to the andreev addition energies . the height of these steps is related to the effective coupling strength of the excitations . the finite - frequency noise spectrum hence provides a full spectroscopy of the system . | 1410.6694 | nanostructures with quantum dots in proximity to superconducting electrodes are an ideal playground to study superconducting correlations in systems with few degrees of freedom that exhibit strong coulomb - interaction effects . @xcite intriguingly , these types of structures are at the heart of recent proposals to generate majorana - fermion excitations in quantum dots @xcite and to establish and detect different symmetries of superconducting pairing in a controllable way . @xcite another line of research has focussed on the possibility to use double quantum dots , tunnel coupled to superconductors as a source of entangled electron pairs . @xcite it is therefore of vital importance to get access to the properties of such hybrid quantum - dot systems . here we are interested in a setup , in which the tunnel - coupling rate between the superconductor and the quantum dot is strong ( larger than the coupling to other normal - conducting leads eventually present in the device ) , such that it is possible to establish a bcs - like state in a single - level quantum dot even in the presence of strong coulomb repulsion . @xcite such a state is characterised by a coherent exchange of cooper pairs between the dot and the superconducting lead . if one considers the two - terminal case of a quantum dot , tunnel - coupled to one normal and one superconducting lead , the proximity effect is established by generating a non - equilibrium situation by means of an applied transport voltage . the presence of the proximity effect can be detected by measuring the andreev current and its zero - frequency noise . @xcite in various experimental studies the subgap spectrum of hybrid superconductor - quantum dot devices has been analyzed by andreev level spectroscopy @xcite , which allows to measure the andreev addition energies and the total line width of the resonances via the differential conductance . the coherent dynamics underlying the proximity effect in the dot shows up for example in the waiting time distribution of transport events @xcite in the normal lead , which exhibits an oscillatory behaviour due to the coherent exchange of cooper pairs between dot and superconductor . @xcite however , a direct measurement of the waiting time distribution might be challenging and it is therefore interesting to look at alternative possibilities to reveal the coherent tunnelling of cooper pairs between dot and superconductors . a quantity of high interest to look at is the finite - frequency noise of the andreev current . current noise spectroscopy in mesoscopic systems has become a standard tool to gain information on the transport processes and internal time scales of mesoscopic conductors . @xcite indeed , the non - equilibrium finite - frequency noise of quantum dots in different regimes and setups has previously been at the focus of various theoretical studies . @xcite this manuscript focuses on the finite - frequency current noise for subgap transport through a single - level quantum dot tunnel - coupled to one normal and one superconducting lead . we consider strong coupling between quantum dot and superconductor , while the dot is only weakly coupled to the normal conducting lead . we employ a non - equilibrium real - time diagrammatic perturbation expansion in the tunnel - coupling to the normal lead . @xcite we find that the coherent oscillations between dot states with different particle numbers leads to a resonant feature at the oscillation frequency in the finite - frequency noise spectrum . the magnitude of this feature ( a sharp dip in the spectrum ) is directly related to the pair amplitude in the dot . in the quantum noise regime , @xmath0 , it is possible to extract information on the relative coupling of different bcs - like states to the normal lead . beyond the knowledge of the andreev addition energies , which can also be obtained from andreev level spectroscopy by means of a differential conductance measurement , @xcite the finite - frequency noise additionally provides information on the coherent dynamics of the system and its characteristic time scale , and the effective coupling strengths of the different andreev levels . the remainder of this paper is organised as follows . in sec . [ modelandformalism ] , we introduce the mathematical model used to describe the hybrid quantum - dot system and present the diagrammatic approach employed to calculate its noise spectrum . after a brief overview over the properties of the andreev current in sec . [ sec : current ] , we show the results for the finite - frequency current noise in sec . [ results ] organised by biasing and frequency regimes . the main conclusions of the paper are summarised in sec . [ conclusions ] . in this section we present the model for the interacting single - level quantum dot attached to one normal and one superconducting lead and the real - time diagrammatic approach to obtain the finite - frequency current noise . in this paper , we study subgap transport through a single - level quantum dot tunnel - coupled to one normal and one superconducting lead . we consider strong coupling between quantum dot and superconductor , while the dot is only weakly coupled to the normal conducting lead . we restrict ourself to the case when the the temperature is larger than the tunnel - coupling strength of the normal - conducting lead ( @xmath1 , where @xmath2 will be defined in terms of tunnel amplitude and the density of states of the lead later in this section ) . in this particular regime the kondo correlations due to the coupling with the normal lead are negligible and we can treat the tunnelling with the normal lead to lowest non - vanishing order . a large body of theoretical work regards the interplay of superconductivity and kondo physics@xcite . for the subgap transport characteristics of the system , the superconductor can be described by means of an effective hamiltonian which becomes exact in the regime of infinite superconducting gap . however the effective hamiltonian still describes well the subgap transport features even for finite values of the gap as long as the temperature is larger than the kondo temperature related to the kondo screening by the quasiparticle excitations in the superconductor . a detailed study of the reliability of this approximation can be found in ref . . the hamiltonian of the system can be written as the sum of three terms : @xmath3 . the first term , @xmath4 , is the hamiltonian for the hybrid system composed of the dot and the superconductor in the limit of large superconducting gap , the second term , @xmath5 , describes the non - interacting normal lead and the third one , @xmath6 , the tunnel - coupling between the proximized dot and the normal lead . we model the quantum dot tunnel - coupled to the superconducting lead by means of the following effective hamiltonian , which becomes exact in the limit of very large superconducting gap @xcite @xmath7 here , @xmath8 is the dot level energy , @xmath9 the on - site coulomb repulsion and @xmath10 the tunnel - coupling strength between dot and superconducting lead . all energies are measured with respect to the chemical potential of the superconductor , i.e. @xmath11 . here , @xmath12 is the annihilation ( creation ) operator for an electron on the dot with spin @xmath13 and @xmath14 the corresponding number operator . the hilbert space of the proximized dot is spanned by the states : @xmath15 ( empty ) , @xmath16 ( singly occupied ) and @xmath17 ( doubly occupied ) . the normal lead is described by the non - interacting hamiltonian @xmath18 , where @xmath19 is the annihilation ( creation ) operator for an electron with spin @xmath20 in the single - particle state of the lead characterised by the momentum quantum number @xmath21 with energy @xmath22 . the normal - conducting lead has an electrochemical potential @xmath23 , which in general differs from zero . the tunnel - coupling between the dot and the normal lead is modelled by means of the tunnelling hamiltonian , @xmath24 with the tunnelling amplitude @xmath25 . we assume the density of states , @xmath26 , of the normal lead to be constant and spin- and momentum independent ; we define the tunnel - coupling strength as @xmath27 . the effective dot - superconductor subsystem coupled to a normal - conducting lead is sketched in fig . [ fig : model ] . when discussing the results of this paper , we will always assume @xmath28 . we proceed to discuss the eigenstates of the effective hamiltonian in the absence of coupling to the normal lead . the singly - occupied states @xmath29 are not affected by the proximity effect and are eigenstates of @xmath30 with eigenenergy @xmath8 . due to the tunnel - coupling to the superconductor the states @xmath31 and @xmath32 form andreev bound states ( abs ) @xmath33 with the eigenenergies of the effective hamiltonian , @xmath34 . here , @xmath35 is the detuning between the empty and the doubly - occupied states and @xmath36 is the energy splitting between the @xmath37 and @xmath38 states . the excitation energies of the dot are the so called andreev addition energies , which are given by the differences of the eigenenergies of those states which have occupation numbers differing by one @xmath39 with @xmath40 . when an electron leaves or enters the normal lead , its energy must account for the energy difference between the initial and final state of the dot - superconductor subsystem , which are the andreev addition energies , represented by the energy levels in the sketch , fig . [ fig : model ] . at this stage it is useful to introduce effective coupling strengths which describe the coupling between the electronic reservoir and the dot resonances , namely the andreev levels . these effective coupling strengths turn out to be essential to understand the form of the finite - frequency noise spectrum . the effective tunnel - coupling strengths of the andreev levels to the normal - conducting lead @xmath2 corresponding to the transition @xmath29 to @xmath41 are given by @xmath42 while for the opposite transition they read @xmath43 in the following , we use the convention @xmath44 . we aim to study the finite - frequency noise for transport through a quantum dot coupled to a normal- and a superconducting lead as described by the above introduced model . we take into account on - site coulomb interaction of arbitrary magnitude and non - equilibrium conditions without resorting to the linear - response regime . while we are interested in a strong coupling of the quantum dot to the superconducting lead , leading to strong superconducting correlations , we assume the coupling to the normal - conducting lead to be weak ( @xmath45 , where @xmath46 is the boltzmann constant and @xmath47 is the absolute temperature ) . considering these conditions , we make use of a diagrammatic real - time perturbation theory in the tunnel - coupling with the normal lead @xcite and its extension to a system with superconducting electrodes , @xcite in order to derive the current and the finite - frequency current noise . the formalism to obtain the finite - frequency noise using this real - time diagrammatic approach has been introduced previously , where it was applied to the case of normal - conducting electrodes @xcite as well as for the ferromagnetic case . @xcite in this section we review the formalism to obtain the finite - frequency current noise by relating it directly to the system of a quantum dot coupled to one normal and one superconducting lead . the aim is to formulate a method which allows to calculate the reduced density matrix of the proximized quantum dot as well as the current through it and the finite - frequency current noise . the full system is represented by a density matrix describing the normal - conducting lead ( which has many degrees of freedom but is non - interacting ) coupled to the interacting quantum dot proximized by the superconducting condensate ( the latter having just a few degrees of freedom ) . since we are not interested in the degrees of freedom of the normal - conducting reservoir we trace them out , making use of wick s theorem . we are then left with an effective description by means of the reduced density matrix of the quantum dot proximized by the superconductor . this reduced density matrix has the form @xmath48 where the diagonal elements are the probabilities to find the dot singly occupied , @xmath49 , or in a bcs - like state , @xmath50 or @xmath51 . the off - diagonal elements , which we refer to as the coherences , describe coherent superpositions of two eigenstates of the proximized dot . importantly , the time evolution of the coherences between states of single occupation decouples from the one of the diagonal elements due to spin - conserving tunnelling and these coherences will hence be disregarded in the following . in contrast , in order to fully describe the short - time dynamics of the system , it is necessary to consider the off - diagonal elements in the reduced density matrix between the andreev bound states , @xmath52 and @xmath53 . it turns out that these become important for the finite - frequency noise at frequencies @xmath54 . the non - equilibrium time evolution of the reduced density matrix can be depicted on the keldysh contour and expressed in terms of a propagator @xmath55 . an example of the keldysh contour for the calculation of the current is shown in the sketch in fig . [ fig : contour ] ( a ) . the upper ( lower ) horizontal time line stands for the propagation of the individual dot state forward ( backward ) in real - time , indicated by arrows . in frequency space , this full propagator can be written in terms of a dyson equation , @xmath56^{-1}\ , \end{split } \label{eq : full_prop_finite}\ ] ] with the frequency - dependent free propagator on the keldysh contour @xmath57 and the kernel @xmath58 , representing the self energy of the dyson equation due to coupling to the normal - conducting reservoir . the full propagator is broken up in two types of blocks on the keldysh contour , irreducible self energy insertions and free propagation , as depicted in fig . [ fig : contour ] ( a ) . the matrix elements of the free propagator are given by @xmath59 where @xmath60 denote the different dot states at different times @xmath61 . the kernel @xmath62 describes transitions between different reduced - density - matrix elements due to tunnel events between quantum dot and normal - conducting lead . the kernel @xmath62 is defined as the sum of all irreducible diagrams and can be obtained diagrammatically , based on a perturbation expansion in the tunnel - coupling to the normal - conducting lead as displayed in fig . [ fig : contour ] ( b ) . a tunnel event , where an electron hops between dot and normal - conducting lead , is represented by an internal vertex ( black dot ) on the keldysh contour . a directed tunnel line between two vertices indicates the contribution due to the contraction of two lead operators . the transformation into frequency space enters the diagrammatic representation by an additionaly horizontal bosonic line transporting the energy @xmath63 . finally , in the stationary limit , the reduced density matrix , @xmath64 , is found from the solution of a generalised master equation @xmath65\boldsymbol{p}_\mathrm{stat } \label{eq : master_equation}\ ] ] containing the coherent evolution of the reduced system described by the zero - frequency contribution to the free propagator and the dissipative coupling to the normal - conducting lead described by the zero - frequency contribution to the kernel . with the help of the solution for @xmath64 , we will in the following be able to determine the expectation values of the current and the current - current correlator yielding the finite - frequency noise . in the results part of this manuscript , we will restrict ourselves to the weak - coupling regime , performing a perturbation expansion with respect to the tunnel coupling to the normal lead . the explicit expression for the kernel can be obtained by using diagrammatic rules , see appendix [ app : rules ] . the current through the hybridised quantum dot is given by the operator representing the rate of change of the number of electrons in the normal lead : @xmath66 $ ] , where @xmath67 . when calculating the time - dependent expectation value of the current operator , the latter acts as an external vertex on the keldysh contour . in order to calculate the expectation value of the charge current , @xmath68 , the current operator is placed at the rightmost point of the keldysh contour , see fig . [ fig : noise ] ( a ) , and contracted to an internal tunnel vertex via a tunnelling line . it turns out that the current can be expressed as @xmath69\quad . \label{eq : current_average}\ ] ] the kernel @xmath70 can be obtained from @xmath62 by replacing one of the internal tunnelling vertices ( black dot ) by an external current vertex ( open circle ) . @xcite the current kernel @xmath70 takes into account whether an electron enters or leaves the dot through the normal lead . the diagrammatic rules to compute the kernel @xmath70 are summarised in appendix [ app : rules ] in lowest order in the tunnel coupling . the symmetrized finite - frequency current noise is defined as the fourier transform of @xmath71 , namely of the current - current correlator at different times , @xmath72\left(e^{-i\omega t}+e^{+i\omega t}\right ) \\ & \quad-4\pi\delta(\omega)\langle\hat{i}\rangle^{2}\quad . \end{split } \label{eq : noise_ft}\ ] ] by construction , the finite - frequency current noise eq . ( [ eq : noise_ft ] ) , also referred to as the power spectral density , is symmetric with respect to frequency , @xmath73 . it represents a real quantity , which can be measured by a classical detector.@xcite experimentally , the current noise can be measured in the normal lead . however , at finite - frequencies so - called displacement currents appear and the tunneling current @xmath74 is not equal to the measured currents . the displacement current can be included in the calculation by means of the ramo - shockley theorem @xcite . a derivation of the ramo - shockley theorem for the effective hamiltonian eq . can be found in appendix [ app : diss_current ] . in the following we assume the capacitance of the superconducting junction to be much larger than the capacitance of the normal junction . in this case the displacement current in the normal - conducting lead can be neglected , as discussed in appendix [ app : diss_current ] . this assumption is consistent with @xmath75 . in order to calculate the current correlator , two current operators at different times have to be placed on the keldysh contour . diagrammatically , this means that two internal tunnelling vertices have to be replaced by external current vertices . the contributions to the current - current correlator can be grouped into two different classes , as shown in fig . [ fig : noise ] ( b ) and ( c ) . either both current vertices are placed in the same irreducible block or in two different ones separated by a propagator . these external operators are connected by additional bosonic ( dashed ) lines , carrying the frequency @xmath63 of the fourier transform . the symmetrized finite - frequency noise can be written as @xcite @xmath76\nonumber\\ & & -2\pi\delta(\omega)\langle\hat{i}\rangle^{2}+(\omega\rightarrow -\omega)\quad . \label{eq : noise_diagram}\end{aligned}\ ] ] here , the kernels @xmath77 and @xmath78 are the sum of all diagrams , where one tunnel vertex ( black dot ) is replaced by a current vertex ( open circle ) and a frequency line @xmath63 is attached to the current vertex . the indices @xmath79 and @xmath80 indicate whether the frequency line leaves the diagram to the right or enters it from the left as shown in fig [ fig : noise ] ( c ) . the kernel @xmath81 contains diagrams with both current vertices in the same irreducible block ( see fig . [ fig : noise ] ( b ) ) . and the chemical potential of the normal lead @xmath82 both in units of the coulomb interaction strength @xmath9 . the other parameters are @xmath83 , @xmath84 and @xmath85.,scaledwidth=40.0% ] before discussing the finite - frequency current noise , we give a brief overview over the properties of the andreev current . when a finite bias voltage is applied across the quantum dot with one superconducting and one normal - conducting lead , a so called andreev current flows across the structure , which is due to cooper - pair tunnelling between quantum dot and superconductor caused by andreev reflection processes . @xcite we determine the current through the single - level quantum dot by using eq . , and show the result in fig . [ fig : current ] as a function of the chemical potential of the normal lead , @xmath82 , and the detuning , @xmath86 . the current is largest when superconducting correlations on the dot are strong and it furthermore shows features at the andreev addition energies . the andreev addition energies ( dashed lines in fig . [ fig : current ] ) are symmetric around zero bias voltage , @xmath87 , and with respect to zero detuning , @xmath88 . in the region around zero bias voltage the system is mainly in one of the singly - occupied states , @xmath29 , since the charging energy suppresses transitions from @xmath29 to the @xmath41 states . the andreev current is thus zero . only when the bias voltage is large enough , such that one of the conditions @xmath89 or @xmath90 is fulfilled , the quantum dot has a finite probability to be either empty or doubly occupied and the andreev current sets in . a further increase of the andreev current is observed , when also the other two addition energies , @xmath91 and @xmath92 , enter the bias window . outside the region where the current is suppressed due to the charging energy , the current is largest for @xmath93 , the regime of strongest superconducting correlation . we obtain a simple analytic result for the andreev current in the unidirectional transport regime , namely when @xmath94 , where the applied bias voltage to the normal conducting lead is much larger than all other energy scales in the system apart from the superconducting gap @xmath95 , @xmath96 as required , this result matches the current displayed in fig . [ fig : current ] , for @xmath97 . indeed , the current in the unidirectional regime , eq . , is maximal for @xmath98 , namely when @xmath99 , the splitting between the abss , is minimal and just given by the coupling strength to the superconducting lead @xmath100 . in this situation , the empty @xmath15 and doubly - occupied state @xmath101 are nearly degenerate and hence the mixing between them is maximal ( the proximity effect is on resonance ) . if the detuning @xmath86 becomes large , i.e. @xmath102 , the superconducting correlations on the dot are almost zero ( the proximity effect is off - resonance ) and the andreev current goes to zero as shown in fig . [ fig : current ] . the value of the andreev addition energies can however only roughly be extracted from the current . we will show in secs . [ sec : unidirectional ] , [ sec : quantum_noise ] and [ sec : high_bias ] that they lead to sharp features in the noise spectrum . in this section we come to the actual focus of the present paper , the finite - frequency noise associated to the current flow through the hybrid quantum - dot system , which we calculate based on the diagrammatic real - time approach introduced before . the discussion of the finite - frequency noise is divided into three parts : the unidirectional transport regime , where the applied bias voltage @xmath82 is chosen such that no back tunnelling to the normal lead is allowed , the finite bias regime , where the applied bias voltage can be of the same order of the andreev addition energies and the noise frequency , and the low - bias regime , where the current through the dot is suppressed . all regimes are shown to provide direct access to the internal dynamics of the system . depending on the applied bias voltage different frequency regimes of the noise are accessible . table [ table : regimes ] gives an overview over the different types of noise depending on the characteristic energy scales of the system , the thermal energy @xmath103 , the energy related to the noise frequency @xmath63 , and the applied bias voltage @xmath82 , which we are going to discuss within the different parts of this section . .table sumarising different noise regimes depending on noise frequency ( increasing from left to right ) and applied bias voltage ( decreasing from up to down ) . [ cols="<,<,<,<",options="header " , ] at low and intermediate frequencies the noise is dominated by time - dependent fluctuations in the conductance . in the limit of zero frequency ( @xmath104 ) the noise spectrum exhibits the information of a long - time measurement . in this frequency range , the equilibrium thermal noise , due to thermal fluctuations in the occupation number of the leads , is dominant in the spectrum if @xmath105 . in contrast , the so - called non - equilibrium shot noise , which is due to charge quantization , is dominant for @xmath106 . the quantum noise , which arises from zero - point fluctuations in the device , is dominant for high frequencies @xmath107 . it is a measure of the ability of the system to absorb or to emit a certain energy @xmath63 , @xcite and will therefore allow to visualize transport processes which are enabled or blocked by energy absorption or emission . although we here consider the symmetrized noise , we refer to the regime of high frequencies @xmath107 as quantum noise , as discussed e.g. in ref . . we start our analysis with the unidirectional transport regime , where we set the chemical potential of the normal lead @xmath82 to be much larger than all relevant energy scales of the system ( apart from the superconducting gap @xmath95 ) . in particular , since @xmath94 , all andreev levels are in the transport window and sufficiently far away from the chemical potential @xmath82 , thus allowing no back tunnelling from the dot into the normal conducting lead . a sketch of this situation is shown in fig . [ fig : model ] . the finite - frequency noise in the unidirectional transport regime is given by @xmath108 . \end{split}\ ] ] fig . [ fig : fano_high_coherences ] shows the finite - frequency noise in units of @xmath2 as a function of the noise frequency @xmath63 . two limiting cases are displayed , the one where the proximity effect on the dot is on resonance , @xmath98 ( red dashed line ) , and the one where the proximity effect on the dot is off resonance , @xmath109 ( blue solid line ) . in different frequency regimes , the spectrum shows sharp features , which depend on the strength of the detuning , @xmath86 . in the following subsections we discuss first the low - frequency noise , followed by a discussion of the intermediate - frequency regime . in the unidirectional transport regime for @xmath83 , @xmath84 both for the case where the proximity effect is on resonance , @xmath88 , and off resonance , @xmath110 . the peaks and dips are located at @xmath111 and at @xmath112 , the oscillation frequency of the cooper pairs . here and in the following figures we concentrate on the positive - frequency part of the spectrum.,scaledwidth=40.0% ] the first part of eq . is the low - frequency contribution to the current noise , @xmath114 it is indeed the only contribution to the noise , when the noise frequency is of the order of the coupling strength to the normal - conducting lead , @xmath113 . note that in the low - frequency noise for unidirectional transport only shot noise is present . on resonance ( @xmath115 ) , when the andreev current is maximal in the high - bias regime , see fig . [ fig : current ] , the noise is frequency independent and is given by two times the andreev current @xmath116 , which in the limit of zero - detuning discussed here equals @xmath117 . this means that the noise equals the long - time measurement result ( @xmath118 ) over the whole low - frequency range . this effect has previously been discussed in ref . : if the proximity effect is on resonance , the superconducting correlations on the dot are maximal and cooper pairs oscillate rapidly between dot and superconductor . this oscillation of cooper pairs is only interrupted by single - electron tunnel events from the normal conducting lead to the dot . it is these independent charge injections which give rise to a poissonian transfer of single electrons . when the proximity effect is off resonance ( @xmath109 ) , the low - frequency noise can be approximated by @xmath119 the noise spectrum shows a lorentzian dependence on the frequency @xmath63 , as shown by the low - frequency contribution of fig . [ fig : fano_high_coherences ] ( solid blue line ) . this maximum has a width given by the coupling strength @xmath2 and a height scaling with the magnitude of the andreev current . except for this maximum , the noise is overall suppressed with respect to the case on resonance . the reason for this is that in the unidirectional transport regime if @xmath120 , the andreev current becomes negligibly small with increasing detuning , as depicted in fig . [ fig : current ] . this behaviour is similar to the case of a quantum dot with normal conducting leads only , @xcite as presented in the appendix [ subsec : unidirectional_anderson_dot ] . in this purely normal - conducting case the low - frequency noise shows a lorentzian behaviour if the coupling to the two leads is asymmetric ( similar to what we observe in the case of finite detuning in the hybrid system ) . the noise is frequency - independent , when the coupling to the leads is symmetric . however the constant is only half as big as in the hybrid case discussed here , due to the absence of cooper pairs in the system . the remaining part of eq . carries the information of the noise in the intermediate - frequency regime , @xmath121 , where the noise frequency becomes larger than the coupling strength to the normal - conducting lead . in this regime , the noise starts to reveal the internal dynamics of the quantum dot . indeed , the finite - frequency noise shows resonance dips , whose position and size depend on the splitting of the andreev bound states ( equal to the frequency of the coherent oscillation between the empty and doubly - occupied dot states @xcite ) , @xmath99 , and hence on the strength of the proximity effect , see fig . [ fig : fano_high_coherences ] . such dips are characteristic for the noise spectrum of multilevel quantum dots . @xcite in the case of a hybrid quantum - dot system studied here , the dips arise from a coherent destructive interference between the abs , leading to features at @xmath122 . in the following we describe the properties of these dips . similar to the noise - enhancing peak in the low - frequency regime , the shape of these noise - suppressing resonance dips is lorentzian , see eq . , with a width given by @xmath2 . the depth of the resonance dip depends on the strength of the proximity effect and is equal to @xmath123 , the prefactor of the intermediate - frequency contribution in eq . . the resonance dip becomes most prominent if the proximity effect is on resonance and its depth is maximally equal to @xmath2 for @xmath115 . in contrast , the dip vanishes if the detuning @xmath86 becomes much larger than the coupling strength to the superconducting lead @xmath100 , namely when the superconducting correlations on the dot are almost zero , see fig . [ fig : fano_high_coherences ] . the resonance dip in the spectrum hence indicates the strength of the proximity effect . it is a signature of the coherent oscillation of cooper pairs between dot and superconductor . for even higher frequencies , @xmath124 , the noise in the unidirectional transport regime is given by two times the andreev current , which depends strongly on the detuning @xmath86 , see figs . [ fig : current ] and [ fig : fano_high_coherences ] . in the unidirectional transport regime the noise spectrum can be used to extract the splitting of the abs , but not the individual andreev addition energies . in order to get the information of the excitation energies and the effective coupling strengths of the abss , also back tunnelling to the normal - conducting lead must be allowed . hence , in the next two subsections we will consider the regime where the electrochemical potential of the normal lead is not the largest energy scale any more and the frequency can become larger than the distance between the andreev addition energies and the transport voltage , @xmath125 . in this regime quantum noise can become the dominant contribution to noise . [ sec : zero_bias ] if the transport voltage of the normal lead goes to zero , @xmath127 , the dot is in the singly - occupied state @xmath29 , see fig . [ fig : v0 ] ( a ) for a sketch of the energy landscape of the system . hence the transport excitation energies @xmath128 are outside the bias window and transport is blocked , meaning that the andreev current is zero as shown in fig . [ fig : current ] . in this limit shot noise is negligible and quantum noise is dominant in the spectrum , since @xmath129 , @xmath82 . also the thermal noise , which is expected to be dominant at zero bias up to a noise frequency of @xmath130 , is here suppressed , since no dot excitation energy is close enough to the fermi energy to allow for thermal excitations of the system . we show the finite - frequency noise in this regime in fig . [ fig : v0 ] and we observe that the noise spectrum has steps at frequencies @xmath131 equal to the andreev addition energies of the system . this behaviour is typical for the high - frequency noise of a system in which transport is blocked . at certain noise frequencies the effect of new `` noisy '' channels becomes visible , leading to steps increasing the noise . the steps thus reflect the internal structure of the energy levels on the quantum dot . here they occur at frequencies equal to the andreev addition energies of the dot - superconductor subsystem as described in detail in the following . an analogy to the normal - conducting case with symmetrically and asymmetrically coupled leads can be found in appendix [ app_lowbias ] . in the limit @xmath127 , the noise is suppressed until the noise frequency is equal to the energy which is necessary to excite the dot from the singly - occupied state @xmath29 to the @xmath132 state , @xmath133 . equally , also the inverse process , namely the excitation from the abs @xmath132 into the singly - occupied state @xmath29 yields a contribution to the noise . it takes place at @xmath134 . these excitation energies are however degenerate in the zero - bias limit , see fig . [ fig : v0 ] ( a ) , and consequently only one step occurs at the noise frequency @xmath135 . the step height is given by the sum of the respective effective coupling strengths , see eqs . and , @xmath136 . a second step takes place at @xmath137 , the energy necessary for the excitation between a singly - occupied state and the abs @xmath138 . the height of the second step is again given by @xmath2 , the sum of the effective coupling strengths , @xmath139 . consequently , when the noise frequency is larger than the energy which needs to be provided to excite between any of the singly - occupied states and the abss , the noise is constant and given by the sum of all four effective coupling strengths , @xmath117 . @xcite if a small ( positive ) bias voltage is applied to the normal - conducting lead , but the bias is still smaller than the energy @xmath141 , necessary to excite from the singly - occupied state @xmath29 into the abs @xmath132 , the dot is singly occupied and the system is hence still in the region where the current is suppressed ( see figs . [ fig : v02 ] ( a ) and [ fig : v02d ] ( a ) for the energy landscape of the system and fig . [ fig : current ] for the respective behaviour of the current ) . in this bias regime the noise spectrum exhibits four steps at noise frequencies @xmath142 as shown in figs . [ fig : v02 ] ( b ) and [ fig : v02d ] ( b ) . the reason for this is that a finite transport voltage breaks the degeneracy between the excitation energies that is present for @xmath143 . this is indicated by the different lengths of the green dashed arrows in the energy - landscape sketches of the system . the height of each of the four steps is given by the respective effective coupling strength , eq . and . in the limit of zero detuning , @xmath88 , the effective coupling to all four levels is equal , @xmath144 . the step heights shown in the noise spectrum in fig . [ fig : v02 ] ( b ) are therefore all equal to @xmath145 . in contrast , if the detuning @xmath86 is finite , the effective coupling strengths differ . for @xmath146 , the coupling for the excitation to go from the singly - occupied state @xmath29 to the @xmath138 state and to excite from the @xmath132 state into the singly - occupied state @xmath29 is stronger than for the other two excitations , @xmath147 . consequently these first excitations give a larger contribution to the noise than the latter ones and the noise spectrum exhibits steps with different heights , see fig . [ fig : v02d ] ( b ) . this behavior holds only as long as the detuning does not become much larger than the coupling strength to the superconducting lead . as soon as @xmath120 , the noise spectrum exhibits again only two steps because the superconducting correlations on the quantum dot vanish . the effective coupling strengths of the andreev levels corresponding to the excitation from @xmath29 to @xmath132 and from @xmath138 to @xmath29 go to zero , @xmath148 , while the other two excitations are coupled with the effective tunnel - coupling strength @xmath2 for @xmath120 , see eq . and . when setting the bias voltage to 0 , we find the previous result , as shown in fig . [ fig : v0 ] ( b ) . this is due to the fact that the coupling strength is here twice as large as in the case of zero detuning , however only half of the excitations contribute to the current and to the noise when @xmath120 . the finite - frequency noise spectrum in this low - bias regime provides a spectroscopy of the andreev levels as well as the effective coupling strengths . note , that in this case , namely when the andreev levels are outside the bias window , the noise steps always lead to an increase of the noise , regardless of whether the noise step is related to a tunnel process between the reservoir and a strongly or a weakly coupled dot resonance . the reason for this is that we here observe features at noise frequencies , which always correspond to energies necessary to _ excite otherwise blocked transport channels _ between dot resonance and reservoir . this is different if some of the andreev levels are in the bias window , as we will discuss in the next section . we consider in this section two different situations for a finite transport voltage applied to the normal - conducting lead : first , the voltage is applied such , that all andreev energies are in the bias window ( high - bias regime ) and second , such that part of the excitation energies lie outside the transport window ( intermediate - bias regime ) . this allows us to study the full noise spectrum with its different contributions , similar to what we observed separately in the previous sections , sec . [ sec : unidirectional ] and sec . [ sec : quantum_noise ] . the finite - frequency noise spectrum will be shown to provide a full spectroscopy of the system . we now assume the dot to be in a regime where the bias , @xmath150 , is chosen such that all andreev energies are in the bias window , see the sketches in fig . [ fig : finite_bias ] ( a ) , ( b ) and ( c ) . we consider the case where the superconducting correlations on the dot are strong , which is realized for the detuning being smaller than the coupling to the superconductor , @xmath151 . the results for the noise spectrum in this regime are shown in fig . [ fig : finite_bias ] , for @xmath152 ( d ) , @xmath88 ( e ) , or @xmath153 ( f ) . in this regime the andreev current is close to maximal , corresponding to the upper edge of the density plot in fig . [ fig : current ] . in the low- and intermediate - frequency regimes , @xmath154 , the noise spectrum shows lorentzian - shaped features , as discussed in sec . [ sec : unidirectional ] . these are a peak for a non - zero detuning @xmath86 in the low - frequency regime , @xmath155 , and dips at a noise frequency equal to the splitting of the abss , @xmath156 , due to a coherent destructive interference of the abss . the high - frequency part of the noise spectrum of fig . [ fig : finite_bias ] ( d ) and ( f ) exhibits quantum noise steps at frequencies @xmath142 , similar to what was discussed in sec . [ sec : quantum_noise ] . however , in contrast to the previous section , where all steps lead to an increase of the noise , the quantum noise steps found here show different signs _ depending on the effective coupling strength_. a noise process related to a strongly coupled andreev level leads to an increase of the noise , while a process between the electronic reservoir and a weaker coupled andreev level decreases the noise . the steps occurring in the finite - bias regime at high frequencies can be understood from an analogy to the ones obtained for a quantum dot coupled to normal - conducting leads . in this case an asymmetric coupling to the two normal - conducting leads takes the role of the differently coupled andreev levels in the hybridised dot . see appendix [ sec : anderson_noise ] for a detailed discussion of this simpler case . from this we deduce that the varying directions of the noise steps namely originate from the competition between different noise contributions of opposite sign . when a certain noise frequency @xmath63 is reached , an ensemble of transport processes becomes visible which can in principle involve different types of transitions , @xmath157 , due to tunneling with the normal lead . the contribution to the noise which stems from correlations of tunnelling processes involving only one type of transitions tends to increase the noise , while the noise contribution stemming from correlations of tunnelling processes with different transitions tends to decrease the noise . depending on which of these contributions has the larger magnitude , the step is positive or negative . the magnitude of the correlations in turn depends on the coupling strength of the involved processes . the addition of these two noise contributions hence yields the noise spectra shown in fig . [ fig : finite_bias ] ( d ) , ( e ) and ( f ) . in the following we discuss the implications of this effect for different magnitudes of the detuning @xmath86 . in fig . [ fig : finite_bias ] ( d ) , we observe a large increase of the noise at the noise frequency @xmath158 , because the related andreev level is coupled strongly to the normal - conducting lead . this noise frequency corresponds to the energy which an electron on the dot needs to absorb in order to tunnel out of the dot . the second step at @xmath159 occurs when the the noise frequency provides the energy for an electron to tunnel out of the weaker coupled andreev level @xmath141 . the noise step stemming from the process with the weaker coupled andreev level has a negative sign . in fig . [ fig : finite_bias ] ( d ) the noise increases again at @xmath160 and decreases at @xmath161 . for even higher frequencies , @xmath162 , the noise is always given by the sum of the effective coupling strengths , @xmath117 , since the noise frequency provides enough energy to excite from the singly - occupied state into either one of the abss and vice versa . if we invert the order of the excitation energies or the strength of their effective couplings , the broad maxima are transformed into troughs . this can for example be achieved by inverting the bias @xmath163 or the detuning @xmath164 . figure [ fig : finite_bias ] ( c ) shows the energy landscape and ( f ) the finite - frequency noise spectrum for @xmath153 . the detuning is reversed compared to the previously discussed case in fig . [ fig : finite_bias ] ( d ) . consequently , the noise is first suppressed , when a process with a weakly coupled andreev level takes place , and then enhanced to the value @xmath117 at a noise frequency related to a strongly coupled andreev level , see fig . [ fig : finite_bias ] ( f ) . the quantum steps in the noise spectrum occur only in a regime of intermediate detuning @xmath86 . if the detuning @xmath86 becomes larger than the coupling to the superconducting lead @xmath100 , the step structure as shown in fig . [ fig : finite_bias ] gets suppressed , because the superconducting correlations become weaker . if the proximity effect is on resonance and the detuning is exactly equal to zero ( @xmath88 ) the effective coupling strengths to the different andreev levels are equal . the finite - frequency noise spectrum for this case is displayed in fig . [ fig : finite_bias ] ( e ) with the corresponding energy landscape of the proximized dot , fig . [ fig : finite_bias ] ( b ) . in this regime , when the probabilities of the dot to be in any of the abss are equal , no steps but only shallow dips appear in the spectrum due to an almost complete compensation of the different noise contributions . the high - frequency noise spectrum is given by the sum of the effective coupling strengths , @xmath117 . only small features at @xmath165 remain as shown in the inset of fig . [ fig : finite_bias ] ( e ) . we finally also address the case of the intermediate bias regime , where only a part of the levels is in the bias window . in fig . [ fig : v05 ] ( a ) , we show the energy landscape of the dot considered here , with @xmath167 and @xmath152 , where the excitation energy @xmath168 is outside the bias window . the noise spectrum of the intermediate - bias regime , see fig . [ fig : v05 ] ( b ) , shows a mixture of the previously observed effects in the unidirectional , low - bias and high - bias regime . we can identify in the spectrum the features discussed in sec . [ sec : unidirectional ] , namely the lorentzian dependence in the low - frequency regime and the resonance dips at @xmath169 , which are a signature of the coherent transfer of cooper pairs between dot and superconducting lead . furthermore , the quantum noise steps show an overlap of the features discussed in the previous subsections : the steps at @xmath170 and @xmath161 both lead to an increase of the noise but with a different step height due to the fact that the corresponding andreev levels couple with different effective coupling strengths to the reservoir . the first of these steps overlaps with the intermediate - frequency regime , namely where the resonance dips due to the internal dynamics occur . the spectrum furthermore shows two steps at frequencies @xmath160 and @xmath159 , where the direction of the steps tells us if it is a noise process between the normal conducting lead and a strongly or weakly coupled andreev level . note that also the resulting trough is here partly found in the intermediate - frequency regime . at even higher frequencies , @xmath162 , the noise is again given by the sum of the effective coupling strengths . we have presented a calculation of the noise spectrum of a system composed by a a single - level quantum dot tunnel - coupled to a superconductor and a normal - conducting lead . we found that the noise spectrum reflects the internal spectrum of the proximized dot . resonance dips occur at a frequency equal to the splitting of the abss , @xmath169 . this feature is a signature of the coherent oscillation of cooper pairs between quantum dot and superconductor . the effect is strongest if the superconducting correlations on the dot are maximal , which happens when the proximity effect is on resonance , @xmath88 . in oder to observe this effect experimentally , the frequency of the resonance dip , approximately @xmath100 close to resonance , needs to be within the ghz frequency range . for example , in recent experiments @xcite the coupling to the superconductor is @xmath171 ghz . the high - frequency regime of the noise spectrum shows quantum - noise steps at frequencies @xmath142 . the quantum - noise steps provide not only information on the andreev addition energies of the system , but also on the effective coupling strength of the andreev levels to the normal conducting lead . the height ( in the low bias regime ) and sign ( in the finite - bias regime ) of the steps tell the strength of the effective coupling of each andreev level to the reservoir . therefore , we conclude that the finite - frequency noise spectrum provides a full spectroscopy of the proximized quantum dot . we thank christina pltl for helpful discussions . part of this work has been done at institute for theorie of statistical physics , rwth aachen university , with a financial support by the ministry of innovation , nrw . s.d . appreciates the hospitality at the rwth aachen university during a long - term visit . in this section we summarize the rules to determine diagrammatically the different contributions to the kernel and current kernel as given in refs . and . we adapt the rules for the system studied here , which has only one normal lead . the rules for the kernel @xmath58 are : 1 . draw all topologically different diagrams with @xmath173 directed tunnelling lines connecting pairs of vertices containing lead electron operators . assign spin index @xmath20 and energy @xmath174 to every tunnelling line . additionally , assign state index @xmath175 and the corresponding energy @xmath176 to each element of the keldysh contour connecting two vertices . also , add an external horizontal bosonic energy line transporting the energy @xmath63 to each diagram , which results from the fourier transform . 2 . for each time segment between two adjacent vertices write a resolvent @xmath177 with @xmath178 being the difference between all backward - going minus forward - going energies , including tunnelling lines transporting the energy @xmath174 as well as the external line transporting the energy @xmath63 . each vertex containing a dot operator @xmath179 gives rise to a matrix element @xmath180 where @xmath175 ( @xmath181 ) is the dot state entering ( leaving ) the vertex with respect to the keldysh contour . consequently , for each vertex connecting a doubly - occupied state @xmath182 to the up state @xmath183 , the diagram acquires a factor ( -1 ) . 4 . each tunnelling line contributes with a factor @xmath184 for a backward - going line with respect to the closed time path and a factor @xmath185 $ ] for a forward - going contribution . each diagram has an overall prefactor @xmath186 , where @xmath187 is the total number of vertices on the backward propagator and @xmath188 is the number of crossings of tunnelling lines . finally , sum over the spin @xmath20 and integrate over the energies @xmath174 of tunnelling lines and sum over all diagrams that contribute to the same kernel element . + + as a next step we provide the additional rules to determine the blocks containing one or two current operators @xmath189 and @xmath81 . replace one ( two ) tunnel vertex by a current vertex to calculate diagrams contributing to the kernels @xmath189 ( @xmath81 ) . note , that the current vertex ( open circle ) might also be placed on the start or end point of the diagram . 8 . multiply each diagram by a prefactor , determining the position of the current vertex inside the diagram : we have to multiply each diagram by a factor of @xmath190 for a current vertex on the upper ( lower ) keldysh time branch and a particle tunnelling into ( out of ) the normal lead . in the two other cases multiply the diagram with a factor of @xmath191 . 9 . the diagrams contributing to @xmath192 , @xmath193 have open external frequency lines to the right or left side attached to the current vertex . diagrams with frequency lines leaving the diagram to the right contribute to the kernel @xmath192 , while diagrams with frequency lines coming from the left contribute to @xmath193 . in this section of the appendix we present results for the current and the noise in a noninteracting single - level quantum dot coupled to two normal - conducting leads . the presentation of these known results , see refs . , is helpful as a comparison for the understanding of the more complex results for the interacting proximized dot studied in this paper . in several cases , the effect of the differently coupled andreev levels studied in the main part of this paper can be mimicked by considering asymmetric coupling of the dot to the two normal - conducting leads . the system studied in this appendix is shown in fig . [ fig : app ] ( a ) . note , that in this section , we consider the total current and its noise rather than that in a single contact . in the unidirectional transport regime , when the applied bias voltage is such that @xmath194 , the current is given by @xmath195 with @xmath196 . the finite - frequency noise in this unidirectional transport regime , where also @xmath197 is fulfilled , is given by @xmath198\ . \label{eq : fano_low_b}\ ] ] the noise shows a lorentzian dependence on the noise frequency @xmath63 . @xcite for a symmetric coupling of the dot to the normal conducting leads @xmath199 , the noise equals @xmath200 and is hence independent of the noise frequency . in order to get an insight into the parameters controlling the height of the steps occurring in the quantum noise regime , we here analyse the high - frequency noise spectrum in the regime of low and finite bias , where quantum noise is dominant . for @xmath201 the noise in the high - frequency regime is found to be given by @xmath202\nonumber\\ & + & \frac{1}{2}\frac{\gamma^{2}_\text{r}}{\gamma}\left[f^{+}_\text{r}(\epsilon)f^{-}_\text{r}(\epsilon+\omega)+f^{+}_\text{r}(\epsilon-\omega)f^{-}_\text{r}(\epsilon)\right]\nonumber\\ & + & \frac{1}{2}\frac{\gamma_\text{l}\gamma_\text{r}}{\gamma}\left[f^{+}_\text{l}(\epsilon+\omega)f^{-}_\text{r}(\epsilon)+f^{+}_\text{l}(\epsilon)f^{-}_\text{r}(\epsilon-\omega)\right]\nonumber\\ & + & \frac{1}{2}\frac{\gamma_\text{l}\gamma_\text{r}}{\gamma}\left[f^{+}_\text{r}(\epsilon)f^{-}_\text{l}(\epsilon+\omega)+f^{+}_\text{r}(\epsilon-\omega)f^{-}_\text{l}(\epsilon)\right]\nonumber\\ & + & \omega\rightarrow -\omega , \label{eq : high_freq_n}\end{aligned}\ ] ] with the fermi function @xmath203 for the two leads @xmath204 and @xmath205 . while the first two contributions result from correlations in the same lead , the latter two are related to correlations between different leads . in the following we will analyse the noise systematically for the different bias regimes and investigate the effect of an asymmetric coupling to the reservoirs ( @xmath206 ) on the noise spectrum . in the limit of @xmath207 , shot noise is negligible . thermal noise , which is generally cut off at @xmath208 , is here suppressed due to @xmath209 . hence quantum noise is dominant in this regime . the noise spectrum , fig . [ fig : app ] ( b ) , has one step at @xmath210 . since at zero bias all factors in eq . ( [ eq : high_freq_n ] ) containing fermi functions are equal , the step height is given by @xmath200 . an asymmetric coupling to the leads does hence not influence the shape of the noise spectrum . when a finite transport voltage is applied , but the transport level is outside the bias window , quantum noise is still the dominant noise contribution and the noise spectrum exhibits two steps . for the situation shown in fig . [ fig : app ] ( c ) , when @xmath212 , the quantum dot is unoccupied in the stationary regime and all factors in eq . ( [ eq : high_freq_n ] ) containing @xmath213 are zero . then the first step stems from the contributions of the first and the third term of eq . ( [ eq : high_freq_n ] ) for which the factor containing fermi functions is equal . it occurs at @xmath214 when the excitation of the dot from the left lead becomes visible and it has the height @xmath215 . analogously , the second step at @xmath216 has height @xmath217 . in both places an increase of the noise is observed as long as the dot level is outside the bias window , because in both cases the effect of an otherwise blocked transport channel becomes visible . [ fig : app ] ( c ) shows two noise spectra , for a symmetrically coupled quantum dot @xmath199 ( red dashed line ) and an asymmetrically coupled dot @xmath218 ( black solid line ) . we finally consider the case , where the energy level lies inside the bias window and shot noise as well as quantum noise is present . when choosing asymmetric coupling to the leads , we find a situation which can be compared to the proximized quantum dot with finite detuning as discussed in the main text . in fig . [ fig : app ] ( e ) the noise is displayed for a situation where the left lead is coupled much stronger to the quantum dot @xmath219 . the lorentzian behaviour of the low - frequency noise , @xmath220 , is described with the expression given in eq . ( [ eq : fano_low_b ] ) . furthermore , steps occur at @xmath221 and @xmath222 . the first step at @xmath221 increases the noise . at this frequency back - tunnelling of an electron to the strongly coupled left lead , emptying the quantum dot , becomes visible . its height is given by @xmath223 . the second step at @xmath216 , which occurs when an electron can tunnel back onto the dot from the right lead , results in a decrease of the noise . its depth is given by @xmath224 . the high frequency noise is again given by @xmath200 . whether a step in the noise leads to an increase or a decrease of the total noise depends on the coupling strength of the different excitations . the reason for this is that , in the regime where the level is in the bias window , the part of the noise stemming from correlations in the same lead increases at the frequencies equal to the excitation energies with an amount given by the square of the respective coupling strength , while the noise due to correlations between different leads decreases by an amount which is always proportional to @xmath225 . this can be directly read off from eq . ( [ eq : high_freq_n ] ) . depending on whether the positive contribution multiplied by @xmath226 or the negative contribution multiplied by @xmath225 is larger the step hence changes direction . [ fig : app ] ( f ) shows the noise for the same applied bias voltage but with the right reservoir coupled stronger than the left reservoir , @xmath227 . the noise spectrum therefore shows a reversed order of the steps with respect to the result shown in fig . [ fig : app ] ( e ) , leading to an occurrence of troughs rather than plateaus in the noise . the situation for a symmetrically coupled quantum dot ( @xmath228 ) where the probabilities of the quantum dot to be empty or singly occupied are equal , is shown in fig . [ fig : app ] ( d ) . the noise spectrum shows no quantum noise steps at @xmath229 due to an almost complete compensation of the noise stemming from correlations in the same lead and correlations between the two leads . the sum of these contributions to the noise leads to the shallow dip structure in the symmetrized noise spectrum as displayed in fig . [ fig : app ] ( d ) . for frequencies larger than the bias voltage , the noise takes again the value @xmath200 . the case of a quantum dot symmetrically coupled to normal - conducting leads is equivalent to the situation of zero detuning in the high bias regime in the main text . we consider a simple capacitive model and denote with @xmath230 and @xmath231 the capacitances of the tunnel barriers with the normal and superconducting lead , respectively . the number of electrons in the dot is @xmath232 . within this model the displacement current in the normal lead is@xcite @xmath233 the total current is simply the sum of the tunnelling current and the displacement current and reads @xmath234 with the tunneling current given by @xmath235 \\ & = \frac{i}{\hbar}\sum_{k,\sigma}\left(t_\mathrm{n}c^{\dagger}_{k\sigma}d_{\sigma}-t^{*}_\mathrm{n}d^{\dagger}_{\sigma}c_{k\sigma}\right ) \ . \end{split}\ ] ] note , that the tunneling current @xmath236 is equal the current @xmath74 introduced in sec . [ sec : current ] . the current @xmath237 is positive when flowing out of the normal lead . at this stage , it is worth mentioning that if @xmath238 , the displacement current in the normal lead can be neglected . this assumption is consistent with @xmath239 . now , we proceed to evaluate @xmath240 : @xmath241=\hat{i}_{\text{n},\text{tunn}}+\frac{1}{i\hbar}[\hat{n},h_{\text{eff}}]\\ & = \hat{i}_{\text{n},\text{tunn}}+\hat{i}_{\text{s},\text{tunn } } , \end{split}\ ] ] where the tunnelling current with the superconductor reads @xmath242 putting everything together we obtain the ramo - shockley theorem for the n - dot - s system : @xmath243 74ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevb.86.134528 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.104.026801 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) http://dx.doi.org/10.1038/nphys1811 [ * * , ( ) ] http://dx.doi.org/10.1038/nphys1911 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) http://dx.doi.org/10.1038/nnano.2013.267 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/revmodphys.82.1155 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) |
we briefly review the hopf algebra structure arising in the renormalization of quantum field theories . we construct the hopf algebra explicitly for a simple toy model and show how renormalization is achieved for this particular model . | 1505.04765 | the underlying hopf algebraic strucuture of the process of renormalization was discovered first by kreimer in @xcite . further progress was made in formulating the renormalization procedure in the language of hopf algebra and doing explicit computations using this algebraic structure , in @xcite . the purpose of this article is to briefly review this algebraic structure . for simplicity , we avoid many of the technicalities of the quantum field theory by considering a simple toy model containing only nested divergences . issues related to overlapping divergences and more realistic field theoretic models have been discussed in literature ( see refs . @xcite and @xcite ) . for a more detailed review of this subject see @xcite . this article closely follows the conventions and notation used in @xcite . for a detailed treatment of renormalization procedure , see @xcite . for a mathematically rigorous introduction to hopf algebra , see @xcite . the article is organized as follows . in section [ sec : pre ] , we describe necessary notation and conventions . in section [ sec : halgebra ] we explicitly construct the hopf algebra structure . for clarity of our arguments and construction , most of the proofs have been relegated to appendices . the article is concluded with a very simple example in appendix [ app : example ] a brief summary of the bphz renormalization procedure and the derivation of the forest formula is given in the appendix [ app : bphz ] . the key result of bphz renormalization is an iterative formula ( forest formula ) which gives a renormalized feynman graph in terms of the divergent graph , its subgraphs and the corresponding counter terms . forest formula can be written in a schematic form as follows : @xmath0 where @xmath1 and @xmath2 are bare and renormalized graphs respectivley . @xmath3 is the graph with all the subdivergences removed . the sum is over all non - empty proper forests of @xmath1 . @xmath4 and @xmath5 are counter terms . @xmath6 is a renormalization scheme dependent operator , which removes the overall divergence associated with graph @xmath1 . to make the notion of forest precise , let @xmath7 be all 1pi , non overlapping divergent subgraphs of @xmath1 , then a proper forest of @xmath1 is any subset of the following set : @xmath8 we would like to represent feynman graphs in a more algebraic fashion such that their forest structure and subdivergences become manifest . this would be done by representing them as ` parenthesized words ' . parentheses encode information about the nestedness or the disjointness of the subdivergences and letters appearing in these words correspond to graphs without subdivergences . parenthesized words can be assigned to a graph by the following procedure : * for every forest we write down a pair of brackets respecting the forest structure , i.e. , if a forest @xmath9 is inside a forest @xmath10 then the pair of brackets corresponding to the forest @xmath9 are contained inside the pair of brackets corresponding to @xmath10 . * consider a given pair of brackets , if we shrink all the brackets / forests inside it to a point the remainder is a graph @xmath11 without any subdivergences . we write the letter corresponding to @xmath11 next to the right closing bracket of the pair of brackets under consideration . * rest of what is contained in the pair under consideration is written to the left of this letter . for an example , consider the diagram in figure [ fig : twodiv ] . it has two disjoint subdivergences and is overall divergent when the subdivergences are shrunk to a point . the two subdivergences are contained in rectangular boxes . these subdivergences themselves are both 1pi and do not contain any subdivergences . in the figure we have also shown the letters corresponding to these subdivergences . it is easy to see that , using our rules above , this diagram corresponds to the parenthesized word @xmath12 . important features of this construction are following . * disjoint forests and configurations inside disjoint pair of brackets commute in this construction . i.e. , @xmath13 * only the forest structure of the graph is made manifest in this construction and we lose information about to which propagator or to which vertex of a graph @xmath14 another graph @xmath15 is attached . several different attachment can yield the same forest structure . hence any feynman diagram belongs to a class given by a parenthesized word . for example , the two diagrams in figure [ fig : samepw ] belong to the class represented by the parenthesized word @xmath16 . + ] * a letter @xmath17 has one and only one closing bracket on its right side while it can have more than one opening brackets . * we include the empty graph as @xmath18 which would act as a unit element ( not to be confused with the unit map ) in the construction of the hopf algebra . * an important characteristic of a parenthesized word is its length , which is simply the total number of letters @xmath17 appearing in it . for example , in collection ( [ coll ] ) , the parentheized words have lengths @xmath19 respectively . * in general we will have a class of feynman graphs represented by the notion of parenthesized words constructed out of letters @xmath17 . some examples are : @xmath20 * a parenthesized word , whose left most bracket is matched with its right most bracket is called an irreducible parenthesized word and corresponds to a 1pi feynman graph . examples are : @xmath21 an arbitrary irreducible parenthesized word can be represented as @xmath22 , where @xmath23 is an any parenthesized word . * a parenthesized word , whose left most and the right most brackets do not match with each other is called a reducible parenthesized word and can be written as product of irreducible parenthesized words . for example @xmath24 is a reducible parenthesized word and is written as a product of two irreducible parenthesized words @xmath25 and @xmath26 . r0.4 hh & & & & hh & + m & + h & & k & & h & + m & + hh & & _ s & & hh & + a detailed discussion of the mathematical properties of hopf algebra will lead us off topic . in this subsection , we will give the formal definition of a hopf algebra and different elements appearing in the definition . we will also give a rough sketch of how the procedure of renomalization can be described by an underlying hopf algebra structure . these notions will be made more precise in the next section . formally a hopf algebra is defined as following . [ defhopf ] a hopf algebra is an associative and co - associative bialbegra @xmath27 over a field @xmath28 with a k - linear map @xmath29 , called antipode such that the diagram [ fig : hopf ] commutes . @xmath30 are called unit , co - unit , product and co - product maps respectively . the condition for the commutativity of the diagram can be written algebraically as : @xmath31\right]=m\left[\left({\mathbf{1_d}}\otimes s\right)\delta\left[x\right]\right]=e\circ \overline{e}\left[x\right]\label{hopfcom}\end{aligned}\ ] ] where @xmath23 is an element of hopf algebra and @xmath32 is the identity map . now we will give a brief overview of how renormalization would turn out to be related to the hopf algebra structure . * basic objects of the hopf algebra are feynman graphs @xmath1 which will be represented by the corresponding parenthesized word @xmath33 . representatives of the overall divergent graphs without subdivergences will be identified as the primitive elements of the hopf algebra . all other elements @xmath33 can be built out of these primitive elements . * the co - product resolves the graph into its forests . @xmath34=\sum_{\text{all forests } \gamma}\ x_{\gamma}\otimes x_{\gamma/\gamma}.\end{aligned}\ ] ] * we have a renormalization map @xmath35 , which extracts the divergent parts of a graph ( depending on the renormalization scheme ) . * the antipode @xmath36 gives the counter term @xmath5 through the renormalization map . @xmath37=-r\left[x_{\gamma}\right]-\ \sum_{\text{all non - empty proper forests } \gamma}r\left[s_r\left[x_{\gamma}\right]x_{\gamma/\gamma}\right].\end{aligned}\ ] ] * the renormalized feynman graph will related to the term @xmath38\right]$ ] , appearing in the condition of the commutativity . we would indeed see that @xmath38\right]=0 $ ] , expressing the fact that the we get a finite result . in this section we will construct the hopf algebra related to renormalization . this will be done by explicitly defining the all the maps and elements appearing in the definition ( [ defhopf ] ) . we will proceed in several steps , establishing algebra , co - algebra , bialgebra and finally hopf algebra structure . as discussed in the previous section , we will represent feynman diagrams by parenthesized words . we will arrange these parenthesized words into an algebra structure here . let @xmath39 be the set of all parenthesized words . we regard this as a @xmath40 vector space . it is easy to see that @xmath39 is a vector space over @xmath40 . now , we introduce a bilinear product map as follows : @xmath41\ \equiv \ \ xy \equiv yx,\ \ \ \forall\ x , y \in \mathcal{a}.\label{defprod}\end{aligned}\ ] ] also we have an identity element @xmath42 which satisfies : @xmath43 to understand the product ( [ defprod ] ) consider the example with @xmath44 and @xmath45 then @xmath46 is a well defined product given by @xmath47 , i.e. , the product of two parenthesized words give a reducible parenthesized word . by introducing the product we have furnished @xmath39 with an algebra structure . now we define a homomorphism ( the unit map ) from @xmath40 to the set @xmath39 as follows : @xmath48&\ \equiv e,\ \ \ \forall\ \text{rational numbers } q.\label{defunit}\end{aligned}\ ] ] now , by definition , the bilinear product @xmath49 is associative , our algebra @xmath39 has an identitiy element @xmath50 and we have constructed a homomorphism from the field of rational numbers @xmath40 to algebra @xmath39 , this means that the set @xmath39 is a unital associative algebra . in this subsection , we furnish @xmath39 with the structure of a coalgebra . let us first give the formal definition of a coalgebra . [ defcoalg ] a coalgebra , @xmath51 over a field @xmath28 is a vector space @xmath51 over @xmath28 together with linear maps @xmath52 ( counit ) and @xmath53 ( coproduct ) such that @xmath54 where @xmath32 is the identity map on @xmath51 , or quivalently , the two diagrams in figure [ fig : coalg ] commute . in the second diagram , we have identified the naturally isomorphic spaces @xmath51 , @xmath55,@xmath56 . the second equation above is also called the coassociativity condition for the coproduct @xmath57 . c & ^&cc & & & c&^ & & cc + & & & & + cc&_&ccc & & & cc & _ & & kccck now , we will define the counit and the coproduct maps for the set @xmath39 under consideration . we define a counit by : @xmath58&\equiv \ 1,\\ & \overline{e}\left[x\right]&\equiv \ 0,\ \ \forall\ x\neq e , \in \mathcal{a}.\ \ \ \ \ \ \ \ \ \ \ \label{defcounit}\end{aligned}\ ] ] this definition is motivated by the fact that there is no rational number which should be assigned naturally to an arbitrary parenthesized word and thus the counit annihilates feynman graphs . on the other hand we assign the rational number @xmath59 to the empty graph @xmath50 . the definition of the coproduct is more involved as compared to the elements defined so far . roughly speaking , coproduct yields a sum of terms @xmath60 , where the first terms , @xmath61 , are to be identified with divergent subgraphs and the second terms , @xmath62 , correspond to the remainder of the graph obtained by reducing @xmath63 to a point . to give a rigorous definition of the coproduct , it will be useful to define a projection map @xmath64 as follows : @xmath65 it is easy to confirm the following properties of the map @xmath64 by explicit computation . @xmath66&=&0,\ \ \forall\ x\in \mathcal{a},\\ p\left[x\otimes y\right]&=&x\otimes y,\ \ \ \ \forall\ x\neq e,\ y , \in \mathcal{a},\\ p^2&=&p.\end{aligned}\ ] ] we also define a useful endomorphism @xmath67 , which is parametrized by a single letter @xmath17 , corresponding to a primitive graph . @xmath68&\equiv \left(xx_i\right).\end{aligned}\ ] ] for example , @xmath69=\left(\left(x_2\right)x_1\right)$ ] . with the help of the maps @xmath64 and @xmath10 , we are now in a position to define the coproduct as follows . @xmath70&\ \equiv\ e\otimes e,\\ & \delta\left[\left(xx_i\right)\right]&\ \equiv\ \left(xx_i\right)\otimes e+e\otimes\left(xx_i\right)+\left({\mathbf{1_d}}\otimes b_{\left(x_i\right)}\right)\left[p\left[\delta\left[x\right]\right]\right].\label{coprod}\end{aligned}\ ] ] this definition of the coproduct is complete . it is easy to use the above definition to show an important property of the coproduct . @xmath71&\ e + e\otimes\left(x_i\right).\label{coprodprim}\end{aligned}\ ] ] another important property of the coproduct is : @xmath72&\ \equiv \ \delta\left[x\right]\delta\left[y\right].\label{compatcoprod}\end{aligned}\ ] ] this can also be shown by using the definition ( [ coprod ] ) , however the proof is a bit involved . the proof is based on the standard induction argument on the length of the words @xmath23 and @xmath73 . another way to write the coproduct is by using the sweedler s notation , @xmath74=\sum_x x_1\otimes x_2 $ ] , where the sum is over the subwords @xmath75 of @xmath23 and @xmath76 . proof of this assertion is given in appendix [ app : sweed ] . using this notation and the properties of the map @xmath64 , we can write the equation ( [ coprod ] ) of the coproduct as : @xmath77=\left(xx\right)\otimes e + \left({\mathbf{1_d}}\otimes b_{\left(x\right)}\right)\left[\sum_x x_1\otimes x_2\right]\label{coprodsweed}.\end{aligned}\ ] ] let us now consider a few example to explain how the coproduct acts on the elements of the set @xmath39 @xmath78}&=&{\left({\left(x_i\right)}x_j\right)}\otimes e + e\otimes { \left({\left(x_i\right)}x_j\right)}+{\left({\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}p{\left(\delta{\left[{\left(x_i\right)}\right]}\right)},\\ & = & { \left({\left(x_i\right)}x_j\right)}\otimes e + e\otimes { \left({\left(x_i\right)}x_j\right)}+{\left({\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}{\left[{\left(x_i\right)}\otimes e\right]},\\ & = & { \left({\left(x_i\right)}x_j\right)}\otimes e + e\otimes { \left({\left(x_i\right)}x_j\right)}+{\left(x_i\right)}\otimes { \left(x_j\right)}.\label{eq : exampledel1}\end{aligned}\ ] ] 2 . using the similar method ( but after more tedious algebra ) we can also compute : @xmath79}=&&{\left({\left(x_i\right)}{\left(x_j\right)}x_k\right)}\otimes e + e\otimes { \left({\left(x_i\right)}{\left(x_j\right)}x_k\right)}+{\left(x_i\right)}\otimes { \left({\left(x_j\right)}x_k\right)}\nonumber \\ & & + { \left(x_j\right)}\otimes { \left({\left(x_i\right)}x_k\right)}+{\left(x_i\right)}{\left(x_j\right)}\otimes { \left(x_k\right)}.\label{eq : exampledel2}\end{aligned}\ ] ] we have defined the counit and the coproduct maps for @xmath39 , but in order to furnish the coalgebra structure on @xmath39 we need to show that these maps satisfy the equations ( [ coalg1 ] ) and ( [ coalg2 ] ) . the first of these relations is trivial to show due to the definition of the counit as : @xmath80}=x={\left(\overline{e}\otimes{\mathbf{1_d}}\right)}\delta{\left[x\right]}.\end{aligned}\ ] ] next , we want to show the equation ( [ coalg2 ] ) holds . this can be proved using induction on the length of the words . a detailed proof is given in the appendix [ app : coalg ] . after successfully defining a counit and a coproduct on @xmath39 , we have completed the construction of the coalgebra structure on @xmath39 . we have already established the fact that @xmath39 is a unital coassociative algebra . the property ( [ compatcoprod ] ) ensures that the algebra and the coalgebra structures are compatible . this implies that @xmath39 is actually a bialgebra . to complete the construction of the hopf algebra , what remains to find is an antipode . it turns out that antipode is actually the object which achieves the renormalization , it combines the terms generated by the coproduct and combines them in a way which is similar to the forest formula . we define the antipode as follows : @xmath81}&=&e,\\ s{\left[{\left(x_i\right)}\right]}&=&-{\left(x_i\right)},\\ s{\left[xy\right]}&=&s{\left[y\right]}s{\left[x\right]},\\ s{\left[{\left(xx_i\right)}\right]}&=&-{\left(xx_i\right)}-m{\left[{\left(s\otimes{\mathbf{1_d}}\right)}p_2{\left(\delta{\left[{\left(xx_i\right)}\right]}\right)}\right]},\label{eq : sdef1}\\ s{\left[{\left(xx_i\right)}\right]}&=&-{\left(xx_i\right)}-m{\left[{\left({\mathbf{1_d}}\otimes s\right)}p_2{\left(\delta{\left[{\left(xx_i\right)}\right]}\right)}\right]},\label{eq : sdef2}\\ p_2&\equiv & { \left({\mathbf{1_d}}-e\circ\overline{e}\right)}\otimes { \left({\mathbf{1_d}}-e\circ\overline{e}\right)}\equiv p_1\otimes p_1.\end{aligned}\ ] ] this completely defines the antipode . however , we need to show that this antipode is actually well defined and induces a hopf algebra structure . this amounts to showing that equations ( [ eq : sdef1 ] ) and ( [ eq : sdef2 ] ) are equivalent and also the condition ( [ hopfcom ] ) is satisfied . equivalence of the two definitions follow from the associativity of the product @xmath49 and the coassociativity of the coproduct @xmath57 . the detailed proof is given in appendix [ app : antipodedef ] . the proof that the condition ( [ hopfcom ] ) is satisfied , is given in appendix [ app : hopfcheck ] . we have now completely furnished the set of all feynman diagrams , @xmath39 with the structure of a hopf algebra . we have not yet discussed precisely how the renomalization is achieved by this structure . this will be the subject of the next section . in this section , we describe how the hopf algebra constructed above produces the forest formula , generates counter terms and the renormalized feynman graphs . we will see that an important ingredient in this regard is the renormalization map , @xmath35 , which is renormalization scheme dependent . given a feynman graph @xmath1 , we associate a parenthesized word @xmath33 to it . using the feynman rules we obtain an integral expression associated with the graph @xmath1 , denote it by @xmath82 @xmath83 , where @xmath84 is a vector space , endowed with suitable structure which is not important for our considerations . for example , it could be the space of laurent polynomials in the regularization parameter . these feynman integrals are subject to some renormalization conditions which are described by renormalization map @xmath85 . the renormalization map depends on the renormalization scheme , for example , in the case of minimal subtraction , @xmath35 picks out the only the divergent part of @xmath82 . the map @xmath86 , the renormalization map @xmath35 and the antipode of the hopf algebra @xmath36 give rise to a map @xmath87 at the level of the feynman integrals , which is written as : @xmath88}=-r{\left[\phi{\left({\left(xx\right)}\right)}\right]}-r{\left[m{\left[{\left(s_{r}\otimes\phi\right)}p_2{\left(\delta{\left[{\left(xx\right)}\right]}\right)}\right]}\right]}.\label{eq : antipode2}\end{aligned}\ ] ] with @xmath89}=e$ ] . this map @xmath87 gives the counter terms for a given graph depending on the particular renormalization scheme @xmath35 . consider the following examples where , for simplicity , we omit writing @xmath86 explicitly : 1 . @xmath90}=-r{\left[{\left({\left(x_i\right)}x_j\right)}\right]}-r{\left[m{\left[{\left(s_r\otimes { \mathbf{1_d}}\right)}p_2{\left(\delta{\left[{\left({\left(x_i\right)}x_j\right)}\right]}\right)}\right]}\right]}.\end{aligned}\ ] ] + we can use @xmath91}$ ] as computed in equation ( [ eq : exampledel1 ] ) . since , @xmath92 annihilates @xmath50 , we finally find that : @xmath90}=-r{\left[{\left({\left(x_i\right)}x_j\right)}\right]}+r{\left[r{\left[{\left(x_i\right)}\right]}{\left(x_j\right)}\right]}.\end{aligned}\ ] ] 2 . similarly , after a straightforward but tedious computation one can find that : @xmath93}&=&-r{\left[{\left({\left(x_i\right)}{\left(x_j\right)}x_k\right)}\right]}+r{\left[r{\left[{\left(x_i\right)}\right]}{\left({\left(x_j\right)}x_k\right)}\right]}+r{\left[r{\left[{\left(x_j\right)}\right]}{\left({\left(x_i\right)}x_k\right)}\right]}\nonumber \\&&-r{\left[r{\left[{\left(x_i\right)}\right]}r{\left[{\left(x_j\right)}\right]}{\left(x_k\right)}\right]}.\label{eq : example}\end{aligned}\ ] ] let us now proceed further to show that the forest structure in equations ( [ eq : for1],[eq : for2],[eq : for3 ] ) emerges from the hopf the algebra structure . let @xmath94 be a subword of @xmath23 , then by using the representation of the coproduct in sweedler s notation and the fact that @xmath92 annihilates @xmath50 , the antipode can be written as : @xmath95}=- x - \sum_{u\neq e , x } s{\left[u\right]}{\left(x / u\right)}.\end{aligned}\ ] ] if the parenthesized word @xmath23 is associated to a feynman graph @xmath1 then the subwords @xmath96 are associated to the proper forest @xmath97 of the graph @xmath1 . using this fact , we can now write the map @xmath87 in the following way : @xmath98}&=&-r{\left[\gamma\right]}-\sum_{\text{proper forest } \gamma\subset \gamma } r{\left[s_{r}{\left[\gamma\right]}\gamma/\gamma\right]},\\ & = & -r{\left[\gamma+\sum_{\text{proper forest } \gamma\subset \gamma } s_{r}{\left[\gamma\right]}\gamma/\gamma\right]}.\end{aligned}\ ] ] now , if we identify @xmath99}$ ] with counter term associated to the subgraph @xmath97 , then the argument of the map @xmath35 in the above equation is just @xmath3 , the graph @xmath1 with all its subdivergences renormalized as defined in equation ( [ eq : for2 ] ) . we can also identify the renormalization map @xmath35 with the operator @xmath6 , both are renormalizatio scheme dependent operators and picks out just the divergent part of a feynman integral in ms scheme . with this identification , we see that @xmath100}$ ] just gives the counter term @xmath5 and we recover the forest structure of equations ( [ eq : for1],[eq : for2],[eq : for3 ] ) . the renormalized feynman graph @xmath101 is obtained as follows . let @xmath23 be the parenthesized word associated with the graph @xmath1 ( we will use the parenthesized word @xmath23 , the corresponding graph @xmath1 and the corresponding feynman integral @xmath102 interchangeably ) then : @xmath103}\delta{\left[x\right]}&=&m{\left[s_{r}\otimes \phi\right]}{\left(e\otimes x + x\otimes e+ \sum_{\text{subwords } u\neq e , x}u\otimes { \left(x / u\right)}\right)},\\ & = & \phi{\left[x\right]}+s_r{\left[x\right]}+\sum_{\text{subwords } u\neq e , x } s_{r}{\left[u\right]}\phi{\left[x / u\right]},\end{aligned}\ ] ] in the last equation , the first term is just the feynman integral associated with graph @xmath1 , the second term is the counter term @xmath5 and the last term just removes the subdivergences as we have seen earlier . now , we omit writing @xmath86 and replace parenthesized words with the respecting graphs to find : @xmath103}\delta{\left[x\right]}=\gamma + z_{\gamma}+\sum_{\text{proper forests } \gamma\subset \gamma}z_{\gamma}\gamma/\gamma=\overline{\gamma}+z_{\gamma}=\gamma_{ren}.\end{aligned}\ ] ] earlier , we showed that at the hopf algebra level , the operator @xmath104}$ ] annihilates any parenthesized word other than the unit @xmath50 . this expresses the fact that at the level of the feynman integrals we will get essentially a finite result . in this section we will briefly summarize the key results of this article . by representing the feynman diagrams as parenthesized words , we furnished them into a set @xmath39 . we also included the empty graph , represented by the unit element @xmath50 , in that set . then we introduced an algebra structure on @xmath39 by defining a bilinear product @xmath105 . we also defined a unit map @xmath106 , furnishing @xmath39 into a unital associative algebra . next , we introduced the coalgebra structure on @xmath39 by defining the counit map and the coproduct map . the coproduct was defined in such a way that it was compatible with the product @xmath49 and hence we obtained a bialgebra structure on @xmath39 . to complete the construction of the hopf algebra , we defined an antipode map @xmath107 . we also showed that the struture of the forest formula is recovered if we identify the antipode with the counter term of a specific graph . to make this notion precise , we defined a map @xmath108 , which assigns a parenthesized word an analytic expression ( feynman integral ) using the feynman rules . we defined the renormalization map @xmath35 which gives the divergent part of a feynman integral . it turned out antipode @xmath36 induced the counter term for a graph via @xmath35 . the most important result we obtained is the equivalence of the antipode and the forest formula . this equivalence followed by making a set of identifications between the elements of the hopf algebra and the objects of the standard renormalization theory . we list these identifications here . * 1pi feynman graph @xmath1 with subdivergences are identified with irreducible parenthesized word @xmath109 whose bracket structure matches the forest structure of @xmath1 , and the letters label the components of @xmath1 obtained after reducing the subdivergences to a point . * the counter term @xmath5 is identified with @xmath100}$ ] . * the feynman graph , with all its subdivergences renormalized , @xmath3 is identified with the object : @xmath110}\right]},\ \ \text{where}\ \ \ p_r={\mathbf{1_d}}\otimes{\left({\mathbf{1_d}}-e\circ\overline{e}\right)}.\end{aligned}\ ] ] * the renormalized feynman graph @xmath111 is identified with : @xmath112}\right]}.\end{aligned}\ ] ] consider a feynman graph @xmath1 . by using feynman rules we can obtain the corresponding analytic expression @xmath113 . in general this expression can be written as a laurent series in the regularization parameter @xmath114 . if we consider @xmath86-cubed theory in @xmath115 spacetime dimensions and use dimensional regularization then @xmath116 where @xmath117 are some coefficients and the integer @xmath118 is bounded above by the number of loops in the graph @xmath1 , which can be shown explicitly . we stress here that in the general argument for the bphz renormalization nothing depends crucially on the particular toy model chosen here . let us now define a ` subtraction ' operator associated with the graph @xmath1 as follows @xmath119 i.e. , it picks out the divergent part of @xmath113 . in general , the subtraction operator is renormalization scheme dependent , here we have chosen the minimal subtraction scheme . the finite part of the graph can now be written as : @xmath120 so , we see that the term ` @xmath121 ' provides the counter term for the graph @xmath1 and @xmath122 removes the divergence associated with graph @xmath1 and makes it finite in the @xmath123 limit . now , consider the graph @xmath1 to have proper 1pi subgraphs @xmath124 for simplicity , we assume that all these subgraphs are overall divergent , if they are not divergent , there is no need for renormalization . we order these graphs such that if @xmath125 then @xmath126 . now we define the following : @xmath127 where the product in the second equality needs to be ordered . since the operator ` @xmath128 ' removes the divergence associated with the subgraph @xmath129 , we see that equation ( [ subdivr ] ) is nothing but the graph @xmath1 with all its subdivergences renormalized . now we define the ` bogoliubov @xmath35 operator ' which removes the over all divergence associated with @xmath1 and renders it finite : @xmath130 let us now define a restricted graph @xmath131 as the graph obtained by reducing @xmath27 to a point inside @xmath1 , then it is easy to see that @xmath132 , i.e. , we can replace the subgraph @xmath27 in @xmath1 by @xmath133 and multiply by the counter term which makes @xmath27 finite . we can write equation ( [ rengam ] ) as : @xmath134 where the sum is taken over all subgraphs of @xmath1 ( i.e. , all non empty subsets ( denoted by @xmath86 ) of the set @xmath135 ) . we will also need the following theorem due to hepp @xcite , which we state here without proof . [ hepp ] let @xmath136 be overlapping 1pi subgraphs of @xmath1 . then consider a subgraph @xmath137 such that @xmath138 then @xmath139 i.e. , the finite part of the graph left after replacing the overlapping subdivergences is zero . courtesy this theorem we can restrict the @xmath86 in equation ( [ renf2 ] ) to be the subset of non overlapping 1pi divergences . since @xmath140 provides the counter term associated with the subgraph @xmath86 we can write : @xmath141 where @xmath142 is the counter term which makes the subgraph @xmath86 finite . the subgraph @xmath86 is formally defined as : @xmath143 and is called a ` forest ' of graph @xmath1 . in the above expression the term inside second set of parenthesis is the graph @xmath1 with all non - overlapping subdivergences renormalized . the remaining divergence is then removed by the operator @xmath144 . equation ( [ renf3 ] ) is called ` zimmermann s forest formula ' . we can write the forest formula in a schematic fashion as follows : @xmath145 where @xmath1 and @xmath2 are bare and renormalized graphs respectivley . @xmath3 is the graph with all the subdivergences removed . @xmath97 denotes all proper forests of @xmath1 . @xmath4 and @xmath5 are the counter terms . we now consider an example which explains some important aspects of the forest structure of a feynman graph and the application of the forest formula . let us look at the diagram in figure [ fig : twodiv ] . this graph ( say @xmath146 ) has only two non overlapping 1pi subgraphs , say @xmath147 and @xmath148 , as labeled and boxed in the diagram . the corresponding proper forests are : @xmath149 so we find that : @xmath150 we shoowed earlier that this diagram corresponds a parenthesized word @xmath151 . if we compare the structure of the counter term @xmath5 obtained here with equation ( [ eq : example ] ) ( which computes @xmath152}$ ] ) , we see that the two objects have exactly the same structure after the identifications described in the section [ sec : sum ] . let @xmath94 be any subword of a parenthesized word @xmath23 , then our coproduct is defined in such a way that : @xmath153=\sum _ { u } u\otimes { \left(x / u\right)}.\end{aligned}\ ] ] this assertion is easy to prove using the induction on length of the words . it is obviously true for words of length @xmath59 . assume that it is true for word @xmath23 of length @xmath154 and then induce . let us consider an irreducible parenthesized word @xmath155 of length @xmath156 . @xmath157}&=&{\left(xx\right)}\otimes e + e\otimes { \left(xx\right)}+{\left({\mathbf{1_d}}\otimes b_{{\left(x\right)}}\right)}p\delta{\left[x\right]},\\ & = & { \left(xx\right)}\otimes e + e\otimes { \left(xx\right)}+{\left({\mathbf{1_d}}\otimes b_{{\left(x\right)}}\right)}{\left(\sum_{\text{all subwords } u\neq e \text { of x}}u\otimes { \left(x / u\right)}\right)},\\ & = & { \left(xx\right)}\otimes e + e\otimes { \left(xx\right)}+{\left(\sum_{\text{all subwords } u\neq e \text { of x}}u\otimes { \left(x / u x\right)}\right ) } , \\ & = & { \left(xx\right)}\otimes e + { \left(\sum_{\text{all subwords } u \text { of x}}u\otimes { \left(x / u x\right)}\right ) } , \\ & = & \sum_{\text{all subwords } u \text { of } { \left(xx\right)}}u\otimes { \left(xx\right)}/u \end{aligned}\ ] ] which proves our assertion for irreducible word of length @xmath156 . for an arbitrary word @xmath46 of length @xmath156 , the assertion follows by using the induction assumption and the fact that @xmath158}=\delta{\left[x\right]}\delta{\left[y\right]}$ ] . this completes our proof . here we prove that the coproduct defined in equation ( [ coprod ] ) is coassociative and satisfies the following condition : @xmath159}&=&{\left({\mathbf{1_d}}\otimes\delta\right)}\delta{\left[x\right]},\ \ \ x\in\mathcal{a}.\label{coass}\end{aligned}\ ] ] we will prove this using induction on the length of the parenthesized words . it is trivial to see that @xmath57 is coassociative when acting on the words of length @xmath59 . for the induction we assume that it is coassociative acting on words of length @xmath154 . first , we show that it is coassociative on irreducible parenthesized words of length @xmath156 and then we prove the assertion for arbitrary parenthesized words . we use the sweedler s notation and also drop the summation sign @xmath160 to simplify the notation further . let @xmath23 be a parenthesized word of length @xmath154 then : @xmath161}&=&x_1\otimes x_2,\label{eq : dx}\\ { \left(\delta\otimes{\mathbf{1_d}}\right)}\delta{\left[x\right]}&=&{\left({\mathbf{1_d}}\otimes\delta\right)}\delta{\left[x\right]}\label{eq : indass},\end{aligned}\ ] ] where equation ( [ eq : dx ] ) is just the simplified sweedler s notation and equation ( [ eq : indass ] ) is the induction assumption . now , consider the parenthesized word @xmath162 of length @xmath156 . a straightforward computation gives : @xmath163}&=&{\left(\delta\otimes{\mathbf{1_d}}\right)}{\left({\left(xx_j\right)}\otimes e+{\left({\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}{\left(x_1\otimes x_2\right)}\right)},\\ & = & \delta{\left[xx_j\right]}\otimes e + \delta{\left[x_1\right]}\otimes { \left(x_2x_j\right)},\\ & = & { \left(xx_j\right)}\otimes e\otimes e + x_1\otimes { \left(x_2x_j\right)}\otimes e + \delta{\left[x_1\right]}\otimes { \left(x_2x_j\right)}.\label{eq : pr1}\end{aligned}\ ] ] now , let us compute the rhs of equation ( [ coass ] ) . by using the definition ( [ coprodsweed ] ) @xmath164}&=&{\left(xx_j\right)}\otimes e\otimes e + x_1\otimes { \left(x_2x_j\right)}\otimes e + x_1\otimes{\left[{\left({\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}\delta{\left[x_2\right]}\right]}.\ \ \ \label{eq : pr2}\end{aligned}\ ] ] first two terms in the above equation are the same as in equaton ( [ eq : pr1 ] ) . let s focus on the third term . an important result in this regard is the following . @xmath165 } & = & { \left({\mathbf{1_d}}\otimes{\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}{\left({\mathbf{1_d}}\otimes\delta\right)}{\left(x_1\otimes x_2\right)},\\ & = & { \left({\mathbf{1_d}}\otimes{\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}{\left(x_1\otimes \delta{\left[x_2\right]}\right)},\\ & = & x_1\otimes { \left[{\left({\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}\delta{\left[x_2\right]}\right]}.\end{aligned}\ ] ] using this we can write : @xmath166}\right]}&=&{\left({\mathbf{1_d}}\otimes{\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}{\left({\mathbf{1_d}}\otimes\delta\right)}\delta{\left[x\right]},\\ & = & { \left({\mathbf{1_d}}\otimes{\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}{\left(\delta\otimes{\mathbf{1_d}}\right)}\delta{\left[x\right ] } , \\ & = & { \left({\mathbf{1_d}}\otimes{\mathbf{1_d}}\otimes b_{{\left(x_j\right)}}\right)}{\left(\delta\otimes{\mathbf{1_d}}\right)}{\left[x_1\otimes x_2\right]},\\ & = & \delta{\left[x_1\right]}\otimes { \left(x_2x_j\right)},\end{aligned}\ ] ] where , the second equality just follows from the induction assumption ( [ eq : indass ] ) . this is precisely the third term in equation ( [ eq : pr1 ] ) and this complete the proof of coassociativity for parenthesized words of length @xmath156 an of the form @xmath167 . now , for a general parenthesized word @xmath46 of length @xmath156 , we use the property of the coproduct ( [ compatcoprod ] ) to get : @xmath168}&=&{\left({\mathbf{1_d}}\otimes\delta\right)}{\left(\delta{\left[x\right]}\delta{\left[y\right]}\right)},\\ & = & { \left({\left({\mathbf{1_d}}\otimes\delta\right)}\delta{\left[x\right]}\right)}{\left({\left({\mathbf{1_d}}\otimes\delta\right)}\delta{\left[y\right]}\right)},\\ & = & { \left({\left(\delta\otimes{\mathbf{1_d}}\right)}\delta{\left[x\right]}\right)}{\left({\left(\delta\otimes{\mathbf{1_d}}\right)}\delta{\left[y\right]}\right)},\\ & = & { \left(\delta\otimes{\mathbf{1_d}}\right)}{\left(\delta{\left[x\right]}\delta{\left[y\right]}\right)},\\ & = & { \left(\delta\otimes{\mathbf{1_d}}\right)}\delta{\left[xy\right]},\end{aligned}\ ] ] where the first and second lines follow from property ( [ compatcoprod ] ) , third equality follows fromt the induction assumption . fourth and fifth lines again follow from ( [ compatcoprod ] ) . this complete the proof of coassociativity for the coproduct . in the definition of the antipode , two definitions , ( [ eq : sdef1 ] ) and ( [ eq : sdef2 ] ) , were given . for the antipode to be well defined , these two definitions should be equivalent . we prove this equivalence in the following . we can strip off the parenthesized word @xmath22 from the argument of the antipode in equations ( [ eq : sdef1 ] ) and ( [ eq : sdef2 ] ) and represent antipode as an operator acting on @xmath39 . then , we need to show that : @xmath169}\ & = & -{\mathbf{1_d}}-\ m{\left[{\left({\mathbf{1_d}}\otimes s\right)}p_2\delta\right]},\\ -{\mathbf{1_d}}-m{\left[{\left(s p_1\otimes p_1\right ) } \delta\right]}\ & = & -{\mathbf{1_d}}-\ m{\left[{\left(p_1\otimes sp_1\right)}\delta\right]}.\label{eq : claim1}\end{aligned}\ ] ] both sides still involve the antipode @xmath36 , let us do one more iteration on the both sides . for the left hand side we get : @xmath170}\right)}p_1\otimes p_1\right ) } \delta\right]},\\ & = & -{\mathbf{1_d}}+m{\left[{\left(p_1\otimes p_1\right ) } \delta\right]}+m{\left[{\left({\left(m{\left[{\left(s p_1\otimes p_1\right ) } \delta\right]}\right)}p_1\otimes p_1\right ) } \delta\right]},\\ & = & -{\mathbf{1_d}}+m{\left[{\left(p_1\otimes p_1\right ) } \delta\right]}\nonumber \\ & & + m{\left[{\left(m\otimes{\mathbf{1_d}}\right)}{\left(s\otimes{\mathbf{1_d}}\otimes{\mathbf{1_d}}\right)}{\left(p_1\otimes p_1\otimes{\mathbf{1_d}}\right)}{\left(\delta\otimes{\mathbf{1_d}}\right)}{\left(p_1\otimes p_1\right)}\delta\right]},\\ & = & -{\mathbf{1_d}}+m{\left[{\left(p_1\otimes p_1\right ) } \delta\right]}\nonumber \\ & & + m{\left[{\left(m\otimes{\mathbf{1_d}}\right)}{\left(s\otimes{\mathbf{1_d}}\otimes{\mathbf{1_d}}\right)}{\left(p_1\otimes p_1\otimes p_1\right)}{\left(\delta\otimes{\mathbf{1_d}}\right)}\delta\right]},\label{lhs1}\end{aligned}\ ] ] where the last equality follows because of the fact that @xmath171 , which is easy to confirm . for the right hand side , a similar computation yields : @xmath172}\nonumber \\ & & + m{\left[{\left({\mathbf{1_d}}\otimes m\right)}{\left({\mathbf{1_d}}\otimes{\mathbf{1_d}}\otimes s\right)}{\left(p_1\otimes p_1\otimes p_1\right)}{\left({\mathbf{1_d}}\otimes \delta\right)}\delta\right]}.\label{rhs1}\end{aligned}\ ] ] from equations ( [ lhs1 ] ) and ( [ rhs1 ] ) , we deduce that , to show the equivalence of the two definitions we need to prove the following : @xmath173}\nonumber \\ & = & m{\left[{\left(m\otimes{\mathbf{1_d}}\right)}{\left(s\otimes{\mathbf{1_d}}\otimes{\mathbf{1_d}}\right)}{\left(p_1\otimes p_1\otimes p_1\right)}{\left(\delta\otimes{\mathbf{1_d}}\right)}\delta\right]}\end{aligned}\ ] ] this is very easy to show using the previously established properties of the coproduct @xmath57 and the product @xmath49 . using the coassociativity @xmath174 , we can freely make the following change in the left side of the above equation : @xmath175 similarly , now we make use of the associativity of the product , this implies that @xmath176 . using this , we can again move the last two operators in the direct product to the first two places , yielding : @xmath177 this completes the proof for the equivalence of the two definitions . here , we show that the antipode defined earlier in this article actually satisfies the condition ( [ hopfcom ] ) . we will do this using induction . for a parenthesized word of length 1 , @xmath178 , it is easy to see that : @xmath179}=0,\end{aligned}\ ] ] and @xmath180}\right]}&=&m{\left[{\left(s\otimes{\mathbf{1_d}}\right)}{\left({\left(x\right)}\otimes e + e\otimes { \left(x\right)}\right)}\right]},\\ & = & m{\left[-{\left(x\right)}\otimes e + e\otimes { \left(x\right)}\right]}=0.\end{aligned}\ ] ] a similar computation yields @xmath181}\right]}=0.\end{aligned}\ ] ] let us now assume that the assertion holds for parenthesized words of length @xmath154 , consider an irreducible parenthesized word @xmath155 of length @xmath156 . since the map @xmath92 annihilates @xmath50 , using the sweedler s notation we can write the antipode of @xmath155 as follows : @xmath182}=-{\left(xx\right)}-\sum_{x_{1}\neq e}s{\left[x_1\right]}{\left(x_{2}\ x\right)}.\end{aligned}\ ] ] now , @xmath157}&=&{\left(xx\right)}\otimes e + \sum_{x_1}x_{1}\otimes { \left(x_2\ x\right)},\\ { \left(s\otimes { \mathbf{1_d}}\right)}\delta{\left[{\left(xx\right)}\right]}&= & s{\left[{\left(xx\right)}\right]}\otimes e + \sum_{x_1}s{\left[x_{1}\right]}\otimes { \left(x_2\ x\right)},\\ & = & -{\left(xx\right)}\otimes e-\sum_{x_{1}\neq e}s{\left[x_1\right]}{\left(x_{2}\ x\right)}\otimes e+ \sum_{x_1}s{\left[x_{1}\right]}\otimes { \left(x_2\ x\right)},\\ m{\left(s\otimes { \mathbf{1_d}}\right)}\delta{\left[{\left(xx\right)}\right]}&=&-{\left(xx\right)}-\sum_{x_{1}\neq e}s{\left[x_1\right]}{\left(x_{2}\ x\right)}+ \sum_{x_1}s{\left[x_{1}\right ] } { \left(x_2\ x\right)},\\ & = & -{\left(xx\right)}+s{\left[e\right]}{\left(xx\right)}=0,\end{aligned}\ ] ] where the last line follows from the fact that then when @xmath183 , @xmath184 . now , let us consider the case for @xmath185}\right]}$ ] . due to the equivalence of two definitions ( [ eq : sdef1 ] ) and ( [ eq : sdef2 ] ) , and the properties of @xmath186 we have the following identity : @xmath187 using this , we find that : @xmath188}&=&{\left(xx\right)}\otimes e + { \left({\mathbf{1_d}}\otimes s\right)}\sum_{x_1}x_1\otimes { \left(x_2 x\right ) } , \\ & = & { \left(xx\right)}\otimes e + { \left(s\otimes { \mathbf{1_d}}\right)}\sum_{x_1\neq e}x_1\otimes { \left(x_2 x\right)}+{\left({\mathbf{1_d}}\otimes s\right)}e\otimes { \left(xx\right ) } , \\ & = & { \left(xx\right)}\otimes e + \sum_{x_1\neq e}s{\left[x_1\right]}\otimes { \left(x_2 x\right)}+{\left({\mathbf{1_d}}\otimes s\right)}{\left[e\otimes { \left(xx\right)}\right]},\\ & = & { \left(xx\right)}\otimes e + \sum_{x_1\neq e}s{\left[x_1\right]}\otimes { \left(x_2 x\right)}-e\otimes { \left(xx\right)}\nonumber \\ & & -\sum_{x_{1}\neq e}e\otimes s{\left[x_1\right]}{\left(x_2x\right)},\end{aligned}\ ] ] which implies @xmath189}\right]}&= & 0.\end{aligned}\ ] ] since the counit annihilates any parenthesized word we finally conclude that : @xmath190}\right]}=0=e\circ\overline{e}{\left[{\left(xx\right)}\right]}.\end{aligned}\ ] ] for an arbitrary parenthesized word @xmath46 , due to the induction assumption and the property ( [ compatcoprod ] ) of the coproduct , the assertion holds trivially . this completes our proof . here , we will work out an elementary example which elucidates how all the different elements of the hopf algebra fit together to give a finite result for a divergent integral . we will use a very simple toy model , defined below : @xmath191}&\equiv & \int_{c}^{\infty}dy y^{-1-j\epsilon}\equiv i_j,\\ { \left(xx_{j}\right)}{\left[c\right]}&\equiv & \int_{c}^{\infty}dy y^{-1-j\epsilon}x{\left[y\right]},\\ { \left(x_j\right)}{\left(x_k\right)}{\left[c\right]}&= & i_ji_k,\\ r{\left[x{\left[c\right]}\right]}&\equiv & x{\left[1\right]}.\end{aligned}\ ] ] it is easy to see that @xmath192 is divergent as @xmath193 . we call the subscript @xmath194 in @xmath195 , the loop order of @xmath195 . this toy model is the simplest realization of our hopf algebra . let us consider the divergent graph @xmath196 . our claim is that the expression @xmath197}\right]}\right]}$ ] is a finite integral as expected from our hopf algebra construction . by making use of the already worked out examples for @xmath198}$ ] , @xmath199}$ ] , @xmath200}$ ] and the fact @xmath201}=s_r{\left[x\right]}s_r{\left[y\right]}$ ] we find that : @xmath202}-{\left(x_1\right)}{\left[1\right]}{\left({\left(x_2\right)}x_1\right)}{\left[c\right]}-{\left(x_2\right)}{\left[1\right]}{\left({\left(x_1\right)}x_1\right)}{\left[c\right]}+{\left(x_1\right)}{\left[1\right]}{\left(x_2\right)}{\left[1\right]}{\left(x_1\right)}{\left[c\right]}\nonumber \\ & & -{\left(\text{first four terms with } c\text { replaced by } 1\right)}.\label{eq : xr}\end{aligned}\ ] ] now , @xmath203}&=&\int_{c}^{\infty}dx x^{-1-\epsilon}\int_{x}^{\infty}dy y^{-1 - 2\epsilon}\int_{x}^{\infty}dz z^{-1-\epsilon},\\ t_2\equiv - { \left(x_1\right)}{\left[1\right]}{\left({\left(x_2\right)}x_1\right)}{\left[c\right]}&=&- \int_{c}^{\infty}dx x^{-1-\epsilon}\int_{x}^{\infty}dy y^{-1 - 2\epsilon}\int_{1}^{\infty}dz z^{-1-\epsilon},\\ t_3\equiv - { \left(x_2\right)}{\left[1\right]}{\left({\left(x_1\right)}x_1\right)}{\left[c\right]}&=&-\int_{c}^{\infty}dx x^{-1-\epsilon}\int_{1}^{\infty}dy y^{-1 - 2\epsilon}\int_{x}^{\infty}dz z^{-1-\epsilon},\\ t_4\equiv { \left(x_1\right)}{\left[1\right]}{\left(x_2\right)}{\left[1\right]}{\left(x_1\right)}{\left[c\right]}&=&\int_{c}^{\infty}dx x^{-1-\epsilon}\int_{1}^{\infty}dy y^{-1 - 2\epsilon}\int_{1}^{\infty}dz z^{-1-\epsilon}.\end{aligned}\ ] ] the first two terms can be combined to get : @xmath204 the third term can be written as @xmath205 so that the sum of the four terms is : @xmath206 plug this in equation ( [ eq : xr ] ) we finally obtain the expression : @xmath207 which is clearly well defined and finite in the @xmath208 limit . although this was a very simple example , there should be no hinderance in generalizing this to more realistic qft examples . if we consider some realistic feynman graph , our hopf algebra will renormalize it with the same ease by applying the operator @xmath104}$ ] . d. kreimer , _ on the hopf algebra structure of perturbative quantum field theories _ , http://arxiv.org/pdf/q-alg/9707029v4.pdf[arxiv:q-alg/9707029 ] . a. connes and d. kreimer , _ hopf algebra , renormalization and non commutative geometry _ , http://arxiv.org/pdf/hep-th/9808042.pdf[arxiv:hep-th/9808042 ] . d. j. broadhurst , d. kreimer , _ renormalization automated by hopf algebra _ , http://arxiv.org/pdf/hep-th/9810087.pdf[arxiv:hep-th/9810087 ] . d. kreimer , r. delbourgo , _ using the hopf algebra structure of qft in calculations _ , http://arxiv.org/pdf/hep-th/9903249.pdf[arxiv:hep-th/9903249 ] . a. connes , d. kreimer , _ lessons from quantum field theory _ , http://arxiv.org/pdf/hep-th/9904044.pdf[hep-th/9904044 ] . d. kreimer , _ on overlapping divergences _ , http://arxiv.org/abs/hep-th/9810022[arxiv:hep-th/9810022 ] . w. v. suijlekom , _ the hopf algebra of feynman graphs in qed _ , http://arxiv.org/abs/hep-th/0602126[hep-th/0602126 ] e. panzer , _ hopf algebraic renormalization of kreimer s toy model _ , http://arxiv.org/abs/1202.3552[arxiv:1202.3552[math-ph ] ] . d. kreimer , _ knots and feynman diagrams _ , 2000 , cambrige university press . j. c. collins , _ renormalization _ , 1986 , cambridge university press . s. majid , _ foundations of quantum group theory _ , 1996 , cambridge university press . k. hepp , _ proof of the bogoliubov - parasiuk theorem on renormalization _ , phys . , 1966 , springer . |
we analyze line - of - sight atomic hydrogen ( ) line profiles of 31 nearby , low - mass galaxies selected from the very large array - acs nearby galaxy survey treasury ( vla - angst ) and the nearby galaxy survey ( things ) to trace regions containing cold ( t 1400 k ) from observations with a uniform linear scale of 200 pc beam . our galaxy sample spans four orders of magnitude in total mass and nine magnitudes in m . we fit single and multiple component functions to each spectrum to isolate the cold , neutral medium given by a low dispersion ( km s ) component of the spectrum . most spectra are adequately fit by a single gaussian with a dispersion of 8 - 12 km s . cold is found in 23 of 27 (% ) galaxies after a reduction of the sample size due to quality control cuts . the cold contributes% of the total line - of - sight flux when found with warm . spectra best fit by a single gaussian , but dominated by cold emission ( i.e. , have velocity dispersions km s ) are found primarily beyond the optical radius of the host galaxy . the cold is typically found in localized regions and is generally not coincident with the very highest surface density peaks of the global distribution ( which are usually areas of recent star formation ) . we find a lower limit for the mass fraction of cold - to - total gas of only a few percent in each galaxy . | 1208.0983 | dwarf irregular ( dirr ) galaxies in the nearby universe are laboratories for studying the most fundamental properties of gas evolution and star formation . large , multi - wavelength studies of these systems are only just beginning . however , recent , high resolution surveys of large , spiral galaxies in the ultraviolet ( e.g. , ngs - @xcite ) , optical ( e.g. , nfgs - @xcite ; angst - @xcite ) , infrared ( e.g. , sings - @xcite ) , and radio ( e.g. , things - @xcite ; heracles - @xcite ) have given us insight into where and when stars form in a galaxy . they have also provided clues to the gas conditions surrounding sites of current star formation ( @xcite ; @xcite ) . the conclusion is that stars form in regions rich in cold , dense , molecular material @xcite . these studies mainly focus on high metallicity systems , however . how these results translate to low metallicity environments has not been fully explored , mainly due to the fact the most commonly observed tracer of molecular material , co , is notoriously difficult to observe at low metallicity ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ) . nearby dirr galaxies offer the opportunity to study the star formation process in low metallicity systems at high spatial resolution . studies of the star formation rates in dirr galaxies defined from uv , 24@xmath5 m , and/or h@xmath6 emission ( e.g. , @xcite ; @xcite ; @xcite ) have found a non - linear correlation with the local atomic hydrogen ( ) distributions . however , the dependence of star formation on molecular gas content is apparent in the large spiral galaxies observed in @xcite and @xcite . the star formation rate surface density is shown to correlate linearly with the molecular gas surface density in these galaxies . these results can be understood if one assumes that efficient star formation requires molecular hydrogen ( e.g. , @xcite and references therein ) . based on theoretical studies , the very character of the interstellar medium ( ism ) is likely to change at low metallicity ( e.g. , @xcite ) . @xcite modeled the so called x - factor which measures the relationship between amount of detected co emission and the abundance of molecular hydrogen . they found that the amount of co drops substantially at low metallicities , consistent with prior predictions ( e.g. , @xcite ) . this deficit of co at low metallicities is supported by observations showing relatively low or no detections of co emission in nearby dirr galaxies ( e.g. , @xcite ) . consequently , this lack of metals also results in lower amounts of polycyclic aromatic hydrocarbon ( pah ) emission ( e.g. , @xcite ) . the pahs are important to the formation of the cold ism since they are critical in the shielding of uv and soft x - ray photons . the long standing problem of studying the molecular component in the low metallicity dwarfs continues . we expect dirr galaxies to contain molecular hydrogen since they _ are _ forming stars . however , since direct observations of the molecular gas responsible for forming stars in large samples of dirr galaxies via co are currently not feasible , we must find a different tracer of star - forming gas . one intriguing idea is to find the immediate precursors of the molecular gas . the assembly of star forming molecular clouds is generally believed to require cold ( e.g. , @xcite ; @xcite ) . in our galaxy , cold clouds have been observed to surround and even intermix with molecular clouds ( e.g. , @xcite ) through studies of narrow self absorption ( hinsa ; @xcite ) . one promising technique to find cold gas was pioneered by @xcite in a sample of nearby , star forming dirr galaxies . these authors used high angular and spectral resolution data of two nearby dirr galaxies to decompose line - of - sight spectra into narrow and broad gaussian components . the narrow - lined gas ( @xmath7 @xmath4 4.5 km s@xmath1 ) was found only in specific regions within the disk while the broad - lined gas ( @xmath7 @xmath4 10 km s@xmath1 ) was found along every line - of - sight observed . the narrow gaussian component was attributed to cold ( t@xmath01000 k ) gas while the broad lined gas was assumed to be warm , neutral hydrogen ( t@xmath85000 k ) . @xcite studied three more dirr galaxies , finding that they , too , contained evidence for cold gas . other authors have used this technique to discover cold in a small number of other galaxies . @xcite investigated the distribution in ngc 6822 and also found regions rich with cold gas . recent work by @xcite found evidence of cold gas in six dirr galaxies from the faint irregular galaxies gmrt survey ( figgs ; @xcite ) . to date , cold in emission has been discovered in 10 nearby dirr galaxies . within the limited data available , different measurements of the cold do appear to give comparable results . two galaxies , ddo 210 and gr8 , from the sample of @xcite overlapped with that of @xcite . @xcite used the very large array ( vla ) at a linear scale of @xmath4200 pc beam@xmath1 with a velocity resolution of 1.3 km s@xmath1 while @xcite used the giant metrewave radio telescope ( gmrt ) at a linear resolution of @xmath4300 pc beam@xmath1 with a velocity resolution of 1.65 km s@xmath1 . despite the use of different facilities and different spatial / spectral resolutions , both studies find similar results for both the spatial distributions and velocity dispersions of the narrow component this agreement between the two studies suggests that the measurement of cold is generally robust , and not extremely sensitive to the exact observing parameters . the technique of identifying a cold neutral medium has also been used in nearby , high metallicity spiral galaxies . @xcite isolated cold gas using relatively low spectral resolution ( 6 km s@xmath1 ) imaging from the vla for 11 of the closest spiral galaxies . he combined spectra from discrete radial bins and found that these combined spectra all had a narrow gaussian core ( fwhm @xmath0 6 km s@xmath1 ) superposed onto broad ( fwhm @xmath4 30 km s@xmath1 ) lorentzian wings . he also found that the cold gas is filamentary and is found in clumps , preferentially in the spiral arms . since the majority of star formation in spiral galaxies occurs in molecular clouds within the spiral arms ( e.g. , @xcite ) , it is not surprising to find the bulk of the cold associated with these features . the purpose of our study is to build upon the previous results of @xcite , @xcite , @xcite , and @xcite in order to provide limits to the locations and amount of cold gas in a large sample of 31 nearby , low - mass galaxies using a common spatial resolution . we describe our galaxy sample in [ obs ] and our methods of signal extraction in [ idents ] . our results are described in [ results ] and we end with a summary of our conclusions in [ conclusions ] . galaxies for this work were taken from two major surveys of nearby galaxies : the nearby galaxy survey ( things ; aw0605 ; @xcite ) and the very large array - advanced camera for surveys nearby galaxy survey treasury ( vla - angst ; ao0215 ; ott et al . _ in preparation _ ) . our sample consists of 24 galaxies from vla - angst and 7 dwarf galaxies in the m81 group from things . the vla - angst observations all have a high spectral resolution of 0.65 - 2.6 km s@xmath1 . the 7 galaxies from the things sample have velocity resolutions of 1.3 - 2.6 km s@xmath1 . the high velocity resolution is crucial for detecting narrow velocity components . other galaxies from the things sample were not used because they had velocity resolutions of 5 km s@xmath1 , too coarse for this type of analysis . our final sample contains 31 galaxies with distances ranging from 1.3 - 5.3 mpc ( average 2.9 mpc ) and masses between 4@xmath910@xmath10 - 1@xmath910@xmath11 @xmath12 . all observations were made using the very large array ( vla ) observatory . we briefly outline the basic reduction procedure here , but note that both surveys followed similar recipes , described fully in @xcite and ott et al . ( _ in preparation _ ) . standard aips processing of spectral line data was followed . phase , bandpass , and flux corrections were made using typical vla calibrators . two sets of data cubes were produced : a natural weighted cube with a typical beam size of @xmath410@xmath13 and a robust weighted cube with a beam size of @xmath47@xmath13 . both sets of data have rms noise values of @xmath41 mjy beam@xmath1 in a single line - free channel . we use the robust weighted cubes in this work , which allows us to use the smallest possible uniform linear scale . further , robust weighting produces minimal side - lobes allowing for a clean - beam which better approximates the dirty beam structure than that of the more complex structure seen in the dirty - beam of natural weighting @xcite . the better dirty - beam approximation of the robust weighting is ideal for the extended emission observed in data . the main difference between the data for the two surveys is due to the inclusion of updated expanded vla ( evla ) antennas into the array for the vla - angst survey ( the evla has since been renamed the karl g. janksy very large array " ) . unfortunately , the conversion of the digital evla signal to an analog signal ( to be compatible with the vla signal ) aliased power into the first 0.5 mhz of the baseband . this only affected the evla - evla baselines and they were subsequently removed from the data ( see ott et al . _ in preparation _ for full details ) . aliasing did not correlate into the mixed vla - evla baselines and the vla - vla baselines were unaffected . to compensate for the loss of the evla - evla baselines , additional observation time was spent on each source . table [ galprop ] gives an overview of the major observational properties of our galaxy sample . the table columns are : 1 ) galaxy name , 2,3 ) ra and dec , 4 ) distance in mpc ( from trgb measurements in @xcite unless otherwise noted ) , 5 ) total mass , 6 ) absolute @xmath14-band magnitude , 7 ) the major - axis length in arcminutes at the @xmath425 mag arcsec@xmath15 level ( except for kkh 98 and kk 230 , which use the holmberg system of @xmath426.5 mag arcsec@xmath15 ; @xcite ) , 8) inclination @xcite , 9 ) the 200 pc beam size ( see [ methods ] ) , 10 ) the spectral resolution , and 11 ) the rms of the noise in the line - free channels of the data cube . while some metallicity estimates exist for the most massive galaxies in our sample , most do not have measured abundances . using the relation in @xcite which relates m@xmath2 to the oxygen abundance ( see also @xcite ) , we expect a range of 7.0@xmath012+log(o / h)@xmath08.2 with a median value of 7.7 for the low mass galaxies in our sample . we analyze line - of - sight spectra following methods similar to those outlined in @xcite and @xcite . first , we smoothed each image cube with a gaussian kernel to a common linear scale of 200 pc beam@xmath1 . to do this we computed the angular size equivalent of 200 pc for each galaxy using the distances given in table [ galprop ] . we then used the convl task in aips which incorporates the initial beam size , beam position angle , and requested output beam size to compute the required convolving kernel . we note that our circular beam sizes can sample slightly larger linear scales depending on galaxy inclination . spectra were then extracted through every 1@xmath165 pixel ( 1@xmath13 pixels for ngc 247 and ngc 3109 ) . extracting spectra from every pixel over - samples the beam , but for our purposes , does not influence the results . also , the beam sizes reported in table [ galprop ] and the typical galaxy rotation speeds of 10 - 20 km s@xmath1 are small enough that kinematic broadening is minimal , regardless of galaxy inclination . note that spectral broadening would work to hide a narrow component , not create one . each extracted spectrum was successively fitted using our own routines with two different functional forms : a single gaussian and a double gaussian . we also attempted to fit the spectra with a fourth order gauss - hermite polynomial of the form @xmath17\ ] ] where @xmath18 @xcite , @xmath19 and @xmath20 measure the amplitudes of an asymmetric and a symmetric deviation from an underlying gaussian profile with amplitude @xmath21 , central velocity @xmath22 , and standard deviation @xmath23 . gauss - hermite polynomials offer useful characterizations of non - gaussian line profiles , but as @xcite point out , the direct relationship between the gauss - hermite variables and the physical conditions of the gas is not obvious . in general , the results of the double gaussian and gauss - hermite polynomial fits were in excellent agreement , although the detection efficiency of the gauss - hermite polynomials is much lower . however , we expect the observed spectra to have a gaussian profile if the gas has an isothermal density distribution @xcite . since the relationship between the gauss - hermite polynomial parameters and the physical nature of the gas is not obvious , we only report the results of the single and double gaussian fits . a spectrum was identified as containing narrow - line only if multiple criteria were fulfilled . first , we only fit extracted spectra with a s / n greater than 10 ( see [ sims ] ) . we define s / n in the same manner as @xcite , that is , the peak in the spectrum divided by the rms noise in the line free channels of the data cube . we also required each individual component of the double gaussian fit to have a minimum s / n of 3.1 . this minimum s / n requirement was also used by @xcite , @xcite , and @xcite and is motivated by our desire to obtain significant detections . also , each spectrum ( single or multiple component ) was required to have a velocity dispersion greater than the velocity resolution of the data ( 0.65 , 1.3 , or 2.6 km s@xmath1 ) beyond the 1@xmath7 errors in the fits . requiring the velocity dispersion plus error to be greater than the velocity resolution ensures we are not fitting noise spikes in the data . fortunately , none of the fitted spectra were affected by this criterion . typical errors in the velocity dispersion of each single and double gaussian component are roughly 0.7 km s@xmath1 . we also do not enforce any restrictions on the central velocity offsets between individual components , leaving this as a free parameter . to quantify which function best fitted each spectrum , we first computed the variance of the residuals for each fit . the ratio of the variances between the single and double gaussian fits were then compared statistically via a single - tailed f - test . a double gaussian fit was determined to be statistically significant if the probability of improvement over a single gaussian fit was 95% or greater . we were as conservative with our approach as possible in order to provide secure lower limits to the total amount of narrow - line in each galaxy . @xcite , @xcite , and @xcite were less conservative in their approach , requiring only a 90% significance of improvement over a single gaussian profile . because of our relatively high s / n cut - off , the results of our fitting are relatively independent of the f - test criterion that we use , as long as it is above roughly 70% . since the number of detections of double gaussian profiles increases very slowly as the cut - off is reduced from 95% down to 70% , we chose 95% confidence to establish a very firm lower limit on the total cold content . establishing a realistic upper limit is more problematic and is highly dependent upon assumptions . we further adopt the criteria of @xcite that cold has a velocity dispersion of less than 6 km s@xmath1 . a velocity dispersion of 6 km s@xmath1 results in an upper limit to the gas temperature of @xmath41400 k ( assuming ( 3/2)kt = ( 1/2)@xmath24@xmath7@xmath25 ) . thus , we did not ascribe a second component as cold when the fits detected multiple components with similar velocity dispersions between 6 - 10 km s@xmath1 or one component with a velocity dispersion of 6 - 10 km s@xmath1 and another component with a much higher dispersion . the vast majority of spectra that were best fit by double gaussian profiles with both velocity dispersions greater than 6 km s@xmath1 were double peaked ( in both low and high inclination galaxies ) which is normally associated with expanding structures ( e.g. , @xcite ) . furthermore , these higher velocity dispersions are not typical for the cold described in the literature with velocity dispersions of 2 - 6 km s@xmath1 ( @xcite ) ; we therefore rejected these locations as having narrow line emission . @xcite also noted that the bulk of the spectra within this category for ngc 6822 were double - peaked , leading them to make a cut at 6 km s@xmath1 . thus , having a cut at 6 km s@xmath1 ensures that we are both reporting cold gas and eliminating a large fraction of the double peaked spectra . the double - peaked spectra eliminated by this criteria represent less than 1% of the total spectra and are not included in any of the following analyses . we also identify spectra that were sufficiently fit by a single gaussian profile with a velocity dispersion of less than 6 km s@xmath1 as cold . limiting our cold identifications to gas with velocity dispersions less than 6 km s@xmath1 ensures we are not misidentifying warm as cold . one possible scenario is that the broad component is the combination of multiple cold gas ( narrow gaussian ) clouds . given that the typical line - of - sight spectrum is best fit by a single gaussian profile ( see [ results ] ) , the multiple cold clouds would have to conspire in such a way to maintain this gaussianity . while this scenario is physically possible , it requires a large number of cold hi clouds with distributions in velocities and amplitudes that would systematically combine to maintain a total gaussian velocity dispersion equal to the observed velocity dispersion for the warm hi phase . high resolution absorption line studies would be required to distinguish whether the broad component is truly broad or a composite of many narrow components . however , if we assume that the galaxies can be approximated by a thin disk we would expect that the line centroid of the cold to be at or very near the line centroid of the warm . warm gas scale heights in dwarf galaxies tend to be a few hundred pc ( e.g. , @xcite ) compared to their few kpc diameters , so this is not a bad assumption . offset velocity gas due to inclination effects would then contribute to the broadening of the warm ( broad gaussian ) component and less so to the cold ( narrow ) component gas . not surprisingly , the average central velocity offset between the components of the best fit double gaussian profiles in our sample is 0.35 km s@xmath1 , which is less than the velocity resolution of our data . thus , the observations are most simply interpreted as a combination of cold gas near the central velocity ( when observed ) and nearly ubiquitous warm gas . we find it highly unlikely that the broad component is a combination of multiple narrow components . this leads us to the conclusion that the is dominated by the warm phase and that the cold phase represents a much smaller fraction of the ism in our sample of dwarf galaxies . the top panel of figure [ spectrum ] shows an extracted spectrum ( black ) from the vla - angst observations of sextans a with single gaussian profile ( blue dashed ) , double gaussian profile ( solid red ) , and gauss - hermite polynomial fits ( dotted magenta ) . the bottom panel shows the residual to the fit . this spectrum has a clear narrow peak that is not fit well by the single gaussian profile . both the double gaussian profile and the gauss - hermite polynomial fit the spectrum statistically better than the single gaussian profile at the 99.9% confidence level . although the gauss - hermite polynomial is a better fit than the single gaussian profile , the residuals clearly show that the double gaussian profile is even better . in summary , we identify cold by the following process : 1 ) only spectra with a s / n greater than 10 were fit , 2 ) each spectrum was successively fitted with a single and double gaussian function , 3 ) each individual component of the double gaussian fit was required to have a minimum s / n of 3.1 , 4 ) the velocity dispersion plus error had to be greater than the velocity resolution of the data , 5 ) double gaussian fits were accepted if they improved the fit over a single gaussian profile at the 95% or greater confidence level in a single tailed f - test , and 6 ) cold can be described by single or double gaussian profiles with velocity dispersions less than 6 km s@xmath1 . there are two factors which could lead to non - detections of existing cold . cold could be missed simply because the column density is too low to reach our s / n requirement , or it could be missed because a higher s / n is required to deblend the cold from the warm . as a test of our ability to extract multiple components from a given spectrum , we produced three sets of 50,000 uniformly distributed synthetic spectra that spanned the s / n of our sample ( 10 @xmath26 s / n @xmath26 60 ; one set each at 0.65 , 1.3 , and 2.6 km s@xmath1 velocity resolution ) . each spectrum contained two gaussian profiles that are representative of the warm and cold gas ; a broad " gaussian with @xmath27 between 8 and 12 km s@xmath1 and a narrow " gaussian with @xmath28 between 3 and 6 km s@xmath1 . we allowed for an offset of @xmath292.5 km s@xmath1 between the line centroids . velocity offsets larger than 2.5 km s@xmath1 allow for much higher detection fractions since the spectra become more asymmetric . the observed cold line centroids are also very similar to the warm line centroids ( as discussed in [ methods ] ) . for each s / n , we randomized the amplitude of the cold gaussian . we then chose the warm gaussian such that the s / n was conserved . as for the analysis with our observations , each individual component was required to have a s / n @xmath30 3.1 . a new set of gaussians was produced if each component did not have a s / n @xmath30 3.1 . we next added random noise to each spectrum and then passed each randomly generated spectrum through our fitting routine to gauge our recovery rate . the top row of figure [ mcarlo ] shows histograms of the total number of synthetic spectra ( solid black line ) and the total number identified as requiring multiple components by a double gaussian fit ( solid red line ) as a function of s / n . the red histograms also include the locations best fit by a single gaussian with a velocity dispersion of less than 6 km s@xmath1 . the bottom row shows the detection efficiency as a function of s / n for the double gaussian fits ( including the best fit single gaussians with a velocity dispersion of less than 6 km s@xmath1 ) . figure [ mcarlo ] shows that we never identify all double gaussian profiles within the range of s / n values scrutinized and our recovery efficiency worsens as the velocity resolution decreases . the differences in recovery efficiency between the 0.65 and 1.3 km s@xmath1 simulations are less pronounced than from 1.3 to 2.6 km s@xmath1 . table [ simtab ] shows the results of the simulations for each velocity resolution for the double gaussian profiles that were identified . column ( 1 ) is the velocity resolution , column ( 2 ) is the ratio between the input and extracted broad gaussian amplitude ( a@xmath31/a@xmath32 ) , columns ( 3 ) and ( 4 ) are the average differences between the simulated and extracted broad gaussian central velocities ( @xmath33v@xmath34 ) and velocity dispersions ( @xmath33@xmath7@xmath34 ) , column is the ratio between the input and extracted narrow gaussian amplitude ( a@xmath35/a@xmath36 ) , and columns ( 6 ) and ( 7 ) are the average differences between the simulated and extracted narrow gaussian central velocities ( @xmath33v@xmath37 ) and velocity dispersions ( @xmath33@xmath7@xmath37 ) , respectively . these results show that when we do recover two gaussian profiles , our routines accurately reproduce the input gaussian parameters . in figure [ mcarlo2 ] we plot the amplitude ratio of the broad and narrow gaussian components ( a@xmath38/a@xmath39 ) versus the ratio of the velocity dispersions ( @xmath40/@xmath41 ) for the 1.3 km s@xmath1 velocity resolution simulation . we show these ratios in four different s / n bins to understand where our fitting routines have trouble identifying the two gaussian components . the grey points represent all of the simulated spectra , the red points are those spectra identified as containing two gaussian components , and the blue points are the simulated spectra best fit by a single gaussian with a velocity dispersion of less than 6 km s@xmath1 ( even though the input spectra contains both a broad and narrow gaussian as described above ) . as is shown in figure [ mcarlo ] , the recovery efficiency is dependent upon the s / n . the best fit single gaussians are predominately in a regime where the narrow component dominates the spectra ( amplitude ratios less than 1 and velocity dispersion ratios less than 2.5 ) . we also have difficulty identifying the narrow component when the broad component dominates the spectra in all s / n bins . lastly , our routines do not pick up multiple components when the velocity dispersions are similar . these plots show some of the complexities in simulating and recovering multiple gaussian components . when we do identify the spectra as containing multiple components , however , our routines accurately reproduce the input parameters . the simulations indicate that we can be confident of our detections in the observed data . however , the simulations also show that we can only reliably compute lower limits to the cold content . a complete census of the cold would require both higher s / n spectra and a detailed knowledge of the intrinsic distributions in a@xmath38/a@xmath39 and @xmath40/@xmath41 ( in order to make incompleteness corrections ) . we also produced 50,000 synthetic spectra of single gaussian profiles to quantify our false positives . these spectra were generated using a velocity dispersion between 8 and 12 km s@xmath1 with s / n values between 10 and 60 . we used a velocity resolution of 2.6 km s@xmath1 since this gives the worst recovery results . with our acceptance criteria , we failed to find a single false identification of multiple components in our synthetic spectra . we do not start to see the potential for false detections until the confidence level is lowered to @xmath470% . these simulations give us confidence that our fitting results are robust to false detections . each galaxy in our sample varies in terms of the total number of spectra and observed minimum , maximum , and average s / n ( and column density ) . in total , we have analyzed roughly 4,100 independent lines - of - sight . table [ fitres ] summarizes the observed spectral properties of each galaxy . the columns indicate 1 ) galaxy name , 2 ) total number of independent lines - of - sight scrutinized ( n@xmath42 ) , 3 ) average s / n of the extracted spectra ( @xmath3s / n@xmath43 ) , 4 ) the approximate column density at a s / n of 10 ( n@xmath44 ) , 5 ) the peak column density ( n@xmath45 ) , and 6 ) the average column density ( @xmath3n@xmath46@xmath30 ) . all of the spectra in ddo 82 , kdg 73 , kk 230 , and kkh 98 have s / n values below our threshold and as a result had no spectra analyzed , reducing our final sample to 27 galaxies . table [ indres ] summarizes the results of our fitting routine . the columns are described as follows : 1 ) galaxy name , 2 ) the average s / n of the spectra where cold is found ( @xmath3s / n@xmath47 ) , 3 ) the areal filling factor of the cold defined as the ratio of the number of cold detections and the total number of scrutinized spectra ( @xmath48 ) , 4 ) the cold - to - total mass fraction defined as the ratio of the total summed column densities of the cold gaussian profiles and the total summed column densities in the areas scrutinized ( @xmath49 ) , 5 ) an estimate of the upper limit to the cold - to - total mass fraction assuming each line - of - sight contains cold ( @xmath50 ; see [ amfrac ] ) , 6 & 7 ) the average velocity dispersion of the narrow and broad gaussian components ( @xmath3@xmath51@xmath30 and @xmath3@xmath52@xmath30 ) , and 7 ) the average velocity dispersion of the locations where a single gaussian profile was sufficient to describe the spectrum ( @xmath3@xmath53@xmath30 ) . column 5 also includes all of the best fit single gaussian profiles with a velocity dispersion of less than 6 km s@xmath1 , while column 7 excludes them . we detected cold in 23 out of the 27 galaxies in our final sample . figure [ 2plots ] shows the spatial distribution of the cold for each galaxy . the left panels show the total integrated intensity maps with contours of 10@xmath54 and 10@xmath55 cm@xmath15 overlaid . the right panels have the surface density contours overlaid on the locations of the cold emission . the majority of the cold is clumped in localized regions . typically , the cold is not spatially coincident with the very highest peaks in the total distribution in a given galaxy , although it is mainly concentrated in locations where the total column density exceeds the canonical threshold for star formation of @xmath410@xmath55 cm@xmath15 @xcite . cold described in each of the previous studies also shows a preference for being located in regions of high column density , but not necessarily associated with the highest columns . the concentration of the cold in these regions is not unexpected given these location have the highest s / n values . the four galaxies in which we do not detect cold ( ddo 99 , mcg09 - 20 - 131 , ngc 3741 , and ugca 292 ) show no distinguishing characteristics to give us insight as to why they are non - detections . each of their total masses , m@xmath2 values , distances , stellar disk sizes , and star formation rates ( see [ h2sfr ] ) are similar to other galaxies in our sample that do show cold detections . fortunately , our sample of 27 galaxies had some overlap with previous studies using similar methods to detect cold . @xcite detected cold in ugca 292 and gr8 using the vla while @xcite detected it in ddo 53 , mcg09 - 20 - 131 , gr8 and m81 dwarf a using the gmrt . these five galaxies are amongst the faintest and smallest galaxies in our sample and as a result , we detected cold in only 3 of these 5 galaxies : ddo 53 , gr8 , and m81 dwarf a. our selection criteria failed to produce a cold signature in ugca 292 or mcg09 - 20 - 131 . the three galaxies in which we did detect cold are included in figure [ 2plots ] . for gr8 , @xcite and @xcite each found cold in both the northern and south - western portions of the galaxy while @xcite also found some evidence in the eastern portion of the galaxy . in contrast , we only found cold in the northern region of gr8 . for m81 dwarf a we detect cold only in the eastern side of the galaxy while @xcite find a hint of cold in the western side of the galaxy as well . for ddo 53 we found good overall agreement with the previous study , although our total detection area is smaller . the data quality between each survey is similar , yet they provide minor differences in their results . our simulations suggest velocity resolution differences may play a role in the detection differences between the studies . the velocity resolutions in our sample for these galaxies are 0.65 , 1.3 , and 2.6 km s@xmath1 for gr8 , m81 dwarf a , and ddo 53 , respectively . the literature velocity resolutions are 1.3 and 1.65 km s@xmath1 for gr8 , 1.65 km s@xmath1 for m81 dwarf a , and 1.65 km s@xmath1 for ddo 53 . the galaxy with the most similar cold distribution , ddo 53 , has the worst velocity resolution in our sample , 0.85 km s@xmath1 worse than the literature value . gr8 has been analyzed with three different velocity resolutions and each study finds slightly different results . we reduced our velocity resolution to 1.3 km s@xmath1 for gr8 to compare to the other studies and obtained no difference in our results . dwarf a was analyzed with similar velocity resolutions , yet there still exists slight differences . despite analyzing data from different observatories with different velocity resolutions , the general results are all similar . it seems likely that the differences in each galaxy arise from small differences in the selection criteria and beam shape . for example , both @xcite and @xcite used naturally weighted data cubes which , on average , produce higher s / n spectra . however , as noted in [ obs ] , the robust weighting for combined array data is ideal for studying extended emission due to the better clean beam approximation of the dirty beam in the clean algorithm . the previous studies also used a cutoff of 90% in the f - test statistic and also allow the narrow gaussian component to be larger than 6 km s@xmath1 , finding values up to @xmath48 km s@xmath1 . if each spectrum consisted of both a broad and narrow component and were not inherently a single gaussian profile , we would expect to find far more locations with a narrow line signature than we actually do . the top panel of figure [ expected ] shows a histogram for every independent line - of - sight ( black line ) , the expected number of identified narrow line detections ( blue ) ( based upon our detection efficiency as calculated in [ sims ] and the assumption that a narrow component exists at each location ) , and the actual number of narrow line detections ( red ) as a function of s / n . the bottom panel shows the detection fraction as a function of s / n . figure [ expected ] demonstrates the large gap between the number of spectra for which we are sensitive to the presence of narrow and the number of spectra for which there are detections . our expected recovery fraction is insensitive to small changes in the reasonable ranges used for the parameters . the lower s / n spectra have fewer fractional detections . as discussed in [ sims ] , the lower fractional detections at lower s / n values are due , in part , to our recovery efficiency but may also be due to a minimum total column density required for the appearance of cold . @xcite recently observed in the milky way and determined a minimum column density threshold of 2@xmath910@xmath54 cm@xmath15 for the formation of a cold phase of , which is just at the column density limit of where our sample begins . the top row of figure [ cdplot ] shows histograms of all of the observed column densities in our galaxy sample ( black ) , and the column densities where we detect cold ( red ) . the bottom row shows our detection fraction as a function of column density . the small fraction of narrow line detections at high s / n compared to what would be expected clearly shows that the cold that can be identified with this technique is inconsistent with a ubiquitous distribution . it should also be noted that there could exist cold at every line - of - sight below a s / n of 3.1 that we would be insensitive to . however , it would be unphysical for the presence of gas to be related to our s / n criterion and not to the column density . furthermore , absorption line studies in the milky way indicate that cold gas is not ubiquitous ( e.g. , @xcite ; @xcite ) . the typical velocity dispersion for the cold is @xmath44.5 km s@xmath1 . the broad components and best - fit single gaussian profiles vary by galaxy , but they generally have similar values . we expect the broad and best fit single components to be similar if they are both tracing the same warm . figure [ veldisp ] shows histograms of the narrow ( red ) , broad ( blue ) , and single ( black ) velocity dispersions for each galaxy . the dotted vertical line denotes our cold cutoff limit of 6 km s@xmath1 . the single component histograms have been scaled in order to show the narrow and broad component histograms more clearly . generally , the velocity dispersions of the broad components overlap with those of the single components . however , the peak in the broad component is offset from the peak in the single component towards higher values . this behavior is also seen for the majority of the cases in the literature , and is probably due to the fact that it is easier to identify a narrow component when the broad component has a much larger value . in figure [ disprad ] , we plot the average velocity dispersion , @xmath3@xmath7@xmath30 , of the warm ( black ) and cold ( red ) gas as a function of radius for each of our different velocity resolutions . we have omitted the locations best fit by a single gaussian with a velocity dispersion of less than 6 km s@xmath1 . the average warm velocity dispersion declines with radius . this trend is similar to the total velocity dispersion versus radius seen by @xcite . the decrease in the velocity dispersion with increasing radius may indicate a decrease in turbulent energy supplied by the underlying stellar population as the radius increases away from the center of the stellar disk . recent work on galaxy simulations of spiral disks have put limits on the predicted volume filling factor of the different gas phases in the ism ( e.g. , @xcite ) . the cold ( t @xmath3 1000 k ) ism occupies @xmath020% of a galaxy s volume , depending upon the star formation ( supernova ) rate . the cold neutral medium , however , may occupy only a few percent of the volume @xcite . the actual value will vary by galaxy and must also be sensitive to the available ism coolants ( e.g. , c@xmath57 , co , dust , etc . ) and the global gravitational potential . without knowing the exact 3-dimensional structure of the galaxies in our sample , we can not calculate the volume filling fractions . instead , we calculate the areal filling fractions ( @xmath48 ) , i.e. , the ratio of the number of lines - of - sight with detected cold and the total number of lines - of - sight with s / n @xmath30 10 ( see table [ indres ] , column 3 ) . we derive an average @xmath48 value of 9% . the true volume filling fractions based upon our detections are most likely even lower than these values since the scale height of the warm will be larger than that of the cold . the low filling fractions we compute are consistent with the ism model of @xcite . we also computed a lower limit to the total mass ( flux ) contribution of the cold for each galaxy using the narrow component of the gaussian fits ( @xmath49 ; see table [ indres ] , column 4 ) . @xcite and @xcite found that roughly 20% of the total mass is in the cold , which is significantly higher than the @xmath43% typical for our values of @xmath58 . it is unclear as to the reason for this large discrepancy since the data are of similar quality . the biggest differences may lie in the higher s / n resulting from the natural weighting these other studies use as compared to the robust weighting we employ . also , @xcite use a different statistical test in defining the acceptance of cold detections . these authors accept a double gaussian fit as statistically better than a single gaussian fit if the @xmath59@xmath25 value is improved by 10% . furthermore , our values are lower limits given our selection criteria and sensitivities . figure [ fracs ] plots @xmath48 and @xmath49 as a function of m@xmath2 ( left ) and @xmath60 ( right ) . no obvious trends arise between the areal filling fraction and mass fraction with absolute @xmath14-band magnitude or total mass . if we assume ( as noted in [ nodets ] ) that each line of sight has cold just below our detection limits , we can estimate an upper limit to the amount of cold in each galaxy . to do this we have used a representative cold gaussian of amplitude 3.1@xmath7 and velocity dispersion of 4.5 km s@xmath1 at each location where we do not detect cold . we then add this extra " mass to our detected cold mass to produce an upper limit to the cold mass fraction ( @xmath50 ; see column 5 in table [ indres ] ) . these values of @xmath61 can range up to 50% . however , since the cold is most frequently associated with higher values of n@xmath46 ( see [ colddets ] ) , it is unlikely that the true values of cold mass get as high as these upper limits . another , perhaps more reasonable way to calculate an upper limit to the cold - to - total mass fraction is to find the typical fractional flux contributed to the cold component along each line - of - sight and assume this fraction holds throughout the sample . for locations where we find both warm and cold components , the cold typically is not the dominant phase in the ism . the top row of figure [ histrat ] shows a normalized histogram of the flux ratio of cold and total defined as the ratio of the column density of the narrow gaussian profile and the total column density along the line - of - sight . the cold constitutes @xmath420% of the total line flux for any given line - of - sight . the bottom panels of figure [ histrat ] show this same ratio as a function of s / n for each velocity resolution . if we assume that _ every _ line - of - sight contains cold , then an upper limit to the cold - to - total mass ratio for each galaxy is @xmath420% . this value is what is reported by @xcite and @xcite . however , this value is also likely too high since it requires the entire galaxy to contain cold gas . @xcite observed the small magellanic cloud in 21 cm absorption and found that the mass fraction of cold to be less than 15% of the total content . it seems likely that the true value of the cold mass fractions of the low mass galaxies in our sample are between our lower limits and @xmath415% . @xcite observed 21-cm milky way absorption towards 12 background radio continuum sources . they were able to determine the flux contribution of the cold along many of these sight lines . they find that the cold contributes from a low of @xmath42% to as high as @xmath469% , of the total flux determined in the same manner , with a median of @xmath430% . while a direct comparison between the optically thick cold observed by @xcite and our optically thin emission is not obvious , we would not expect the fraction of cold in emission to be drastically different . our result of @xmath420% is slightly lower than their median value , but does not appear to be at odds with their result . as discussed in [ methods ] , some lines - of - sight were best fit by a single gaussian profile with a velocity dispersion of less than 6 km s@xmath1 . these best fit single gaussian profiles were found in 6 of the 23 galaxies with cold detections . most of these six galaxies have a large fraction of their at galactocentric radii beyond the @xmath425 mag arcsec@xmath15 optical level , and the majority of these single gaussian fits are also located at larger radii . figure [ muchoblooshto ] shows the radial distribution of the cold locations that lack a warm component . the galaxies have been ordered from low ( m81 dwarf a ) to high ( ngc 247 ) absolute @xmath14-band luminosity . two of these galaxies , m81 dwarf a and holmberg i , have large central holes in their distributions ( see @xcite and references therein ) where much of their gas is pushed beyond the optical radius . only ngc 247 has a majority of these locations within the optical radius . figure [ n42142plot ] shows an example of the difference between those areas that contain a warm component and those that lack one for ngc 4214 . the left panel shows only those locations that contain both a warm and cold component while the right panel also includes the locations that lack a warm component . the contours correspond to the 10@xmath54 and 10@xmath55 cm@xmath15 total column densities and the blue circle approximates the 25 mag arcsec@xmath15 level ( uncorrected for inclination and position angle ) . it is immediately apparent that the majority of the best fit single gaussian profiles fall well beyond the bulk of the stellar population and far from any heat sources . @xcite provides evidence that the warm velocity dispersions inside the optical radius of disk galaxies is driven by stellar processes , while areas beyond the optical radius may be driven by thermal broadening and/or magneto - rotational instabilities . these authors also point out that the velocity dispersions decline as a function of radius , falling below the typical 10 km s@xmath1 beyond the optical radius . we also see this declining trend in the velocity dispersion with radius ( see figure [ disprad ] ) . these results are suggestive that cold can be the dominant constituent of the ism only in the lower uv radiation field beyond the @xmath425 mag arcsec@xmath15 optical level . we next compare the cold gas masses to the molecular gas masses and star formation rates ( sfrs ) in each galaxy . to make these comparisons we compute the sfrs and estimate the amount of molecular hydrogen each galaxy contains . the relationship between the sfr and molecular hydrogen is discussed in detail in [ intro ] . we used the sfrs given by both fuv and h@xmath6 luminosities to estimate the amount of molecular hydrogen . table [ sfrs ] lists the sfrs derived from fuv and h@xmath6 luminosities taken from the literature for each galaxy with a detected cold signature . the columns are : ( 1 ) galaxy name , ( 2 ) mass of the cold ( @xmath62 ) computed using the lower limit mass fraction from table [ indres ] multiplied by the total mass given in table [ galprop ] , ( 3 ) apparent fuv magnitude ( m@xmath63 ; @xcite ) , ( 4 ) the logarithm of the fuv luminosity ( l@xmath63 ) , ( 5 ) the fuv sfr ( sfr@xmath63 ) computed using the relation given in @xcite , ( 6 ) the h@xmath64 gas mass ( @xmath65 ) computed from sfr@xmath63 assuming a 100 myr timescale and a star formation efficiency of 10% ( @xmath65 = sfr@xmath63 @xmath9 100 myr / 0.1 ) , ( 7 ) the approximate fraction of the cold ism attributed to the cold ( @xmath66 ) defined as @xmath62/(@xmath62 + @xmath65 ) , ( 8) the logarithm of the h@xmath6 luminosity ( l@xmath67 ; @xcite ) , ( 9 ) the sfr derived from l@xmath67 using the relation in @xcite , ( 10 ) the h@xmath64 gas mass ( @xmath68 ) computed from sfr@xmath67 assuming a 5 myr timescale and a star formation efficiency of 10% ( @xmath68 = sfr@xmath63 @xmath9 5 myr / 0.1 ) , and ( 11 ) the approximate fraction of the cold ism attributed to the cold ( @xmath69 ) defined as @xmath62/(@xmath62 + @xmath68 ) . using our lower limit mass fraction from the spectra above s / n of 10 and extrapolating it over the entire galaxy provides a lower limit to the total cold mass . the expected relationship between the currently observed cold and the computed molecular gas masses is not immediately clear . the mass fractions of the cold ism attributed to the cold is lower for h@xmath64 masses computed from the longer timescale sfr@xmath63 ( table [ sfrs ] , column 7 ) than they are for those computed from sfr@xmath67 ( table [ sfrs ] , column 11 ) . the time scale to convert the cold into molecular hydrogen is likely dependent upon the metallicity of the environment ( a longer time scale is expected for lower metallicities ) and local uv radiation field ( a shorter time scale is expected for lower uv radiation ) . since dwarf galaxies have both low metallicities and different ranges of uv radiation strengths , it is not obvious which molecular gas mass estimate provides the best comparison . unfortunately , our current data do not allow us to investigate these timescales further . the top and bottom panels of figure [ sfrfig ] show plots of the sfr@xmath63 ( filled circles ) and sfr@xmath67 ( open squares ) as a function of the cold mass and total mass , respectively . if the cold is related to the star formation in any direct way , we would expect to find a trend in this relation . indeed , the higher the cold gas mass , the higher the sfr , although this same trend is seen with the total gas mass and may just be a consequence of the more gas available , the higher the sfr . the middle panel shows the sfr efficiency ( sfe@xmath70 ) defined as the sfr divided by the total gas mass as a function of the cold gas mass . the sfe@xmath70 is roughly constant over our range of cold masses . the differences between the derived fuv and h@xmath6 star formation rates has been previously reported in @xcite and is due at least in part to the fact that the sfrs are averaged over different timescales . environmental conditions may also influence sfr@xmath67 by leakage of the ionizing photons from the star forming regions ( see @xcite for details ) . we can compare our molecular gas mass estimates derived from the sfrs to a direct observational estimate of the molecular gas in ngc 4214 @xcite to see if our values are reasonable . these authors observed co emission in the inner @xmath41@xmath716@xmath91@xmath716 of the star forming disk of ngc 4214 ( an area much smaller than the disk , see [ nowarm ] ) and derived a molecular gas mass of 5.1@xmath910@xmath72 @xmath12 . this value is similar to the molecular gas mass estimate we get from sfr@xmath67 of 6.1@xmath910@xmath72 @xmath12 . the molecular gas mass value computed by @xcite was based on an assumed co to h@xmath64 conversion factor ( the so called x - factor ) derived from observations of galactic environments ( i.e. , high metallicity ) . recent work on the x - factor has shown that the relation changes with decreasing mean visual extinction ( e.g. , @xcite ) . the lower metallicity environment of dirr galaxies results in lower visual extinction values and thus a different x - factor value . depending upon the true mean visual extinction value of ngc 4214 , the actual molecular gas mass could be higher by a factor of a few to an order of magnitude or more . @xcite derive an x - factor value roughly 40 times higher than the galactic value for ngc 4214 . using this value , the @xcite molecular gas mass increases to @xmath42@xmath910@xmath73 @xmath12 , closer to our derived h@xmath64 mass using sfr@xmath63 of 1@xmath910@xmath73 @xmath12 . @xcite also derive a molecular gas mass of @xmath42@xmath910@xmath73 @xmath12 for ngc 4214 . the factor of 2 difference between the values most likely arise from our basic assumptions of a timescale of 100 myr for the fuv emission and/or the star formation efficiency of 10% . unfortunately , the lack of co detections for the few galaxies in our sample that have been observed for co prohibit us from repeating this exercise for other galaxies in our sample ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ) . nevertheless , our estimates of molecular gas masses , although uncertain , are not unreasonable . direct comparisons between the co detections in ngc 4214 and our cold detections is the focus of future work . we have observed line - of - sight emission spectra in 31 nearby , low metallicity dirr galaxies in order to search for cold defined by a velocity dispersion of less than 6 km s@xmath1 . we have detected it in 23 of 27 galaxies after quality control cuts were applied . the cold may be the future sites of molecular cloud and star formation and to date , are the only way to potentially trace star forming gas at low metallicity . based upon our observations , we find : * the cold gas is found in localized regions and usually at total column densities above the canonical threshold of star formation of 10@xmath55 cm@xmath15 @xcite . the cold in a given galaxy is also not typically associated with the very highest surface density gas ( see figures [ 2plots ] & [ cdplot ] ) . * the cold has a typical velocity dispersion of @xmath44.5 km s@xmath1 ( t @xmath0 800 k ) . * we derive an average value to the volume filling fraction of the cold of 9% ( assuming our lower limit detections ) . * we find lower limits to the cold gas mass fractions of a few percent , consistent with some models of the multi - phase ism @xcite . * the sfe@xmath70 is roughly constant as a function of the total cold mass over the observed gas mass ranges . * the cold contributes @xmath420% of the line - of - sight flux in locations where we detect both a cold and warm component . * cold gas that lacks a warm component is typically found at radii larger than the 25 mag arcsec@xmath15 optical radius . barone , l. t. , heithausen , a. , httemeister , s. , fritz , t. , & klein , u. 2000 , , 317 , 649 begum , a. , chengalur , j. n. , karachentsev , i. d. , kaisin , s. s. , & sharina , m. e. 2006 , , 365 , 1220 begum , a. , chengalur , j. n. , karachentsev , i. d. , sharina , m. e. , & kaisin , s. s. 2008 , , 386 , 1667 begum , a. , stanimirovi , s. , goss , w. m. , et al . 2010 , , 725 , 1779 berg , d. a. , skillman , e. d. , marble , a. r. , et al . 2012 , arxiv:1205.6782 bigiel , f. , leroy , a. , walter , f. , et al . 2008 , , 136 , 2846 braun , r. 1997 , , 484 , 637 briggs , d. s. 1995 , ph.d . thesis , new mexico institute of mining and tchnology brinks , e. , & bajaja , e. 1986 , , 169 , 14 dalcanton , j. j. , williams , b. f. , seth , a. c. , et al . 2009 , , 183 , 67 de avillez , m. a. , & breitschwerdt , d. 2004 , , 425 , 899 de blok , w. j. g. , & walter , f. 2006 , , 131 , 363 dickey , j. m. , mebold , u. , stanimirovic , s. , & staveley - smith , l. 2000 , , 536 , 756 engelbracht , c. w. , gordon , k. d. , rieke , g. h. , et al . 2005 , , 628 , l29 gil de paz , a. , boissier , s. , madore , b. f. , et al . 2007 , , 173 , 185 glover , s. c. o. , & mac low , m .- 2011 , , 412 , 337 gordon , k. d. , prez - gonzlez , p. g. , misselt , k. a. , et al . 2004 , , 154 , 215 israel , f. p. , tacconi , l. j. , & baas , f. 1995 , , 295 , 599 jackson , d. c. , cannon , j. m. , skillman , e. d. , et al . 2006 , , 646 , 192 jansen , r. a. 2000 , ph.d . thesis , kapteyn astronomical institute , univ . groningen kanekar , n. , braun , r. , & roy , n. 2011 , , 737 , l33 karachentsev , i. d. , karachentseva , v. e. , huchtmeier , w. k. , & makarov , d. i. 2004 , , 127 , 2031 kennicutt , r. c. , jr . 1998 , , 498 , 541 kennicutt , r. c. , jr . , armus , l. , bendo , g. , al . 2003 , , 115 , 928 kennicutt , r. c. , jr . , lee , j. c. , funes , s. j. , et al . 2008 , , 178 , 247 kennicutt , r. c. , jr . , & evans , n. j. , ii 2012 , arxiv:1204.3552 kro , m. , & goldsmith , p. f. 2010 , , 724 , 1402 krumholz , m. r. , mckee , c. f. , & tumlinson , j. 2009 , , 693 , 216 lee , h. , skillman , e. d. , cannon , j. m. , et al . 2006 , , 647 , 970 lee , j. c. , gil de paz , a. , tremonti , c. , et al . 2009 , , 706 , 599 lee , j. c. , gil de paz , a. , kennicutt , r. c. , jr . , et al . 2011 , , 192 , 6 leroy , a. , bolatto , a. d. , simon , j. d. , & blitz , l. 2005 , , 625 , 763 leroy , a. k. , walter , f. , brinks , e. , et al . 2008 , , 136 , 2782 leroy , a. k. , walter , f. , bigiel , f. , et al . 2009 , , 137 , 4670 li , d. , & goldsmith , p. f. 2003 , , 585 , 823 maloney , p. , & black , j. h. 1988 , , 325 , 389 mckee , c. f. , & ostriker , j. p. 1977 , , 218 , 148 merrifield , m. r. 1993 , , 261 , 233 relano , m. , kennicutt , r. c. , jr . , eldridge , j. j. , lee , j. c. , & verley , s. 2012 , arxiv:1204.4502 roychowdhury , s. , chengalur , j. n. , begum , a. , & karachentsev , i. d. 2009 , , 397 , 1435 salim , s. , rich , r. m. , charlot , s. , et al . 2007 , , 173 , 267 schruba , a. , leroy , a. k. , walter , f. , et al . 2012 , arxiv:1203.4231 skillman , e. d. 1987 , in lonsdale persson c.j . , ed . , star formation in galaxies . nasa conference publication , 2466 , 263 spaans , m. , & norman , c. a. 1997 , , 483 , 87 tamburro , d. , rix , h .- w . , leroy , a. k. , et al . 2009 , , 137 , 4424 taylor , c. l. , kobulnicky , h. a. , & skillman , e. d. 1998 , , 116 , 2746 van der marel , r. p. , & franx , m. 1993 , , 407 , 525 verter , f. , & hodge , p. 1995 , , 446 , 616 walter , f. , taylor , c. l. , httemeister , s. , scoville , n. , & mcintyre , v. 2001 , , 121 , 727 walter , f. , brinks , e. , de blok , w. j. g. , et al . 2008 , , 136 , 2563 warren , s. r. , weisz , d. r. , skillman , e. d. , et al . 2011 , , 738 , 10 wolfire , m. g. , mckee , c. f. , hollenbach , d. , & tielens , a. g. g. m. 2003 , , 587 , 278 young , l. m. , & lo , k. y. 1996 , , 462 , 203 young , l. m. , & lo , k. y. 1997 , , 490 , 710 young , l. m. , van zee , l. , lo , k. y. , dohm - palmer , r. c. , & beierle , m. e. 2003 , , 592 , 111 /a@xmath39 ) as a function of the ratio of the broad and narrow velocity dispersions ( @xmath40/@xmath41 ) for the 1.3 km s@xmath1 velocity resolution simulation in four different s / n bins . the grey dots are all of the simulated spectra . the red dots are those spectra identified as containing two components by our fitting routines , and the blue dots are the simulated spectra that were identified only containing a single gaussian component with a velocity dispersion of less than 6 km s@xmath1 . the values in the parentheses are the recovery percentages for the best fit double ( left ) and single ( right ) gaussians . we clearly identify more of the spectra at higher s / n values . our routines have a harder time recovering the two gaussian components when the velocity dispersions are similar . [ mcarlo2],width=661 ] . the velocity resolution ( @xmath7 ) is listed below the galaxy names . typically , the broad component velocity dispersions overlap with the single gaussian profile velocity dispersions indicating they arise from the same gas phase . [ veldisp],width=661 ] @xmath7@xmath30 , as a function of radius for the warm ( black ) and cold ( red ) gas for each of our velocity resolutions . we have omitted the locations best fit by a single gaussian with a velocity dispersion of less than 6 km s@xmath1 . the averages are taken over 200 pc bins ( the beam size ) . the error bars are the dispersions of the values in the bins . the velocity of the warm component decreases with radius indicating a possible decrease in turbulence as the radius increases from the main stellar body . some of the errorbars at large radii for the 2.6 km s@xmath1 velocity resolution overlap with values below our 6 km s@xmath1 cutoff . this is due to the dispersions being affected by a few larger values in these bins . [ disprad],width=661 ] ) and the lower limit to the fraction of gas in the cold phase ( @xmath49 ) as a function m@xmath2 ( left ) and @xmath60 ( right ) . the areal filling fraction and mass fractions do not correlate with m@xmath2 or @xmath60 . [ fracs],width=661 ] km s@xmath1 ) component as a function of radius . the x - axis has been normalized by the 25 mag arcsec@xmath15 radius ( vertical dashed line ) . the galaxies have been ordered faint ( m81 dwarf a ) to bright ( ngc 247 ) absolute @xmath14-band magnitude . a significant fraction of the cold is outside of the optical radius of each galaxy , except for ngc 247 . [ muchoblooshto],width=661 ] only . the contours represent the 10@xmath54 and 10@xmath55 cm@xmath15 total column densities . the green circle approximates the 25 mag arcsec@xmath15 optical level . the beam is shown at lower left . the majority of the cold detections described by a single gaussian profile are outside of the bulk of the stellar distribution . [ n42142plot],width=661 ] ( filled circles ) and sfr@xmath67 ( open squares ) as a function of the cold mass . _ middle : _ the sfr efficiency ( sfe@xmath70 ) defined as the sfr divided by the total gas mass as a function of the cold gas mass . _ bottom : _ the sfr as a function of the total gas mass . the sfr shows a linear trend with both cold and total mass . [ sfrfig],width=661 ] |
we propose a simple spectral model for the seyfert 1 galaxy mcg-6 - 30 - 15 that can explain most of the 1 40 kev spectral variation by change of the partial covering fraction , similar to the one proposed by miller et al . ( 2008 ) . our spectral model is composed of three continuum components ; ( 1 ) a direct power - law component , ( 2 ) a heavily absorbed power - law component by mildly ionized intervening matter , and ( 3 ) a cold disk reflection component far from the black hole with moderate solid - angle ( ) accompanying a narrow fluorescent iron line . the first two components are affected by the surrounding highly ionized thin absorber with and . the heavy absorber in the second component is fragmented into many clouds , each of which is composed of radial zones with different ionization states and column densities , the main body ( , ) , the envelope ( , ) and presumably a completely opaque core . these parameters of the ionized absorbers , as well as the intrinsic spectral shape of the x - ray source , are unchanged at all . the central x - ray source is moderately extended , and its luminosity is not significantly variable . the observed flux and spectral variations are mostly explained by variation of the geometrical partial covering fraction of the central source from 0 ( uncovered ) to.63 by the intervening ionized clouds in the line of sight . the ionized iron k - edge of the heavily absorbed component explains most of the seemingly broad line - like feature , a well - known spectral characteristic of mcg-6 - 30 - 15 . the direct component and the absorbed component anti - correlate , cancelling their variations each other , so that the fractional spectral variation becomes the minimum at the iron energy band ; another observational characteristic of mcg-6 - 30 - 15 is thus explained . | 1202.2797 | the seyfert 1 galaxy mcg-6 - 30 - 15 is the primary `` disk - line '' target , in which a relativistically broadened iron k - emission line is suggested to originate in the innermost region of the accretion disk around the black hole ( e.g. , tanaka et al . mcg-6 - 30 - 15 is also known to exhibit characteristic x - ray spectral variations ; inoue & matsumoto ( 2001 , 2003 ) , fabian et al . ( 2002 ) and matsumoto et al . ( 2003 ) reported significantly small variability in the iron line energy band of mcg 6 - 30 - 15 . it is found that the energy - dependent root mean square ( rms ) variability is suppressed around the iron line energy , particularly in longer timescales . inoue and matsumoto ( 2001 , 2003 ) proposed that the absorbed spectrum due to photo - ionized warm absorbers mimic shape of the strongly red - shifted iron line , and variation of the warm absorbers may explain the apparently small variability in the iron line energy band . on the other hand , miniutti and fabian ( 2004 ) proposed that the suppressed variability in the iron line band may be explained by the general relativistic light - bending effects , which take place in the very vicinity of the black hole . miniutti et al . ( 2007 ) claim that the suzaku spectrum of mcg-6 - 30 - 15 is in fact consistent with the light - bending model . niedwiecki & ycki ( 2008 ) and niedwiecki & miyakawa ( 2010 ) independently re - examined the light - bending model , and concluded that it is possible to explain the suppressed variability in the iron energy band assuming a particular configuration and movement of the illuminating source . ycki et al . ( 2010 ) , on the other hand , claim that the light - bending model is not able to explain the suzaku spectrum , if a broader energy range is adopted than used by miniutti et al . ( 2007 ) . the primary question in mcg-6 - 30 - 15 is whether the iron line is truely broad and its little variability is a consequence of the general relativistic light bending effects or not . the critical point to address the problem is to model the underlying continuum spectrum correctly , since the broad iron line parameters are dependent on the choice of the continuum spectral model . in fact , on one hand , miller , turner , and reeves ( 2008 ) claim that the `` disk - line '' is not required , when the continuum spectrum is modeled with multiple warm absorbers . on the other hand , reynolds et al . ( 2009 ) argue that the absorption - dominated model over - predicts the 6.4 kev fluorescent line emission , and the disk - line model is more physically reasonable . miller et al . ( 2009 ) counterargue that the calculation in reynolds et al . ( 2009 ) neglects opacity at the 6.4 kev line energy , and conclude that variation in the partial - covering fraction may dominate the observed x - ray spectral variability . however , the partial covering scenario by miller et al . ( 2009 ) requires a number of free parameters to model the observed spectrum . therefore , in order to examine the partial covering scenario , reducing the number of parameters , if possible , and constructing a concrete physical picture to explain the spectral variation is certainly intriguing . recently , using the suzaku data taken in january 2006 , miyakawa et al . ( 2009 ) ( hereafter paper i ) found a clear correlation between the intensity in the 6 10 kev band and the spectral ratio of 0.5 3.0 kev / 6 10 kev in a time - scale of @xmath9 sec . essentially the same spectral correlation was confirmed for longer timescales up to @xmath814 years in the rxte data ( miyakawa 2010 ) . paper i adopted a spectral model composed of a direct power - law component , its reflection component by a neutral matter ( with the solid - angle @xmath10 ) , two warm absorbers with different ionization states . the observed spectral variation requires change of the apparent slope of the direct component , whereas the shape and intensity of the reflection component being invariable . a mildly broad iron emission line was required at @xmath11 kev with an intrinsic width of 0.29 kev ( 1 @xmath12 ) and the equivalent - width 100 ev ; such an extremely broadened `` disk line '' as is claimed by miniutti et al . ( 2007 ) was not required . in this paper , following paper i , we attempt to comprehensively understand spectral variability of mcg-6 - 30 - 15 using suzaku and chandra archival data . our goal is to find a spectral model which can naturally explain the observed spectral variation of mcg-6 - 30 - 15 with a minimum set of free - parameters . if we are successful to construct such a reasonable spectral model , we will see if the relativistically distorted iron emission line is truely present or not in the energy spectrum . we believe this is an important step forward to fully understand x - ray spectral properties of mcg-6 - 30 - 15 , seyfert galaxies , and black holes . we use the same suzaku data used in paper i , taken in 2006 from january 9 to 14th ( 143 ksec exposure ) , from 23 to 26 ( 99 ksec ) and from 27 to 30 ( 97 ksec ) . we use the xis data in 1 10 kev , and pin data in 10 40 kev . in the present paper , we are interested in spectral variations above 1 kev , so we did not use the data below 1 kev . all of the spectral fits were made with xspec v11 ( arnaud 1996 ) . in the following , the xspec model names used in the spectral analysis are explicitly given . a constant factor to adjust normalizations between xis and pin is fixed at 1:1.086 ( ishida et al . 2007 ) . for more details about suzaku data reduction , see paper i. the chandra satellite observed mcg-6 - 30 - 15 several times . in this paper , we use the high energy transmission grating spectrometer ( hetgs ) data taken between 2004 may 19 and 27 , when the source was observed four times resulting in a good exposure time of 522 ksec . for data reduction we used the ciao 4.0 software package ; `` tgextract '' is used to produce pha2 spectral files from the level2 event file . hetgs is composed of high energy grating ( heg ) and medium energy grating ( meg ) . the @xmath13 and @xmath14 orders of the heg were combined , and so were the @xmath13 and @xmath14 orders of the meg . the heg and meg spectra were fitted separately . errors quoted in this paper are at statistical 90@xmath15 confident level . in paper i , spectral variation of mcg-6 - 30 - 15 was studied assuming the `` two - component '' spectral model given as , @xmath16 where @xmath17 is the intrinsic power - law spectrum with a high - energy cut - off ( fixed at 160 kev ; guainazzi et al . 1999 ) , @xmath18 is normalization factor of the directly observed component , @xmath19 is normalization factor of the reflected component , @xmath20 is reflection albedo by the cold optically thick matter ( `` pexrav '' in xspec : magdziarz @xmath21 zdziarski 1995 ) , @xmath22 and @xmath23 represent attenuation by a high- and a low - ionized warm absorbers , respectively , and @xmath24 is an iron @xmath25 emission line . effect of interstellar absorption is always taken into account in the spectral fitting ( `` phabs '' ; balucinska - church @xmath21 mccammon 1992 ) , but its term is omitted for simplicity in the above expression and hereafter . amount of the disk reflection is measured with @xmath26 , solid angle of the reflector seen from the central source . ratio of the normalization of the reflection component to that of the cutoff power - law component corresponds to @xmath26/2@xmath27 , assuming isotropic emission of the central source . we used xstar version 2.1kn8 ( kallman et al . 2004 ) to model the warm absorbers . the temperature , pressure and density of the warm absorbers are assumed to be @xmath28 k , 0.03 dyne @xmath29 and @xmath30 , respectively . the incident photo - ionizing spectrum is assumed to be a power - law with the index 2.0 , and the solar abundance by greeves , noels and sauval ( 1996 ) is adopted for the ionized material . we made a grid model by running xstar for different values of @xmath31 and @xmath32 ; the log @xmath31 values are from 0.1 to 5 ( erg cm s@xmath33 ) and the @xmath32 values are from @xmath34 to @xmath35 ( @xmath2 ) . the number of steps for log @xmath31 and @xmath32 are both 20 , thus our grid model has @xmath36 grid - points . the redshift of a warm absorber and the source are fixed at 0.001 and 0.00775 , respectively ( young et al . 2005 , fisher et al . 1995 ) . in paper i , we found that the spectral variation at various timescales is successfully explained by variations of two parameters , normalization factor of the directly observed component ( @xmath18 ) and apparent slope of the direct component , such that the slope gets steeper for greater normalizations . the apparent slope change is described by either change of the index of the power - law component ( @xmath17 ) , ionization state of the low - ionized warm absorber , or column density of the low - ionized warm absorber ( @xmath23 ) . below , we scrutinize the spectral model in paper i carefully , and introduce a more elaborated model . using the chandra hetgs , young et al . ( 2005 ) resolved a narrow emission line at 6.4 kev in mcg-6 - 30 - 15 , which we confirmed via re - analysis of the archival data . the equivalent width of the line is measured as @xmath37 ev , which requires a cold reflector with a solid angle of @xmath26/2@xmath27 @xmath8 0.3 viewed from the central x - ray source ( e.g. , george @xmath21 fabian 1991 ) . on the other hand , the best - fit values of @xmath18 and @xmath19 obtained in paper i indicates that @xmath38 . consequently , in order to account for the observed disk reflection spectrum of an amount of @xmath26/2@xmath27 @xmath8 1 including the outer disk reflection with @xmath26/2@xmath27 @xmath8 0.3 , we need another disk reflection component or a different spectral component which is similar to disk reflection , corresponding to an amount of @xmath26/2@xmath27 @xmath8 0.7 . thus , we separate the @xmath19 term in ( [ eq1 ] ) into two terms . one is the reflection continuum component from a cold reflector emitting a narrow iron @xmath39 line , solid angle of which is @xmath26/2@xmath27 @xmath8 0.3 . the other is introduced to represent the remaining part of the @xmath19 term ; for this , we adopt a heavily absorbed component by warm absorber , since the absorbed component is similar to the reflection component in shape , and does not have to accompany the fluorescent iron emission line . in fact , a partial covering model has been proposed to explain the continuum spectral shape and variation of mcg-6 - 30 - 15 ( e.g. , matsuoka et al . 1990 ; mckernan and yaqoob 1998 ; miller , turner and reeves 2008 , 2009 ) . resultantly , we introduce a new spectral model expressed as @xmath40 where @xmath41 is normalization of the direct component , @xmath42 is that for the heavily absorbed component , @xmath43 stands for another warm absorber that is for the ionized optically thick partial absorber . the normalization factor of the outer reflection component , @xmath44 is constant and related to the average of @xmath41 so that @xmath45 , and @xmath24 is a narrow iron emission line with a fixed equivalent width of @xmath46 ev . hereafter , this model is referred as the `` three - component '' model . we try the three - component model to the suzaku average spectrum in 1 40 kev . we found the fit improves if we additionally put a weak iron neutral edge at 7.11 kev and an absorption line at 7.0 kev . the former is considered to be due to a slight difference of the assumed @xmath26 from the exact value , and the latter due to deficiency of the h - like iron absorption line equivalent - width already included in the xstar model . figure @xmath47 and table [ miyakawa_model_ave_suzaku ] show the fitting result for the suzaku average spectrum with the three - component model in 1 40 kev . parameters of the high - ionized and low - ionized warm absorbers are similar to those in paper i. for the newly introduced absorption component ( @xmath43 ) , we obtain @xmath48 1.6 @xmath49 10@xmath50 @xmath2 and @xmath51 . normalizations of the heavily absorbed component and the direct component are 0.60 and 1.44 ( @xmath52 photons s@xmath33@xmath2 at 1 kev ) , respectively , which suggests that the source is partially covered by the ionized thick matter by @xmath53 30 % . we note that the absorbing matter is marginally compton thick , so that the compton scattering may not be negligible for precise spectral calculation ( e.g. , yaqoob 1997 ) . this is not considered in the present paper , and will be a subject for future work . cc + @xmath32 ( @xmath54 @xmath29 ) & 1.5@xmath55 + + @xmath32 ( @xmath56 @xmath29 ) & 2.4@xmath57 + log @xmath31 & [email protected] + + @xmath32 ( @xmath54 @xmath29 ) & [email protected] + log @xmath31 & 1.54@xmath59 + + 10@xmath60 ph / s/@xmath61 at 1 kev & [email protected] + + 10@xmath60 ph / s/@xmath61 at 1 kev & [email protected] + + @xmath32 ( @xmath35 @xmath29 ) & 1.6@xmath62 + log @xmath31 & 1.57@xmath63 + + photon index & 1.91@xmath64 + @xmath65 ( kev ) & 160(fixed ) + + & @xmath660.3 + + cosincl & 0.866(fixed ) + + line e ( kev ) & 6.35 ( fixed ) + sigma ( kev ) & 0.01 ( fixed ) + norm ( @xmath67ph / s/@xmath61 ) & [email protected] + ew ( ev ) & 27@xmath587 + edge e ( kev ) & 7.11 ( fixed ) + maxtau & [email protected] + line e ( kev ) & 7.0 ( fixed ) + sigma ( kev ) & 0.01(fixed ) + norm ( @xmath68ph / s/@xmath61 ) & @xmath692.4 + line e ( kev ) & [email protected] + sigma ( kev ) & 0.01(fixed ) + norm ( @xmath67ph / s/@xmath61 ) & @xmath700.5 + reduced chi - square ( d.o.f ) & 1.12 ( 203 ) + + [ miyakawa_model_ave_suzaku ] ( 90mm,90mm)figure1.eps cc + @xmath32 ( @xmath54 @xmath29 ) & 3.5@xmath71 + + e ( kev ) & [email protected] + index & @xmath72 ( fixed ) + @xmath73 ( @xmath74 ) & 220@xmath75 + @xmath76 ( @xmath74 ) & 400 ( fixed ) + inclination ( deg ) & 30 ( fixed ) + norm ( @xmath67 ) & 1.3@xmath77 + ew ( ev ) & 40@xmath5820 + + @xmath32 ( @xmath56 @xmath29 ) & 2.9@xmath78 + log @xmath31 & [email protected] + + @xmath32 ( @xmath54 @xmath29 ) & 3.7@xmath79 + log @xmath31 & 1.52@xmath80 + + 10@xmath60ph / s/@xmath61 at 1 kev & 1.44@xmath81 + + 10@xmath60ph / s/@xmath61 at 1 kev & 0.55@xmath82 + + @xmath32 ( @xmath35 @xmath29 ) & 1.7@xmath83 + log @xmath31 & 1.65@xmath84 + + photon index & 1.91@xmath64 + @xmath65 ( kev ) & 160(fixed ) + + & @xmath850.3 + + cosincl & 0.866(fixed ) + + line e ( kev ) & 6.35 ( fixed ) + sigma ( kev ) & 0.01 ( fixed ) + norm ( @xmath68ph / s/@xmath61 ) & 5.4@xmath86 + ew ( ev ) & 12@xmath87 + edge e ( kev ) & 7.11 ( fixed ) + maxtau & [email protected] + line e ( kev ) & 7.0 ( fixed ) + sigma ( kev ) & 0.01(fixed ) + norm ( @xmath68ph / s/@xmath61 ) & @xmath88 + line e ( kev ) & [email protected] + sigma ( kev ) & 0.01(fixed ) + norm ( @xmath67ph / s/@xmath61 ) & @xmath700.5 + reduced chi - square ( d.o.f ) & 1.07 ( 200 ) + + [ miyakawa_diskline_model_ave_suzaku ] ( 90mm,90mm)figure2.eps several authors claim presence of an extremely broad `` disk line '' in the energy spectrum of mcg-6 - 30 - 15 ( e.g. , miniutti et al . therefore , we examine if presence of the disk line is reconciled with our three - component model . we analyze the time - average spectrum of mcg-6 - 30 - 15 adding a disk line model from a fast - rotating kerr black hole ( laor 1991 ) . thus , the model we adopt here has both the narrow line from very far from the black hole , and a putative broad line which is from the region very close to the black hole . we try to fit the spectrum with laors disk line model with varying the inner disk radius , @xmath73 . table [ miyakawa_diskline_model_ave_suzaku ] shows the best - fitting result for the suzaku average spectrum adding laor s model to the three - component model , where the fit is slightly improved with @xmath89 200 @xmath90 and the reduced chi - square is 1.07 ( @xmath91/d.o.f = 214/200 ) . note that the additional line is only mildly broadened with @xmath89 200 @xmath90 , and the equivalent width of the broad disk line is just 40 @xmath58 20 ev . figure @xmath92 shows the chi - squares of fitting to the average spectrum as a function of the inner disk radius . as seen from this figure , @xmath93 chi - squares become greater as @xmath73 gets less than @xmath94 200 @xmath90 , and it exceeds 2.71 when @xmath73 is less than 9 @xmath90 . namely , our three - component model is only reconciled with a mildly broad disk line from the region @xmath95 with 90 % confidence . the reason we do not require a strong , extremely distorted disk line is primarily because that the newly introduced heavily absorbed component in our three - component model ( the @xmath96 term in equation [ eq2 ] ) has an ionized iron edge feature which resembles the broad disk line shape ( figure [ miyakawa_model_fit_suzaku ] ) . secondary reason is that the warm absorbers ( @xmath23 and @xmath22 in equation [ eq2 ] ) have spectral curvatures below @xmath85 kev without which the residual may look like a low - energy tail of the disk line . consequently , we see that the disk line feature is dependent on the choice of the continuum spectral models , and an extremely distorted disk line is not required in our three - component model . next , we apply the three - component model to the eight `` intensity - sliced spectra '' ( the same spectral sets as used in paper i ) to investigate for spectral variations . the method of creating the intensity - sliced energy spectra is as follows : ( 1 ) we created a light curve ( the average of xis0 , xis2 , and xis3 ) , with a bin - width of 128 s in the 0.2 12 kev band . we found that the counting rate varies from @xmath97 to @xmath98 cts s@xmath33 . ( 2 ) we chose eight time - periods when the source intensity is in the ranges of 11.75 , 1.752.50 , 2.503.25 , 3.254.00 , 4.004.75 , 4.755.50 , 5.506.25 , 6.257.00 cts s@xmath33 . these intensity ranges are chosen so that the exposure time for each intensity bin be approximately equal . ( 3 ) from the eight time - periods corresponding to the different source flux levels , we created eight intensity - sliced energy spectra ( the sum of xis0 , xis2 , and xis3 ) . here , all the parameters are made the same for the eight spectra except the following three parameters ; the direct power - law normalization , @xmath41 , the absorbed power - law normalization , @xmath42 , and ionization degree of the low - ionized warm absorber , @xmath23 . attenuation by a warm absorber is expressed as @xmath99 , where @xmath100 is energy and ionization - degree dependent cross - section and @xmath101 is the hydrogen column density of the low - ionized absorber . variation of @xmath23 is due to either a variation of @xmath100 or @xmath101 , or both . here , we investigate two cases , one with a variation of the cross - section @xmath100 due to change of the ionization parameter @xmath31 , the other with a variation of @xmath102 . eight spectra are fitted simultaneously , and the both cases equally exhibit acceptable fits with reduced chi - square values 1.09 and 0.95 ( d.o.f = 1135 ) , respectively . thus , we could not tell if the ionization parameter or column density of the low - ionized warm absorber is more variable from spectral fitting alone . when @xmath101 is fixed , the top and bottom panel of figure @xmath103 shows the relation between @xmath41 and @xmath104 and that between @xmath41 and @xmath42 , respectively . similarly , when @xmath31 is fixed , the top and bottom panel of figure @xmath105 shows the relation between @xmath41 and @xmath101 and that between @xmath41 and @xmath42 , respectively . in either case , correlations between each two of the three parameters are obvious . note that we see two independent correlations among three free parameters . this indicates that there should be a single principle parameter which is primarily responsible for the observed spectral variations ( section [ sec : vpcmodel ] ) . ( 90mm,90mm)figure3a.eps ( 90mm,90mm)figure3b.eps ( 90mm,90mm)figure4a.eps ( 90mm,90mm)figure4b.eps we have seen that the three - component model is effective to describe the suzaku spectral variation , and that the ionization degree of the low - ionized warm absorber is variable with intensities . next , we will see if the three - component model is valid to describe the chandra hetg spectra and their variations . strength of chandra hetg compared to suzaku is that it can study variation of individual low - energy absorption lines corresponding to change of the ionization degree . first , we fit the chandra / hetgs spectra with the three - component model in 1.0 7.5 kev . since the energy range is much smaller than that of suzaku , some parameters are not constrained and thus fixed to the suzaku best - fit values . in figure @xmath106 and table [ miyakawa_model_ave_chandra ] , we show the fitting result for the average spectrum . we see that the chandra hetg spectra are successfully modeled with the three - component model with similar parameters to those of suzaku . next , we study possible ionization degree variations in the chandra / hetgs data by extracting the `` bright spectrum '' and `` faint spectrum '' as follows : 1 ) create a light curve with a time - bin - width of 128 sec , and calculate average counting rates . 2 ) create the `` bright spectrum '' from the period when the meg and heg count rates ( 0.410 kev ) are higher than the average , and the `` faint spectrum '' when the count rates are lower than the average . we fit the bright / faint spectra in the 1.151.55 kev band with a single power - law and two negative gaussians for the mgxi ( he - like 1.34 kev ) and mgxii ( h - like 1.47 kev ) absorption lines . we also fit the bright / faint spectra in the 1.752.10 kev band separately with a single power - law and two negative gaussians to investigate for the sixiii ( he - like 1.851.86 kev ) and sixiv ( h - like 2.00 kev ) lines . in figure @xmath107 , we show the fitting result for the bright / faint spectra in both energy bands . we could fit the bright and faint spectra with the common photon index , while only the power - law normalization and the normalization of absorption lines are varied . the best fit parameters are shown in table @xmath108 . we notice that the mgxi absorption line equivalent width is larger in the faint state , while that of the mgxii line is larger in the bright state . this is understood as due to change of the ionization state of the warm absorber , such that mgxi is more abundant in the faint state , while mgxii is more abundant in the bright state . in fact , ion fraction of mgxi and that of mgxii become equal at @xmath109 ( e.g. , kallman @xmath21 bautista 2001 ) , that is about the best - fit value of the low - ionized warm absorber ( tables [ tab : chandra3comp ] ) . as for si absorption lines , the sixiii line equivalent width is oppositely larger for the bright spectrum , which is explained by the low - ionized warm absorber with @xmath110 , since ion fraction of sixiii increases when @xmath104 varies from @xmath111 to @xmath112 ( kallman @xmath21 bautista 2001 ) . fraction of sixiv should increase more dramatically when @xmath104 increases accordingly ( kallman @xmath21 bautista 2001 ) , but the observed equivalent widths of the sixiv line is not much different . this result is explained by the presence of the high - ionized warm absorber at @xmath113 , which produces a significant amount of sixiv but not sixiii . consequently , observed variation of the mg and si absorption line equivalent - widths is primarily explained by change of the ionization degree of the low - ionized warm absorber according to the flux changes . ( 90mm,90mm)figure5.eps ( 90mm,90mm)figure6a.eps ( 90mm,90mm)figure6b.eps [ around_mg ] [ tab : chandra3comp ] cc + @xmath32 ( @xmath54 @xmath29 ) & 1.5(fixed ) + + @xmath32 ( @xmath56 @xmath29 ) & 2.4(fixed ) + log @xmath31 & [email protected] + redshift ( @xmath114 ) & 1.8@xmath115 + + @xmath32 ( @xmath54 @xmath29 ) & 3.7(fixed ) + log @xmath31 & [email protected] + redshift ( @xmath114 ) & [email protected] + + 10@xmath60 ph / s/@xmath61 at 1 kev & [email protected] + + 10@xmath60 ph / s/@xmath61 at 1 kev & @xmath1160.2 + + @xmath32 ( @xmath35 @xmath29 ) & 1.6(fixed ) + log @xmath31 & 1.57(fixed ) + redshift ( @xmath114 ) & 1(fixed ) + + photon index & [email protected] + @xmath65 ( kev ) & 160(fixed ) + + & @xmath660.3 + + cosincl & 0.866(fixed ) + + line e ( kev ) & 6.34 ( fixed ) + sigma ( kev ) & 0.018 ( fixed ) + norm ( @xmath67ph / s/@xmath61 ) & [email protected] + ew ( ev ) & 25@xmath588 + edge e ( kev ) & 7.11 ( fixed ) + maxtau & 0.13@xmath117 + line e ( kev ) & 6.98@xmath118 + sigma ( kev ) & 0.01(fixed ) + norm ( @xmath68 ph / s/@xmath61 ) & @xmath119 + reduced chi - square ( d.o.f ) & 0.92 ( 5099 ) + + [ miyakawa_model_ave_chandra ] lccc & bright & faint + + line e ( kev ) & + sigma ( 10@xmath120 kev ) & + norm ( 10@xmath121ph / s/@xmath61 ) & @xmath1220.3 & @xmath1230.2 + ew ( ev ) & @xmath1240.5 & @xmath1250.4 + + line e ( kev ) & + sigma ( 10@xmath120 kev ) & + norm ( 10@xmath121ph / s/@xmath61 ) & @xmath1230.2 & @xmath1260.2 + ew ( ev ) & @xmath1270.3 & @xmath1230.5 + photon index & + norm ( 10@xmath120ph / s/@xmath61 at 1 kev ) & [email protected] & [email protected] + reduced chi - square & + & bright & faint + + line e ( kev ) & + sigma ( 10@xmath128kev ) & + norm ( 10@xmath121 ph / s/@xmath61 ) & @xmath1230.2 & [email protected] + ew ( ev ) & @xmath1290.5 & @xmath1300.7 + + line e ( kev ) & + sigma ( 10@xmath128kev ) & + norm ( 10@xmath121 ph / s/@xmath61 ) & @xmath1310.4 & @xmath1320.4 + ew ( ev ) & @xmath1331.0 & @xmath1341.4 + photon index & + norm ( 10@xmath120 ph / s/@xmath61 at 1 kev ) & [email protected] & [email protected] + reduced chi - square & + [ table_chandra][table_chandra2 ] ccc & bright & faint + + @xmath32 ( @xmath54 @xmath29 ) & + + @xmath32 ( @xmath56 @xmath29 ) & + log @xmath31 & [email protected] & @xmath135 + redshift ( @xmath114 ) & + + @xmath32 ( @xmath54 @xmath29 ) & + log @xmath31 & [email protected] & [email protected] + redshift ( @xmath114 ) & + + 10@xmath60 ph / s/@xmath61 at 1 kev & [email protected]&@xmath136 + + 10@xmath60 ph / s/@xmath61 at 1 kev & + + @xmath32 ( @xmath35 @xmath29 ) & + log @xmath31 & + redshift ( @xmath114 ) & + + photon index & + @xmath65 ( kev ) & + + & + + cosincl & + + line e ( kev ) & + sigma ( kev ) & + norm ( @xmath68ph / s/@xmath61 ) & + ew ( ev ) & 14@xmath588 & @xmath137 + edge e ( kev ) & + maxtau & + line e ( kev ) & + sigma ( kev ) & + norm ( @xmath68 ph / s/@xmath61 ) & + reduced chi - square ( d.o.f ) & + + [ miyakawa_model_slice_chandra ] ccc & bright & faint + + @xmath32 ( @xmath54 @xmath29 ) & + + @xmath32 ( @xmath56 @xmath29 ) & + log @xmath31 & [email protected] & [email protected] + redshift ( @xmath114 ) & + + @xmath32 ( @xmath54 @xmath29 ) & [email protected] & [email protected] + log @xmath31 & + redshift ( @xmath114 ) & + + 10@xmath60 ph / s/@xmath61 at 1 kev & [email protected]&@xmath138 + + 10@xmath60 ph / s/@xmath61 at 1 kev & + + @xmath32 ( @xmath35 @xmath29 ) & + log @xmath31 & + redshift ( @xmath114 ) & + + photon index & + @xmath65 ( kev ) & + + & + + cosincl & + + line e ( kev ) & + sigma ( kev ) & + norm ( @xmath68ph / s/@xmath61 ) & + ew ( ev ) & 17@xmath588 & @xmath139 + edge e ( kev ) & + maxtau & + line e ( kev ) & + sigma ( kev ) & + norm ( @xmath68 ph / s/@xmath61 ) & + reduced chi - square ( d.o.f ) & + + [ miyakawa_model_slice_chandra_columnb ] we further simultaneously fit the bright / faint spectra in the 1.07.5 kev band with the three - component model ( table @xmath140 and [ miyakawa_model_slice_chandra_columnb ] ) by varying normalization of the direct component and ionization degree of the low - ionized warm absorber . in stead of allowing the normalization of the absorbed power - law component to be free , we fixed it to the value of a constant total normalization minus the direct power - law normalization , in order to confirm the anti - correlation between the direct component and absorbed component as observed in suzaku ( bottom panels in figures [ relation_for_the_slice_spectra ] and [ relation_for_the_slice_spectra_nh ] ) . in table @xmath140 , we show results of allowing the ionization degrees of the low - ionized warm absorber to be free while the column density is fixed ; we see the parameter is significantly variable ( @xmath104 = [email protected] and [email protected] ) . also , ionization degree of the high - ionized warm absorber is made a free parameter , but these values are hardly variable ( @xmath104 = [email protected] and [email protected] ) . because of the presence of constant high - ionized warm absorber , varying column density of the low - ionized warm absorber effectively changes the average ionization states . therefore , in stead of varying ionization degree of the low - ionized warm absorber , the bright and faint spectral variation can be equally explained by varying its column density . this result is shown in table [ miyakawa_model_slice_chandra_columnb ] . in summary , chandra / hetgs data confirm that the spectral variation is described by change of only two spectral parameters in the framework of the three - component model ( equation [ eq2 ] ) , normalization of the direct component ( @xmath41 ) and either ionization degree or column - density of the low - ionized warm absorber(@xmath23 ) . next , we create a time - sequence of the energy spectra from suzaku data , and apply the three - component model . we have made energy spectra every 20 ksec , and performed spectral fits to all the sequential 20 ksec spectra with the three - component model , to investigate time variation of the spectral parameters . we chose a 20 ksec interval , since variation amplitudes in the 110 kev band is maximum at around 2080 ksec ( inoue , miyakawa and ebisawa 2011 ) . in the spectral fits , @xmath41 , @xmath42 and @xmath23 are treated as time - variable in equation ( [ eq2 ] ) . the fits are made for the three observation sequences separately ( obsid:700007010 , 700007020 and 700007030 ) . we have seen that observed variation of @xmath23 is equally represented with change of the column density @xmath102 or ionization parameter , @xmath31 ( section [ sec : sliced ] ) . hereafter , we vary @xmath102 while @xmath31 is assumed to be invariable throughout the observation . in fact , as we shall see below , variation of @xmath102 is more physically reasonable than that of @xmath31 . figure [ direct_absorbed_20ksec ] shows a correlation between @xmath41 ( direct power - law normalization ) and @xmath42 ( absorbed power - law normalization ) , where an inverse correlation is seen between them , as well as in the bottom panels in figure [ relation_for_the_slice_spectra ] and [ relation_for_the_slice_spectra_nh ] . the correlation coefficient between @xmath41 and @xmath42 is @xmath141 . if we fit with a linear relation , @xmath142 , we obtain @xmath143 , @xmath144 and @xmath145 ( d.o.f.=35 ) . on the other hand , fits with a constant @xmath41 ( @xmath146 with @xmath147 for d.o.f.=36 ) or @xmath42 ( @xmath148 for d.o.f.=36 ) are not acceptable . we found that slope of the anti - correlation between @xmath41 and @xmath42 is not very far from @xmath14 , which suggests that sum of the two parameters may be rather constant . therefore , we introduce the `` total normalization '' , @xmath149 , as , @xmath150 similarly , we introduce the `` covering fraction '' , @xmath151 , as @xmath152 ( 90mm,90mm)figure7.eps ( 90mm,90mm)figure8a.eps ( 90mm,90mm)figure8b.eps ( 90mm,90mm)figure8c.eps ( 90mm,90mm)figure9a.eps ( 90mm,90mm)figure9b.eps ( 90mm,90mm)figure9c.eps obviously , @xmath153 thus , instead of @xmath154 , we can equivalently use @xmath155 to describe the observed spectral variation . in figure [ correlation_20ksec ] , we show correlations among these three parameters . cccccc & & & + @xmath156 & @xmath157 & cc@xmath158 & @xmath159 & @xmath160 & @xmath161 + @xmath149 & @xmath151 & @xmath162 & @xmath163 & @xmath164 & @xmath165 + @xmath149 & @xmath101 & @xmath166 & @xmath163 & n.a . & @xmath167 + @xmath151 & @xmath102 & @xmath168 & @xmath169 & n.a . & @xmath170 + + [ tab : correlation ] we study variations of and correlations among the three parameters , @xmath155 , by calculating the correlation coefficients between the two , as well as fitting the correlation with a constant parameter or a linear - function . results are summarized in table [ tab : correlation ] . consequently , we can see a good `` orthogonality '' between @xmath149 and @xmath151 , and @xmath149 and @xmath102 ( i.e. , the fitting parameter @xmath171 is almost zero , where @xmath172 or @xmath173 ) . namely , the total power - law normalization @xmath149 is not significantly variable , while the covering fraction , @xmath151 , and column - density of the low - ionized absorber , @xmath101 , are more variable . furthermore , @xmath151 and @xmath102 indicate a good correlation with the correlation - coefficient 0.64 . therefore , spectral variation of mcg-6 - 30 - 15 at a timescale of 20 ksec is primarily described by a single parameter , either @xmath151 or @xmath101 . we remind , however , there is a hint of weak anti - correlation between @xmath149 and @xmath151 . further sophistication of the model may explain this correlation ( section [ sec : blocker ] ) . figure [ 20ksec_par_var ] indicates another representation of the spectral parameter variations . as a function of time , the three parameters , @xmath174 , respectively normalized by their average values , are indicated . the @xmath175 values around the average ( 1.0 ) are 2.4 , 2.9 and 10.3 for @xmath176 respectively ( d.o.f.=36 ) . we can see @xmath149 is the least variable , while @xmath151 and @xmath101 are in sync and more significantly variable . in the following , we will discuss physical meaning of these spectral variations . [ phenomenological_inte_section ] we have seen that suzaku and chandra energy spectra of mcg6 - 30 - 15 are successfully described by the three - component model introduced with equation ( [ eq2 ] ) . we found @xmath42 , normalization of the heavily absorbed component , is anti - correlated with normalization of the direct component , @xmath41 ( bottom panels of figures [ relation_for_the_slice_spectra ] and [ relation_for_the_slice_spectra_nh ] , and figure [ direct_absorbed_20ksec ] ) . the spectral variation indicates that the `` total normalization '' , @xmath149 , ( equation [ eq3 ] ) is not so variable , while the covering fraction , @xmath151 , ( equation [ eq4 ] ) and column - density of the low - ionized warm absorber are in sync and more significantly variable ( figure [ correlation_20ksec ] and [ 20ksec_par_var ] ) . how should we interpret this characteristic spectral variation ? since @xmath177 and @xmath178 represent amounts of the non - absorbed and absorbed fluxes emitted from the central x - ray source , respectively , the anti - correlation is naturally understood that a rather constant x - ray source is partially obscured by absorbing matters with variable covering fraction , @xmath151 . the `` difference variation function analysis '' introduced by inoue , miyakawa and ebisawa ( 2011 ) also confirms that the ( @xmath149 , @xmath151 ) parameter - set is more orthogonal and fundamental than the ( @xmath41 , @xmath42 ) parameter - set . correlation between the low - ionized absorber @xmath101 and @xmath151 ( bottom panel of figure [ correlation_20ksec ] ) is understood as follows : the low - ionized absorber represented by @xmath179 is optically thin ( table [ miyakawa_model_ave_suzaku ] ) , so that @xmath180 on the other hand , the same absorption effect can be expressed by a partial covering model with a fixed column - density as , @xmath181 @xmath182 namely , @xmath183 that is seen in the bottom panel of figure [ correlation_20ksec ] . as we have seen , the observed spectral variation of mcg-6 - 30 - 15 is primarily explained by variation of the partial covering fraction , @xmath151 . below , we designate our model as the `` variable partial covering model '' . in the framework of the variable partial covering model , the suzaku intensity - sliced spectra in 1 40 kev is explained mostly by change of the covering fraction . using the total normalization @xmath149 and the covering fraction @xmath151 , our three - component model ( [ eq2 ] ) is rewritten as follows : @xmath184 in equations ( [ eq9 ] ) and ( [ eq6 ] ) , only variable parameters are @xmath151 and @xmath149 . in figure [ covering_fraction_change ] and table [ table1_slice ] , we show an example of the simultaneous fit of the suzaku intensity - sliced spectra with variable @xmath151 ( the same dataset as used in section [ sec : sliced ] ) . the eight intensity - sliced spectra are successfully fitted with the constant normalization , @xmath185 photons s@xmath33 @xmath2 at 1 kev ( except the brightest one where @xmath149 is 1.5 times higher ) , while @xmath151 is variable from 0.63 ( dimmest ) to null ( brightest ) . the well - known small variability in the iron energy band ( section 1 ) is also nicely explained in the framework of the variable partial covering model . in order to see energy dependence of the variation at a specific timescale , we took the following method to calculate the `` difference variation function '' , @xmath186 ( inoue , miyakawa and ebisawa 2011 ) : ( 1 ) we choose a time - interval of interest , @xmath187 , to investigate for spectral variations ; here , we take @xmath18840 ksec . ( 2 ) energy spectra are calculated for every @xmath189 , namely , 20 ksec . ( 3 ) for every two contiguous energy spectra , we recognize the `` brighter spectrum '' and the `` fainter spectrum '' , @xmath190 and @xmath191 , respectively . ( 4 ) we average all the brighter spectra and all the fainter spectra . ( 5 ) thus , for a given time - period @xmath187 , we create a single `` averaged bright spectrum '' @xmath192 and a single `` averaged faint spectrum '' @xmath193 . ( 6 ) we calculate difference of the averaged bright spectrum and the averaged faint spectrum relative to the average , as @xmath194 in figure [ cov_fac_simulation ] , we show the difference variation function thus calculated at a timescale of 40 ksec . together , we show a model prediction of the difference variation function calculated in the framework of the variable partial covering model ( eq . 29 in inoue , miyakawa and ebisawa 2011 ) , where the covering fraction @xmath151 is assumed to be the major variable parameter , while the total normalization @xmath149 adds minor fluctuation . agreement of the observation and model calculation is obvious . next , we estimate physical parameters of the warm absorbers in equation ( [ eq6 ] ) , @xmath22 , @xmath23 and @xmath43 , which are high - ionized warm absorber , thin low - ionized warm absorber and thick partial absorber , respectively . we provide suffixes , @xmath195 and @xmath196 to distinguish column densities and ionization parameters of these warm absorbers . in the following , @xmath197 is distance from the central x - ray source to the absorber , @xmath198 is a representative thickness of the absorbing region along the line of sight , and @xmath199 is density . ( 90mm,90mm)figure10.eps ( 90mm,90mm)figure11.eps ccccccccc state & slice 1 & slice 2 & slice 3 & slice 4 & slice 5 & slice 6 & slice 7 & slice 8 + + @xmath32 ( @xmath54 @xmath29 ) & + + @xmath32 ( @xmath56 @xmath29 ) & + log @xmath31 & + + @xmath151 & 0.632 & 0.356 & 0.269 & 0.195 & 0.181 & 0.166 & 0.122 & @xmath200 + + @xmath201 ( @xmath202 ) & + @xmath32 & + log @xmath31 & + + 10@xmath60ph / s/@xmath61 at 1 kev & & 2.71 + + & + + & + + @xmath32 ( @xmath35 @xmath29 ) & + log @xmath31 & + + photon index & + @xmath65 ( kev ) & + + pexrav k ( 10@xmath120 ) & + + cosincl & + + line e ( kev ) & + sigma ( kev ) & + norm ( @xmath67ph / s/@xmath61 ) & + edge e ( kev ) & + maxtau & + line e ( kev ) & + sigma ( kev ) & + norm ( @xmath68ph / s/@xmath61 ) & + line e ( kev ) & + sigma ( kev ) & + norm ( @xmath67ph / s/@xmath61 ) & + reduced chisq ( d.o.f ) & + + [ table1_slice ] first , in general , we note the following relations hold : @xmath203 @xmath204 for the luminosity , we take the representative value @xmath205 erg / s , which is valid for order estimation , since the flux variation is much less than an order of magnitude . from model fitting ( table [ table1_slice ] ) , @xmath206 @xmath207 10@xmath208 erg cm / s , @xmath209 @xmath207 10@xmath210 @xmath2 . hence , @xmath211 our analysis has indicated that the high - ionized warm absorber is constant , while partial covering fraction is variable . this suggests the variable partial absorbers be embedded in the largely extended , static high - ionized warm absorber . hence , we assume @xmath212 @xmath8 @xmath213 , and take @xmath214 10@xmath215 cm and @xmath216 10@xmath217 cm@xmath120 for typical parameters of the high - ionized warm absorbers . from model fitting ( table [ table1_slice ] ) , @xmath219 10@xmath220 erg cm / s , @xmath221 10@xmath222 @xmath2 . hence , @xmath223 here , we may further constrain the parameters of the partial absorber from variation timescale of the partially absorbed component . in our variable partial covering model , typical variation timescale ( @xmath224 sec ; figure [ 20ksec_par_var ] ) corresponds to the crossing time of the x - ray absorbing clouds in front of the x - ray source . we will call these clouds as the `` low - ionized clouds '' , hereafter . assuming that size of the x - ray source and that of a low - ionized cloud have similar dimensions , the crossing timescale is expressed as @xmath225 , where @xmath226 is the velocity of a low - ionized cloud . hence , @xmath227 where @xmath226 is normalized to a typical velocity of the broad line region ( blr ) clouds . combining @xmath228 and @xmath229 , we estimate parameters of the partial absorbers as follows : @xmath230 @xmath231 note that a low - ionized cloud may have internal ionization structures , which we will see below . in our variable partial covering model , the low - ionized warm absorber ( @xmath23 ) and the partial absorber ( @xmath43 ) are closely associated . presumably , they are different parts of the low - ionized clouds . we point out that presence of such internal structures in the blr clouds is also suggested in mrk 766 ( risaliti et al . 2011 ) . in our model , the low - ionized warm absorber has a fixed column - density @xmath201 with a variable partial covering factor ( equation [ eq9 ] ) ; we simply write this column density as @xmath101 below . from model fitting ( table [ table1_slice ] ) , @xmath232 @xmath2 and @xmath233 10@xmath234 erg cm / s . if we take the distance of the low - ionized clouds @xmath235 , @xmath236 @xmath237 note that values of @xmath238 and @xmath239 are between those of the high - ionized absorber ( @xmath240 cm@xmath120 and @xmath241 ) and the partial absorber ( @xmath242 @xmath2 and @xmath243 ) , respectively . the thickness @xmath244 is much smaller than that of the partial absorber @xmath245 cm . therefore , the low - ionized absorber is considered to be in the boundary layer between the low - ionized cloud and the high - ionized warm absorber . we may call this putative layer as the `` cloud envelope '' . meanwhile , it is known that there takes place thermal instability in the range of @xmath246 erg cm / s ( reynolds @xmath21 fabian 1995 ) . therefore , it is reasonable that the cloud envelope continuously connects the partial absorber in the low - ionized cloud and the surrounding constant high - ionized warm absorber . we notice the thomson optical depth ( @xmath247 ) of the partial absorber is about unity . presumably , this is not just a coincidence , but suggests that the low - ionized cloud has internal density gradients extending up to more than unit thomson optical depth , such that attenuation with continuously different optical depths is approximated by unit thomson optical depth . thus , the low - ionized cloud may well have central , cold thomson thick cores ( @xmath248 @xmath2 ) , which are completely opaque to incoming x - rays . we may not see these `` x - ray blockers '' directly , but if we take into account the effect of the x - ray blockers , the weak anti - correlation between the total normalization @xmath149 and the covering fraction @xmath151 ( top - panels of figure [ correlation_20ksec ] ) might be explained . [ cols="<,^,^,^,^,^,^,^",options="header " , ] finally , we present a physical picture surrounding the x - ray source to explain the observations . in table [ tab : absorbers ] , we summarize parameters of the absorbers in our model . basic points of our variable partial covering model are the following : 1 . observed x - ray flux and spectral variability is primarily caused by random passages of the low - ionized clouds in front of the central x - ray source . a typical size of each low - ionized cloud is @xmath245 cm and a typical velocity is @xmath249 cm / s ( equation [ sizeofcloud ] ) . central x - ray source is not always fully blocked , nor fully exposed . the covering fraction is variable between null to @xmath250 ( table [ table1_slice ] ) , and it fluctuates by an amplitude of @xmath251 % ( figure [ 20ksec_par_var ] ) . we may estimate the average covering fraction @xmath252 from intensity - sliced spectra as @xmath252 = @xmath253 , where @xmath254 and @xmath255 are exposure time for each spectrum and the total observation , respectively . we thereby obtain @xmath256 0.3 . the integrated covering fraction is considered to be about 0.1 for blr ( e.g. , blandford et al . thus , our results suggest that the partial absorption takes place in the blr , and the low - ionized clouds corresponds to the blr blobs . we may estimate number of the absorbing clouds in the field of view , @xmath257 , and size of the central x - ray source , @xmath258 , as follows . assuming that fluctuation of the covering fraction , @xmath259 , is determined by statistical fluctuation of the number of clouds , @xmath260 , @xmath257 is estimated as @xmath261 . the average covering fraction , @xmath262 , may be written as @xmath263 , thus , @xmath264 cm . typically , blr blobs have a velocity of @xmath226 @xmath8 10@xmath265 cm / sec . assuming the kepler velocity has a similar value , @xmath266 500 @xmath267 , where @xmath267 is the schwarzschild radius . identifying this with @xmath268 10@xmath269 cm ( equation [ eq : r2 ] ) , the central black hole mass of mcg-6 - 30 - 15 is @xmath270 . thus , the central x - ray source ( @xmath271 cm ) is extended to @xmath272 around the black hole . besides variation of the blr coluds , the x - ray source is variable on a timescale down to @xmath8 1 ksec ( inoue , miyakawa and ebisawa 2011 ) , which may be compared with the free fall time , @xmath273 . then , the intrinsic source variation arises at a few times the schwarzshild radius , around the innermost stable circular orbit . we point out that solid angle of the low - ionized clouds seen from the x - ray source should be much smaller than 4@xmath27 , otherwise we will have to see strong fluorescent lines from the clouds . if a thick absorber with @xmath274 @xmath2 covers the power - law component completely ( @xmath275 ) , equivalent width of the iron line to the reflection component is @xmath8 1 kev . considering @xmath276 0.3 and the ratio of the reflection component to the direct component , 10 to 1 , the expected iron line equivalent width is @xmath8 30 ev . as mentioned in subsection 5.5.2 , we found the fit improves if a mildly broad disk - line is added , where the best - fit inner radius and the equivalent width are @xmath277 200 @xmath90 and @xmath278 ev , respectively ( table [ miyakawa_diskline_model_ave_suzaku ] ) . these mildly broadened , weak iron emission lines may well be expected from fluorescence in the blr clouds . figure @xmath279 gives a schematic view of internal structure of the low - ionized cloud and the variable partial covering model surrounding the black hole . the low - ionized cloud ( blr cloud ) has internal ionization structure . a putative blocker in the core would block the incoming x - rays completely . the main body of the low - ionized cloud is the thick partial absorber surrounding the core . still the outer cloud envelope corresponds to the thin low - ionized warm absorber . central x - ray emission region around the black hole is mildly extended , and partially covered by these low - ionized clouds . variation of the covering fraction explains most of the observed flux and spectral variations . finally we comment that the idea of partial covering is not new for mcg-6 - 30 - 15 to explain its spectral shape and varaition ( e.g. , matsuoka et al . 1990 ; mckernan and yaqoob 1998 ; miller , turner and reeves 2008 , 2009 ) . recently , similar partial covering models are proposed to explain spectral variations of ngc 3516 ( turner et al . 2008 ) , mrk 766 ( risaliti et al . 2011 ) and other seyfert galaxies ( turner & miller 2009 ) . presumably , partial covering of rather constant x - ray source is a common mechanism in seyfert galaxies . it is of interest to study spectral variations of seyfert galaxies systematically to see if the observed x - ray variablity is truely due to intrinsic luminosity variation or explained by partial covering of the constant source . we have analyzed suzaku and chandra archival data of mcg-6 - 30 - 15 and constructed a new model to explain the observed spectral variation . our main conclusions are summarized below : 1 . we have shown that the observed energy spectra and spectral variation can be explained by the `` three - component model '' ( equation [ eq2 ] ) that includes ( 1 ) a direct power - law component , ( 2 ) a heavily absorbed power - law component by thich photoionized material , and ( 3 ) a cold disk reflection component far from the black hole . the first two components are affected by two warm abosrbers having different ionization states . the ionized iron k - edge of the heavily absorbed component and spectral curvature due to the warm absorbers explain most of the seemingly broad `` disk line '' spectral feature . consequently , general relativistic interpretation of the broad iron emission line feature in mcg-6 - 30 - 15 is not confirmed . the claim that mcg-6 - 30 - 15 is a kerr black hole with nearly extreme rotation assuming the relativistic line - broadening ( e.g. , miniutti et al . 2007 ) is questionable . 3 . we propose the `` variable partial covering model '' , in which a central , moderately extended x - ray source is partially covered by variable low - ionized clouds in the line of sight . in this model , observed x - ray spectral variation is primarily caused by change of the partial covering fraction of the central source . these absorbing clouds presumably correspond to the fast - moving broad line region ( blr ) clouds . the authors would like to thank profs . p. ycki , y. terashima , y. ueda and k. makishima for fruitful discussion . this research has made use of public suzaku data obtained through the data archives and transmission system ( darts ) , provided by the institute of space and astronautical science ( isas ) at japan aerospace exploration agency ( jaxa ) , public chandra data obtained through the high energy astrophysics science archive research center ( heasarc ) at nasa / goddard space flight center . for data reduction , we used the software provided by the high energy astrophysics science archive research center ( heasarc ) at nasa / goddard space flight center and chandara x - ray center ( cxc ) at harvard - smithsoniann center for astrophysics . k.e . acknowledges ms . akiko sato for helping us illustrate our model and producing a beautiful artwork . |
the superposition of the quantum rotational motion ( tunneling ) of the encapsulated scc complex with the classical rotational motion of the surrounding c molecule in a powder crystal of scc@c fullerite is investigated by theory . since the quantum rotor is dragged along by the c molecule , any detection method which couples to the quantum rotor ( in casu the c bond of the scc complex ) also probes the thermally excited classical motion ( uniaxial rotational diffusion and stochastic meroaxial jumps ) of the surrounding fullerene . the dynamic rotation - rotation response functions in frequency space are obtained as convolutions of quantum and classical dynamic correlation functions . the corresponding raman scattering laws are derived , the overall shape of the spectra and the width of the resonance lines are studied as functions of temperature . the results of the theory are confronted with experimental low - frequency raman spectra on powder crystals of scc@c [ m. krause et al . , phys . rev . lett . * 93 * , 137403 ( 2004 ) ] . the agreement of theory with experiment is very satisfactory in a broad temperature range . | cond-mat0702598 | the existence of an endohedral fullerene , i.e. one or several atoms encapsulated in a fullerene molecule , was originally inferred from an analysis of mass spectra of lacl@xmath2-impregnated graphite and did lead to the proposal la@c@xmath3 @xcite . at present the study of endohedral metallofullerenes m@xmath4@c@xmath5 , @xmath6 and @xmath7 , where m are group ii and iii metals such as sc , y , @xmath8 or lanthanides ce , @xmath8 , lu , is a subject of interdisciplinary research @xcite in physics , chemistry and materials sciences . by now one is able to produce materials where not only single atoms but clusters of atoms are encapsulated @xcite . due to charge transfer between the cluster and the surrounding carbon cage it is possible to obtain molecular - like complexes which do not exist otherwise ( i.e. in absence of encapsulation ) and which have unusual properties . not only clusters with metal atoms of a same kind , such as the sc@xmath2 trimer in sc@xmath2@c@xmath9 are produced , but also clusters composed of different kinds of atoms . a remarkable case is the production of sc@xmath0c@xmath0@c@xmath1 in crystalline powder form @xcite . the powder crystal is composed of crystallites where the sc@xmath0c@xmath0@c@xmath1 units are arranged with average space group symmetry @xmath10 . from spectroscopic and structural characterization by nmr- and synchrotron x - ray diffraction experiments @xcite it follows that the sc@xmath0c@xmath0 complex is encaged as a rigid unit with point group symmetry @xmath11 in a c@xmath1 fullerene of symmetry d@xmath12 ( isomer iii , number 23 @xcite ) . the center of mass of sc@xmath0c@xmath0 coincides with the center of mass of the molecule . the two sc atoms are located at a distance of @xmath13 on the long @xmath14 ( @xmath15 ) axis of c@xmath1 . the two c atoms of sc@xmath0c@xmath0 are located in the plane containing the two @xmath14 axes perpendicular to the long axis of the c@xmath1 molecule and have a calculated distance of @xmath16 . this distance lies between those of typical double and triple carbon bonds , and is consistent with the experimental and calculated c c stretching frequency of the c@xmath0 unit ( exp . @xmath17 @xmath18 , calc . @xmath19 @xmath18 ) @xcite . in the following we will speak of this c c bond as a c@xmath0 unit or molecule . indeed low energy raman spectra @xcite on powder samples of sc@xmath0c@xmath0@c@xmath1 in a temperature range @xmath20 @xmath21 k ( kelvin ) have revealed the existence of quantized rotational states of the c@xmath0 unit . the raman lines positions reflect transitions between energy levels of a c@xmath0 planar quantum rotor in a fourfold static potential due to the surrounding c@xmath1 cage . therefore , one can speak of a quantum gyroscope . in ref . the potential parameters of the encaged quantum rotor were obtained from density functional calculations using the vasp ( vienna ab initio simulation package ) code @xcite . the energy levels were then determined from the solution of the schrdinger equation . within this approach , the raman spectra consist of infinitely sharp lines while experimentally the lines are broadened and have a characteristic temperature behavior . a reason for this shortcoming is the restriction of the role of the encapsulating c@xmath1 molecule to a purely static body . since the measured transition frequencies are in the range of @xmath22 @xmath23 @xmath18 and since the line broadening is of the order of a few @xmath18 , any involvement of internal vibrational modes of the c@xmath1 cage as well as of stretching modes sc c@xmath1 can be excluded . the latter are of higher frequencies and have been measured in c@xmath1 and in sc@xmath0@c@xmath1 by infrared and raman techniques @xcite . however the low - frequency external rotational modes of the c@xmath1 molecule and their superposition with the transitions of the quantum rotor should be retained : indeed the encapsulated sc@xmath0c@xmath0 gyroscope is dragged by the classical rotational motion of the c@xmath1 molecule and this dragging will affect the raman spectrum of the c@xmath0 unit . it follows that an experimental probe such as raman scattering which couples to the encapsulated species in an endohedral complex , in casu sc@xmath0c@xmath0 , also yields information on the dynamics of the encapsulating molecule , in casu c@xmath1 . in the present paper we will extend the theoretical interpretation given in ref . and develop a unified theory where the quantum mechanical motion of the sc@xmath0c@xmath0 complex is coupled to the thermally excited classical rotational motion of the c@xmath1 fullerene . the coupling results from the fact that the long axis of the quantum gyroscope coincides with the @xmath15 axis of the surrounding c@xmath1 molecule . in a given crystallite the c@xmath1 molecules are randomly oriented with their long @xmath24 axis in equivalent @xmath25 directions of the face - centered cubic ( fcc ) unit cell @xcite . we call this situation meroaxial disorder ( this terminology seems to be more appropriate than merohedral disorder ) . we start from a model where at low temperature the meroaxially oriented c@xmath1 molecules in the fcc crystal perform uniaxial rotational diffusions about their long axis . such a classical motion can be seen as a time - dependent modulation of the fourfold potential experienced by the quantum rotor and causes a temperature - dependent broadening of the quantum levels . an additional broadening effect is to be expected from the stochastic reorientations of the c@xmath1 molecules among the meroaxial directions which should become increasingly important at higher temperature . in addition , the stochastic reorientations lead to the appearance of a temperature - dependent quasi - elastic peak in the raman spectrum . the content of the paper is as follows . in section ii we write down the raman scattering law for the c@xmath0-unit belonging to the sc@xmath0c@xmath0 complex encapsulated by the c@xmath1 molecule in sc@xmath0c@xmath0@c@xmath1 fullerite . we start from a single crystal with @xmath10 structure and static meroaxial disorder of the c@xmath1 molecules . assuming a quantum mechanical rotational motion of the sc@xmath0c@xmath0 complexes and a classical rotational motion of the c@xmath1 molecules , the dynamic polarizability - polarizability correlation function is decoupled in a product of correlation functions for the rotational dynamics of sc@xmath0c@xmath0 and c@xmath1 respectively . the scattering law is obtained as a convolution of these correlation functions in fourier space . next ( sect . iii ) we calculate the correlation functions , using a quantum mechanical tunneling model for the sc@xmath0c@xmath0 complex and a uniaxial rotational diffusion model for the surrounding c@xmath1 molecule . the rotational diffusion motion of the encapsulating c@xmath1 molecule leads to a linear temperature - dependent broadening of the energy transition lines of the c@xmath0 planar rotor . in sect . iv we extend the theory to a powder crystal which consists of arbitrarily oriented crystallites , each with @xmath10 space group symmetry and static meroaxial disorder . in the following ( sect . v ) we consider the case of dynamic meroaxial disorder , describing the reorientations of c@xmath1 molecules among the three meroaxial directions by a stochastic jump model . this model yields an exponential temperature - dependent broadening of the transition lines . in the last section vi we give a numerical evaluation of the raman scattering law for a sc@xmath0c@xmath0@c@xmath1 powder crystal where quantum mechanical tunneling of the encapsulated sc@xmath0c@xmath0 units is superimposed by uniaxial rotational diffusion and dynamic meroaxial disorder of the c@xmath1 molecules . the temperature dependence of the line intensities and of the line broadenings is discussed . we will derive the raman scattering law where we limit ourselves to the interaction of the incident laser light with the plane rotational motion of the induced dipole of the c c bond belonging to the sc@xmath0c@xmath0 complex of sc@xmath0c@xmath0@c@xmath1 . this means that we consider the low frequency part of the spectrum ( say @xmath26 @xmath18 ) . the sc@xmath0c@xmath0 complex is centered in the origin ( center - of - mass position ) of c@xmath1 . the long axis of sc@xmath0c@xmath0 coincides with the @xmath15 axis of c@xmath1 . the c c bond of sc@xmath0c@xmath0 lies in the plane containing the secondary @xmath24 axes of c@xmath1 and rotates about the @xmath15 axis . in that respect we will consider the c c bond as a c@xmath0 planar rotor which experiences a fourfold potential inside the c@xmath1 molecule . our formulation of the raman scattering law is an extension of the conventional theory @xcite in as much as we describe a situation where the rotational motion of the induced dipole with respect to the laboratory - fixed frame is a superposition of the quantum motion of the c@xmath0 planar rotor inside the c@xmath1 molecule and of the classical motion of the c@xmath1 molecule in the laboratory frame . we start with considering a single crystal of sc@xmath0c@xmath0@c@xmath1 units with static meroaxial disorder . we assume that the sc@xmath0c@xmath0@c@xmath1 units are statistically independent , hence it will be sufficient to consider one single representative unit . the cubic crystal axes @xmath27 are chosen to coincide with the laboratory - fixed cubic coordinate system @xmath28 . we consider a cubic system of axes @xmath29 fixed in the c@xmath1 molecule such that the @xmath30 axis coincides with the @xmath15 axis while @xmath31 and @xmath32 coincide with the secondary twofold axes ( fig . [ figc84 ] ) . the meroaxial orientations of the c@xmath1 molecules correspond to the situation where the @xmath30 axes are randomly oriented along the @xmath33 , @xmath34 or @xmath35 crystal axes . the c@xmath0 units then rotate in the planes @xmath36 , @xmath37 or @xmath38 respectively ( fig . [ figaxes3d ] ) . we say that the c@xmath1 molecule is in standard orientation if the @xmath15 axis coincides with the laboratory - fixed @xmath39 axis and the plane containing the secondary @xmath24 axes coincides with the laboratory @xmath40 plane . the @xmath32 axis forms an angle @xmath41 with the @xmath42 axis , while the c c bond forms an angle @xmath43 with the @xmath32 axis . hence the polar angle @xmath44 of the c c bond with the laboratory @xmath42 axis ( fig . [ figaxes2d ] ) is a sum of two terms : @xmath45 since the c@xmath0 rotor is confined to the @xmath40 plane , the azimuthal angle @xmath46 measured away from @xmath39 has value @xmath47 . the distinction of two contributions to the angle @xmath44 is essential . in the following we will assign the angular variable @xmath43 to the quantum mechanical tunneling of the sc@xmath0c@xmath0 complex about its long axis inside the c@xmath1 cage and the angular variable @xmath41 to the thermally excited classical rotation of the c@xmath1 molecule about the @xmath15 axis . the assumption of classical uniaxial rotational diffusion motion as a first approximation to the dynamics of the c@xmath1 molecule at low temperature is motivated by the structural results of meroaxial disorder @xcite . it is also inspired from the dynamics of solid c@xmath48 in the rhombohedral and monoclinic phases . there the importance of uniaxial rotational diffusion about the long axis of the c@xmath48 molecule has been probed by muon spin spectroscopy @xcite , nuclear magnetic resonance @xcite and inelastic neutron scattering @xcite . we treat the c c bond of sc@xmath0c@xmath0 as a rigid cylindrical rod with longitudinal and transverse static polarizability @xmath49 and @xmath50 respectively . the raman scattering law for incident and scatterd radiation in @xmath42 direction is given by the fourier transform of the time - dependent autocorrelation function of the polarizability @xmath51 : @xmath52 here @xmath53 is the number of sc@xmath0c@xmath0@c@xmath1 units and @xmath54 is the frequency difference of incident and scattered radiation . the polarizability has to be understood as an average over the three meroaxial molecular orientations . in the following we label these orientations by a superscript @xmath55 , @xmath56 . if the c@xmath1 molecule is in standard orientation ( @xmath30 axis @xmath57 ) , or in orientation @xmath58 , the corresponding orientation - dependent polarizabilities @xmath59 and @xmath60 are equal and given by @xcite @xmath61 while with @xmath62 one has @xmath63 independent of @xmath44 . in the case of meroaxial disorder , the average polarizability is given by @xmath64 where we have defined 2 a = , & & b = . hence the time - dependent correlation function reads @xmath65 similarly we obtain for incident radiation in @xmath42 direction and scattered radiation in @xmath66 direction @xmath67 if the c@xmath1 molecule is in standard orientation , @xmath68 while for @xmath58 , @xmath69 , and @xmath70 , @xmath71 . the average polarizability for the case of meroaxial disorder reads @xmath72 and the correlation function becomes @xmath73 the problem of determining the scattering laws @xmath74 and @xmath75 consists in the calculation of the orientation - orientation thermal correlation functions @xmath76 taking into account the basic relation eq . ( [ angles ] ) , we expand in terms of @xmath77 , @xmath78 , @xmath79 and @xmath80 thereby obtaining correlation functions of the form @xmath81 and similarly for @xmath82 and @xmath83 . observing that @xmath43 refers to quantum dynamics of c@xmath0 and @xmath41 to classical dynamics of c@xmath1 , we decouple the thermal averages : 2 c^ = q^(t)f^(t ) , & & c^ = q^(t)f^(t ) , + s^ = q^(t)f^(t ) , & & s^ = q^(t)f^(t ) . here the correlation functions @xmath84 describe the quantum dynamics of the c@xmath0 unit while the correlation functions @xmath85 describe the classical dynamics of the surrounding c@xmath1 molecule . finally quantum and classical dynamics occur as products of correlation functions : @xmath86 defining fourier transforms @xmath87 and using eqs . ( [ ct ] ) , ( [ st ] ) , ( [ avalphaalphazz ] ) and ( [ avalphaalphazy ] ) , we rewrite the raman scattering law in terms of convolutions of fourier - transformed quantum and classical correlation functions , thereby obtaining @xmath88 , \label{rzzzz2}\end{aligned}\ ] ] with scattering function @xmath89 , \label{comega}\end{aligned}\ ] ] and @xmath90 with scattering function @xmath91 . \label{somega}\end{aligned}\ ] ] the first term in brackets on the right - hand side of eq . ( [ rzzzz2 ] ) corresponds to the unshifted rayleigh line of the spectrum while the function @xmath92 ( as also @xmath93 in eq . ( [ rzyzy2 ] ) ) accounts for the inelastic part . expressions ( [ comega ] ) and ( [ somega ] ) which are convolutions in fourier space show that the quantum motion of the c@xmath0 rotor is modulated by the classical rotational motion of the surrounding c@xmath1 cage . this is an example of direct coupling " of two motions through the detection process @xcite , in contradistinction to the indirect coupling " through a hamiltonian . the origin of the direct coupling here is due to the fact that the detection angle @xmath44 is a sum of two terms , eq . ( [ angles ] ) . in the next section we will calculate the quantum mechanical and classical orientational correlation functions for c@xmath0 and c@xmath1 respectively . the quantum mechanics of a diatomic molecular rotor in crystals goes back to pauling @xcite . a still valid review of the subject of single particle rotations in molecular crystals has been given by w. press @xcite . we will calculate the orientational autocorrelation functions @xmath94 and @xmath95 by starting from the model of the c@xmath0 planar quantum rotor in the fourfold potential due to the c@xmath1 cage . we will refer to this motion as rotational tunneling @xcite . we will show that the resonances of the correlation functions @xmath96 and @xmath97 are due to transitions between tunneling energy levels . the sole degree of freedom is the angle @xmath43 which accounts for the rotatory motion of c@xmath0 with respect to the cage . the corresponding schrdinger equation reads @xmath98\psi(\tau ) = e\psi(\tau ) . \label{schroedinger}\end{aligned}\ ] ] here @xmath99 is the rotational constant and @xmath100 the moment of inertia of c@xmath0 , @xmath101 is the barrier height of the potential . the rotational constant has the dimension of an energy , from experiment @xcite one deduces @xmath102 @xmath18 ( wave number units ) and @xmath103 . these values are supported by ab initio density functional calculations @xcite . equation ( [ schroedinger ] ) which is an extension of mathieu s equation @xcite is also called hill s equation @xcite . with the definitions 2 = ( e- ) , & & q = , eq . ( [ schroedinger ] ) reads @xmath104\psi(\tau ) = 0 . \label{schroedinger2}\end{aligned}\ ] ] from symmetry considerations ( nuclear spin is zero for @xmath105c , electron wave function of c@xmath106 is totally symmetric ) , it follows that the rotational wave function @xmath107 must be symmetric with respect to the operation @xmath108 . for even periodic solutions one makes the ansatz @xmath109 @xmath110 . equation ( [ schroedinger2 ] ) then leads to an infinite system of homogeneous equations for the coefficients @xmath111 . truncation of this system for a given value @xmath112 leads to @xmath113 equations which separate into two systems : a first one for @xmath114 and a second one for @xmath115 ( we take @xmath53 even ) . solving for the two discriminants yields the roots @xmath116 for @xmath117 and @xmath118 . in case of zero potential , i.e. @xmath119 , these solutions reduce to the free planar rotor energies @xmath120 with normalized eigenfunctions @xmath121 in the interval @xmath122 . the ansatz for odd periodic solutions reads @xmath123 proceeding as before one determines the roots @xmath124 . in case of zero potential the eigenfunctions are @xmath125 in the following we will label the energy eigenfunctions and eigenvalues by the double index @xmath126 , @xmath127 , also in the case of nonzero potential . in fig . [ tunneling ] we show plots of @xmath128 . we next perform a spectral decomposition of the correlation functions @xmath129 and @xmath130 in terms of eigenfunctions and eigenvalues of the schrdinger equation ( [ schroedinger ] ) . in general form the result reads @xmath131 where @xmath132 here the label @xmath133 ( @xmath134 ) stands for the double index @xmath126 of the solutions of the schrdinger equation . we calculate the matrix elements @xmath135 and @xmath136 with the free planar rotor energies . symmetry implies that only functions of a same parity @xmath137 or @xmath138 contribute to c@xmath139 while @xmath136 differs from zero only for functions @xmath140 with different parity . for instance @xmath141 these matrix elements imply selection rules for transitions between energy levels . we take fourier transforms of eqs . ( [ qcct ] ) and ( [ qsst ] ) , using the identity @xmath142 we insert the energies @xmath143 and take into account the selection rules ( [ sr1 ] ) ( [ sr3 ] ) . defining the frequency transfer @xmath144 we obtain @xmath145 \\ + \frac{e^{-e^-_{2}/t}}{2}\bigl[\delta(\omega - \omega^{--}_{12 } ) + \sum_{m=2}^\infty\frac{e^{-e^-_{2m}/t}}{2}\bigl[\delta(\omega - \omega^{--}_{mm+1 } ) + \delta(\omega - \omega^{--}_{mm-1})\bigr]\biggr\ } \label{qccomega},\end{gathered}\ ] ] with @xmath146 similarly we get @xmath147 \\ + \sum_{m=1}^\infty\frac{e^{-e^-_{2m}/t}}{2}\bigl[\delta(\omega - \omega^{-+}_{mm+1 } ) + \delta(\omega - \omega^{-+}_{mm-1})\bigr]\biggr\}. \label{qssomega}\end{gathered}\ ] ] we notice that in absence of the uniaxial rotation of the c@xmath1 cage , i.e. for @xmath148 , the correlation functions eqs . ( [ fcct ] ) and ( [ fsst ] ) reduce to constants : @xmath149 , @xmath150 . hence the spectral functions @xmath92 and @xmath93 entering the raman scattering laws eqs . ( [ rzzzz2 ] ) and ( [ rzyzy2 ] ) reduce to @xmath151 and @xmath152 and exhibit infinitely sharp @xmath153-peaks which account for transitions between quantized planar rotor states . in table i we have quoted some values ( @xmath154 ) of @xmath155 for @xmath156 ( free rotor ) and @xmath157 ( value of the potential strength taken from experiment in ref . ) . rrrrrrrrr @xmath158 & @xmath159 & @xmath160 & @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 & @xmath166 + + @xmath167 & @xmath168 & & @xmath168 & & & & & + @xmath169 & @xmath170 & @xmath170 & @xmath170 & @xmath170 & @xmath171 & & & @xmath171 + @xmath172 & @xmath173 & @xmath173 & @xmath173 & @xmath173 & @xmath174 & @xmath174 & @xmath174 & @xmath174 + @xmath175 & @xmath176 & @xmath176 & @xmath176 & @xmath176 & @xmath177 & @xmath177 & @xmath177 & @xmath177 + @xmath178 & @xmath179 & @xmath179 & @xmath179 & @xmath179 & @xmath180 & @xmath180 & @xmath180 & @xmath180 + + @xmath167 & @xmath181 & & @xmath182 & & & & & + @xmath169 & @xmath183 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & & & @xmath188 + @xmath172 & @xmath189 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 + @xmath175 & @xmath197 & @xmath198 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 + @xmath178 & @xmath203 & @xmath203 & @xmath203 & @xmath203 & @xmath204 & @xmath205 & @xmath205 & @xmath204 + in order to calculate the classical correlation functions @xmath206 and @xmath207 we treat the c@xmath1 molecule as a classical uniaxially diffusing rotor with rotation axis @xmath15 in coincidence with a cubic crystal axis , in casu the @xmath33 axis . the corresponding rotation angle @xmath41 is measured away from the @xmath35 axis . equivalently one considers the @xmath15 axis along @xmath34 and @xmath35 ( meroaxial disorder ) . given the @xmath15 axis it would be tempting to study this motion in a crystal field potential of fourfold symmetry . such a study can be carried out along the lines of ref . and leads to a continued fraction expansion in terms of frequency moments of the orientational variables . it is adequate in the case of a strong crystal field potential since then one can limit the continued fraction to a few steps . however this approximation is not valid in the case of weak potentials . since the equator of the c@xmath1 molecule for rotations about @xmath15 deviates only slightly from circular shape , we prefer to consider the rotator about the @xmath15 axis in the rotational - diffusion approximation . this model has the obvious advantage of simplicity and leads to a linear temperature - dependent broadening of the tunneling transition lines . within the uniaxial diffusion model the c@xmath1 molecule experiences a random rotational torque ( also called brownian motion torque ) about its @xmath15 axis . this torque is caused by the thermal motion of the surrounding lattice ( heat bath ) . in that respect the present problem is different from the situation of the heavy symmetrical top with gravitational torque since on the molecular scale the effect of gravitation is negligible in comparison with the heat bath . the idea of rotational diffusion goes back to debye @xcite who applied the concept of rotational brownian motion to the theory of dielectric relaxation ( see also @xcite and @xcite ) . in appendix [ appa ] we give some details for the present problem . as a result we obtain @xmath208 here the rotational diffusion coefficient @xmath209 is given by the einstein relation @xmath210 where @xmath32 is the friction coefficient and @xmath211 the temperature . the equality of @xmath212 and @xmath213 is a consequence of our neglect of the crystal field potential within the large - friction approximation . from eqs . ( [ ct ] ) and ( [ st ] ) one sees that then @xmath214 in the following we will neglect the superscripts @xmath215 and @xmath216 on @xmath217 and @xmath218 and write just @xmath219 . the fourier transform is obtained from eq . ( [ fsstb ] ) with the result @xmath220 . \label{resfccomega}\end{aligned}\ ] ] we rewrite the scattering functions eqs . ( [ comega ] ) and ( [ somega ] ) as @xmath221f(\omega ' ) . \label{integral}\end{aligned}\ ] ] using expressions ( [ qccomega ] ) and ( [ qssomega ] ) we carry out the integral over @xmath222 and obtain @xmath223 where @xmath224\right\ } , \label{cpp } \\ c_{--}(\omega ) & = \frac{1}{2z}\left\{\frac{e^{-e_2 ^ -/t}}{2}f(\omega - \omega_{12}^{-- } ) + \sum_{m=2}^\infty \frac{e^{-e_{2m}^-/t}}{2}\bigl[f(\omega - \omega_{mm+1}^{-- } ) + f(\omega - \omega_{mm-1}^{--})\bigr]\right\ } , \\ c_{+-}(\omega ) & = \frac{1}{2z}\left\{e^{-e_0^+/t}f(\omega - \omega_{01}^{+- } ) + \frac{e^{-e_2^+/t}}{2}f(\omega - \omega_{12}^{+-})\right . \nonumber \\ & \phantom{=\frac{1}{2z}\left\{\right.}\left . + \sum_{m=2}^\infty \frac{e^{-e_{2m}^+/t}}{2}\bigl[f(\omega - \omega_{mm+1}^{+- } ) + f(\omega - \omega_{mm-1}^{+-})\bigr]\right\ } , \\ c_{-+}(\omega ) & = \frac{1}{2z}\left\ { \sum_{m=1}^\infty \frac{e^{-e_{2m}^-/t}}{2}\bigl[f(\omega - \omega_{mm+1}^{-+ } ) + f(\omega - \omega_{mm-1}^{-+})\bigr]\right\}. \label{cmp}\end{aligned}\ ] ] we see that @xmath92 is a sum of weighted lorentzians @xmath225 \label{lorentzian}\end{aligned}\ ] ] centered around the allowed frequency transfers @xmath226 and of width @xmath227 ( full width half maximum ) . since @xmath209 has dimension s@xmath228 , it follows from eq . ( [ diffcoeff ] ) that @xmath32 has the dimension of an action . we write @xmath229 , where @xmath230 is a dimensionless number taken as parameter . we then obtain @xmath231 s@xmath228 where @xmath211 is measured in kelvin . equivalently , @xmath232 @xmath18 . since to our knowledge there are so far no direct measurements of the orientational dynamics of the c@xmath1 molecule in sc@xmath0c@xmath0@c@xmath1 , we will choose a value of @xmath209 such that the correlation time @xmath233 has a value that is intermediate between the values of @xmath172 ns and @xmath234 ps measured by nmr experiments for the c@xmath48 molecule in the low - temperature monoclinic and high - temperature fcc phases of solid c@xmath48 , respectively @xcite . assuming that @xmath235 is a realistic value ( then @xmath236 s@xmath228 at @xmath237 k ) , we have plotted the scattering function @xmath92 for several temperatures in fig . [ comegaplot ] . the resonances are centered at the frequency transfers @xmath238 for the potential strength @xmath239 . the spectra reflect the characteristic assymetries for @xmath240 and @xmath241 due to anti - stokes and stokes processes , respectively . in our calculations , we have included transitions with the values @xmath242 . + + we notice that if one artificially excludes the tunneling motion of the c@xmath0 unit by taking a fixed value , say @xmath167 , for the angle @xmath43 , one finds that @xmath243 and @xmath244 . then eq.([integral ] ) becomes @xmath245 . since the c@xmath0-unit is dragged along with the classical rotational diffusion of the encapsulating c@xmath1 molecule , its polarizability is changing accordingly with time and the raman scattering laws @xmath74 and @xmath75 will exhibit a lorentzian @xmath246 of width @xmath247 centered at @xmath248 . in sect . [ secdynamic ] we have considered a cubic crystal with crystal axes @xmath27 in coincidence with the laboratory - fixed cubic axes @xmath28 . since experiments are performed on powder samples , we will extend the previous results . the powder sample consists of a large number of arbitrarily oriented cubic crystallites , each crystallite has symmetry @xmath10 where the sc@xmath0c@xmath0@c@xmath1 units are meroaxially disordered @xcite . we first will consider one single crystallite where the crystal - fixed system of axes is related to the laboratory - fixed system of axes by the euler angles @xmath249 . the c@xmath0 rotors are now moving in planes normal to the @xmath33 , @xmath34 , @xmath35 axes of the rotated coordinate system . this means that the polarizabilities @xmath250 or @xmath251 measured in the laboratory - fixed coordinate system will depend on the euler angles of the given crystallite . in appendix [ appb ] we have calculated the polarizability components @xmath252 which are obtained from @xmath253 by application of the rotation operation @xmath254 : @xmath255 the meroaxial average @xmath256 is obtained as @xmath257 . \label{alphazzrmeroaxial}\end{aligned}\ ] ] where @xmath258 refers to @xmath259 , @xmath260 to @xmath261 and @xmath262 to @xmath263 . the coefficients @xmath264 , @xmath265 and @xmath266 are derived in appendix [ appb ] , they are found to depend on only two euler angles . in the present section we assume that the meroaxial disorder is static or equivalently there are no reorientations of the c@xmath1 molecules among the three meroaxial directions in a given crystallite . the angle @xmath44 is then the sole dynamic quantity . the time - dependent polarizability correlation function per molecule in the given crystallite is obtained as @xmath267 . \label{alphazzrmeroaxialt}\end{aligned}\ ] ] the correlation functions @xmath268 and @xmath269 , defined by eqs . ( [ ctdef ] ) and ( [ stdef ] ) respectively , have been calculated in sects . ii and iii . the powder average for a function @xmath270 is defined as @xmath271 the results for the products @xmath272 are quoted in appendix [ appb ] . the powder - averaged polarizability correlation function per molecule reads @xmath273 . \label{refazzr}\end{aligned}\ ] ] taking into account @xmath274 , eq . ( [ alsosees ] ) , we obtain the raman scattering law for a powder - averaged sample with meroaxial disorder : @xmath275 the expression for a single crystal with meroaxial disorder has been given by eq . ( [ rzzzz2 ] ) . in an analogous way one calculates @xmath276 with the result @xmath277 . \label{alphazyrmeroaxial}\end{aligned}\ ] ] the coefficients @xmath278 are given in appendix [ appb ] . the time - dependent polarizability correlation function per molecule reads @xmath279 . \label{alphazyrmeroaxialt}\end{aligned}\ ] ] the powder average of a function @xmath280 is defined by @xmath281 taking into account the powder averages @xmath282 etc . , calculated in appendix [ appb ] , we obtain @xmath283 . \label{refazyr}\end{aligned}\ ] ] the raman scattering law then reads @xmath284 where again we have used @xmath274 , eq . ( [ alsosees ] ) . we see that the powder - averaged expression is the same as the one for a single crystal with meroaxial disorder , eq . ( [ rzyzy2 ] ) . so far we have assumed that the orientation of the long axis ( @xmath15 ) of the c@xmath1 molecule in a given cubic crystallite along the equivalent @xmath285 directions is random but static . the sole effect of the heat bath was the uniaxial rotational diffusion studied in sect . [ s3b ] . while this situation of static meroaxial disorder is realistic at temperatures inferior to say @xmath286 k , it becomes less valid at higher @xmath211 . here again we refer to the situation in solid c@xmath48 where with increasing temperature it is found that the uniaxial rotation axis flips between different symmetry equivalent orientations such that the rotational motion becomes more and more isotropic @xcite . we therefore will extend the previous model and take into account the situation where a molecule at a given lattice site in one crystallite changes orientation with the @xmath15 axis jumping randomly between equivalent potential minima in @xmath25 directions . here the heat bath causes stochastic torques about axes perpendicular to the long axis of the c@xmath1 molecule or equivalently perpendicular to the rotation axis of the encapsulated sc@xmath0c@xmath0 quantum gyroscope . we recall that accordingly the normal to the plane of the c@xmath0 quantum rotor will change its orientation . within a simple three sites stochastic jump model ( see e.g. @xcite ) , the conditional probability @xmath287 to find a c@xmath1 molecule in an orientation @xmath56 at time @xmath288 when it was in orientation @xmath289 at time @xmath167 is obtained by solving a system of three linear differential equations . one obtains @xmath290 where @xmath291 is the transition rate for a molecular reorientation . we associate the transition rate with the inverse of a relaxation time : @xmath292 here we have assumed an arrhenius - type law , known from reaction rate theory @xcite , where @xmath293 is an attempt frequency and @xmath294 an activation energy for meroaxial reorientations of the sc@xmath0c@xmath0@c@xmath1 complex as a whole . the equilbrium value of the conditional probability is independent of the initial and final orientation and corresponds to an a priori probability : @xmath295 in the previous section the meroaxial orientations within a given crystallite have been characterized by the coefficients @xmath296 , @xmath297 in eqs.([alphazzrmeroaxial ] ) and ( [ alphazyrmeroaxial ] ) of the polarizabilities @xmath250 and @xmath251 . treating these coefficients as dynamic stochastic quantities we obtain instead of eqs.([alphazzrmeroaxialt ] ) and ( [ alphazyrmeroaxialt ] ) @xmath298 , \label{azz } \\ \bigl\langle\alpha_{zy}^r(t)\alpha_{zy}^r(0)\bigr\rangle & = b^2\bigl [ \bigl\langle a_{zy}(t)a_{zy}(0)\bigr\rangle \nonumber \\ & \phantom{b^2\bigl [ } + \bigl\langle b_{zy}(t ) b_{zy}(0)\bigr\rangle s(t ) + \bigl\langle c_{zy}(t ) c_{zy}(0)\bigr\rangle c(t ) \bigr ] . \label{azy}\end{aligned}\ ] ] the correlation functions @xmath299 which refer to meroaxial reorientations are calculated within the frame of the stochastic jump model . for instance for a given set @xmath300 one has @xmath301 where the conditional probabilities @xmath287 are given by eqs . ( [ ieqj ] ) and ( [ ineqj ] ) , while the factor @xmath302 accounts for the equilibrium initial probability . since the coefficients @xmath303 depend on the euler angles which specify the orientation of a given crystallite ( sect . [ secpowder ] ) , the powder - averaged correlation functions are obtained by averaging over the euler angles : @xmath304 taking into account the values of the powder - averaged products given in appendix [ appb ] , we obtain : @xmath305 , \\ \overline{\bigl\langle c_{zz}(t)c_{zz}(0)\bigr\rangle } & = \frac{12}{135}\left[1 + 3e^{-3t/\tau}\right].\end{aligned}\ ] ] the powder average of eq . ( [ azz ] ) then reads @xmath306 where the function @xmath307 is given by @xmath308 . \label{dt}\end{aligned}\ ] ] here we have used again @xmath309 , eq . ( [ alsosees ] ) . the first term on the right - hand side @xmath268 accounts for the superposition of the quantum motion ( tunneling ) of the c@xmath0 rotor and the uniaxial rotational diffusion of the encapsulating c@xmath1 molecule , the second term @xmath310 accounts for the classical motion of the c@xmath0 rotor when its plane of motion is changing with the meroaxial reorientations of the encapsulating c@xmath1 molecule , finally the third term @xmath311 accounts for the interference of the two motions of the c@xmath1 molecule with the tunneling of the c@xmath0 rotor . similarly , using again eq . ( [ ata0 ] ) and the powder - averaged products @xmath312 etc . in appendix [ appb ] , we find @xmath313 , \\ \overline{\bigl\langle c_{zy}(t)c_{zy}(0)\bigr\rangle } & = \frac{1}{135}\left[4 + 17e^{-3t/\tau}\right],\end{aligned}\ ] ] and hence @xmath314 with @xmath307 again given by eq . ( [ dt ] ) . in the limit of small relaxation time we recover eq . ( [ refazyr ] ) for static meroaxial disorder . the raman scattering laws are given by @xmath315 and @xmath316 the fourier transform of @xmath307 leads to the scattering function @xmath317 the spectral function @xmath92 is given by eqs . ( [ rescomega ] ) ( [ lorentzian ] ) while @xmath318 is the fourier transform of the relaxation function @xmath319 . the fourier transform of the interference term @xmath320 is rewritten as @xmath321 using eqs . ( [ rescomega ] ) ( [ lorentzian ] ) and ( [ jomega ] ) we obtain the scattering function @xmath322 the functions @xmath323 have the same structure as @xmath324 , eqs . ( [ cpp ] ) ( [ cmp ] ) , respectively , but where the lorentzians @xmath325 , eq . ( [ lorentzian ] ) , are replaced by @xmath326.\end{aligned}\ ] ] similarly to @xmath92 , eq . ( [ rescomega ] ) , the function @xmath327 is a sum of weighted lorentzians centered around @xmath328 but of width @xmath329 where @xmath330 the broadening of the transition frequencies of the quantum rotor with increasing temperature is now due to the uniaxial rotational diffusion and the meroaxial reorientations of the encapsulating c@xmath1 molecule . notice that both contributions depend on temperature . in fig . [ fig6 ] we have plotted the function @xmath327 for several temperatures . the parameters describing the dynamics of the c@xmath1 molecule are @xmath235 for the rotational diffusion model and @xmath331 s@xmath228 ( attempt frequency ) and @xmath332 k ( activation energy ) for the thermally activated meroaxial reorientations . comparable values of the activation energy , i.e. @xmath333 mev and @xmath334 mev have been deduced from neutron scattering studies in solid c@xmath48 @xcite and solid c@xmath3 @xcite , respectively . while for @xmath335 k and @xmath336 k the contribution of @xmath337 to the half width @xmath338 is negligible in comparison to @xmath247 , both ( additive ) contributions become comparable at @xmath21 k. at higher @xmath211 the thermally - activated reorientations are dominant and lead to a smearing out of the low - frequency resonances in the scattering function @xmath327 . + + + + if one would artificially exclude the tunneling motion , the function @xmath339 entering the raman scattering laws of the c@xmath0 unit would reduce to a superposition of lorentzians centered at @xmath248 : @xmath340 the first term on the right - hand side [ given by eq . ( [ resfccomega ] ) ] accounts solely for the rotational uniaxial diffusion , the second term for the meroaxial reorientations and the last term for the interference of these classical motions of the encapsulating c@xmath1 molecule . it has been shown that the low - frequency ( rotational ) part of the raman scattering spectrum of a powder crystal of sc@xmath0c@xmath0@c@xmath1 fullerite reflects the superposition of the quantum tunneling motion of the encapsulated sc@xmath0c@xmath0 complex about its long axis and the random classical rotational motion of the surrounding c@xmath1 molecule . the effect of the c@xmath1 molecule on the dynamics of sc@xmath0c@xmath0 is twofold . firstly , since the long axis of sc@xmath0c@xmath0 gyroscope coincides with the @xmath15 axis of the molecule , the rotation of sc@xmath0c@xmath0 about this axis corresponds to the motion of the c@xmath0 bond as a planar quantum rotor in a fourfold potential @xcite . secondly , any rotation of the c@xmath1 molecule caused by torques due to the thermal lattice environment leads to a dragging of the enclosed sc@xmath0c@xmath0 unit and hence affects the spectrum of the c@xmath0 quantum rotor seen in the laboratory frame . the low - frequency raman spectra resulting from the interaction of the scattering radiation with the induced dipole of the c@xmath0 rotor reflect these features . in analogy with the dynamics of the c@xmath48 molecule in solid c@xmath48 @xcite , we have assumed that the rotational motion of the c@xmath1 molecule at a lattice site in a given crystallite is composed of two parts : uniaxial rotational diffusion about the @xmath15 axis and stochastic jumps of the @xmath15 orientation among @xmath341 directions . the superposition of the tunneling motion of the planar quantum rotor with the classical rotations of the c@xmath1 molecule leads to the spectral function @xmath342 given by eq . ( [ domega ] ) , in the raman scattering laws @xmath74 and @xmath75 . the function @xmath92 , defined by eqs . ( [ integral ] ) ( [ lorentzian ] ) , accounts for tunneling transitions between the energy levels of the encapsulated c@xmath0 rotor . the spectrum consists of a series of resonances described by lorentzians centered at the transition frequencies ( table [ hbaromega ] , @xmath239 ) and broadened by the uniaxial rotational diffusion ( half width @xmath247 ) of the surrounding c@xmath1 molecule . since the hindering potential for the rotational diffusion about the @xmath15 axis is weak , this motion affects the spectrum already at low @xmath211 . the temperature dependence of the spectrum has been studied in fig . [ comegaplot ] . the term @xmath343 in eq . ( [ 61 ] ) accounts for the raman spectrum of the radiation - induced c@xmath0 dipole while the sc@xmath0c@xmath0 unit is dragged along by the classical reorientations of the c@xmath1 molecule among its three meroaxial directions . this motion which reflects the changes of the orientation of the c@xmath0 rotor plane is described by a three sites stochastic jump model , characterized by a thermally activated relaxation time @xmath344 . notice that @xmath343 leads to a central resonance of half width @xmath345 in the raman scattering law even in absence of any quantum mechanical tunneling of c@xmath0 . the width of this central resonance ( quasi - elastic peak ) becomes appreciable at @xmath346 k. in the scattering law @xmath347 , eq . ( [ f518 ] ) , this quasi - elastic peak is present in addition to the elastic rayleigh peak . we suggest that in future low - energy raman experiments additional attention will be given to the possible identification of the temperature - dependent quasi - elastic peak . the last term @xmath327 in eq . ( [ 61 ] ) is due to the interference between the uniaxial diffusion - modulated tunneling motion described by @xmath92 and the stochastic jump model accounted for by @xmath343 . the function @xmath327 is a convolution of @xmath348 and @xmath349 [ see eq . ( [ gomega ] ) ] . while at low @xmath211 the spectra of @xmath92 and @xmath327 are very similar ( compare the plots for @xmath335 k , @xmath350 k in fig . [ comegaplot ] and [ fig6 ] ) they become different at higher @xmath211 ( see the @xmath351 k plots ) where the increasing influence of the stochastic jumps adds to the line broadening . the width @xmath329 of the individual resonances , eq . ( [ gammaeq ] ) , increases from @xmath352 @xmath18 at @xmath335 k to @xmath353 @xmath18 at @xmath350 k and @xmath354 @xmath18 at @xmath355 k. this broadening leads to an overlap of the low - frequency resonances with increasing @xmath211 . finally the sum @xmath339 of these contributions which corresponds to the low - frequency raman response function is shown in fig . the quasi - elastic peak centered at @xmath248 becomes important with increasing temperature . in addition the growing importance of @xmath327 smears out the low - frequency resonances with increasing @xmath211 while the higher frequency resonances remain prominent . the overall shape of the spectral function @xmath339 and its temperature evolution agree very well with the low - frequency raman scattering results of ref . . there is quantitative agreement with the position of the resonance lines . the smearing out of the low - frequency resonances and the prominence of the higher - frequency resonances with increasing @xmath211 ( fig . 3 of ref . ) are well reproduced by the present theory . in addition to the positions of the resonance lines , the theory accounts for their temperature - dependent broadening . in fig . [ figexpthe ] we confront the theoretical spectra @xmath339 with the experimental raman spectra , for both @xmath350 k and @xmath356 k. we notice that the experimental spectra are contaminated by a plasma line at @xmath357 @xmath18 @xcite . note that the central parts of the experimental spectra have been omitted in order to remove the effect of the unshifted rayleigh peak . on the other hand the theoretical spectrum exhibits a quasi - elastic peak which is an intrinsic effect due to the meroaxial stochastic reorientations of the sc@xmath0c@xmath0@c@xmath1 complex [ contribution @xmath343 in @xmath339 ] . complementary to the present work it would be useful to measure the dynamics of the c@xmath1 molecule in solid sc@xmath0c@xmath0c@xmath1 directly say by nmr , neutron or @xmath359-spin spectroscopy . the theoretical work has been supported by the bijzonder onderzoeksfonds , universiteit antwerpen ( bof - ua ) . b.v . is a postdoctoral fellow of the research foundation - flanders ( fwo ) . the experimental work has been supported by the eu project nanotemp and by the austrian fwf ( 17345-phy ) . we recall that @xmath41 is the angle of rotation of the c@xmath1 molecule about its @xmath15 axis . treating the molecule as a brownian rotor we have the one - dimensional langevin equation @xmath360 here @xmath361 is the moment of inertia about the @xmath15 axis , @xmath362 is the friction torque , @xmath363 a white noise driving torque and @xmath364 is a torque due to the orientational crystal field potential @xmath365 . under the assumption @xcite that the friction torque is dominant in comparison to the inertial term and that the variation of @xmath366 with @xmath41 is weak , one can use standard methods @xcite to derive a smoluchowski equation for the angular distribution function @xmath367 : @xmath368 where @xmath369 . the functions @xmath370 @xmath371 are orthonormal eigenfunctions of the operator @xmath372 with eigenvalues @xmath373 . we calculate the conditional probability distribution @xmath374 to find the molecule at an angle @xmath41 at time @xmath375 when it was at angle @xmath376 at @xmath377 . the initial condition can be written as @xmath378 where the second member equality is just the closure relation . on the other hand in the long - time limit the orientation of the molecule should be random which corresponds to the condition @xmath379 a particular solution of ( [ smoluchowski ] ) subject to these boundary condition is of the form @xmath380 the correlation function @xmath381 , eq . ( [ fcct ] ) , is rewritten as a thermal average : @xmath382 here @xmath383 is the joint probability of finding the c@xmath1 molecule with orientation angle @xmath384 in the interval @xmath385 initially and orientation @xmath386 in @xmath387 at time @xmath375 . one has @xmath388 where the conditional probability @xmath374 is given by @xmath389 @xmath371 , and where @xmath390 is the initial equilibrium probability . carrying out the integrals in eq . ( [ a7 ] ) gives as a result eq . ( [ fcctb ] ) . in a similar way we obtain eq . ( [ fsstb ] ) . here we give details about the calculation of the powder averages in sect . [ secpowder ] . we start from the situation where the c@xmath1 molecule is in standard orientation , which corresponds to the polarizabilities @xmath59 and @xmath391 given by eqs . ( [ alphazz1 ] ) and ( [ alphazy1 ] ) respectively . in order to apply the rotation operation @xmath254 we rewrite the polarizabilities in terms of spherical harmonics @xmath392 . we use the notations and conventions of bradley and cracknell @xcite . with @xmath393 we get @xmath394\right|_{\phi = \pi/2}. \label{alphazya}\end{aligned}\ ] ] the condition @xmath395 ensures that the c@xmath0 rotor initially moves in the @xmath40 plane . although the function @xmath396 does not depend on @xmath46 , the condition @xmath395 has to be taken into account after the application of a rotation operation . the transformation law of spherical harmonics under a rotation @xmath254 reads : @xmath397 here @xmath398 are the wigner rotator functions defined by @xmath399 where @xmath400 @xcite . the functions @xmath401 are polynomials in @xmath402 and @xmath403 . they satisfy the relations @xmath404 we recall that the angles @xmath44 and @xmath46 on the right - hand side of eq . ( [ rotation ] ) refer to the values before the application of the rotation . applying @xmath254 to @xmath59 , eq . ( [ alphazz1a ] ) , we have to evaluate @xmath405 in addition to the definitions ( [ y20 ] ) and ( [ y21 ] ) we quote @xmath406 we further use @xmath407 we then find after some bookkeeping @xmath408,\end{aligned}\ ] ] where @xmath409 , \\ b_{zz}^{(1)}(\beta,\gamma ) & = \sin 2\beta\sin\gamma , \\ c_{zz}^{(1)}(\beta,\gamma ) & = \frac{1}{2}\left[\sin^2\beta\cos 2\gamma + 3\cos^2\beta - 1\right].\end{aligned}\ ] ] hence @xmath410 . \label{alphazzr}\end{aligned}\ ] ] with the change @xmath411 we find the coefficients entering @xmath412 @xmath413 , \\ b_{zz}^{(2)}(\beta,\gamma ) & = -\sin 2\beta\sin\gamma , \\ c_{zz}^{(2)}(\beta,\gamma ) & = \frac{1}{2}\left[-\sin^2\beta\cos 2\gamma + 3\cos^2\beta - 1\right],\end{aligned}\ ] ] and with @xmath414 the coefficients entering @xmath415 @xmath416 , \\ b_{zz}^{(3)}(\beta,\gamma ) & = 0 , \\ c_{zz}^{(3)}(\beta,\gamma ) & = \frac{1}{2}\left[\cos^2\beta + 3\sin^2\beta - 1\right].\end{aligned}\ ] ] the average polarizability @xmath250 is then given by eq . ( [ alphazzrmeroaxial ] ) . applying the definition of powder average eq . ( [ fbetagammaaverage ] ) we get @xmath417 ^ 2 } = \frac{4}{45}$ ] , @xmath56 ; @xmath418 ^ 2 } = \frac{4}{15}$ ] , @xmath419 ; @xmath420 ^ 2 } = \frac{4}{15}$ ] , @xmath419 ; @xmath421 ^ 2 } = \frac{8}{15}$ ] ; @xmath422 ^ 2 } = -\frac{2}{45}$ ] , @xmath423 ; @xmath424 ^ 2 } = 0 $ ] , @xmath423 ; @xmath425 ^ 2 } = -\frac{2}{15}$ ] , @xmath419 , @xmath426 ; @xmath427 ^ 2 } = \frac{2}{15}$ ] . and of the relations ( [ relations ] ) we find @xmath429\right|_{\phi = \pi/2 } \\ = \left(\frac{15}{8\pi}\right)^{1/2}\bigl[a_{zy}^{(1)}(\alpha,\beta,\gamma ) + b_{zy}^{(1)}(\alpha,\beta,\gamma)\sin 2\theta + c_{zy}^{(1)}(\alpha,\beta,\gamma)\cos 2\theta\bigr]\end{gathered}\ ] ] where @xmath430 , \\ b_{zy}^{(1)}(\alpha,\beta,\gamma ) & = \cos\beta\cos\gamma\cos\alpha - \cos 2\beta\sin\gamma\sin\alpha , \\ c_{zy}^{(1)}(\alpha,\beta,\gamma ) & = \frac{1}{2}\left[-\sin\beta\sin2\gamma\cos\alpha - \frac{\sin 2\beta\cos 2\gamma\sin\alpha}{2 } + \frac{3\sin 2\beta\sin\alpha}{2}\right].\end{aligned}\ ] ] hence @xmath431.\end{aligned}\ ] ] with the change of angles @xmath432 we get the coefficients entering @xmath433 : @xmath434 , \\ b_{zy}^{(2)}(\alpha,\beta,\gamma ) & = \cos\beta\sin\gamma\cos\alpha + \cos 2\beta\cos\gamma\sin\alpha , \\ c_{zy}^{(2)}(\alpha,\beta,\gamma ) & = \frac{1}{2}\left[-\sin\beta\sin 2\gamma\cos\alpha + \frac{\sin 2\beta\cos 2\gamma\sin\alpha}{2 } + \frac{3\sin 2\beta\sin\alpha}{2}\right],\end{aligned}\ ] ] and with @xmath435 the coefficients entering @xmath436 : @xmath437 with the definition of powder average eq . ( [ falphabetagammaaverage ] ) we get @xmath438 ^ 2 } = \frac{1}{15}$ ] , @xmath56 ; @xmath439 ^ 2 } = \frac{1}{5}$ ] , @xmath419 ; @xmath440 ^ 2 } = \frac{1}{3}$ ] ; @xmath441 ^ 2 } = \frac{1}{5}$ ] , @xmath419 ; @xmath442 ^ 2 } = \frac{1}{15}$ ] ; @xmath443 , @xmath423 ; @xmath444 , @xmath423 ; @xmath445 , @xmath419 ; @xmath446 . t. kai , t. tomiyama , t. yoshida , y. kobayashi , e. nishibori , m. takata , m. sakata , and h.shinohara , angew . chem . * 113 * , 411 ( 2001 ) ; c .- r . wang , m. inakuma , h. shinohara , chem . 300 * , 379 ( 1999 ) . |
we show that charged - particles decaying in the early universe can induce a scale - dependent or ` running ' spectral index in the small - scale linear and nonlinear matter power spectrum and discuss examples of this effect in minimal supersymmetric models in which the lightest neutralino is a viable cold - dark - matter candidate . we find configurations in which the neutralino relic density is set by coannihilations with a long - lived stau , and the late decay of staus partially suppresses the linear matter power spectrum . nonlinear evolution on small scales then causes the modified linear power spectrum to evolve to a nonlinear power spectrum similar ( but different in detail ) to models parametrized by a constant running by redshifts of 2 to 4 . thus , lyman- forest observations , which probe the matter power spectrum at these redshifts , might not discriminate between the two effects . however , a measurement of the angular power spectrum of primordial 21-cm radiation from redshift might distinguish between this charged - decay model and a primordial running spectral index . the direct production of a long - lived charged particle at future colliders is a dramatic prediction of this model . | astro-ph0410714 | while recent cosmological observations provide convincing evidence that nonbaryonic dark matter exists @xcite , we do not know the detailed particle properties of the dark matter , nor the particle spectrum of the dark sector . there has been considerable phenomenological effort towards placing model - independent limits on the possible interactions of the lightest dark - matter particle ( ldp ) in an attempt to try and identify candidates within detailed particle theories or rule out particular candidate theories . for instance , models with stable charged dark matter have been ruled out @xcite , while significant constraints have been made to dark - matter models with strong interactions @xcite , self - interactions @xcite , and a millicharge @xcite . recently , it was shown that a neutral dark - matter particle with a relatively large electric or magnetic dipole moment remains a phenomenologically viable candidate @xcite . concurrent with this phenomenological effort , theorists have taken to considering physics beyond the standard model in search of a consistent framework for viable dark - matter candidates . the leading candidates , those that produce the correct relic abundance and appear in minimal extensions of the standard model ( often for independent reasons ) , are the axion @xcite and weakly interacting massive particles ( wimps ) , such as the neutralino , the lightest mass eigenstate from the superposition of the supersymmetric partners of the @xmath4 and @xmath5 neutral gauge bosons and of the neutral higgs bosons @xcite . however , other viable candidates have also been considered recently , such as gravitinos or kaluza - klein gravitons produced through the late decay of wimps @xcite . these latter candidates are an interesting possibility because the constraints to the interactions of the ldp do not apply to the next - to - lightest dark - matter particle ( nldp ) , and the decay of the nldp to the ldp at early times may produce interesting cosmological effects ; for instance , the reprocessing of the light - element abundances formed during big - bang nucleosynthesis @xcite , or if the nldp is charged , the suppression of the matter power spectrum on small scales and thus a reduction in the expected number of dwarf galaxies @xcite . in this paper we describe another effect charged nldps could have on the matter power spectrum . if all of the present - day dark matter is produced through the late decay of charged nldps , then , as discussed in ref . @xcite , the effect is to essentially cut off the matter power spectrum on scales that enter the horizon before the nldp decays . however , if only a fraction @xmath6 of the present - day dark matter is produced through the late decay of charged nldps , the matter power spectrum is suppressed on small scales only by a factor @xmath7 . this induces a scale - dependent spectral index for wavenumbers that enter the horizon when the age of the universe is equal to the lifetime of the charged particles . what we show below is that , for certain combinations of @xmath6 and of the lifetime of the charged particle @xmath8 , this suppression modifies the nonlinear power spectrum in a way similar ( but different in detail ) to the effect of a constant @xmath9 . although these effects are different , constraints based on observations that probe the nonlinear power spectrum at redshifts of 2 to 4 , such as measurements of the lyman-@xmath1 forest , might confuse a running index with the effect we describe here even if parametrized in terms of a constant @xmath10 . this has significant implications for the interpretation of the detection of a large running of the spectral index as a constraint on simple single - field inflationary models . the detection of a unexpectedly large spectral running in future observations could instead be revealing properties of the dark - matter particle spectrum in conjunction with a more conventional model of inflation . we note that the sloan digital sky survey lyman-@xmath1 data @xcite has significantly improved the limits to constant-@xmath10 models compared to previous measurements alone @xcite . a detailed study of the ( @xmath8,@xmath6 ) parameter space using these and other cosmological data would also provide interesting limits to the models we discuss here . while , even with future lyman-@xmath1 data , it may be difficult to discriminate the effect of a constant running of the spectral index from a scale - dependent spectral index due to a charged nldp , other observations may nevertheless discriminate between the two scenarios . future measurements of the power spectrum of neutral hydrogen through the 21cm - line might probe the linear matter power spectrum in exquisite detail over the redshift range @xmath11 at comoving scales less than 1 mpc and perhaps as small as @xmath12 mpc @xcite ; such a measurement could distinguish between the charged - particle decay scenario we describe here and other modifications to the primoridal power spectrum . if , as in some models we discuss below , the mass of these particles is in reach of future particle colliders the signature of this scenario would be spectacular and unmistakable the production of very long - lived charged particles that slowly decay to stable dark matter . although we describe the cosmological side of our calculations in a model - independent manner , remarkably , there are configurations in the minimal supersymmetric extension of the standard model ( mssm ) with the right properties for the effect we discuss here . in particular , we find that if the lsp is a neutralino quasi - degenerate in mass with the lightest stau , we can naturally obtain , at the same time , ldps providing the correct dark matter abundance @xmath13 @xcite and nldps with the long lifetimes and the sizable densities in the early universe needed in the proposed scenario . such configurations arise even in minimal supersymmetric schemes , such as the minimal supergravity ( msugra ) scenario @xcite and the minimal anomaly - mediated supersymmetry - breaking ( mamsb ) model @xcite . this implies that a detailed study of the ( @xmath8,@xmath6 ) parameter space using current and future cosmological data may constrain regions of the mssm parameter space that are otherwise viable . furthermore , we are able to make quantitative statements about testing the scenario we propose in future particle colliders or dark matter detection experiments . the paper is organized as follows : we first review in section [ sec : charged ] how the standard calculation of linear perturbations in an expanding universe must be modified to account for the effects of a decaying charged species , calculate the linear matter power spectrum and discuss the constraints to this model from big bang nucleosynthesis ( bbn ) and the spectrum of the cosmic microwave background ( cmb ) . in section [ sec : nonlinear ] we briefly discuss how we estimate the nonlinear power spectrum from the linear power spectrum and present several examples . in section [ sec:21 cm ] we discuss how measurements of the angular power spectra of the primordial 21-cm radiation can be used to distinguish this effect from other modifications to the primordial power spectrum . in section [ sec : partmodel ] we describe how this scenario can be embedded in a particle physics model , concluding that the most appealing scheme is one where long lifetimes are obtained by considering nearly degenerate ldp and nldp masses . in section [ sec : lifetimes ] we compute the lifetimes of charged next - to - lightest supersymmetric particles ( nlsps ) decaying into neutralino lsps , and show that , in the mssm , the role of a nldp with a long lifetime can be played by a stau only . in section [ sec : fraction ] we estimate what fraction of charged to neutral dark matter is expected in this case , while in section [ sec : msugra ] we describe how consistent realizations of this scenario can be found within the parameter space of msugra and mamsb . finally , in section [ sec : colliders ] we discuss the expected signatures of this scenario at future particle colliders , such as the large hadron collider ( lhc ) , and prospects for detection in experiments searching for wimp dark matter . we conclude with a brief summary of our results in section [ sec : conclusion ] . in this section we discuss how the decay of a charged particle @xmath14 ( the nldp ) to a neutral particle @xmath15 ( the ldp ) results in the suppression of the linear matter power spectrum on small scales . as the @xmath14 particles decay to @xmath15 particles , their comoving energy density decays exponentially as @xmath16 increasing the comoving energy density of @xmath15 particles as @xmath17 here @xmath18 is the comoving number density of dark matter , and @xmath19 is the comoving number density of dark matter produced through the decay of @xmath14 particles , @xmath20 is the scale factor , and @xmath21 is the cosmic time . since the @xmath14 particles are charged , they are tightly coupled to the ordinary baryons ( the protons , helium nuclei , and electrons ) through coulomb scattering . it is therefore possible describe the combined @xmath14 and baryon fluids as a generalized baryon - like component @xmath22 as far as perturbation dynamics is concerned . we thus denote by @xmath23 the total charged - particle energy density at any given time . at late times , after nearly all @xmath14 particles have decayed , @xmath24 . the relevant species whose perturbation dynamics are modified from the standard case are the stable dark matter ( subscript @xmath15 ) , the charged species ( subscript @xmath22 ) , and the photons ( subscript @xmath25 ) . by imposing covariant conservation of the total stress - energy tensor , accounting for the compton scattering between the electrons and the photons , and linearizing about a friedmann - robertson - walker ( frw ) universe we arrive at the equations describing the evolution of linear fluid perturbations of these components in an expanding universe . in the synchronous gauge , for the ` @xmath22 ' component , the perturbation evolution equations are @xmath26 and @xmath27 here and in what follows , @xmath28 is the fractional overdensity and @xmath29 is the divergence of the bulk velocity in fourier space of a given species @xmath30 . an overdot represents a derivative with respect to the conformal time . the number density of electrons is @xmath31 , while @xmath32 is the thomson cross section . because the @xmath14 particles and the baryons share a common bulk velocity and overdensity , will generally evolve away from zero in an arbitrary gauge , even when starting with the adiabatic initial condition @xmath33 , due to gradients in the proper time . it is a special simplifying property of the synchronous gauge that @xmath34 = 0 for all time for adiabatic initial conditions . ] these equations are identical to the standard perturbation equations for the baryons with the replacement @xmath35 ( see , for example , ref . @xcite ) . for the dark matter we find that @xmath36 and @xmath37 where @xmath38 . the modifications to the photon perturbation evolution are negligibly small because @xmath14 decays during the radiation - dominated epoch when @xmath39 and , as discussed below , for viable models @xmath40 to prevent unreasonably large spectral distortions to the cmb . combining these equations with the ( unmodified ) equations for the neutrino perturbations we can solve for the linear power spectrum of matter fluctuations in this model . we have solved these equations using a modified version of cmbfast @xcite . in fig . [ fig : linear ] we show the linear matter power spectrum in this model for several values of the @xmath14 lifetime @xmath8 and fraction @xmath6 . as shown in this figure , the small - scale density modes that enter the horizon before the @xmath14 particles decay ( when the age of the universe is less than @xmath8 ) are suppressed relative to the standard case by a factor of @xmath7 . since the decaying particles are charged , the production of the ldp will always be accompanied by an electromagnetic cascade . the latter could in principle reprocess the light elements produced during bbn , or induce unreasonably large spectral distortions to the cmb . we show here that in fact these effects are small for the models discussed in this paper . the energy density released by the decay of @xmath14 particles can be parametrized as @xmath41 where @xmath42 is the average electromagnetic energy released in a @xmath14 decay and @xmath43 is the dark - matter to photon ratio . in the specific models we discuss below , @xmath44 , and @xmath45 this yields @xmath46 in the models we discuss below , @xmath47 , and @xmath48 , giving @xmath49 ev . the limit derived from too much reprocessing of the bbn light element abundances is @xmath50 ev @xcite where @xmath51 , so we are safely below this bound . for @xmath52 , electromagnetic energy injection will result in a chemical - potential distortion to the cmb of @xcite @xmath53 and so we expect @xmath54 @xmath55 for lifetimes between @xmath56 @xmath57 , below the current limit of @xmath58 @xcite . as density perturbations grow under the influence of gravity , linear evolution ceases to describe their growth and nonlinear effects must be taken into account . on large scales , where density perturbations have had insufficient time to become nonlinear , the linear matter power spectrum describes the statistics of density fluctuations . however , on small scales the full nonlinear matter power spectrum is required . in order to calculate the nonlinear power spectrum for a given model we have used the recently devised halofit method @xcite . this method uses higher - order perturbation theory in conjunction with the halo model of large - scale structure to determine the nonlinear power spectrum given a linear power spectrum . it has been shown to accurately reproduce the nonlinear power spectra of standard n - body simulations and , unlike the earlier mappings such as the peacock and dodds formula @xcite , it is applicable in cases ( like we consider here ) when @xmath59 is not a monotonic function . in particular we have checked that it approximately reproduces the shapes of the nonlinear power spectra determined in ref . @xcite through n - body simulations in models where the linear power spectrum is completely cut off on small scales . as we are discussing here less drastic alterations to the linear power spectrum , we believe the halofit procedure provides an estimate of the nonlinear power spectrum adequate for illustrating the effect we describe in this paper . any detailed study would require a full n - body simulation . in figs . [ fig : nonlinear_t20][fig : nonlinear_t1 ] we show both the linear and nonlinear matter power spectra at redshift @xmath60 ( data from measurements the lyman-@xmath1 forest probe redshifts 24 at wavenumbers @xmath61@xmath62 ) for a charged - decay model , and for a model with a running spectral index . although these models have different linear power spectra , nonlinear gravitational evolution causes these models to have nearly identical nonlinear power spectra over interesting ranges of wavenumbers . the lifetimes shown were chosen because they produce effects on scales that can be probed by measurements of the lyman-@xmath1 forest , while the values @xmath63 were chosen because they arise in the supersymmetric models we discuss below . in fig . [ fig : sdss_limits ] we show the current constraints on the ( @xmath64,@xmath10 ) parameter space from wmap and sdss lyman-@xmath1 forest data . the constant-@xmath10 models we have compared our charged - decay model to lie within the wmap+lya contours shown in the figure . since , in each case , the charged - decay model tends to interpolate between the standard @xmath65cdm model on larger scales ( in particular on the scales probed by the cmb ) and a constant-@xmath10 running - index model on smaller scales , and both of them are allowed by these data , we expect the charged - decay models we have considered in figs . [ fig : nonlinear_t20][fig : nonlinear_t1 ] to be consistent with current data as well . we leave for future work the task of using a combined analysis of lyman-@xmath1 forest and other cosmological measurements to put limits directly on the ( @xmath6,@xmath8 ) parameter space ( and thus on the parameter space of the mssm models we discuss below ) . after recombination and the formation of neutral hydrogen , the gas in the universe cools with respect to the cmb temperature @xmath66 starting at at a redshift @xmath67 . the spin temperature @xmath68 of the gas , which measures the relative populations of the hyperfine levels of the ground state of hydrogen separated by the 21-cm spin - flip transition , remains collisionally coupled to the temperature @xmath69 of the baryons until a redshift @xmath70 when collisions become inefficient and the spin temperature rises to @xmath71 . there is thus a window between @xmath72@xmath3 in which neutral hydrogen absorbs the cmb at a wavelength of @xmath73 . it has recently been suggested that the angular fluctuations in the brightness temperature of the 21-cm transition within this window may be measured with future observations and used to constrain the matter power spectrum at these very high redshifts @xcite . at redshifts @xmath74 , the matter power spectrum on the scales @xmath75 of interest here are still in the linear regime . as discussed in ref . @xcite , due to the unprecedented wealth of potential information contained in these @xmath76-cm measurements , models with a running index or other small - scale modifications of the matter power spectrum can in principle be distinguished from each other . even if the charged - particle lifetime is so small that no significant modifications to the power spectrum occur on scales probed by other cosmological observations ( like the model shown in fig . [ fig : nonlinear_t1 ] ) , 21-cm observations might detect or constrain such effects on the linear matter power spectrum . from the particle - physics point of view , the setup we have introduced may seem _ ad hoc_. we need a pair of particles that share a conserved quantum number and such that the lightest , the ldp , is neutral and stable , while the other , the nldp , is coupled to the photon and quasi - stable , in order to significantly contribute to the cosmological energy density at an intermediate stage in the structure - formation process . such a picture requires three ingredients : _ ( i ) _ the relic abundance of the ldp must be compatible with the cdm component ; _ ( ii ) _ the abundance of the nldp must be at the correct level ( namely , @xmath77 the total dark - matter density ) ; and _ ( iii ) _ the nldp must have the proper lifetime ( i.e. , @xmath78 yr ) . let us start with the last requirement . one way to get the required lifetime is to introduce a framework with strongly - suppressed couplings . one such possibility is , for instance , to assume that the ldp is a stable super - weakly interacting dark - matter particle , such as a gravitino lsp in r - parity conserving supersymmetric theories or the kaluza - klein first excitation @xmath79 of the graviton in the universal extra - dimension scenario @xcite . the nldp can have non - zero electric charge , but at the same time a super - weak decay rate into the ldp ( with the latter being the only allowed decay mode ) . in models of gauge - mediated supersymmetry breaking , this might indeed be the case with a stau nlsp decaying into a gravitino lsp . in this specific example , we have checked that , to retrieve the very long lifetimes we introduced in our discussion , we would need to impose a small mass splitting between the nlsp and lsp , as well as to raise the mass scale of the lsp up to about 100 tev . this value is most often considered uncomfortably large for a susy setup , and it also makes thermal production of ldps or nldps unlikely , being near a scale at which the unitary bound @xcite gets violated . without thermal production one would need to invoke first a mechanism to wipe out the thermal components and then provide a viable non - thermal production scheme that fixes the right portion of ldps versus nldps . finally , such heavy and extremely weakly - interacting objects would evade any dark - matter detection experiment , and would certainly not be produced at the forthcoming cern large hadron collider ( lhc ) . this would then be a scheme that satisfies the three ingredients mentioned above , but that can not be tested in any other way apart from cosmological observations . an alternative ( and to us , more appealing ) scenario is one where long lifetimes are obtained by considering nearly degenerate ldp and nldp masses . in this case , decay rates become small , without suppressed couplings , simply because the phase space allowed in the decay process gets sharply reduced . sizeable couplings imply that , in the early universe , the ldp and nldp efficiently transform into each other through scattering from background particles . the small mass splitting , in turn , guarantees that the thermal - equilibrium number densities of the two species are comparable . to describe the process of decoupling and find the thermal relic abundance of these species , the number densities of the two have to be traced simultaneously , with the nldp , being charged , playing the major role . this phenomenon is usually dubbed as _ coannihilation _ @xcite and has been studied at length , being ubiquitous in many frameworks embedding thermal - relic candidates for dark matter , including common susy schemes . we will show below that sufficiently long lifetimes may be indeed obtained in the minimal supersymmetric standard model ( mssm ) , when the role of the nldp is played by a stau nearly degenerate in mass with the lightest neutralino ( with the former being the stable lsp and the thermal - relic cdm candidate we will focus on in the remainder of our discussion ) . a setup of this kind appears naturally , e.g. , in minimal supergravity ( msugra ) @xcite . this is the susy framework with the smallest possible parameter space , defined by only four continuous entries plus one sign , and hence also one of the most severely constrained by the requirement that the neutralino relic density matches the value from cosmological observations . neutralino - stau coannihilations determine one of the allowed regions , on the border with the region where the stau , which in the msugra scheme is most often the lightest scalar susy particle , becomes lighter than the neutralino . although a stau - neutralino mass degeneracy is not `` generic '' in such models , this scenario is economical in that the requirements of a long nldp lifetime and of comparable ldp and nldp relic abundances are both consequences of the mass degeneracy . in this sense , evidence for a running spectral index or any of the other observational features we discuss would simply help us sort out which configuration ( if any ) nature has chosen for susy dark matter . we refer to a mssm setup in which the lightest neutralino @xmath80 is the lightest susy particle . the charged particles that could play the role of the nldp include : ( 1 ) scalar quarks , ( 2 ) scalar charged leptons , and ( 3 ) charginos . we now discriminate among these cases by the number of particles in the final states for the decay of nlsps to neutralino lsps . scalar quarks and leptons have as their dominant decay mode a prompt two - body final state ; i.e. , @xmath81 , where we have labeled @xmath82 the susy scalar partner of the standard - model fermion @xmath83 . a typical decay width for this process is @xmath84(1 ) gev , corresponding to a lifetime @xmath84@xmath85 s. this holds whenever this final state is kinematically allowed ; i.e. , if @xmath86 . if it is _ kinematically forbidden _ , there are two possibilities : squarks may either decay through ckm - suppressed flavor - changing processes or through four - body decays . for instance , the stop decay may proceed through @xmath87 or @xmath88 . on the other hand , within the same minimal flavor - violation framework , scalar leptons are not allowed to decay in flavor - changing two - body final states , and only the four - body decay option remains . the case for the chargino is different because this nlsp decay has a three - body final state , either with two quarks bound in a meson state i.e. , @xmath89 or with a leptonic three body channel i.e. , @xmath90 . the latter final state becomes dominant , in particular , for electron - type leptons , @xmath91 , if the mass splitting between nlsp and lsp becomes small . we have listed all decay topologies as these are especially relevant when discussing the limit in which we force a reduction of the allowed decay phase - space volume ; i.e. , the limit in which the nlsp and lsp are quasi - degenerate in mass . here we can also safely assume that the masses of the final - state particles , apart from the neutralino , are much smaller than the mass of the decaying particle . we can then consider the limit of a particle of mass @xmath92 decaying into a @xmath80 and @xmath93 massless final states , and derive an analytical approximation to the behavior of the final - state phase space @xmath94 and of the decay width @xmath95 as functions of @xmath96 . in the case of two - body decays , the phase space reads @xmath97 on the other hand , a recursive relation between @xmath98 and @xmath99 based on the invariant mass of couples of final states yields @xmath100 the dependence of the decay width @xmath95 on @xmath96 must , however , take into account not only the phase - space dependence , but also the behavior of the amplitude squared @xmath101 of the processes as a function of @xmath102 . the occurrence of a massless final state yields , in the amplitude squared , a factor that scales linearly with the momenta circulating in the feynman diagram . one therefore has the further factor @xmath103 finally , we have @xmath104 i.e. , the lifetime to decay to a two - body final state scales like @xmath105 , while for a four - body decay we have @xmath106 . reducing the nlsp mass splitting @xmath96 one may hope to obtain `` cosmologically relevant '' nlsp lifetimes . for scalar quarks , this is not quite the case , as two - body final states always dominate , and even for amplitudes that are ckm - suppressed , the scaling with @xmath96 is too shallow , and the resulting lifetimes are rather short , even for very small mass splittings . the situation is slightly more favorable for charginos , whose three - body decay width is approximately equal to @xmath107 at tree level and in the limit of pure higgsino - like or wino - like states , the lightest neutralino and lightest chargino are perfectly degenerate in mass . however , when one takes into account loop corrections to the masses , @xmath96 usually turns out to be larger than a few tens of mev ; see , e.g. , @xcite . this translates into an absolute upper limit on the chargino lifetime of about @xmath108 s. finally , in the case of sleptons @xmath109 , the interesting regime is when @xmath110 and only four - body decays are allowed . for example , for the lightest stau @xmath111 , the processes @xmath112 with @xmath113 where , depending on @xmath96 , only final states above the kinematic threshold are included . in this case , @xmath96 can be safely taken as a free parameter of the theory : in most scenarios the gaugino mass parameter setting the neutralino mass for bino - like neutralinos , and the scalar soft mass parameter setting the stau mass , are usually assumed to be independent ; analogously , there are models in which the @xmath114 parameter setting the mass for a higgsino - like neutralino , and scalar soft mass parameter are unrelated . a similar picture applies to the smuon , though lifetimes start to be enhanced at much smaller mass splittings ( @xmath115 instead of @xmath116 ) , and it is theoretically difficult to figure out a scenario in which the lightest smuon is lighter than the lightest stau . the scalings we have sketched are summarized in fig . [ fig : lifetimes ] , where we plot lifetimes for a stop ( ckm - suppressed ) , a chargino , and a stau as a function of @xmath96 . the decaying - particle masses have been set to 1 tev , and the @xmath96 range is between the electron and the tau mass . the lifetimes of the stop and of the chargino have been computed with sdecay @xcite . some details on the computation of the stau lifetime are given below . notice that the scaling of eq . ( [ eq : scaling ] ) is accurately reproduced for all cases . the bottom line is that indeed _ the lightest stau can play the role of the nldp with a cosmologically relevant lifetime _ , and that the stau is the only particle in the mssm for which this can be guaranteed by adjusting only the ldp - nldp mass splitting . the next step is to determine the relic density of the nlsp and lsp in the early stages of the evolution of the universe . as we have already mentioned , we are going to consider thermal production . we briefly review here how to compute the evolution of number densities with coannihilation @xcite . consider a setup with @xmath117 supersymmetric particles @xmath118 , @xmath119 , ... @xmath120 , each with mass @xmath121 and number of internal degrees of freedom @xmath122 . the ordering is such that @xmath123 . in the evolution equations , the processes that change the number density of susy particles @xmath124 are of three kinds : @xmath125 where @xmath126 , @xmath127 , @xmath128 , and @xmath129 are ( sets of ) standard - model ( sm ) particles . in practice , the relevant processes one should include are those for sm particles that are in thermal equilibrium . assuming the distribution function for each particle @xmath130 is the same as for the equilibrium distribution function , @xmath131 and invoking the principle of detailed balance , the boltzmann equation for the evolution of the number density of susy particle @xmath124 , @xmath132 , normalized to the entropy density of the universe , @xmath133 , as a function of the variable @xmath134 ( with @xmath135 the universe temperature ; this is equivalent to describing the evolution in time ) is given by @xmath136}{h } \left(\frac{y_i}{y_i^{\rm eq}}-\frac{y_j}{y_j^{\rm eq}}\right ) \\ \nonumber & & + \sum_{j > i } \frac{\gamma_{j \rightarrow i}}{h } \left(\frac{y_j}{y_i^{\rm eq}}-\frac{y_i \ , y_j^{\rm eq}}{(y_i^{\rm eq})^2}\right ) - \sum_{j < i } \frac{\gamma_{i \rightarrow j}}{h } \left(\frac{y_i}{y_i^{\rm eq}}-\frac{y_j}{y_j^{\rm eq}}\right ) \;. \label{eq : b3}\end{aligned}\ ] ] in this equation , analogously to @xmath137 , we have defined @xmath138 , the ratio of the equilibrium number density of species @xmath130 ( at temperature @xmath139 ) to the entropy density . on the left - hand side , we introduced @xmath140^{-1}$ ] with @xmath141 being the effective degrees of freedom in the entropy density . the function @xmath142 is close to 1 except for temperatures at which a background particle becomes nonrelativistic . on the right - hand side , the last two terms contain factors in @xmath143 that label the partial decay width of a particle @xmath130 in any final state containing the particle @xmath144 ; in our discussion they play a role just at late times when nlsps decay into lsps , giving the scaling we have used in eq . ( [ eqn : rhophi ] ) . the first two terms refer , respectively , to processes of the kind @xmath145 and @xmath146 in eq . ( [ eq : list ] ) , including all possible sm final and initial states . the symbol @xmath147 indicates a thermal average of the cross section @xmath148 ; i.e. , @xmath149 as is evident from the form we wrote the boltzmann equation , interaction rates have to be compared with the expansion rate @xmath150 of the universe . in general , over a large range of intermediate temperatures , the @xmath151 terms will be larger than the @xmath152 terms , since we expect the cross sections to be of the same order in the two cases , but the scattering rates will be more efficient as long as they involve light background particles @xmath30 with relativistic equilibrium densities @xmath153 that are much larger than the nonrelativistic maxwell - boltzmann suppressed equilibrium densities @xmath154 for the more massive particles @xmath155 . this implies that collision processes go out of equilibrium at a smaller temperature , or later time , than pair - annihilation processes . writing explicitly the expression for @xmath156 , in which maxwell - boltzmann suppressed terms and terms in mass splitting over mass scale can be neglected , one finds explicitly that in the limit that collisional rates are much larger than the expansion rate , for any @xmath124 and @xmath130 , @xmath157 , or equivalently , @xmath158 with @xmath159 and @xmath160 . at the temperature @xmath161 , when @xmath162 , collision processes decouple and the relative number densities become frozen to about @xmath163 up to the time ( temperature ) at which heavier particles decay into lighter ones . the sum @xmath164 of the number densities has instead decoupled long before . ( [ eq : ratio ] ) is the relation that is implemented to find the usual boltzmann equation @xcite for the sum over number densities of all species compared to the sum of equilibrium number densities , and that shows that the decoupling for @xmath165 occurs when the total effective annihilation rate becomes smaller than the expansion rate , at a temperature @xmath166 that , as mentioned above , is much larger than @xmath161 . to get an estimate for @xmath161 , we can take , whenever a channel is kinematically allowed , the ( very ) rough s - wave limit , @xmath167 where an approximate relation between annihilation rate and relic density has been used @xcite . there are then two possibilities depending on whether _ ( i ) _ the background particle @xmath30 enforcing collisional equilibrium has a mass much larger than the mass splitting @xmath96 between the susy particles involved , or _ ( ii ) _ the opposite regime holds . in the first case we find that @xmath168 , roughly the temperature at which @xmath30 itself ( except for neutrinos ) decouples from equilibrium . from eq . ( [ eq : ratio ] ) we find that @xmath169 ; i.e. , they have comparable abundances . in the opposite case , we find instead that @xmath170 and hence @xmath171 $ ] ; i.e. , the abundance of the heavier particle is totally negligible compared to the lighter one . long - lived stau nlsps are kept in collisional equilibrium with neutralinos by scattering on background @xmath172 and emission of a photon . in this case we are clearly in the limit _ ( i ) _ , as @xmath173 . since the number of internal degrees of freedom for both staus and neutralinos is 2 , we find that the fraction of charged dark matter in this model is @xmath174 more carefully , though , this is not a strict prediction of our mssm setup for a stau nldp , and should be interpreted just as an upper limit . in fact , if we have other susy particles that are quasi - degenerate in mass with the lsp , and if they then coannihilate and decouple from the neutralino at a later time than the stau , either through mode _ ( i ) _ and then immediately decaying into neutralinos ( in all explicit examples , the decay into staus is strongly suppressed compared to the decay into neutralinos ) , or through mode _ ( ii ) _ , then @xmath175 is reduced to @xmath176 where the sum in the denominator involves the neutralino , the stau and all susy particles with @xmath161 lower than the @xmath161 for staus . in particular , if the lightest neutralino is a nearly - pure higgsino , then the next - to - lightest neutralino will also be a higgsino very nearly degenerate in mass , and the lightest chargino will also be nearly degenerate in mass , with mass splittings possibly smaller than @xmath177 . in this case , charginos and neutralinos will be kept in collisional equilibrium through scatterings on @xmath178 pairs . at the same time , the collisional decoupling of staus might be slightly delayed because of chargino - stau conversions through the emission of a photon and absorption of a tau neutrino ; however this second process has a yukawa suppression ( as we are considering higgsino - like charginos ) compared to the first , and hence it is still guaranteed that the stau decoupling temperature is larger than the chargino @xmath161 temperature . since the mass splitting between neutralino and chargino and that between lightest neutralino and next - to - lightest neutralino can not be smaller than few tens of mev ( due to loop corrections to the masses ) , decoupling will always happen in mode _ ( ii ) _ and we do not have to worry about possible stau production in their decays . since @xmath179 , applying the formula in eq . ( [ eq : smallerfrac ] ) we find @xmath180 . in the case of a wino - like lightest neutralino , instead , the only extra coannihilating partner would be the lightest chargino , yielding @xmath181 . finally , adding to this picture , e.g. , a quasi - degenerate smuon and selectron , @xmath182 could be obtained . we provide a few examples of well motivated theoretical scenarios where neutralino - stau degeneracy occurs , possibly in connection with further coannihilating partners driving low values of @xmath175 . in surveying the possible models , the criterion we take here is that of the _ composition _ of the lightest neutralino in terms of its dominating gauge eigenstate components . we thus outline _ bino- _ , _ higgsino- _ and _ wino-_like lightest - neutralino benchmark scenarios . as we mentioned above , in the framework of minimal supergravity ( msugra ) @xcite one of the few cosmologically allowed regions of parameter space is the tail where the neutralino and the stau are quasi - degenerate . in this case , coannihilations reduce the exceedingly large bino - like neutralino relic abundance to cosmologically acceptable values for neutralino masses up to around 600 gev . coannihilation effects depend on the relative mass splitting between the two coannihilating species . requiring a mass splitting as small as those found above amounts , as far as the neutralino relic density is concerned , to effectively setting @xmath184 . this , in turns , sets the mass of the neutralino - stau system once a particular value of the relic abundance is required . we plot in fig . [ fig : ps ] points fulfilling at once @xmath184 and @xmath185 , the latter being the central value as determined from the analysis of cmb data @xcite . result are shown in the ( @xmath186-@xmath187 ) plane , with @xmath187 the ratio of the vacuum expectation values of the two neutral components of the @xmath5 higgs doublets , and at a fixed value of the trilinear coupling @xmath188 ( this latter quantity is , however , not crucial here ) . the solid line corresponds to negative values of the higgsino mass parameter @xmath114 , while the dashed line corresponds to positive values of @xmath114 . notice that at @xmath189 the accidental overlap of the heavy higgs resonance with the coannihilation strip , around @xmath190 , shifts the neutralino masses to larger values . we point out that the two requirements of mass degeneracy and of the correct relic density determine , at a given value of @xmath187 , the required mass of the neutralino - stau system , thus solving the residual msugra parameter space degeneracy , and making the present framework testable and predictive . in the minimal configuration we have considered , only the lightest neutralino and the lightest stau are playing a role , hence @xmath183 . however , since the mass splitting between the lightest stau and lightest smuon and selectron is rather small , assuming a slight departure from universality in the scalar sector , two additional quasi - degenerate scalar particles can be obtained and the fraction of charged dark matter reduced to @xmath191 . when the @xmath114 term is lighter than the gaugino masses @xmath193 and @xmath194 , the lightest neutralino gets dominated by the higgsino component . this situation occurs , again within the msugra model , in the so - called hyperbolic branch / focus point ( hb / fp ) region @xcite , where large values of the common soft breaking scalar mass @xmath195 drive @xmath114 to low values . in this region , scalars are naturally heavy , at least in the minimal setup ; however , the occurrence of non - universalities in the scalar sector @xcite may significantly affect the sfermion mass pattern . in particular , in a susy - gut scenario , soft breaking sfermion masses get contributions from @xmath196-terms whenever the gut gauge group is spontaneously broken with a reduction of rank @xcite . light staus may naturally occur , for instance when the weak hypercharge @xmath196-term dominates and features negative values . in this case the hierarchy between diagonal entries in the soft supersymmetry - breaking scalar mass matrices is @xmath197 . the @xmath198 term may also be lowered in presence of additional @xmath196-terms originating from the breaking of further @xmath4 symmetries . the relic neutralino abundance is here fixed by the interplay of multiple chargino - stau - neutralino coannihilations . in the limit of pure higgsino , the dynamics of these processes fixes the value of @xmath114 yielding a given relic neutralino abundance . on the other hand , a mixing with the bino component along the borders of the hb / fp region may entail a larger spread in the allowed mass range , affecting the @xmath80 higgsino fraction . we sketch the situation in fig . [ fig : higgsino ] , where we resort , for computational ease , to a low - energy parameterization of the above outlined scenario . the smaller frame shows the points on the @xmath199 plane that produce the required amount of relic neutralinos . the larger frame reproduces the values of the chargino - neutralino and neutralino next - to - lightest - neutralino mass splitting ; the lines end in the pure - higgsino regime . suitable models , in the present framework , must also fulfill the mass splitting requirement @xmath200 . this enforces the allowed neutralino mass range , at @xmath201 , between 870 and 880 gev . had we lowered the value of @xmath187 , the corresponding @xmath186 range would only have shifted to masses just a few tens of gev lighter . as we have already mentioned , since we are dealing with a case with two neutralinos , a chargino and a stau quasi - degenerate in mass , we find @xmath192 . again , a smuon and a selectron can be added to this to shift the charged particle fraction to @xmath202 . a benchmark case where the lightest neutralino is wino - like is instead provided by the _ minimal anomaly mediated susy breaking _ ( mamsb ) scenario @xcite . in this framework , tachyonic sfermion masses are cured by postulating a common scalar mass term , @xmath195 . the lightest sfermion turns out to correspond to the lightest stau @xmath111 again . the latter , at suitably low @xmath195 values , may be degenerate in mass with the lightest neutralino . we performed a scan of the mamsb parameter space , requiring @xmath203 , and found that the correct relic abundance requires the neutralino masses to lie in the range @xmath204 gev , with the lower bound holding for small values of @xmath187 , while the higher for larger ones . the chargino - neutralino mass splitting is always below the @xmath8 mass , since @xmath80 in mamsb is always a very pure wino , and it has large masses . we finally remark that here , as in the case of higgsinos , the occurrence of stau coannihilations _ raises _ the neutralino relic abundance , contrary to the standard result with a bino - like lsp . since , in this case , we have one neutralino , a chargino and a stau quasi - degenerate in mass , we find @xmath191 . the @xmath111 four - body decay proceeds through diagrams of the types sketched in fig . [ fig : feyn ] . they come in three sets : those with @xmath205 exchange , those with @xmath206 exchange , and those with a sfermion exchange . however , since @xmath207 is much smaller than any supersymmetric - particle mass , the virtuality of all diagrams except those featuring a @xmath8 exchange ( diagrams ( a ) and ( b ) in the figure ) is extremely large . hence all diagrams but those with a @xmath8 exchange will be suppressed by a factor @xmath208 , and the interferences with the dominating diagrams by a factor @xmath209 . of the two diagrams with a @xmath8 exchange , however , the one with the @xmath206 exchange has a yukawa suppressed @xmath210 vertex , which gives a suppression , with respect to the @xmath205-exchange diagram , @xmath211 in the most favorable muonic final channel . notice that the chirality structure of the couplings entails that no interference between these two diagrams is present . diagrams with a charged higgs or a sfermion exchange are moreover further suppressed with respect to those with a @xmath205 exchange by a factor @xmath212 which , depending on the susy spectrum , can also be relevant . in this respect , a very good approximation to the resulting stau lifetime is obtained by considering only the first diagram , whose squared amplitude reads @xmath213,\end{aligned}\ ] ] with @xmath214 the left- and right - handed coupling of the @xmath111 in the @xmath215 vertex . the sum is extended over the final states of eq . ( [ eq : fs ] ) . for numerical purposes , the four - particles phase space is integrated with the use of the monte carlo routine rambo , with final - state finite - mass corrections for all final states . [ fig : staulife ] shows the stau lifetime for a sample of the supersymmetric scenarios outlined in the preceding sections . we fully account for threshold effects in the phase space , with the four contributions from electronic , muonic , and first- and second - generation quarks . the quark masses have been set to their central experimental values @xcite . for the case of higgsinos , we reproduce the two extreme regimes when the @xmath198 term is large ( no left - right mixing ) and when @xmath216 ( maximal left - right mixing ) . the differences in the lifetimes are traced back to overall mass effects and to the values of the @xmath217 and @xmath218 couplings ( for instance , @xmath219 in the wino case , while the opposite regime holds for the case of higgsinos and no lr - mixing ) . in any case , we conclude that lifetimes of the order of @xmath220 years are obtained with a mass splitting @xmath221 mev . unlike other charged long - lived nldp scenarios , the framework we outlined above has the merit of being , in principle , detectable at dark - matter detection and collider experiments . we show in fig . [ fig : si ] the spin - independent neutralino - proton scattering cross section for the msugra parameter - space points of fig . [ fig : ps ] , and for the higgsino and wino ( mamsb ) cases , as discussed in sec . [ sec : higgsino ] and [ sec : wino ] , together with the current exclusion limits from the cdms experiment @xcite , and the future reach of the xenon 1-ton facility @xcite . for binos , at @xmath222 most points lie within less than one order of magnitude with respect to the future projected sensitivity , making it conceivable that this scenario may be tested in next - generation facilities . for negative @xmath114 , destructive interference among the lightest- and heavy - higgs contributions lead instead to cancellations in @xmath223 . finally , higgsino and wino detection rates respectively lie one and two orders of magnitude below the future expected sensitivity . indirect - detection experiments look less promising , even for winos and higgsinos , which always feature uncomfortably large neutralino masses . we checked , for instance , that the expected muon flux from the sun , generated by neutralino pair annihilations , is at most @xmath224 muons per @xmath225 per year , far below the sensitivity of future neutrino telescopes like icecube @xcite . turning to collider experiments , considering as the searching tool the usual missing transverse - energy channels , dedicated studies have shown that the msugra coannihilation strip will be within lhc reach , mainly through in a mass range that extends up to @xmath226 gev along the coannihilation strip , quite independently of @xmath187 @xcite . concerning higgsinos and winos , instead , the relevant mass range we study here appears to be beyond standard lhc searches @xcite . even at high - energy colliders , the peculiar and distinctive feature of this scenario is however represented by the long - lived stau . the production of what have sometimes been dubbed long - lived _ charged massive particles _ ( champ s ) @xcite has been repeatedly addressed . exclusion limits were determined by the cdf collaboration @xcite , and the future reach of the lhc and of a future linear collider has been also assessed for this broad class of exotic new particles @xcite . the case of a stable stau has been , in particular , considered several times , since it occurs in the context of various susy - breaking models , like gauge - mediated susy - breaking ( gmsb ) scenarios @xcite ; see also , e.g. , ref . we remark that , contrary to gmsb , in the present framework the production of long lived staus at accelerators is expected to come along together with the production of neutralinos , thus making the two scenarios distinguishable , at least in principle . a long - lived stau would behave as a highly penetrating particle , appearing in the tracking and muon chambers of collider detectors with a small energy deposit in calorimeters . staus , depending on their velocities , would produce either highly - ionizing tracks in the low-@xmath22 regime , or , if quite relativistic , they would appear similar to energetic muons . in this latter case , one could look at excesses of dimuon or multi - lepton events as a result of superparticle production ; for example , considering the ratio @xmath227 . in case additional particles have masses close to the neutralino - stau system , the total production cross section of superparticles would be greatly enhanced . long - lived charginos may also give interesting accelerator signals @xcite . while the reach of the tevatron appears to be insufficient to probe the parameter space of the models we are considering here @xcite , the discovery of this kind of scenarios at the lhc , though challenging , looks quite conceivable , particularly in the case of binos . in fact , even in the less promising case in which the particle spectrum does not feature any particle close in mass with the @xmath228 system , the highly - ionizing track channel should cover a mass range widely overlapping that indicated in fig . [ fig : ps ] . on the other hand , excess dimuon events could provide an independent confirmation , although the 5-@xmath229 lhc reach for champ s in this channel alone has been assessed to lie around 300 gev @xcite . moreover , if additional coannihilating particles ( charginos , smuons , or selectrons ) are present , the discovery at the lhc would certainly look even more promising @xcite . a further recently proposed detection technique for long - lived staus is represented by trapping these particles into large water tanks placed outside the lhc detectors @xcite . following the results of ref . @xcite , a 10-kton water tank may be capable of trapping more than 10 staus per year , if the mass @xmath230 gev . the subsequent decays could then be studied in a background - free environment . more than twice as many sleptons would also get trapped in the lhc detectors , although in this case a study of the stau properties would look more challenging @xcite . we have examined a scenario in which a fraction @xmath6 of the cold - dark - matter component in the universe is generated in the decay of long - lived charged particles and shown that a scale - dependent ( or ` running ' ) spectral index is induced . the power spectrum on scales smaller than the horizon size when the age of the universe is equal to the charged - particle lifetime gets suppressed by a factor @xmath7 . such a feature might be singled out unambiguously by future measurements of the power spectrum of neutral hydrogen through the 21-cm line , obtaining direct information on the charged particle lifetime and @xmath6 . on the contrary , current and future tests for departures from a scale - invariant power spectrum based on lyman-@xmath1 data may fail to uniquely identify the scenario we propose . in fact , we have estimated the modifications to the non - linear power spectrum at the redshifts and wavenumbers currently probed by lyman-@xmath1 forest data , and shown that these resemble ( but are different in detail ) those in models with a constant running of the spectral index @xmath10 . we expect , based on this resemblance , models with @xmath6 in the range @xmath231 ( as predicted in some explicit models we have constructed ) to be compatible with current cosmological data for lifetimes in the range @xmath232 yr . we have also verified that constraints from the primordial light - element abundances and distortions to the cmb spectrum are not violated . from the particle - physics point of view , we have shown that the proposed scenario fits nicely in a picture in which the lightest neutralino in susy extensions of the standard model appears as the cold - dark - matter candidate , and a stau nearly degenerate in mass with the neutralino as the long - lived charged counterpart . a small mass splitting forces the stau to be quasi - stable , since the phase space allowed in the its decay process gets sharply reduced . at the same time , it implies that neutralino and stau are strongly linked in the process of thermal decoupling , with the charged species playing the major role . owing to these coannihilation effects , the current neutralino thermal relic abundance is compatible with the value inferred from cosmological observatations and , at early times , the stau thermal relic component is at the correct level . we have described several explicit realizations of this idea in minimal supersymmetric frameworks , including the minimal supergravity scenario , namely the supersymmetric extension of the standard model with smallest possible parameter space . we have pointed out that charged dark matter fraction from 1/2 to 1/7 ( or even lower ) can be obtained and that stau lifetimes larger than 1 yr are feasible . we have also shown that some of the models we have considered may be detected in future wimp direct searches , and discussed the prospects of testing the most dramatic feature of the model we propose , i.e. the production of long - lived staus at future high energy particle colliders . we thank r. r. caldwell and u. seljak for useful discussions . sp acknowledges fruitful help and elucidations from y. mambrini about the sdecay package . ks acknowledges the support of a canadian nserc postgraduate scholarship . this work was supported in part by nasa nag5 - 9821 and doe de - fg03 - 92-er40701 . p. de bernardis _ et al . _ , nature * 404 * , 955 ( 2000 ) ; a. lange _ et al . _ , d * 63 * , 042001 ( 2001 ) ; s. hanany _ et al . _ , * 545 * , l5 ( 2000 ) ; n. w. halverson _ et al . _ , j. 568 , 38 ( 2002 ) ; b. s. mason _ et al . _ , j. * 591 * , 540 ( 2003 ) ; a. benoit _ et al . _ , . astrophys . * 399 * , l2 ( 2003 ) . d. n. spergel _ et al . _ , astrophys . j. suppl . * 148 * , 175 ( 2003 ) . a. gould , b. t. draine , r. w. romani , and s. nussinov , phys . b * 238 * , 337 ( 1990 ) . g. d. starkman , a. gould , r. esmailzadeh , and s. dimopoulos , phys . d * 41 * , 3594 ( 1990 ) . e. d. carlson , m. e. machacek , and l. j. hall , astrophys . j. * 398 * , 43 ( 1992 ) . d. n. spergel and p. j. steinhardt , phys . * 84 * , 3760 ( 2000 ) . s. davidson , s. hannestad , and g. raffelt , jhep * 0005 * , 003 ( 2000 ) . s. l. dubovsky , d. s. gorbunov , and g. i. rubtsov , jetp lett . * 79 * , 1 ( 2004 ) [ pisma zh . fiz . * 79 * , 3 ( 2004 ) ] . k. sigurdson , m. doran , a. kurylov , r. r. caldwell , and m. kamionkowski , phys . d * 70 * , 083501 ( 2004 ) . m. s. turner , phys . * 197 * , 67 ( 1990 ) ; g. raffelt , phys . rept . * 198 * , 1 ( 1990 ) ; k. van bibber and l. rosenberg , phys . rept . * 325 * , 1 ( 2000 ) . g. jungman , m. kamionkowski , and k. griest , phys . rept . * 267 * , 195 ( 1996 ) . l. bergstrom , rept . phys . * 63 * , 793 ( 2000 ) . j. l. feng , a. rajaraman , and f. takayama , phys . * 91 * , 011302 ( 2003 ) . j. l. feng , s. f. su , and f. takayama , arxiv : hep - ph/0404198 . k. jedamzik , arxiv : astro - ph/0405583 . j. l. feng , a. rajaraman , and f. takayama , phys . rev . d * 68 * , 063504 ( 2003 ) . k. sigurdson and m. kamionkowski , phys . * 92 * , 171302 ( 2004 ) . u. seljak _ et al . _ , arxiv : astro - ph/0407372 . u. seljak , p. mcdonald , and a. makarov , mon . not . soc . * 342 * , l79 ( 2003 ) . m. tegmark _ et al . _ [ sdss collaboration ] , phys . d * 69 * , 103501 ( 2004 ) . a. loeb and m. zaldarriaga , phys . lett . * 92 * , 211301 ( 2004 ) . a. h. chamseddine , r. arnowitt , and p. nath , phys . lett . * 49 * , 970 ( 1982 ) ; r. barbieri , s. ferrara , and c. a. savoy , phys . b * 119 * , 343 ( 1982 ) ; l. j. hall , j. lykken , and s. weinberg , phys . d * 27 * , 2359 ( 1983 ) ; p. nath , r. arnowitt , and a. h. chamseddine , nucl . b * 227 * , 121 ( 1983 ) . l. randall and r. sundrum , nucl . b * 557 * , 79 ( 1999 ) ; g. f. giudice , m. a. luty , h. murayama , and r. rattazzi , jhep * 9812 * , 027 ( 1998 ) ; t. gherghetta , g. f. giudice , and j. d. wells , nucl . b * 559 * , 27 ( 1999 ) . c. p. ma and e. bertschinger , astrophys . j. * 455 * , 7 ( 1995 ) . u. seljak and m. zaldarriaga , astrophys . j. * 469 * , 437 ( 1996 ) . r. h. cyburt , j. r. ellis , b. d. fields , and k. a. olive , phys . d * 67 * , 103521 ( 2003 ) . k. jedamzik , phys . lett . * 84 * , 3248 ( 2000 ) . w. hu and j. silk , phys . lett . * 70 * , 2661 ( 1993 ) . j. fixsen , e. s. cheng , j. m. gales , j. c. mather , r. a. shafer , and e. l. wright , astrophys . j. * 473 * , 576 ( 1996 ) . r. e. smith _ et al . _ [ the virgo consortium collaboration ] , mon . not . * 341 * , 1311 ( 2003 ) . j. a. peacock and s. j. dodds , mon . not . soc . * 280 * , l19 ( 1996 ) . m. j. white and r. a. c. croft , astrophys . j. * 539 * , 497 ( 2000 ) . g. servant and t. m. p. tait , nucl . b * 650 * , 391 ( 2003 ) . k. griest and m. kamionkowski , phys . lett . * 64 * , 615 ( 1990 ) . k. griest and d. seckel , phys . d * 43 * , 3191 ( 1991 ) . j. edsjo and p. gondolo , phys . d * 56 * , 1879 ( 1997 ) . j. r. ellis , k. a. olive , y. santoso , and v. c. spanos , phys . b * 565 * , 176 ( 2003 ) . d. m. pierce , j. a. bagger , k. t. matchev , and r. j. zhang , nucl . b * 491 * , 3 ( 1997 ) . h. c. cheng , b. a. dobrescu , and k. t. matchev , nucl . b * 543 * , 47 ( 1999 ) . m. muhlleitner , a. djouadi , and y. mambrini , hep - ph/0311167 . k. l. chan , u. chattopadhyay , and p. nath , phys . d * 58 * , 096004 ( 1998 ) . j. l. feng , k. t. matchev , and t. moroi , phys . d * 61 * , 075005 ( 2000 ) . s. profumo , phys . d * 68 * , 015006 ( 2003 ) . m. drees , phys . b * 181 * , 279 ( 1986 ) . k. hagiwara _ et al . _ [ particle data group collaboration ] , phys . d * 66 * , 010001 ( 2002 ) . d. s. akerib _ et al . _ [ cdms collaboration ] , astro - ph/0405033 . e. aprile _ et al . _ , astro - ph/0207670 . j. edsj , internal amanda / icecube report , 2000 . h. baer , c. balazs , a. belyaev , t. krupovnickas , and x. tata , jhep * 0306 * , 054 ( 2003 ) . d. acosta _ et al . _ [ cdf collaboration ] , phys . rev . lett . * 90 * , 131801 ( 2003 ) . f. abe _ et al . _ [ cdf collaboration ] , phys . d * 46 * , 1889 ( 1992 ) . a. nisati , s. petrarca , and g. salvini , mod . a * 12 * , 2213 ( 1997 ) . g. mercadante , j. k. mizukoshi , and h. yamamoto , phys . d * 64 * , 015005 ( 2001 ) . g. f. giudice and r. rattazzi , phys . * 322 * , 419 ( 1999 ) . j. l. feng and t. moroi , phys . d * 58 * , 035001 ( 1998 ) . h. baer , p. g. mercadante , x. tata , and y. l. wang , phys . d * 62 * , 095007 ( 2000 ) . j. l. feng , t. moroi , l. randall , m. strassler , and s. f. su , phys . lett . * 83 * , 1731 ( 1999 ) . h. baer , j. k. mizukoshi , and x. tata , phys . b * 488 * , 367 ( 2000 ) . a. j. barr , c. g. lester , m. a. parker , b. c. allanach , and p. richardson , jhep * 0303 * , 045 ( 2003 ) . j. l. feng and b. t. smith , hep - ph/0409278 . |
the random walk to be considered takes place in the-spherical dual of the group , for a fixed finite dimensional irreducible representation of . the transition matrix comes from the three term recursion relation satisfied by a sequence of matrix valued orthogonal polynomials built up from the irreducible spherical functions of type of . one of the stochastic models is an urn model and the other is a young diagram model . | 1010.0720 | around 1770 d. bernoulli studied a model for the exchange of heat between two bodies . this model can also be seen as a description of the diffusion of a pair of incompressible gases between two containers . this model was independently analyzed by s. laplace around 1810 , see the references in @xcite . another model of similar characteristics was introduced by p. and t. ehrenfest in 1907 in connection with the controversies surrounding the work of l. boltzmann in the kinetic theory of gases dealing with reversibility and convergence to equilibrium . boltzmann had apparentlly deduced his h - theorem dictating convergence to equilibrium starting from the time reversible equations of newton . for a nice account of this see @xcite . both of these models are instances of discrete time markov chains with fairly explicit tridiagonal one - step transition probability matrices which are obtained by considering carefully the underlying stochastic mechanism that connects the state of the system at two consecutive values of time . the second model features two urns , i and ii , that share a total of n balls . the state of the system at time @xmath4 is the number of balls in urn i. each ball has a different label from the set @xmath5 . at time @xmath4 a number @xmath6 in the set @xmath5 is chosen with equal probabilities and the ball with this label is moved from the urn where it sits to the other urn . this gives the state of the system at time @xmath7 . writing down the one - step transition probability matrix is now a matter of counting carefully . while it had been possible to obtain interesting answers for these two models for quite some time , it is only much more recently that some very nice connections have been noticed between these models and some basic sets of discrete orthogonal polynomials , namely the krawtchouk and the dual hahn polynomials . moreover although there are many ways of arriving at these polynomials it is relevant to mention here that they can be realized as the spherical functions " for certain finite bihomogeneous spaces . a very good reference for this material is @xcite . we stress the remarkable fact that these two models of old vintage and clear physical significance can be solved in terms of the simplest of all hypergeometric functions , namely @xmath8 and @xmath9 . as many readers certainly know many of the classical special functions of mathematical physics , such as the legendre , the hermite and the laguerre polynomials , could have been obtained for the first time as spherical functions for certain symmetric spaces . a good basic reference here is @xcite . the way that things developed historically is , of course , completely different . the interplay between important physical problems and certain tools that arise naturally in group representation theory constitutes the theme of this paper . the situation described here is the reverse of what has been discussed above for the bernoulli - laplace and the ehrenfest models : we will go from group representation theory to some concrete models that might be of some physical interest . we will start from a matrix that is obtained from group representation theory and try to build a model that goes along with it . the models constructed here are certainly not the only possible ones . more natural ones might be lurking around . in a series of papers including @xcite one considers matrix valued spherical functions associated to a pair @xmath10 arriving at sequences of matrix valued polynomials of one real variable satisfying a three term recursion relation whose semi - infinite block tridiagonal matrix is stochastic , i.e. the entries are non - negative and the sum of the elements in any row is @xmath11 . this matrix depends on a number of free parameters that have a very definite group theoretical meaning . the important point is that the tools developed in the papers just mentioned allow one to give explicit expressions , in terms of some definite integrals , of all the entries of any power of the original matrix . this means that if one could think of a nice markov chain with this matrix as its one - step transition probability matrix one would have an explicit form for the entries of the @xmath4-step transition probability matrix . many readers will recognize that this is exactly what s. karlin and j. mcgregor , see @xcite , proposed as a way of exploiting orthogonal polynomials and the role they play in the spectral analysis of certain finite or semi - infinite tridiagonal matrices . the method advocated in @xcite starts with a so called birth - and - death process whose one - step tridiagonal transition matrix is easily constructed from the given model and one has to look for the corresponding spectral information : the eigenfunctions and the spectral measure . here we travel this road in the opposite direction in a more elaborate set - up . the relation between matrix valued orthogonal polynomials , block tridiagonal matrices and quasi - birth and death processes has been first exploited independently in @xcite as well as in later papers by these authors . we will consider several random walks whose configuration spaces are subsets of @xmath12 , the so call @xmath13-spherical dual of @xmath1 , and whose one - step transition matrices come from the stochastic matrix that appears in @xcite and @xcite , see also @xcite . the dual of @xmath1 is the set @xmath14 of all equivalence classes of finite dimensional irreducible representations of @xmath1 . these equivalence classes are parametrized by the @xmath7-tuples of integers @xmath15 subject to the conditions @xmath16 . if @xmath17 , the @xmath13-spherical dual of @xmath1 is the subset @xmath12 of @xmath14 of the representations of @xmath1 whose restriction to @xmath2 contains the representation @xmath13 . then it is well known , see @xcite , that @xmath12 corresponds to the set of all @xmath18 s as above that satisfy the extra constraints @xmath19 in other words @xmath12 can be visualized as the subset of all points @xmath18 of the integral lattice @xmath20 in the set @xmath21\times\cdots\times[k_{n-1},k_n]\times(-\infty , k_n].\ ] ] an example is given in the figure below . , @xmath22,@xmath23.,scaledwidth=40.0% ] we can now state more precisely the point of this paper : starting from the stochastic matrix @xmath24 that appears in @xcite and @xcite , we describe a random mechanism that gives rise to a markov chain whose state space is the subset of @xmath12 of all @xmath25 such that @xmath26 and @xmath27 ( @xmath28 ) , and whose one - step transition matrix coincides with the one we started from . the construction in @xcite and @xcite deals with the case of @xmath29 but in @xcite and @xcite this was extended to the case of @xmath30 . one step of the markov evolution will consist of two substeps taken in succesion . in the first substep one of the values of @xmath31 increases by one , subject to the constraints ( 1 ) . in the second substep one of the new values of our @xmath31 s decreases by one , again this is subject to the same constraints . thus from the configuration @xmath18 one could for instance go to @xmath32 or one could stay put at @xmath18 . we use the notation @xmath33 for the vector with its @xmath34th component equal to @xmath11 and all the others equal to @xmath35 . any state has a total of at most @xmath36 positions where it can move in one complete step of our process consisting of two simpler steps . keep in mind that the two succesive simpler steps can end up with our random walker in the initial state . we will analyze in detail the simpler substeps that constitute one full step of our process . this will take up most of the analysis in the next sections . we now describe the contents of the paper . in section [ esfericas ] we collect the necessary material to state and explain a three term recursion relation ( with matrix coefficients ) for a sequence of matrix valued orthogonal polynomials , built up from irreducible spherical functions of a fixed type associated to the pair @xmath37 . this should help the reader make the connection between @xcite and the present paper . in section [ factorizacion ] we construct a factorization of the stochastic matrix that define the three term recursion relation for the sequence of matrix valued orthogonal polynomials given in the previous section . this factorization into two stochastic matrices leads to the two substeps mentioned above . before starting the analysis of our general urn model in section [ urnmodelun ] for one of the substeps , we describe in detail in section [ urnmodelu3 ] an urn model for @xmath38 . the definition of the stochastic matrix @xmath24 alluded above , as well as its factorization make sense for any @xmath25 . to each configuration @xmath39 of @xmath4 integer numbers we associate its young diagram , a combinatorial object which has @xmath40 boxes in the first row , @xmath41 boxes in the second row , and so on down to the last row which has @xmath42 boxes . for example the young diagram associated to the configuration @xmath43 is @xmath44{}&\makebox[1.5 mm ] { } & \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c|c| } \makebox[1.5 mm ] { } & \makebox[1.5 mm ] { } & \makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c|c| } \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c| } \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \end{array}\ ] ] young diagrams and their relatives the young tableaux are very useful in representation theory . they provide a convenient way to describe the group representations of the symmetric and general linear groups and to study their properties . in particular young diagrams are in one - to - one correspondence with the irreducible representations of the symmetric group over the complex numbers and the irreducible polynomial representations of the general linear groups . they were introduced by alfred young in 1900 . they were then applied to the study of the symmetric group by georg frobenius in 1903 . their theory and applications were further developed by many mathematicians and there are numerous and interesting applications , beyond representation theory , in combinatorics and algebraic geometry . if we consider the subset all @xmath25 such that @xmath45 it is natural to represent such a state of our markov chain by its young diagram , see section [ young ] . then in the last two sections we describe a random mechanism based on young diagrams that gives rise to a random walk in the set of all young diagrams of @xmath46 rows and whose @xmath47 row has @xmath48 boxes @xmath49 , and whose transition matrix is @xmath50 , see . let @xmath51 be a locally compact unimodular group and let @xmath52 be a compact subgroup of @xmath51 . let @xmath53 denote the set of all equivalence classes of complex finite dimensional irreducible representations of @xmath52 ; for each @xmath54 , let @xmath55 denote the character of @xmath0 , @xmath56 the degree of @xmath0 , i.e. the dimension of any representation in the @xmath0 , and @xmath57 . we choose the haar measure @xmath58 on @xmath52 normalized by @xmath59 . we shall denote by @xmath60 a finite dimensional vector space over the field @xmath61 of complex numbers and by @xmath62 the space of all linear transformations of @xmath60 into @xmath60 . a _ spherical function _ @xmath63 on @xmath51 of type @xmath54 is a continuous function on @xmath51 with values in @xmath62 such that 1 . ( @xmath65= identity transformation ) . 2 . @xmath66 , for all @xmath67 . if @xmath68 is a spherical function of type @xmath0 then @xmath69 , for all @xmath70 , @xmath71 , and @xmath72 is a representation of @xmath52 such that any irreducible subrepresentation belongs to @xmath0 . spherical functions of type @xmath0 arise in a natural way upon considering representations of @xmath51 . if @xmath73 is a continuous representation of @xmath51 , say on a finite dimensional vector space @xmath74 , then @xmath75 is a projection of @xmath74 onto @xmath76 . the function @xmath77 defined by @xmath78 is a spherical function of type @xmath0 . in fact , if @xmath79 we have @xmath80 if the representation @xmath73 is irreducible then the associated spherical function @xmath63 is also irreducible . conversely , any irreducible spherical function on a compact group @xmath51 arises in this way from a finite dimensional irreducible representation of @xmath51 . the aim of this section is to collect the necessary material to state and explain a three term recursion relation for a sequence of matrix valued orthogonal polynomials , built up from irreducible spherical functions of the same type associated to the pair @xmath81 . the irreducible finite dimensional representations of @xmath82 are restriction of irreducible representations of @xmath1 , which are parameterized by @xmath83-tuples of integers @xmath84 such that @xmath85 . different representations of @xmath1 can restrict to the same representation of @xmath86 . in fact the representations @xmath18 and @xmath87 of @xmath1 restrict to the same representation of @xmath3 if and only if @xmath88 for all @xmath89 and some @xmath90 . the closed subgroup @xmath91 of @xmath51 is isomorphic to @xmath2 , hence its finite dimensional irreducible representations are parameterized by the @xmath4-tuples of integers @xmath92 subject to the conditions @xmath93 . let @xmath13 be an irreducible finite dimensional representation of @xmath2 . then @xmath13 is a subrepresentation of @xmath18 if and only if the coefficients @xmath94 satisfy the interlacing property @xmath95 moreover if @xmath13 is a subrepresentation of @xmath18 it appears only once . ( see @xcite ) . the representation space @xmath96 of @xmath13 is a subspace of the representation space @xmath97 of @xmath18 and it is also @xmath52-stable . in fact , if @xmath98 , @xmath99 and @xmath100 we have @xmath101 where @xmath102 and @xmath103 . this means that the representation of @xmath52 on @xmath96 obtained from @xmath18 by restriction is parameterized by @xmath104 let @xmath105 be the spherical function associated to the representation @xmath18 of @xmath51 and to the subrepresentation @xmath13 of @xmath52 . then says that the @xmath52-type of @xmath105 is @xmath106 . [ equivalencia ] the spherical functions @xmath107 and @xmath108 of the pair @xmath10 are equivalent if and only if @xmath109 and @xmath110 . the spherical functions @xmath107 and @xmath108 are equivalent if and only if @xmath18 and @xmath111 are equivalent and the @xmath52-types of both spherical functions are the same , see the discussion in p. 85 of @xcite . we know that @xmath112 if and only if @xmath113 besides , the @xmath52 types are the same if and only if @xmath114 therefore @xmath115 , and now it is easy to see that @xmath116 . the standard representation of @xmath1 on @xmath117 is irreducible and its highest weight is @xmath118 . similarly the representation of @xmath1 on the dual of @xmath117 is irreducible and its highest weight is @xmath119 . therefore we have that @xmath120 for any irreducible representation @xmath18 of @xmath1 the tensor product @xmath121 decomposes as a direct sum of @xmath1-irreducible representations in the following way @xmath122 and @xmath123 where @xmath124 is the cannonical basis of @xmath117 , see @xcite . the irreducible modules on the right hand side of and whose parameters @xmath125 do not satisfy the conditions @xmath126 have to be omitted . starting from and , the following theorem is proved in @xcite . [ multiplicationformula ] let @xmath127 and @xmath128 be , respectively , the one dimensional spherical functions associated to the standard representation of @xmath51 and its dual . then @xmath129 the constants @xmath130 and @xmath131 are given by @xmath132 moreover @xmath133 our lie group @xmath51 has the following polar decomposition @xmath134 , where the abelian subgroup a of g consists of all matrices of the form @xmath135 ( here @xmath136 denotes the identity matrix of size @xmath137 ) . since an irreducible spherical function @xmath63 of @xmath51 of type @xmath0 satisfies @xmath138 for all @xmath70 and @xmath71 , and @xmath139 is an irreducible representation of @xmath52 in the class @xmath0 , it follows that @xmath63 is determined by its restriction to @xmath140 and its @xmath52-type . hence , from now on , we shall consider its restriction to @xmath140 . let @xmath24 be the group consisting of all elements of the form @xmath141 thus @xmath24 is isomorphic to @xmath142 and its finite dimensional irreducible representations are parameterized by the @xmath143-tuples of integers @xmath144 such that @xmath145 . if @xmath146 , then @xmath147 commutes with @xmath148 for all @xmath149 . in fact we have @xmath150 the representation of @xmath2 in @xmath151 , @xmath152 restricted to @xmath142 decomposes as the following direct sum @xmath153 where the sum is over all the representations @xmath154 such that the coefficients of @xmath155 interlace the coefficients of @xmath13 , that is @xmath156 , for all @xmath157 . since each @xmath158 appears only once , by schur s lemma , it follows that @xmath159 , where @xmath160 for all @xmath146 . by using proposition [ equivalencia ] , given a spherical function @xmath107 we can assume that @xmath161 . in such a case the @xmath52-type of @xmath107 is @xmath13 , see . now it is easy to see that if @xmath162 is one of such a pair then @xmath163 where @xmath164 , @xmath165 and @xmath166 for @xmath167 . thus if we assume @xmath168 and @xmath169 for @xmath167 all the conditions are satisfied . we observe that the representations @xmath170 of @xmath24 appearing in the right hand side of are of the form @xmath171 , where @xmath172 and @xmath173 is in the following set @xmath174 in particular the number of @xmath24-modules in the decomposition of @xmath96 is @xmath175 we will identify @xmath147 with the column vector @xmath176 of @xmath177 complex valued functions @xmath178 indexed by @xmath179 , where @xmath180 , @xmath146 . from now on we fix @xmath181 and take @xmath182 as in for all @xmath168 and @xmath183 . also in the open subset @xmath184 of @xmath140 , we introduce the coordinate @xmath185 and define on the open interval @xmath186 the complex valued function @xmath187 and the corresponding matrix function @xmath188 for each @xmath189 we also define the following matrices of type @xmath190 @xmath191 where @xmath192 \\ ( c_w)_{{{\mathbf r}},\mathbf s } & = \begin{cases } a^2_{1}({{\mathbf m}}(w,{{\mathbf r}}))b^2_{n+1}({{\mathbf m}}(w,{{\mathbf r}})+{{\mathbf e}}_{1 } ) ) & \quad\text { if } \mathbf s={{\mathbf r}}\\ a^2_1({{\mathbf m}}(w,{{\mathbf r}}))b^2_{j+1}({{\mathbf m}}(w,{{\mathbf r}})+{{\mathbf e}}_1 ) & \quad \text { if } \mathbf s={{\mathbf r}}-{{\mathbf e}}_j\\ 0 & \quad \text { otherwise } \end{cases } \displaybreak[0 ] \\ ( b_w)_{{{\mathbf r}},\mathbf s } & = \begin{cases } \displaystyle \sum_{1\leq j\leq n+1}a^2_j({{\mathbf m}}(w,{{\mathbf r}}))b^2_j({{\mathbf m}}(w,{{\mathbf r}})+{{\mathbf e}}_j ) ) & \quad\text { if } \mathbf s={{\mathbf r}}\\ a^2_{j+1}({{\mathbf m}}(w,{{\mathbf r}}))b^2_{n+1}({{\mathbf m}}(w,{{\mathbf r}})+{{\mathbf e}}_{j+1 } ) & \quad \text { if } \mathbf s={{\mathbf r}}+{{\mathbf e}}_j\\ a^2_{n+1}({{\mathbf m}}(w,{{\mathbf r}}))b^2_{j+1}({{\mathbf m}}(w,{{\mathbf r}})+{{\mathbf e}}_{n+1 } ) & \quad\text { if } \mathbf s={{\mathbf r}}-{{\mathbf e}}_j\\ a^2_{j+1}({{\mathbf m}}(w,{{\mathbf r}}))b^2_{i+1}({{\mathbf m}}(w,{{\mathbf r}})+{{\mathbf e}}_{j+1 } ) & \quad\text { if } \mathbf s={{\mathbf r}}+{{\mathbf e}}_j-{{\mathbf e}}_i \\ 0 & \quad \text { otherwise } \end{cases } \end{aligned}\ ] ] where @xmath193 , and @xmath194 , @xmath195 for @xmath196 , see . [ multiplicationformula1 ] for each fixed @xmath52-type @xmath197 , for all integers @xmath198 and all @xmath199 we have @xmath200 this result is a consequence of theorem [ multiplicationformula ] and of the appropriate definitions of @xmath201 given in , when we take @xmath202 . we recall that @xmath203 and @xmath204 are the one dimensional spherical functions associated to the @xmath51-modules @xmath117 and @xmath205 , respectively . a direct computation gives @xmath206 and @xmath207 then @xmath208 . if @xmath209 let @xmath210 denote the @xmath211 left upper corner of @xmath212 , and let @xmath213 be the dense open subset of all @xmath214 such that @xmath210 is nonsingular . as in @xcite in order to determine all irreducible spherical functions of @xmath51 of type @xmath215 an auxiliary function @xmath216 is introduced . it is defined by @xmath217 where @xmath218 stands for the unique holomorphic representation of @xmath219 corresponding to the parameter @xmath13 . it turns out that if @xmath220 then @xmath221 where @xmath222 . then instead of looking at a general spherical function @xmath223 of type @xmath13 we look at the function @xmath224 which is well defined on @xmath213 . as before we construct the matrix function @xmath225 where @xmath226 , @xmath227 . let @xmath228 be the transpose of @xmath229 , i.e. @xmath230 . in @xcite the following crucial theorem is proved . if @xmath27 , then @xmath231 , @xmath232 and @xmath233 are polynomial functions on the variable @xmath234 whose degrees are @xmath235 it is important to point out that @xmath236 is a sequence of matrix orthogonal polynomials with respect to a matrix weight function @xmath237 supported in the interval @xmath186 and given in @xcite . from it easily follows that @xmath238 satisfies the following three term recursion relation @xmath239 the above three term recursion relation which hold for all @xmath240 can be written in the following way @xmath241 now we observe that the semi - infinite matrix @xmath24 on the right hand side is a stochastic matrix , i.e. all the entries are nonnegative and the sum of the elements in any row is one . in fact , the elements in the @xmath242 row of the @xmath243 blocks are either zero or @xmath244 , @xmath245 , @xmath246 which are given in . their sum is @xmath247\\ & + \sum_{j=2}^n a^2_{j}({{\mathbf m}})b^2_{1}({{\mathbf m}}+{{\mathbf e}}_{j } ) + \sum_{j=1}^{n+1}a^2_j({{\mathbf m}})b^2_j({{\mathbf m}}+{{\mathbf e}}_j)\displaybreak[0]\\ & + \sum_{j=2}^n a^2_{j}({{\mathbf m}})b^2_{n+1}({{\mathbf m}}+{{\mathbf e}}_{j } ) + a^2_{n+1}({{\mathbf m}})\sum_{j=2}^n b^2_{j}({{\mathbf m}}+{{\mathbf e}}_{n+1})\displaybreak[0]\\ & + \sum_{2\le i\ne j\le n}a^2_{j}({{\mathbf m}})b^2_{i}({{\mathbf m}}+{{\mathbf e}}_{j } ) + a^2_{1}({{\mathbf m}})b^2_{n+1}({{\mathbf m}}+{{\mathbf e}}_{1})\displaybreak[0]\\ & + a^2_1({{\mathbf m}})\sum_{j=2}^n b^2_{j}({{\mathbf m}}+{{\mathbf e}}_1 ) , \end{aligned}\ ] ] where we replaced @xmath248 by @xmath18 . the right hand side can be rewritten to obtain @xmath249\\ & \hskip3.3cm=\sum_{j=1}^{n+1}a^2_{j}({{\mathbf m}})\sum_{i=1}^{n+1}b^2_{i}({{\mathbf m}}+{{\mathbf e}}_{j}).\end{aligned}\ ] ] now by using the assertion @xmath250 follows , proving that the semi - infinite matrix @xmath24 is stochastic . in what follows we will construct a factorization of the stochastic matrix @xmath24 appearing in into the product of two stochastic matrices of the form @xmath251 while the random process given by the matrix @xmath24 leaves invariant the set @xmath252 introduced below , see ( 28 ) , this is not true for its substeps going along with this factorization . this section deals with this complication in great detail . the multiplication formulas given in theorem [ multiplicationformula ] restricted to @xmath202 give @xmath253 we recall that we fixed @xmath13 with @xmath27 and we took @xmath182 as in . also making the change of variables @xmath254 we defined @xmath187 . now we make the following important observation @xmath255 introduce the following scalar functions @xmath256 and the matrix function @xmath257 then the first identity in becomes @xmath258 for each @xmath240 we define the following matrix of type @xmath190 @xmath259 where @xmath260 \\ ( y_w)_{{{\mathbf r}},\mathbf s } & = \begin{cases } a^2_{n+1}({{\mathbf m}}(w,{{\mathbf r } } ) ) & \quad\text { if } \mathbf s={{\mathbf r}},\\ a^2_{j+1}({{\mathbf m}}(w,{{\mathbf r } } ) ) & \quad \text { if } \mathbf s={{\mathbf r}}+{{\mathbf e}}_j,\\ 0 & \quad \text { otherwise}. \end{cases } \displaybreak[0 ] \end{aligned}\ ] ] now the set of scalar identities with @xmath261 can be written as a matrix identity in the following more convenient way @xmath262 for each @xmath240 we define the following matrix of type @xmath190 @xmath263 where @xmath264 \\ ( s_w)_{{{\mathbf r}},\mathbf s } & = \begin{cases } b^2_{n+1}({{\mathbf m}}(w,{{\mathbf r}})+\mathbf e_{n+1 } ) & \quad\text { if } \mathbf s={{\mathbf r}},\\ b^2_{j+1}({{\mathbf m}}(w,{{\mathbf r}})+\mathbf e_{n+1 } ) & \quad \text { if } \mathbf s={{\mathbf r}}-{{\mathbf e}}_j,\\ 0 & \quad \text { otherwise}. \end{cases } \displaybreak[0 ] \end{aligned}\ ] ] if we multiply by @xmath265 and use the second multiplication formula given in we obtain @xmath266 since we claim that @xmath267 indeed we have @xmath268 where we used . on the other hand @xmath269 and @xmath270 then follows easily . finally if we compare with in theorem [ multiplicationformula1 ] we obtain @xmath271 which is equivalent to the factorization . we end by checking that both matrices in the right hand side of are stochastic : @xmath272\\ \sum_{\mathbf s\in\omega}&(r_w)_{{{\mathbf r}},\mathbf s}+\sum_{\mathbf s\in\omega}(s_w)_{{{\mathbf r}},\mathbf s}\\ & \qquad = b_{1}^2({{\mathbf m}}(w,{{\mathbf r}})+\mathbf e_{n+1 } ) + b_{n+1}^2({{\mathbf m}}(w,{{\mathbf r}})+\mathbf e_{n+1})\\ & \hskip1.2cm+\sum_{1\leq j\leq n-1}b_{j+1}^2({{\mathbf m}}(w,{{\mathbf r}})+\mathbf e_{n+1})=1,\end{aligned}\ ] ] where we used that @xmath273 , see . now we want to consider the random walks associated to the probability matrices appearing in , @xmath274 @xmath275 let @xmath276 and @xmath277 denote , respectively , the polynomial functions @xmath278 and @xmath279 . then can be written as follows @xmath280 similarly gives @xmath281 we can now rewrite in matrix form , @xmath282 the state space of the random walks @xmath283 associated , respectively , to the stochastic matrices @xmath284 is the set @xmath285 , and @xmath286 is equal to the composition @xmath287 . we recall that the map @xmath288 defined in is an injection of @xmath285 into the @xmath13-spherical dual @xmath12 of @xmath1 , and its image is @xmath289 where @xmath290 . let us now consider the random walk @xmath291 associated to the stochastic matrix @xmath292 . below we display the entries of @xmath292 at the different sites of its @xmath293-row , @xmath294\ ] ] the appearance of the plus sign in the right hand side of makes it natural to consider instead the random walk @xmath295 obtained from @xmath291 by applying a shift by @xmath296 . thus , if the system is at state @xmath297 at time @xmath234 , then at time @xmath298 it can move in the following ways @xmath299\ ] ] because @xmath300 for @xmath301 , and @xmath302 . this is in accordance with the following formula derived by looking at the @xmath303-entry of , @xmath304 now it is worth to observe that @xmath295 does not leave invariant the subset @xmath252 but extends to a random walk @xmath305 in @xmath12 defined by @xmath306 we proceed similarly with the random walk @xmath307 associated to the stochastic matrix @xmath308 . below we display the entries of @xmath308 at the different sites of its @xmath293-row , @xmath309\ ] ] the appearance of the plus sign in the left hand side of makes it natural to consider instead the random walk @xmath310 obtained from @xmath307 by applying a shift by @xmath311 . thus , if the system is at state @xmath297 at time @xmath234 , then at time @xmath298 it can move in the following ways @xmath312\ ] ] because @xmath313 for @xmath301 , and @xmath314 . this is in accordance with the following formula derived by looking at the @xmath303-entry of , @xmath315 then @xmath310 does not leave invariant the subset @xmath252 but extends to a random walk @xmath316 in @xmath12 defined by @xmath317 for @xmath318 . the transition matrices of @xmath319 and @xmath316 are , respectively , the following block bidiagonal matrices @xmath320 with @xmath321 \\ ( \tilde y_w)_{{{\mathbf m}},{{\mathbf n } } } & = \begin{cases } a^2_{n+1}({{\mathbf m } } ) & \quad\text { if } { { \mathbf n}}={{\mathbf m}},\\ a^2_{j+1}({{\mathbf m } } ) & \quad \text { if } { { \mathbf n}}={{\mathbf m}}+{{\mathbf e}}_j,\\ 0 & \quad \text { otherwise } , \end{cases } \displaybreak[0 ] \end{aligned}\ ] ] @xmath322 \\ ( \tilde s_w)_{{{\mathbf m}},{{\mathbf n } } } & = \begin{cases } b^2_{n+1}({{\mathbf m}}+\mathbf e_{n+1 } ) & \quad\text { if } { { \mathbf n}}={{\mathbf m}},\\ b^2_{j+1}({{\mathbf m}}+\mathbf e_{n+1 } ) & \quad \text { if } { { \mathbf n}}={{\mathbf r}}-{{\mathbf e}}_j,\\ 0 & \quad \text { otherwise}. \end{cases } \displaybreak[0 ] \end{aligned}\ ] ] where @xmath323 are such that @xmath324 , and @xmath301 . moreover , the stochastic matrix @xmath325 corresponding to the composition @xmath326 is equal to @xmath327 , and it is given by @xmath328 with @xmath329 \\ ( \tilde c_w)_{{{\mathbf m}},{{\mathbf n } } } & = \begin{cases } a^2_{1}({{\mathbf m}})b^2_{n+1}({{\mathbf m}}+{{\mathbf e}}_{1 } ) ) & \quad\text { if } { { \mathbf n}}={{\mathbf m}}\\ a^2_1({{\mathbf m}})b^2_{j+1}({{\mathbf m}}+{{\mathbf e}}_1 ) & \quad \text { if } { { \mathbf n}}={{\mathbf m}}-{{\mathbf e}}_j\\ 0 & \quad \text { otherwise } \end{cases } \displaybreak[0 ] \\ ( \tilde b_w)_{{{\mathbf m}},{{\mathbf n } } } & = \begin{cases } \displaystyle \sum_{1\leq j\leq n+1}a^2_j({{\mathbf m}})b^2_j({{\mathbf m}}+{{\mathbf e}}_j ) ) & \quad\text { if } { { \mathbf n}}={{\mathbf m}}\\ a^2_{j+1}({{\mathbf m}})b^2_{n+1}({{\mathbf m}}+{{\mathbf e}}_{j+1 } ) & \quad \text { if } { { \mathbf n}}={{\mathbf m}}+{{\mathbf e}}_j\\ a^2_{n+1}({{\mathbf m}})b^2_{j+1}({{\mathbf m}}+{{\mathbf e}}_{n+1 } ) & \quad\text { if } { { \mathbf n}}={{\mathbf m}}-{{\mathbf e}}_j\\ a^2_{j+1}({{\mathbf m}})b^2_{i+1}({{\mathbf m}}+{{\mathbf e}}_{j+1 } ) & \quad\text { if } { { \mathbf n}}={{\mathbf m}}+{{\mathbf e}}_j-{{\mathbf e}}_i \\ 0 & \quad \text { otherwise } \end{cases } \end{aligned}\ ] ] where @xmath323 are such that @xmath324 , and @xmath193 . the coefficients @xmath330 for @xmath331 are those defined in . if we identify @xmath285 with the subset @xmath252 , defined in , by @xmath332 , then clearly @xmath333 , because @xmath24 become a submatrix of @xmath325 . therefore @xmath334 to conclude , the analysis of the random walk @xmath286 associated to the stochastic matrix @xmath24 is simplified by looking at the decomposition @xmath335 instead of considering @xmath336 . we now give a concrete probabilistic mechanism that goes along with the random walk @xmath319 constructed in section [ factorizacion ] by group theoretical means , see . an entirely similar construction going with @xmath316 can be considered for the other substep of our process . this section is included for the benefit of the reader . it describes in detail , for the simple case of @xmath38 going along with the pair @xmath338 , a construction that will be given in general in section [ urnmodelun ] . a configuration , or state of our system , is now a triple of integers @xmath339 subject to the constrains @xmath340 with two fixed integers @xmath341 , see . we describe a stochastic mechanism whereby one of the three values of the @xmath31 is incresased by one with the following probabilities , see @xmath342 in the general scheme to be considered later this case corresponds to the value @xmath38 , and thus we start with two urns @xmath343 . in urn @xmath344 , @xmath345 , place @xmath346 balls of color @xmath347 and @xmath348 balls of color @xmath349 . these four colors are all different . notice that we could have no balls of colors @xmath350 or @xmath351 and that the total number of balls in urn @xmath344 is @xmath352 it will be useful to consider the following ordered set of urns @xmath353 in view of the notation to be introduced in the general case we denote these urns as @xmath354 we will introduce later on an order among certain collections of urns that will yield , in this particular case , @xmath355 now perform a total of three consecutive experiments . each experiment consists of drawing one ball at random ( i.e. with the uniform distribution ) from an urn in the ordered set of urns above , record the outcome as a letter in a word , and continue to the next experiment making sure to return the ball that has been drawn to its original urn after this experiment has been performed . the first experiment consists of picking one ball from urn @xmath356 this can give a ball of color @xmath357 or @xmath358 record the outcome @xmath357 or @xmath350 as the first letter in a word of three letters , and return the ball to its original urn , @xmath359 . the second experiment consists of picking one ball from urn @xmath360 this can result in a ball of color either @xmath361 or @xmath362 record the result as the second letter in a word that will have a total of three letters ( the colors of the balls chosen in experiments 1,2,3 ) , and return the ball to its original urn , @xmath363 . the last experiment consists of picking one ball from the union of the urns @xmath359 and @xmath363 , i.e urn @xmath364 the color of the ball in question i.e. @xmath365 or @xmath351 is the last letter in our word . this last ball drawn from @xmath366 is then returned to the urn @xmath367 or @xmath368 where it came from . there is a total of sixteen ( @xmath369 ) possible words that can arise in this fashion from an alphabet of four letters . these words constitute the set of all possible outcomes of the experiment made up of these three succesive and properly ordered ones . since we return the chosen ball at the end of each one of these experiments to its original urn , we have that the state of the system has not yet changed . this is about to happen now . we need a rule to decide which of the three values @xmath370 will be increased by one unit as the result of our experiment . to this end we break up the set of sixteen words into three disjoint and exhaustive sets . these sets will be denoted by @xmath371 , @xmath372 and @xmath373 and the sample space @xmath374 of cardinality @xmath375 is given by @xmath376 each set @xmath377 consisting of words with three letters will be obtained by a `` growth process '' starting from the sets we would have if we had considered the previous case , namely @xmath22 , when we have only one box and we were dealing with @xmath378 . in that case the sets are made up of words of one letter , either @xmath379 or @xmath380 to make the connection with the general case we will denote these sets in the case of one urn by @xmath381 and @xmath382 and the sample space by @xmath383 explicitly @xmath384 , @xmath385 . let us come back to the case @xmath38 . the class @xmath371 is formed by including all three letter words that start as those of @xmath381 and whose remaining two letters are such that the last one is not @xmath351 , i.e. either @xmath386 or @xmath387 thus @xmath388 the class @xmath372 is formed by including all three letter words that start as those of @xmath389 and whose remaining two letters are such that the first one is not @xmath351 . explicitly @xmath372 is @xmath390 notice that the meaning of the requirement `` not @xmath351 '' is quite different when it applies to the second urn @xmath363 as above , or to the third urn @xmath391 as in the previous case . finally @xmath392 is obtained by taking the union of all three letter words that start as in @xmath381 and have @xmath351 as their last letter , together with all words that start as in @xmath389 and have @xmath351 as the second letter . notice that @xmath392 is obtained by going over all the classes already built , @xmath371 and @xmath372 , and replacing the condition not @xmath351 by @xmath362 the class @xmath392 is thus made up of two sets of words , namely @xmath393 it takes almost no effort to see that all these @xmath394 words have been classified into three disjoint and exhaustive classes . now we compute the total probability of getting a result that belongs to each class . for the first class @xmath371 we have , @xmath395 for the second class @xmath372 we have that the probability is @xmath396 finally the total probability of the third class @xmath392 is , @xmath397 we are ready to give a rule for changing the state of the system in one unit of time . a result belonging to the subset @xmath377 , @xmath398 , will lead to a transition to a new state @xmath399 , where @xmath400 is increased by one . in terms of balls this will be achieved by removing from each urn containing a ball of color @xmath401 one of these balls , and adding to each urn containing a ball of color @xmath347 one ball of this color from the bath . when @xmath402 we do no removal . in this section we describe a random mechanism that gives rise to a markov chain whose one - step transition matrix is @xmath403 appearing in the factorization and where the matrices @xmath404 are defined in . a configuration is a set of @xmath7 values of the integers @xmath31 , @xmath331 , subject to the constrains @xmath405 where the integers @xmath406 remain unchanged throughout time . we will construct a stochastic process whereby in one unit of time one of the @xmath400 is increased by one with probability given by @xmath407 consider @xmath4 urns @xmath408 . in urn @xmath344 place @xmath346 balls of color @xmath347 and @xmath348 balls of color @xmath349 . we assume that the colors @xmath409 are all different . notice that in urn @xmath344 may be no ball of color @xmath349 , and that the total number of balls in @xmath344 is @xmath410 . consider the following ordered set of urns @xmath411 the union of urns is an urn whose content is the union of the set of balls in each urn in the union . observe that the total number of urns under consideration is @xmath412 . let @xmath413 clearly @xmath414 , and the set of all urns @xmath415 is ordered lexicographically according to : @xmath416 if @xmath417 or if @xmath418 and @xmath419 . we will perform a total of @xmath412 consecutive experiments . each experiment consists of drawing one ball at random ( i.e. with the uniform distribution ) from each urn in the ordered set of urns , record the outcome as a letter in a word , and continue to the next experiment making sure to return the ball to the original urn after this experiment has been performed . one should think of a complete experiment as consisting of these @xmath412 individual experiments . the transition from the present state of the system to the next one takes place after the complete experiment is carried out . the first experiment consists of picking one ball from urn @xmath359 , this can give a ball of color @xmath357 or @xmath350 . the result is recorded and the ball is put back in urn @xmath359 . the second experiment consists of picking one ball from urn @xmath363 , this can result in either a ball of color @xmath361 or @xmath351 . record the result as the second letter in a word that will have a total of @xmath412 letters . put the ball back in urn @xmath363 . keep on going by taking successively at random a ball from an urn @xmath420 and adding the letter corresponding to its color to the right of the word obtained in the previous step . the process finishes once a ball of the last urn @xmath421 is picked and a final word of @xmath412 letters is obtained . the alphabet is the set @xmath422 of @xmath423 letters . then the sample space @xmath424 consists of all words @xmath243 of @xmath412 letters that can be written with such an alphabet with the restriction that the letters allowed in the place @xmath425 correspond to the color of any ball in urn @xmath420 . the cardinality of the sample space is @xmath426 now by induction on @xmath427 we define a partition of @xmath424 into @xmath7 disjoint subsets @xmath428 for the benefit of the reader the construction will be spelled out in detail for small values of @xmath4 after we describe it in the general case and prove proposition 5.2 . we start with @xmath429 where @xmath430 then @xmath431 we make the following convention : the symbol @xmath432 in the @xmath425-place of a word stands for any color of a ball in urn @xmath420 different from @xmath349 , and the letter @xmath433 in the @xmath425-place of a word stands for any possible color of a ball in urn @xmath420 . if @xmath434 we set @xmath435 observe that the number of letters in the word @xmath436 to the right of the word @xmath437 is @xmath4 . similarly we define @xmath438 more generally for @xmath439 we let @xmath440 where the number of @xmath433 s to the right of @xmath441 is @xmath442 . the definition of @xmath443 is more interesting , namely @xmath444 let @xmath434 . then for @xmath439 we have @xmath445 @xmath446 @xmath447 is a partition of the sample space @xmath424 . the proof is by induction on @xmath448 . for @xmath22 we have @xmath449 thus the statement is true for @xmath22 . now assume that @xmath450 is a partition of @xmath451 for @xmath448 . if @xmath452 , then @xmath453 where @xmath454 for a unique @xmath6 . the @xmath433 in the @xmath6-place of the last @xmath4 letters is either @xmath455 or of the form @xmath441 . in the first case @xmath456 and in the second case @xmath457 . thus @xmath458 . at the same time we saw that @xmath457 for a unique @xmath318 . this completes the proof . the construction above is now made explicit for small values of @xmath4 . \1 ) @xmath38 . @xmath459 @xmath460 \2 ) @xmath461 . @xmath462 @xmath463 @xmath464 @xmath465 . @xmath466 @xmath467 @xmath468 @xmath469 @xmath470 [ stochastic ] the probability to obtain a word @xmath457 is @xmath471 for all @xmath318 . given @xmath162 let @xmath472 and @xmath473 . then from we get @xmath474 for all @xmath439 . this result allows us to prove the theorem by induction on @xmath427 . when @xmath22 we have only one urn @xmath367 with @xmath475 balls of color @xmath357 and @xmath476 balls of color @xmath350 . thus the probability to obtain a word in @xmath381 is @xmath477 where @xmath478 and @xmath479 . similarly the probability to obtain a word in @xmath389 is @xmath480 thus the theorem holds for @xmath22 . now assume that the theorem is true for @xmath448 . if @xmath439 we have @xmath440 where the number of @xmath433 s to the right of @xmath441 is @xmath442 . thus the probability to obtain a word @xmath457 is equal to @xmath481 times the probability to obtain the symbol @xmath441 from the urn @xmath482 . now we recall the composition of urn @xmath482 . by definition @xmath483 the total number of balls @xmath484 and the number of balls of color @xmath455 is @xmath485 . therefore the probability to obtain the symbol @xmath441 from urn @xmath482 is @xmath486 hence the probability to obtain a word @xmath457 is @xmath487 which establishes the theorem for all @xmath439 . since @xmath488 ( see ) and @xmath489 is a partition of @xmath424 it follows that the statement of the theorem is also true for @xmath490 . since we return the chosen ball at the end of each individual experiment to its original urn , we have that the state of the system has not yet changed . this is about to happen now . the outcome of a complete experiment produces a word that belongs to one of the subsets @xmath491 in the partition of the sample space @xmath424 . depending on which subset turns up we take a different action , thus obtaining a random walk in the space of configurations @xmath15 which satisfy the constraints @xmath492 imposed by the fixed @xmath4-tuple @xmath493 . this simple process will give for each configuration @xmath18 a total of at most @xmath7 possible nearest neighbours to which we can jump in one transition . a result belonging to the subset @xmath491 , @xmath494 , will lead to a transition to a new state @xmath399 , where @xmath400 is increased by one . in terms of balls this will be achieved by removing from each urn containing a ball of color @xmath401 one of these balls , and adding to each urn containing a ball of color @xmath347 one ball of this color from the bath . notice that all these transitions keep the values of @xmath495 unchanged and any transition that would violate the constrains does not occur because the corresponding probability @xmath496 vanishes . to each configuration @xmath497 we associate its young diagram which has @xmath40 boxes in the first row , @xmath498 boxes in the second row , and so on down to the last row which has @xmath499 boxes . @xmath500{}&\makebox[1.5 mm ] { } & \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5 mm ] { } \\ \hline \end{array}\\ \begin{array}{|c|c|c|c|c|c| } \makebox[1.5 mm ] { } & \makebox[1.5 mm ] { } & \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c|c|c| } \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c| } \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c| } \makebox[1.5mm]{}\\ \hline \end{array}\\ \end{array}\ ] ] we will construct a stochastic process whereby in one unit of time one of the @xmath31 is increased by one with probability @xmath501 see . as in section [ urnmodelun ] this will require running some auxiliary experiments . we start with the case @xmath22 . we perform the following experiment to decide if we will increase @xmath40 or @xmath41 : we choose to insert a box among one of the @xmath502 last boxes of the first row or to delete a box from the @xmath476 last boxes of the second row . an insertion can occur either to the left or to the right of one of the @xmath502 last boxes . we observe that there are @xmath475 possibilities of an insertion and @xmath476 possibilities of a deletion . all these are assigned the same probability . as an output of the experiment we get either a diagram with @xmath503 boxes in the first row , or a diagram with @xmath504 boxes in the second row . here we are implicitly assuming that @xmath505 . if @xmath498 were equal to @xmath41 we would get no young diagram . thus the sample space @xmath506 of our auxiliary experiment consists of two ( or one ) young diagrams which are obtained from the original one by adding one box to its first row or deleting one from its second row . let @xmath507 be the subset of @xmath506 consisting of the diagram with one more box in the first row , and let @xmath508 be the subset of @xmath506 consisting of the diagram with one less box in the second row ( or the empty set ) . then the probability to obtain a diagram in @xmath507 after the experiment is performed is @xmath509 similarly the probability to obtain a diagram in @xmath508 is @xmath510 as we wished . in the first case we go from the state @xmath162 to @xmath511 , and in the second case we go from the state @xmath162 to @xmath512 . now let us assume that @xmath38 . in this case we will perform three consecutive auxiliary experiments . the first experiment consists of inserting a box among one of the @xmath502 last boxes of the first row or of deleting a box from the @xmath476 last boxes of the second row . the second experiment consists of inserting a box among one of the @xmath513 last boxes of the third row or of deleting a box from the @xmath514 last boxes of the fourth row . finally the third experiment consists of inserting or deleting a box in one of the first four rows of the diagram as we did in the previous experiments ; odd rows go along with insertion and even rows with deletion . if @xmath505 and @xmath515 the complete experiment gives rise to a triple @xmath516 of young diagrams : @xmath517 is obtained from the original one by adding one box to its first row or by deleting one box from the second row , @xmath518 is obtained from the original one by adding one box to its third row or by deleting one box from the fourth row , and @xmath519 is obtained by adding one box to the first or to the third rows of the original diagram or by deleting one box from the second or the fourth rows . in what follows we use the following notation : @xmath520 denotes the young diagram corresponding to the original configuration @xmath162 and @xmath521 denotes , respectively , the diagram obtained from @xmath520 by adding or deleting one box to the @xmath6-row of @xmath520 , @xmath522 . observe that the sample space consists of all triples of young diagrams @xmath516 with @xmath523 , @xmath524 , and @xmath525 . @xmath526{}&\makebox[1.5 mm ] { } & \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{+}&\makebox[1.5mm]{}&\makebox[1.5 mm ] { } \\ \hline \end{array}\\ \begin{array}{|c|c|c|c|c|c| } \makebox[1.5 mm ] { } & \makebox[1.5 mm ] { } & \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c|c|c| } \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c| } \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c| } \makebox[1.5mm]{}\\ \hline \end{array}\\ \end{array}\ ] ] @xmath527{}&\makebox[1.5 mm ] { } & \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5 mm ] { } \\ \hline \end{array}\\ \begin{array}{|c|c|c|c|c| } \makebox[1.5 mm ] { } & \makebox[1.5 mm ] { } & \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c|c|c| } \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c|c|c| } \makebox[1.5mm]{}&\makebox[1.5mm]{}&\makebox[1.5mm]{}\\ \hline \end{array}\\ \begin{array}{|c| } \makebox[1.5mm]{}\\ \hline \end{array}\\ \end{array}\ ] ] thus our sample space @xmath374 has generically @xmath528 elements . the cardinality of @xmath374 can be smaller , for example if @xmath529 and @xmath530 , then @xmath531 . let us partition the sample space @xmath374 into the following three classes . @xmath532 we have @xmath533 , @xmath534 and @xmath535 . by simple inspection we see that @xmath374 is the disjoint union of @xmath536 and @xmath392 . then the probability to obtain a diagram in @xmath371 after a complete experiment is performed is @xmath537 similarly the probability to obtain a diagram in @xmath372 is @xmath538 finally the probability to obtain a diagram in @xmath392 is @xmath539 as desired . if @xmath529 and @xmath530 then @xmath540 , @xmath541 and @xmath542 . the probability to obtain a diagram in @xmath371 is @xmath543 the probability to obtain a diagram in @xmath372 is @xmath544 , and the probability to obtain a diagram in @xmath392 is @xmath545 as expected . now the state of our random walk is modified in one unit of time as follows : if the outcome of the complete experiment above belongs to @xmath377 , then we go from @xmath162 to @xmath546 , @xmath398 . in terms of diagrams we move from @xmath520 to @xmath547 , @xmath398 . given a young diagram @xmath520 corresponding to the original configuration @xmath162 , @xmath521 denotes , respectively , the diagram obtained from @xmath520 by adding or deleting one box to the @xmath6-row of @xmath520 , @xmath548 . the stochastic process we are going to construct will have a transition mechanism determined by first performing a sequence of auxiliary experiments @xmath549 to be described now . we start by considering the following set of consecutive pairs of rows of the diagram @xmath520 , @xmath550 the experiment @xmath549 , @xmath551 , consists of inserting at random a box in an odd row @xmath34 among the last @xmath552 last boxes of such a row , or deleting at random a box in an even row @xmath34 from the last @xmath553 last boxes of such a row . the row @xmath34 is also chosen at random in the set of consecutive rows @xmath554 the symbol @xmath557 in the place corresponding to the experiment @xmath549 of an @xmath412-tuple of diagrams , will stand for any possible outcome of @xmath549 except the diagram @xmath558 , respectively . while an @xmath559 in such a place stands for any outcome of @xmath549 . for example in the case @xmath38 considered before , see , we can write @xmath560 now we have a convenient notation to define inductively , for @xmath434 , a growth process similar to the one introduced in section [ urnmodelun ] , to break up the outcomes of the sample space @xmath424 into sets @xmath491 @xmath561 starting from the partition of @xmath451 into sets @xmath562 @xmath563 . let @xmath564 denote any @xmath4-tuple in the set @xmath562 , then we set @xmath565 observe that the number of diagrams in the @xmath566-tuple @xmath567 to the right of the @xmath412-tuple @xmath568 is @xmath4 . similarly we define @xmath569 more generally for @xmath439 we let @xmath570 where the number of @xmath559 s to the right of @xmath571 is @xmath442 . now by induction on @xmath434 it is easy to prove that @xmath447 is a partition of @xmath424 . also by induction on @xmath434 it is possible , as we did to established theorem [ stochastic ] , to prove the following main result . the outcome of a complete experiment produces an @xmath412-tuple of young diagrams that belongs to one of the partition subsets @xmath491 of the sample space @xmath424 . depending on which subset turns up we take a different action , thus obtaining a random walk in the space of configurations @xmath15 which satisfy the constraints @xmath574 imposed by the fixed @xmath4-tuple @xmath493 . this simple process will give for each configuration @xmath18 a total of at most @xmath7 possible nearest neighbours to which we can jump in one transition . f. a. grnbaum , _ random walks and orthogonal polynomials : some challenges , _ probability , geometry and integrable systems , mark pinsky and bjorn birnir editorsmsri publication vol 55 ( 2007 ) , 241 - 260 , see also arxiv math pr/0703375 . r. gangolli , v.s . varadarajan , _ harmonic analysis of spherical functions on real reductive groups , _ springer - verlag , berlin , new york , 1988 . series title : ergebnisse der mathematik und ihrer grenzgebeite * 101*. |
we observed two eclipses of the kepler-13a planetary system , on ut 2014 april 28 and ut 2014 october 13 , in the near - infrared using wide field camera 3 on the hubble space telescope . by using the nearby binary stars kepler-13bc as a reference , we were able to create a differential light curve for kepler-13a that had little of the systematics typically present in hst / wfc3 spectrophotometry . we measure a broadband ( 1.1 m to 1.65 m ) eclipse depth ofppm , and are able to measure the emission spectrum of the planet at with an average precision of 70ppm . our observations do not well sample either the eclipse ingress or egress , and so we are not able to provide meaningful constraints on the eclipse timing offset observed by . we do find that our observations , combined with those of give an average dayside brightness temperature of 3000k , and are consistent with a non - inverted , monotonically decreasing vertical temperature profile at 2.4 . we exclude an isothermal profile and an inverted profile . we also find that the dayside emission of kepler-13ab appears generally similar to an isolated m7 brown dwarf at a similar effective temperature . due to the relatively high mass and surface gravity of kepler-13ab , we suggest that the apparent lack of an inversion is due to cold - trap processes in the planet s atmosphere . using a toy - model for where cold - traps should inhibit inversions , and observations of other planets in this temperature range with measured emission spectra , we argue that with more detailed modeling and more observations we may be able to place useful constraints on the size of condensates on the daysides of hot jupiters . | 1612.06409 | the most immediately measurable property of an exoplanet s emission is its temperature . both broadband and spectroscopic observations are fundamentally measuring the brightness temperature of the atmosphere at a given wavelength , whether directly if the planet is actually imaged , or more commonly relative to the temperature of its host star when a planet goes into eclipse . how we relate the observed brightness temperature , as a function of wavelength , to the physical properties of an exoplanet s atmosphere is the crux of atmospheric characterization . unlike the majority of stars and brown dwarfs , the transformation between observed brightness temperature and physical properties for exoplanets is complicated by the fact that the dominant energy source in the atmosphere is usually external : irradiation from a host star . thus , while with a stellar spectrum one can usually assume that hotter temperatures imply light at those wavelengths emerges from deeper within the stellar atmosphere , for exoplanets this is not the case . for example , earth and the giant solar system planets all possess some sort of temperature inversion in their atmospheres , where the temperature begins to increase with height . for earth , this causes sharp emission features in the centers of the 9.5@xmath0 m o@xmath4 and the 15@xmath0 m co@xmath5 absorption bands . without an understanding of the temperature structure of the earth s atmosphere , both these features would be difficult to interpret . understanding what determines the vertical temperature structure of exoplanets , and hot jupiters in particular , has thus been one of the major observational and theoretical tasks of the last decade . based on early observations ( e.g. , * ? ? ? * ) and expectations based on the solar system planets , it was initially believed that all hot jupiters hotter than approximately 1800k should possess temperature inversions in their atmospheres ( e.g. , * ? ? ? generically , a temperature inversion requires a strong optical absorber that also increases the grayness of thermal opacities in the atmosphere and impedes cooling @xcite . observations of field brown dwarfs show clear signatures of gas - phase tio and vo in the optical both of which meet the necessary criteria and so these two molecules are believed to be the primary drivers of potential inversions in hot jupiters atmospheres @xcite . at the typical pressures of a hot jupiter s stratosphere , both tio and vo become gases at approximately 1800k . however , subsequent observations have revealed no clear evidence for temperature inversions in hot jupiters with daysides cooler than 2500k ( e.g. , * ? ? ? * ; * ? ? ? the lack of inversions in hot jupiters led to suggestions that uv radiation from the host stars may be photodisassociating tio / vo @xcite , or that tio / vo may be condensing and settling out of the atmosphere on the dayside @xcite or the night side @xcite . recently , @xcite suggested that gaseous tio / vo may still be present in hot jupiters atmospheres , but obscured by high altitude clouds . currently , there is no clear consensus to explain the lack of strong inversion signals in the surveyed hot jupiters . the observational effort to understand the temperature structure in hot jupiters atmosphere has been tremendously aided by the installation of wide field camera 3 ( wfc3 ) on the hubble space telescope ( hst ) . the spectral resolution and precision of the wfc3 observations mean that rather than having to infer the presence of spectral features as is necessary using broadband data , the wfc3 spectra show them . this capability allowed @xcite to present the first clear detection of a temperature inversion in the ultra - hot jupiter wasp-33b . the high average brightness temperature of wasp-33b s dayside ( 3300k ) led them to suggest that tio / vo driven inversions may only be present in extremely hot giant planets , which is supported by the apparently isothermal wfc3 emission spectra measured for wasp-103b s dayside ( 2850k , * ? ? ? * ) , and wasp-12b s dayside ( 2930k , * ? ? ? * ; * ? ? ? * ) . to further investigate the vertical temperature structure of ultra - hot jupiters , we therefore observed two secondary eclipses of kepler-13ab @xcite using hst / wfc3 . the kepler-13 system is composed of three stars : the planet host kepler-13a , and the unresolved binary kepler-13bc , with the two components separated by @xmath6 @xcite . kepler-13a and -13b are both nearly equal mass a - dwarfs , while kepler-13c is a fainter k - dwarf . the planet kepler-13ab has been observed in eclipse before , by @xcite , who measured the broadband emission spectrum in the kepler bandpass , the @xmath7-band , and the spitzer 3.6@xmath0 m and 4.5@xmath0 m bands . @xcite measured an average dayside brightness temperature of @xmath8k across all four bands , though the results of these observations did not clearly indicate a preferred temperature structure for kepler-13ab s dayside . we were also interested in kepler-13ab due to its relatively high mass of @xmath9@xmath10 @xcite and correspondingly high surface gravity of 3.3 times that of jupiter . recent observations of kelt-1b s emission spectrum by @xcite , which has a dayside temperature of 3150k and a surface gravity 22 times jupiter s , showed a non - inverted , monotonically decreasing temperature profile , which led those authors to suggest that surface gravity also plays a strong role in setting the vertical temperature of hot jupiters . the similar dayside temperature , but lower surface gravity , of kepler-13ab therefore presented itself as a relevant comparison object . we observed two eclipses of the kepler-13a ( koi-13 , bd+49 2629 ) planetary system , on ut 2014 april 28 and ut 2014 october 13 , using the infrared ( ir ) mode of wfc3 on the hubble space telescope . the two visits were each composed of five hst orbits . at the beginning of each visit we took a single direct image in the f126n filter to allow us to determine an initial wavelength solution before switching to the g141 grism ( 1.1@xmath0 m to 1.7@xmath0 m ) for the remainder of both visits . to decrease the image read - out times we used wfc3 s 256@xmath11256 pixel subarray mode . despite the relative brightness of kepler-13a ( @xmath12 and @xmath13 ) we observed both of the eclipses in staring mode , rather than spatial scan mode . this was due to the presence of the nearly equal brightness companion star system kepler-13bc @xmath6 @xcite , or approximately 8 pixels , away from the planet - host kepler-13a . the kepler-13bc system itself is unresolved in our direct and grism images , and we intended to minimize the blending of kepler-13a and kepler-13bc by not scanning during our exposures . we oriented the detector on the sky such that kepler-13a and kepler-13bc were close to being aligned along the detector s pixel columns , perpendicular to the dispersion direction of our spectra . for the april visit , kepler-13a was above kepler-13bc , while for the october visit the spacecraft s roll was reversed . we used the spars10 and nsamp=3 readout modes , which gave us an exposure time of 7.6 seconds . the first orbit of each visit took 100 grism exposures , while the subsequent four orbits each collected 101 individual grism exposures , for a total of 504 grism exposures per visit . we began our image calibration from the flt images provided by stsci . we first flat - fielded the images using the procedure outline in section 6.2 of the axe handbook @xcite . this necessitated an initial wavelength solution for our images , which we determined using the method described in section 2.2 , but here using the flt images . we then used this wavelength solution and the flat - field coefficients given in the g141 flat - field data cube provided by stsci to determine , and apply , a flat - field correction over all of our grism images . we next defined a bad - pixel mask by manually selecting bad pixels on the flat field image , and we corrected these bad pixels on the flat - fielded images by replacing their values with the median values of the eight pixels immediately around the bad pixels . ultimately , none of the bad pixels lay within our spectroscopic extraction aperture , rendering this step largely unnecessary . to identify and remove cosmic ray hits , we divided each image into two sections : the area around the stellar spectra , and the surrounding area dominated by the sky background . within the area around the stellar spectra , which we defined to be 140 pixels wide and 50 pixels tall and centered on the two spectral traces in the first grism image , we began our cosmic ray rejection by creating a model image by median combining the area in all of our flat - fielded and bad - pixel corrected images . we then divided each individual image by this model , and identified cosmic ray hits as anytime a pixel was more than 2.5 times higher than the median - combined model image . we replaced the pixels in the affected images with the value of that pixel in the model image . for the lower signal - to - noise background - dominated section of the image , we again created a median - combined model image from all of our flat - fielded and bad - pixel corrected images . instead of dividing each exposure by this model image , here we subtracted it after fitting for a scaling factor to account for the changing pedestal value of the background . taking the standard deviation of each median subtracted exposure , we identified cosmic ray hits as pixels that lay more than four standard deviations above zero . again , we replaced the pixels in the affected images with the value of that pixel in the model image . finally , we calculated and subtracted the background from each of our grism exposures . to do so , we defined two background regions across the bottom ( y - pixels 5 to 45 ) and top ( y - pixels 240 to 255 ) of each of the flat - fielded , bad - pixel and cosmic ray corrected images . using these two background stripes , we fit for the background in the central portion of each image assuming the background varied as a 2d plane , and subtracted this plane fit from each image . this process left us with flat - fielded , bad - pixel and cosmic ray corrected , background subtracted images with which we performed our spectral extraction . we used the direct image taken at the beginning of each of the visits to establish an initial wavelength solution for that visit . to do so , we used the daofind routine implemented in the photutils python package to determine the x- and y - pixel locations of kepler-13a and kepler-13bc on the detector subarray . we then used the wavelength calibration method described by @xcite with the adjusted wavelength coefficients determined by @xcite , to calculate a wavelength solution for each star . in the spectra of both kepler-13a and kepler-13bc the paschen@xmath14 line at 1.282@xmath0 m was clearly visible , and we verified the accuracy of our initial wavelength solution using this feature in the first grism exposure of each visit . we assumed that this initial wavelength solution was accurate for the first grism exposures in each of the visits , which were taken immediately after the direct image . for each subsequent grism exposure in a visit , we used the spectral alignment method described by @xcite to estimate a wavelength shift in the spectrum along the dispersion direction . briefly , @xcite s alignment method median combines all of the unshifted out - of - eclipse 1d spectra from a visit , and uses this as a reference spectrum against which one measures the wavelength shift of an individual exposure s spectrum . we did an initial extraction of all our 1d spectra using the extraction method described in section 3 , and thus computed a master spectrum for each of the two visits . we cross - correlated this master spectrum against each exposure in the visit using a range of shifts from -10 pixels to + 10 pixels . we linearly interpolated the master spectrum and fit for a normalization constant at each of the cross - correlation steps . note that we assumed that the dispersion of the wavelength solution did not change over the course of a visit . this procedure gave us wavelength shifts for each individual exposure relative to the median combined master spectrum . based on our assumption that the wavelength solution derived from the direct image is correct for the first grism exposure in a visit , we subtracted the wavelength shift we measured for this first exposure from our cross - correlated master shifts to determine the shift of each individual exposure relative to our initial wavelength solution . as described in the introduction , both kepler-13a and kepler-13b are mid- to late - a dwarfs of nearly equal brightness . @xcite identified a third star in the kepler-13 system , which they determined to be on a 65.831 day orbit about kepler-13b and which they estimated contributes about 1% of the combined system light . lcc @xmath15 ( k ) & @xmath16 & sh14 + @xmath17 & @xmath18 & sh14 + @xmath19}}$ ] & @xmath20 & sh14 + v@xmath21 ( km s@xmath22 ) & @xmath23 & jo14 + @xmath24 ( @xmath25 ) & @xmath26 & sh14 + @xmath27 ( @xmath28 ) & @xmath29 & sh14 + @xmath15 ( k ) & @xmath30 & sh14 + @xmath17 & @xmath18 & sh14 + @xmath19}}$ ] & @xmath20 & sh14 + v@xmath21 ( km s@xmath22 ) & @xmath31 & jo14 + @xmath24 ( @xmath25 ) & @xmath32 & sh14 + @xmath27 ( @xmath28 ) & @xmath33 & sh14 + @xmath34 ( days ) & @xmath35 & sa12 + @xmath36 & @xmath37 & sa12 + @xmath24 ( @xmath25 ) & 0.40@xmath25 0.75@xmath25 & jo14 [ tab : stellarprops ] initial spectroscopic characterization of kepler-13a and -13b by @xcite gave estimated effective temperatures for the two stars of 8200k to 8500k , masses around 2@xmath25 , and radii around 2.5@xmath28 . @xcite subsequently collected an independent set of spectral observations , and their analysis gave effective temperatures of around 7500k , masses near 1.7@xmath25 , and radii of about 1.7@xmath28 . concurrently with @xcite s measurements , @xcite performed a bulk re - analysis of all the kepler target stars , and found that kepler-13a was 9100k , with a mass of 2.5@xmath25 , and a radius of 3@xmath28 . the lack of agreement on the temperature , mass , and radius of kepler-13a has complicated previous studies of the planetary system ( e.g. , * ? ? ? to distinguish between these three sets of stellar parameters , we compared the stellar density implied by their masses and radii for kepler-13a to the stellar density measured for kepler-13a using the kepler transit light curve @xcite . the @xcite properties give a density of 0.17 g @xmath38 ( they gave no uncertainties on their mass and radius estimates ) , the @xcite properties give @xmath39 g @xmath38 , and the @xcite properties give a density of @xmath40 g @xmath38 . the kepler light curve , as measured by @xcite , gives a density of @xmath41 g @xmath38 . since the @xcite stellar properties are the only set that correctly reproduce the stellar density measured via the transit light curve , we adopted the @xcite stellar properties as correct , and do not consider either the @xcite or @xcite estimates in our reduction or analysis . we list the aggregate stellar properties for all three stars in table [ tab : stellarprops ] as mentioned previously , kepler-13a is separated from kepler-13bc by @xmath6 @xcite , or approximately 8 pixels on the wfc3/ir detector . although the similar spectral types and brightness of kepler-13a and kepler-13b led us to expect that there would be relatively little wavelength - dependent dilution in our eclipse observations , we still wished to rigorously account for the effect of the added light from the wings of kepler-13bc s spectrum . to do so , we used the wayne simulator @xcite to generate artificial 2d spectra of both stellar components , using the stellar properties listed in table [ tab : stellarprops ] . we then subtracted the simulation for one member of the stellar system ( e.g. , kepler-13bc ) from our observed images to create an undiluted 2d spectrum for the remaining member ( e.g. , kepler-13a ) . wayne is a simulator for hst / wfc3 spectroscopic images , observed with any of the two infrared grisms ( g102 and g141 ) in both observing modes ( staring and spatial scanning ) . it is able to simulate a number of detector characteristics , such as the read noise , the non - linearity effect , the dark current and the wavelength - dependent flat - field , as well as positioning issues , such as horizontal and vertical shifts and scan speed variations ( * ? ? ? * and references within ) . a key feature in wayne is the field - dependent structure of the spectrum , as it takes into account the changes in the spectrum trace and the wavelength solution when the spectrum moves on the detector @xcite . in addition , to simulate the wavelength - dependent psf , wayne uses a linear combination of a pair of 2d gaussian distributions , which results in point sources consistent with the psf ensquared energy fraction given in the wide field camera 3 instrument handbook @xcite . for our simulation of the kepler-13 system , we created two independent simulations , one for each component , and then combined them to a final data set . the position of the simulated spectra was based on the direct ( non - dispersed ) image of the target and the horizontal and vertical shifts that occurred during the observations . in this simulation we included the photon noise , the read noise , the wavelength - dependent flat - field and the sky background , to simulate the flt images . at the stage of the flt images , the remaining reduction steps have already been applied and for this reason we did not include them . in another set of simulations we did not combine the two companions , and also did not include any source of noise or detector characteristics , in order to use them as models for subtracting each companion from the original frames . as inputs for the simulations we used two synthetic spectra for kepler-13a and kepler-13bc , rotationally broadened to the measured @xmath42 of the stars ( table [ tab : stellarprops ] ) . to create each exposure , the simulator multiplies them by the sensitivity curve of the g141 grism to calculate the expected electron rate as a function of wavelength @xcite . furthermore , the spectra are scaled based on a combined model for the transit and the systematics or ramps ( @xmath43 , @xmath44 , @xmath45 in * ? ? ? * equation 8) . finally , the electrons per wavelength channel are distributed on the detector by randomly sampling the wavelength - dependent psf described above . however , since this scaling is only approximate , in order to subtract the simulated model spectra from the real ones we fit for a scaling factor between them . the parameters used for the models are a combination of the stellar parameters listed in table [ tab : stellarprops ] and the planetary parameters listed in table [ tab : broadbandpriors ] . we extracted light curves for both kepler-13a and kepler-13bc . we followed the same procedure for both light curves ; for clarity we will describe the process in the context of kepler-13a s light curve . for the kepler-13a light curve , we began with our subtracted images with the wayne simulation of kepler-13bc s light removed . we fit for the spectral trace of kepler-13a by fitting a gaussian profile to the spectrum along detector columns , allowing the baseline level , the profile width , and profile center to all be free parameters . we used the profile centers as the locations of the trace within each column , and fit a 9-degree polynomial to the resultant set of points to determine the y - pixel position of the trace as a function of x - pixel position . our measured trace positions , and the resulting polynomial fit , are not quite linear , as one would expect from the relations of @xcite . however the dispersion in our measured trace about a linear relation ( 0.085 pixels ) is consistent with the dispersion quoted in @xcite ( 0.08 pixels ) . 0.00 in -0.0 in we then summed along detector columns to generate a 1d spectrum for kepler-13a using an extraction aperture centered on this polynomial fit to the spectral trace . we used an extraction aperture with a half width of 4.5 pixels about the trace location to sum each of the columns . we chose this aperture size for two reasons . first , though we were confident that our simulated spectra of kepler-13bc is effectively removing the wings of that system s light , we were concerned that the core of the spectral emission was being imperfectly subtracted . we therefore did not wish to have our extraction aperture for kepler-13a extend too close to this imperfectly subtracted core . second , we tested several aperture sizes from 3.0 to 6.0 pixels to determine the optimum aperture size and determine whether the exact choice of aperture significantly affected our final results . we did this by extracting broadband light curves over the range of aperture sizes , and fitting the eclipse using the initial nelder - mead likelihood maximization step of our fitting procedure described in section 3.3 . an aperture size of 4.5 pixels results in a fit with the highest likelihood and the lowest scatter in its residuals . we therefore chose 4.5 pixels as the optimum extraction aperture size . with our extracted 1d spectra for each image , we used the wavelength solution for each image ( section 2.4 ) to extract a light curve within the fixed wavelength range of 1.125@xmath0 m to 1.65@xmath0 m . for the broadband data , we summed this entire wavelength range into a single point for each exposure , while for the spectrally - resolved data we subdivided this wavelength range into 15 bins evenly spaced in wavelength . as usual with hst / wfc3 grism observations , our raw extracted photometry displayed clear correlated noise ( top panel of figure [ detrendplot ] ) . as delineated by @xcite , these are typically a long - term temporal trend over the course of several orbits , an `` l''-shaped hook during the course of an individual orbit , and quick variations within an orbit due to the spacecraft s thermal breathing . all three of these affects were present in our broadband and spectrally - resolved photometry of kepler-13a and -13bc . at this point the presence of kepler-13bc greatly simplified our measurement of kepler-13ab s eclipse , by providing a nearly equal brightness and equal color comparison light curve with which to perform differential photometry . since kepler-13c contributes a negligible amount of light to the -13bc light curve , the broadband response of kepler-13bc to the hst / wfc3 systematics closely matched the response of kepler-13a . we initially attempted to create a differential light curve for kepler-13a by the straight division of the kepler-13bc light curve , similar to how we would treat this in ground - based photometry . to our initial surprise , this gave a differential light curve for kepler-13a that displayed a much higher scatter than the raw light curve , about 700ppm versus about 450ppm , even though the systematic trends seemed to be mostly gone . upon further consideration , we realized that this was because our raw hst / wfc3 photometry unlike ground - based photometry was dominated by photon , rather than systematic noise . thus , while dividing the kepler-13a by the -13bc light curve removed the systematic trends , it also added the photon noise scatter of the kepler-13bc in quadrature to the photon noise scatter in kepler-13a . we therefore applied a low - pass filter to the kepler-13bc light curve , to capture the systematic trends in that system s photometry while removing most of the photon noise . specifically , we used a second - order butterworth filter with a cutoff at 0.02 days . as shown by the red line in the middle panel of figure [ detrendplot ] , the thus filtered kepler-13bc light curve appeared to correctly capture the trends that we were interested in . we then generated our differential light curve for the white light kepler-13a data by dividing the raw kepler-13a photometry by this filtered -13bc light curve . as shown in the bottom panel of figure [ detrendplot ] , this removed most of the systematics and made the broadband eclipse clearly visible during both visits . another advantage of using kepler-13bc to make a differential light curve for kepler-13a is that we were able to make effective use of the first orbit of observations in both of our visits . typically the entire first orbit of a visit is discarded for precise eclipse or transit observations , due to the increased ramp and thermal breathing affects . in our case , it is a relatively simple matter to remove most of these affects using our differential comparison . the use of the first orbit s data was particularly helpful in the april visit , as without the first orbit we would not have had any pre - eclipse baseline observations . for the spectrally - resolved data we also created differential light curves but instead of differencing against kepler-13bc , we differenced against kepler-13a itself . we were motivated to do so out of a concern that the different placement of the two stars on the detector , plus their slightly different intrinsic spectra , would generate spurious spectral signatures if we differenced against kepler-13bc . we applied the same low - pass filter as we used in the broadband data to the broadband kepler-13a photometry , and divided each of our spectrally - resolved light curves by this filtered broadband photometry . in our spectrally - resolved data we therefore did not measure the absolute eclipse of kepler-13ab , but rather the change in eclipse depth as a function of wavelength . in both the broadband and spectrally - resolved differential light curves there was residual systematic noise . we chose to fit these residual correlations using a gaussian process ( gp ) regression model , as the traditional parametric methods for dealing with hst / wfc3 systematics were not applicable to our differential light curves . a gp regression models the observed data as random draws from a multivariate gaussian distribution about some mean function . as a result , gps are able to directly model the possible covariances between data points by populating the non - diagonal elements of the covariance matrix which defines the multivariate gaussian distribution . this is in contrast , for example , to a @xmath46 fitting process , which models the data as random draws from a univariate gaussian distribution and assumes no covariance between data points aside from what is inserted via detrending functions . for more detail , @xcite provide a thorough mathematical overview of gp methods . @xcite introduced them in the context of astronomical time series observations using archival nicmos observations of hd 189733 b , though they have a longer history in the general astronomical community ( e.g. , * ? ? ? recently , cartier et al . ( in prep . ) used a gp regression to model hst / wfc3 eclipse observations of wasp-103b , and we follow a similar approach . we defined our gp model using the notation from @xcite . for each visit , we have a vector of @xmath47 observed fluxes , @xmath48 , and times , @xmath49 . additionally , we recorded @xmath50 state parameters at each time @xmath51 with the state vector @xmath52 . we combined these state parameter vectors for each of our @xmath47 observations in the @xmath53 matrix , @xmath54 . the multivariate gaussian distribution underlying our gp model was defined by a combination of a mean function , which in our case is the physical eclipse model @xmath55 , and a covariance matrix @xmath56 . we used @xmath57 to denote the set of physical parameters describing the eclipse , and @xmath58 for the set of `` hyperparameters '' used to generate the covariance matrix from the @xmath54 state parameters . the joint probability distribution of our observed data @xmath59 was then @xmath60,\ ] ] where @xmath61 represents the multivariate gaussian distribution . our gp model thus depended upon on an eclipse model @xmath55 , and a generating function referred to as the covariance kernel for the covariance matrix @xmath56 . we generated our gp covariance matrices and calculated the gp likelihoods using the george python package @xcite . while we used the same eclipse model to fit both the broadband and spectrally - resolved datasets , we chose slightly different covariance kernels for each set of observations . we used a @xcite eclipse model as the mean function in our gp regression ; specifically the implementation in the batman python package @xcite . for our broadband data , we fit for the time of the secondary eclipse ( @xmath62 ) , the orbital period ( as @xmath63 $ ] ) , @xmath64 , @xmath65 , the cosine of the orbital inclination ( @xmath66 ) , the radius of the planet in stellar radii ( @xmath67 ) , the semi - major axis in units of the stellar radii ( as @xmath68 $ ] ) , and the depth of the secondary eclipse ( @xmath69 ) . this gave our broadband eclipse model eight parameters : @xmath70 lcl @xmath62&predicted @xmath62 ( @xmath71 ) & @xmath72 + @xmath34&orbital period ( days ) & @xmath73 + @xmath64 & & @xmath74 + @xmath65 & & @xmath75 + @xmath76&cosine of inclination & @xmath77 + @xmath78&radius ratio & @xmath79 + @xmath80&scaled semimajor axis & @xmath81 + @xmath82&planet mass ( @xmath10 ) & @xmath9 + @xmath83&planet radius ( @xmath84 ) & @xmath85 [ tab : broadbandpriors ] based on @xcite s measurements of the system , we have strong prior expectations for all of these parameters , except for the eclipse depth @xmath69 ( table [ tab : broadbandpriors ] ) . in our fitting , we implement each of these as gaussian priors on these seven parameters . we do not impose any prior on the eclipse depth , as we lack any observations of kepler-13ab s eclipse at these wavelengths , which implicitly imposes a uniform prior . for the spectrally - resolved observations the implementation of our eclipse model remains the same , but we only fit for the time of the secondary eclipse and the eclipse depth . all of the other parameters we fix to the values we determine from the broadband fit . this gives our spectrally resolved eclipse model only two parameters : @xmath86 again , we impose no prior on the eclipse depths , and we use the measured time of eclipse and its associated uncertainty from the broadband fit as a prior on @xmath62 for the spectrally - resolved fits . to model our broadband observations we used a linear combination of a squared - exponential kernel and a periodic kernel both as a function of time . a squared - exponential kernel is usually regarded as a generic choice for generating a gp covariance matrix , as it leads to smooth variations as a function of the generating variable . our intent was to use it to model the residual background temporal trends . we added on a periodic kernel , which is effectively a squared - exponential of a sine function , to model the repeatable covariances between points within an individual orbit . the point - wise covariances between the observations at times @xmath87 and @xmath88 , which collectively made up our broadband @xmath89 covariance matrix @xmath90 were then @xmath91 where @xmath92,\ ] ] and @xmath93/p)}{l_p^2 } \right].\ ] ] we used @xmath94 to denote the hyperparameters used to compute the covariances . these are covariance amplitudes , @xmath95 and @xmath96 , the covariance length scales @xmath97 and @xmath98 . we set the period of the periodic kernel to be @xmath99minutes . this was the measured orbital period of hst on the dates of our observations , based on archival two - line - elements provided to us by the united states strategic command s joint space operations center . for our spectrally resolved data , we used a slightly simplified covariance kernel with only a periodic component , @xmath100/p)}{l_p^2 } \right],\ ] ] where we have @xmath101 hyperparameters for the spectrally - resolved data , with the same definitions as in the broadband case . we imposed no priors on the members of either set of hyperparameters , other than to require that both @xmath97 and @xmath98 be greater than the eclipse s ingress and egress time of 0.01279 days . this ensured that the gp regression model did not treat the eclipse itself as correlated noise . 0.00 in ppm . this depth uncertainty is approximately 1.25 times what one expects from photon noise statistics.,title="fig : " ] -0.0 in to determine the best fits to the broadband and spectrally - resolved data we maximized a log - likelihood function that consisted of the gp model likelihood and the an additional term based on the priors in table [ tab : broadbandpriors ] . for the gp model the log - likelihood for a given set of parameters @xmath57 and @xmath58 was @xmath102 where @xmath103 is the vector of the residuals of our observed data ( @xmath59 ) from our eclipse model ( @xmath104 ) defined in section 3.1.1 . this log - likelihood follows directly from our definition of the gp model as a multivariate gaussian in equation ( [ eq:3110 ] ) . all of the physical parameters for which we applied a prior based on previous observations of the system we used gaussian priors , and we added the log - likelihood of these priors to our gp log - likelihood to compute the total log - likelihood for a given model as @xmath105 0.00 in with 14 degrees of freedom . this corresponds to a @xmath106 detection of variation . as a check , we also attempt to measure variation in kepler-13bc using the same fitting method . the differential depths for kepler-13bc are consistent with zero , with @xmath107 with 14 degrees of freedom . note that both of these spectra are differential measurements against the absolute broadband depth displayed by kepler-13a ( @xmath108ppm ) and kepler-13bc ( @xmath109ppm).,title="fig : " ] -0.0 in our fitting process was composed of two stages for both the broadband and the spectrally - resolved data : an initial nelder - mead maximization of equation ( [ eq:3220 ] ) , followed by an mcmc exploration of the likelihood space around this maximum to determine parameter uncertainties and to verify that we had identified the global likelihood maximum . to conduct the mcmc fitting we used the emcee python package @xcite . our mcmc runs consisted of a 500 step burn - in , followed by a 5000 step production run using 30 walkers . we initialized the walkers by scattering them about the initial nelder - mead maximum using random draws from gaussian distributions in each parameter with @xmath110 widths equal to corresponding prior widths . at the end of the production runs we calculated the gelman - rubin statistic and judged the mcmc to have converged if the statistics for all the parameters were less than 1.1 . we fit the two visits simultaneously using the same set of physical parameters and hyperparameters . we experimented with using differing hyperparameters for each of the visits , but we found that this did not change the results above our final @xmath110 uncertainties , nor did it substantially alter the uncertainties themselves . for the spectrally - resolved data we did allow for different hyperparameters between different spectral channels . 0.00 in -0.0 in tables [ tab : broadbandresults ] and [ tab : spectralresults ] list the results for the broadband eclipse , and the spectrally - resolved eclipses , respectively . our best fit to the broadband data gave a broadband eclipse depth of @xmath111ppm . the standard deviation in the residuals to the best fit model was 436ppm , which was nearly equal to the median per - point flux uncertainty of 400ppm from photon statistics . the uncertainty on the broadband measurement was 1.5 times what one would expect based on pure photon noise alone . the differential depths we measured in our spectrally - resolved light curves ( figure [ diffplot ] ) had a median uncertainty of 64ppm , which is 1.2 times the photon noise expectation . the median standard deviation of the residuals to the best fit models for the spectrally - resolved data was approximately 1250ppm . the differential eclipse depths showed statistically significant variation away from a flat line , with @xmath112 , or @xmath113 with 14 degrees of freedom . this corresponds to a @xmath114 detection of variation with wavelength . as a verification of our results we also performed the exact same fitting procedures , using the exact same priors , on broadband and spectrally - resolved light curves from kepler-13bc . as expected , the broadband kepler-13bc light curve gave an eclipse depth of @xmath115ppm . as shown in figure [ diffplot ] , the differential spectrally - resolved eclipse depths for kepler-13bc are also all consistent with zero , with @xmath107 with 14 degrees of freedom . we considered the results of our eclipse observations from two perspectives . first , we modeled our measured eclipse depths and those of @xcite using exoplanet atmosphere models . second we compared the dayside emission spectrum of kepler-13ab to isolated field brown dwarfs , since the relatively high mass of kepler-13ab ( 6.5@xmath10 ) and the young age of the system ( about 500 myr ) , means that the planet should have some residual luminosity . we modeled the measured eclipse depth of kepler-13ab using the hot jupiter atmospheric model described in @xcite and @xcite . this model is comprised of a 1d plane parallel atmosphere in hydrostatic equilibrium and local thermodynamic equilibrium ( lte ) . the emergent spectrum is computed using a 1d line - by - line radiative transfer solver in the planetary atmosphere . the atmospheric temperature - pressure ( tp ) profile and chemical composition are free parameters of the model , with 6 parameters for the tp profile and a parameter each for each chemical species included in the model . the range of molecules considered and the sources of opacity are discussed in @xcite . 0.00 in -0.0 in we find that our observations are relatively well reproduced by an atmospheric model with a monotonically decreasing temperature profile ( figure [ modelplot ] ) , with @xmath116 ( @xmath117 , or @xmath118 ) . we exclude the best - fit isothermal model atmosphere , with a temperature of 3000k , with @xmath119 ( @xmath120 , or @xmath121 ) , and we exclude an inverted temperature profile @xmath122 ( @xmath123 , or @xmath124 ) due to the relatively high mass of kepler-13ab , we were interested to compare its dayside emission spectrum to a brown dwarf of similar effective temperature . besides investigating the general spectral differences between the dayside of a hot jupiter and a brown dwarf , we also wished to see if the measured surface gravity of kepler-13ab coincided with the surface gravity one would estimate using spectral matching to brown dwarf templates , or using spectral indicators for surface gravity . compared to its isolated brown dwarf equivalent the dayside of kepler-13ab is , needless to say , considerably hotter . from its age and mass we would expect that kepler-13ab would have a surface effective temperature of approximately 400 to 800k if it did not receive any insolation from its host star @xcite . but due to the propinquity of kepler-13a , the dayside of the planet appears to be closer in temperature to a mid to late m - dwarf . we therefore compared our emission spectrum to a series of spectral templates from the spex library covering m4 to t9 spectra for field objects , and a set of low - gravity templates covering m5 to t9 from @xcite . to do so , we converted the normalized flux measurements in the templates to a set of simulated eclipse depths by assuming kepler-13a was a 7650k blackbody , and then scaling the resulting relative flux based on the estimated @xmath125-band surface brightnesses for kepler-13a and the spectral templates from the bt - settl models @xcite using the @xcite values for solar abundances . we find that the dayside of kepler-13ab has a rough spectral type of m7@xmath126 using both the field and low - gravity templates . using only our fifteen wfc3 measurements , we find @xmath127 ( @xmath128 , or @xmath129 ) for the field template and @xmath130 ( @xmath131 , or @xmath132 ) for the low - gravity template . as shown in figure [ bdplot ] , in both cases this result is driven by the presence of the 1.4@xmath0 m @xmath133 absorption feature . the similarity of the higher - gravity field templates and the low - gravity templates in this temperature range means that we are not able to postdict kepler-13ab s surface gravity using either template matching or the surface gravity indicators suggested by @xcite . nevertheless , the general agreement of both the field and low - gravity templates indicates that despite the heavy insolation kepler-13ab receives , its atmosphere appears similar to isolated brown dwarf atmospheres . since isolated brown dwarfs are primarily heated from within , and have correspondingly monotonically decreasing atmospheric temperature - pressure profiles , this strengthens our conclusion from the atmospheric models that kepler-13ab s dayside is neither isothermal , nor does it possess a stratospheric temperature inversion . 0.00 in -0.0 in @xcite measured a center time for kepler-13ab s eclipse that was very close to being half of an orbital period away from the transit center time , with a displacement of @xmath134s . as they noted , due to the light travel time across the diameter of kepler-13ab s orbit , for a perfectly circular orbit one would expect the eclipse time to actually be delayed by @xmath135s . the apparent earliness of the eclipse implies a slightly eccentric orbit , and is what drives the non - zero value of @xmath64 listed in table [ tab : broadbandpriors ] . as described in section 3.1.1 , we used the predicted eclipse time from @xcite as a prior on our broadband fitting process . since our observations poorly sample the predicted eclipse ingress and egress ( i.e. , figure [ niceplot ] ) , this prior on @xmath62 dominates in our fits and we simply recover it in the posterior distribution for @xmath62 from our mcmc fitting . in our adopted fit , we therefore find that the eclipse center time is @xmath136s earlier than the transit center time plus exactly half the orbital period , which as we expected is effectively identical to the value measured by @xcite . to determine how well our hst / wfc3 can independently constrain the eclipse center time , we conducted an additional broadband fit using no prior on the eclipse center . this gave @xmath137 , which is @xmath138s earlier than the transit center time plus exactly half the orbital period . the large uncertainties on this timing offset make it consistent with the expectations from both @xcite and the perfectly circular case , so our observations are not able to meaningfully determine the eccentricity of kepler-13ab s orbit . one complication in interpreting the optical eclipse depth for kepler-13ab is the relatively high amount of reflected light expected from the planet . specifically , the eclipse depth from reflected light alone should be @xmath139ppm , where @xmath140 is the spherical albedo of the dayside at a particular wavelength . since the kepler - band eclipse measured by @xcite is @xmath141ppm , it is possible though unlikely that the optical eclipse is entirely due to reflected light . for our nir observations , the theoretical expectation is that hot jupiters at @xmath142k should have dayside geometric albedos of effectively zero at these wavelengths @xcite . we therefore did not consider possible signatures of reflected light in our wfc3 spectra , and excluded the kepler eclipse point from our evaluation of our thermal emission models . based on those thermal emission models that assume the wfc3 and further nir data contain no reflected light , we calculate that kepler-13ab should have an eclipse 143ppm deep in the kepler bandpass . the difference between this prediction for the thermal emission and the observed depth of @xmath141ppm implies that kepler-13ab has a spherical albedo of @xmath143 in the kepler bandpass . this is substantially lower than the value found by @xcite , @xmath144 , almost entirely because our models of the planetary thermal emission are higher in the optical . our measured geometric albedo and the expectations from @xcite are thus both consistent with our assumption that the wfc3 and other nir eclipse depths contain no substantial components from reflected light . this is in line with what one would expect from other measurements of hot jupiters eclipses , most of which have @xmath145 in the kepler bandpass @xcite . note , though , that the unknown thermal emission components , and the lower stellar insolation in the @xcite sample make this only a general comparison . indeed , the high stellar insolation that kepler-13ab receives makes it possible that kepler-13ab may have different reflective properties compared to these cooler planets . one way to test this would be to observe an optical eclipse spectrum for the planet . in the case that kepler-13ab has a substantial optical albedo , the optical eclipse spectrum would be a combination of the reflectance spectrum ( expected to increase towards the blue , e.g. * ? ? ? * ) , and the thermal emission spectrum ( expected to decrease towards the blue ) , which would make the observed eclipse spectrum relatively flat . as described in the introduction , the lack of clear stratospheric temperature inversions for planets cooler than approximately 3000k led @xcite to recently suggest that that tio / vo driven inversions may only be present in the atmospheres of extremely hot giant planets . @xcite s spectrally - resolved observations of the @xmath146-band eclipse of the transiting brown dwarf kelt-1b ( 3200k ) showed a monotonically decreasing tp profile . due to the high surface gravity of kelt-1b ( 22 times that of jupiter ) , this led them to suggest that surface gravity also plays a role in the presence of a thermal inversion in hot jupiters . specifically , they argued that this was evidence for the cold - trap methods of sequestering tio / vo described by @xcite and @xcite . briefly , both @xcite and @xcite describe a process of tio / vo gas particles condensing and gravitationally settling out of the upper atmosphere . @xcite envisioned this as a `` vertical '' cold - trap , where tio / vo gas on the dayside of a hot jupiter randomly crosses the condensation boundary in the atmosphere , condenses , and falls into the planetary interior . @xcite considered a `` day - night '' cold - trap , where tio / vo molecules condense on the cooler night side of a hot jupiter , and also settle into the interior . in both cases the efficiency of a cold - trap is determined by the interplay of the rate of gravitational settling and the strength of some vertical lofting mechanism to bring tio / vo condensates back into the upper atmosphere where they can re - vaporize . both analyses found that under reasonable assumptions these two cold - traps should be able to deplete a hot jupiter s stratosphere of gas - phase tio / vo . in the cases of both kepler-13ab and kelt-1b , the dayside temperature profiles both remain too hot at depth to allow for a vertical cold - trap to exist . down to pressures of @xmath147 bar , at no point do either of these profiles cross cooler than the tio or vo condensation curves , which vary between approximately 2200k at @xmath147 bar down to approximately 1600k at @xmath148 bar . the most recent modeling of such a process was done by @xcite , who found that a vertical cold - trap of tio / vo was mostly effective in planets with equilibrium temperatures less than 1900k , which is several hundred degrees cooler than the planets we consider . if a cold - trap process is occurring in these atmospheres - and in the atmospheres of the other extremely hot giant planets - it is very probably caused by a day - night cold - trap . to first order , the rate of gravitational settling in a hot jupiter s atmosphere will be given by the free - fall timescale within that atmosphere . this will be the scale height of the atmosphere divided by the free - fall terminal velocity of condensates , @xmath149 here @xmath150 is the atmospheric temperature , @xmath151 is the mean molecular weight of the atmosphere , @xmath152 is the gravitational acceleration , @xmath153 is the particle radius , @xmath154 is the particle density , @xmath155 is the atmospheric density , and @xmath156 is the viscosity of the gas . if we make the assumption that mean molecular weight and viscosity of hot jupiters atmospheres , as well as the condensate and atmospheric density , are all effectively the same , then the free - fall timescale goes as @xmath157 typically , turbulent diffusion is treated as the dominant vertical lofting mechanism for condensates in the upper portion of a hot jupiter s atmosphere , but the exact efficiency of this process is poorly understood . by analogy to molecular diffusion , the efficiency of vertical mixing is parameterized by the effective diffusion coefficient , @xmath158 @xcite . unlike the molecular diffusion coefficient , however , the value of @xmath158 is the result of inherently chaotic processes that are difficult to model . while mixing length theory can be used to derive analytic estimates for @xmath158 ( e.g. , * ? ? ? * ) in the convective regions of a hot jupiter s atmosphere , these estimates are not appicable to the the upper , radiative , portion of the atmosphere probed by eclipse observations . this make the precise value of @xmath158 difficult to predict , and typical values of @xmath158 used in brown dwarf modeling can cover three to four orders of magnitude ( e.g. , * ? ? ? if we make the assumption that the efficiency of vertical mixing is approximately the same in all hot jupiters atmospheres , then for fixed values of @xmath153 , the condensate size , cold - traps will therefore be less efficient in hotter atmospheres with higher values of @xmath150 , but dramatically more efficient as surface gravity increases . 0.00 in ) from the results of @xcite on hd 189733b . based on existing observations , the red shaded region roughly corresponds to parameter space we would expect to see inverted or isothermal atmospheres , while the blue shaded region should have decreasing tp profiles . the middle purple region is ambiguous , owing to the range of allowable particle sizes . note that it is entirely possible that the mean condensate size changes over this parameter space , or varies individually by planet.,title="fig : " ] -0.0 in though more detailed modeling is necessary to precisely assess the role of cold - trap processes in kepler-13ab s atmosphere , let us proceed from the conclusion in @xcite that for particle sizes greater than `` a few microns '' a day - night cold trap is capable of depleting gas - phase tio / vo in hd 189733b s atmosphere . using the scaling in equation ( [ eq:4120 ] , we may extrapolate from this point to other temperatures and surface gravities , and compare against the other hot jupiters with spectrally - resolved nir emission measurements . based on their 1.1@xmath0 m to 1.7@xmath0 m emission and the modeling in their respective papers , we categorized the other planets with spectrally - resolved nir eclipse observations as having an inverted tp profile ( wasp-33b ) , an isothermal profile ( wasp-12b , wasp-103b ) , or as being ambiguous ( wasp-19b @xcite , wasp-4b @xcite , tres-3b @xcite , and corot-2b @xcite ) . of these three ambiguous planets , tres-3b and corot-2b both show wfc3 spectra consistent with both an isothermal and decreasing tp profile . the temperatures for the isothermal models of both ( 1800k and 1780k , respectively ) are very close to the condensation temperature of tio / vo . depending upon the exact temperature structure of their atmospheres , it is therefore possible that tio and vo have condensed out of their upper atmospheres independently of any possible cold - trap mechanisms . for wasp-4b , the wfc3 measured spectrum agrees with both a 2000k isothermal model and a carbon - rich , monotonically decreasing , model . wasp-19b was observed from the ground by @xcite , and their emission spectrum is not precise enough to distinguish the temperature structure of the planet . we categorized kepler-13ab as having a monotonically decreasing tp profile , as well as kelt-1b . extrapolating from hd 189733 b according to equation ( [ eq:4120 ] ) , we then plotted these eight planets alongside the limits of where we would expect cold - traps to inhibit inversions for particle sizes of @xmath159 m , @xmath160 m , and @xmath161 m . as shown in figure [ logglogt ] , this set of classifications and predictions relatively well represents the planets with hst / wfc3 observations of their dayside emission . in the shaded red region , where we would expect the atmospheres to be inverted or isothermal for reasonable particle sizes , are wasp-33b , wasp-12b , and wasp-103b similarly , both kepler-13ab and kelt-1b lie in the region where we would expect to see decreasing tp profiles assuming the condensate particles grow larger than @xmath159 m . we note that @xcite have also considered the efficiency of cold - traps , but instead choose to compare vertical terminal velocities to vertical mixing velocities to determine if a condensate sinks into the planetary interior , without considering the atmospheric scale height . based on numerical simulations of the vertical mixing velocity in planets cooler than 1750k , this leads @xcite to estimate that the efficiency of cold - trap processes should be roughly independent of temperature , and should scale inversely with condensate size and surface gravity . this nicely contrasts with our equations [ eq:4110 ] and [ eq:4120 ] , which propose a different scaling . which of these is correct - if any - would therefore provide us with good insight into the dominant processes with hot jupiters atmospheres , and could indicate whether vertical mixing velocities are roughly constant , as we have assumed , or if they vary coherent;y with planetary properties , as assumed by @xcite . currently , the global maximum size for condensate particles in the atmospheres of extremely hot jupiters with daysides near 3000k is poorly constrained by atmospheric models . based on the rayleigh scattering signatures seen in transmission spectra the maximum particle size at high altitudes along the planetary limb appears to be on the order of 0.1@xmath0 m @xcite . but 3d models of hd 187933 b s ( @xmath162k ) atmosphere by @xcite show a wide range of possible particle sizes as a function of longitude , latitude , and depth within the atmosphere . @xcite recently analyzed kepler - band albedo estimates from eclipse measurements made by @xcite , in an effort to constrain condensate properties on planetary daysides and at higher pressures than those probed by transmission observations . @xcite found that an average condensate size of 0.1@xmath0 m was able to replicate the reflective properties of the cooler planets ( @xmath163k ) , but that the two hotter planets in the @xcite data set near @xmath164k required a larger average condensate size . what the condensate size is in a hot jupiter with a dayside near 3000k has , therefore , not been extensively modeled . based on the toy model diagrammed in figure ( [ logglogt ] ) , with more spectrally - resolved observations of hot jupiters daysides , it may be possible to observationally constrain the general condensate size in these planets atmospheres by mapping out where stratospheric temperature inversion occur in extremely hot jupiters . though this assumes that cold - traps are the dominant mechanism to inhibit inversion , it will also be necessary to determine if one can even speak of single `` condensate size '' for all hot jupiters . to more adequately assess the role that cold - traps play in hot jupiters atmospheres , we therefore need more two- or three - dimensional atmosphere models , such as those in @xcite and @xcite , that include the variation of temperature structure as a function of longitude and latitude , models of particle and condensate growth processes within hot jupiters atmospheres , and we need spectrally - resolved observations of more systems to validate the results of both . we observed two secondary eclipses of the transiting hot jupiter kepler-13ab using hst / wfc3 on ut 2014 april 28 and ut 2014 october 13 . we were able to separate the two primary components of the stellar system , kepler-13a and kepler-13bc , in our staring - mode grism spectroscopy . using wayne simulations of both components , we extracted dilution - corrected spectrophotometry for each component by subtracting off the simulations from our observed images . the presence of kepler-13bc allowed us to use its broadband light curve as a comparison star , and the resulting differential light light curve of the kepler-13a system showed very little of the systematic trends typically associated with hst / wfc3 spectrophotometry ( figure [ detrendplot ] . together with a gaussian process regression model , this allowed us to measure the broadband eclipse depth as @xmath111ppm ( table [ tab : broadbandresults ] ) . this corresponds to an average dayside brightness temperature of 3000k . as a check , we ran the same extraction and fitting process on kepler-13bc and measured an eclipse depth of @xmath165ppm , which is consistent with zero , as expected . for the spectrally - resolved eclipse depths we made another differential measurement but this time against the broadband light curve of kepler-13a itself . we made this choice so that our results were not influenced by possible differences in the spectral responses of kepler-13a and kepler-13bc . as a result , our spectrally - resolved fits using another gaussian process regression model measure the differential eclipse depth of kepler-13ab as a function of wavelength . we find that the planetary emission spectrum displays significant ( @xmath114 ) variation with wavelength ( table [ tab : spectralresults ] ) . we again performed the same extraction and fitting on the corresponding spectrally - resolved data of kepler-13bc , and measure differential eclipse depths consistent with zero , again as we expected ( figure [ diffplot ] ) . our eclipse spectrum shows the 1.4@xmath0 m @xmath133 feature in absorption , and modeling of this feature and the previous eclipse measurements made by @xcite indicate that the dayside of kepler-13ab possesses a monotonically decreasing temperature - pressure profile . the inference that kepler-13ab s dayside temperature is monotonically decreasing is further supported by the fact that the shape and amplitude of our hst / wfc3 spectrum shows that the dayside of kepler-13ab appears similar to the spectrum of a field m7 dwarf . we contend that the dual facts that kepler-13ab possesses a decreasing temperature - pressure profile and a relatively high surface gravity support the hypothesis of @xcite that both surface gravity and temperature play a role in determining the presence of a stratospheric temperature inversion in hot jupiters . specifically , in high surface gravity planets such as kepler-13ab , the characteristic free - fall time within the atmosphere is substantially shorter ( equation [ eq:4120 ] ) . this should , in turn , substantially increase the efficiency of a day - night @xcite cold - trap process , thereby sequestering the tio / vo molecules available to cause an inversion in the interior of the planet . of the nine hot jupiters with hst / wfc3 eclipse observations and dayside brightness temperatures hotter than 1750k , the scaling of where an inversion should occur implied by equation [ eq:4120 ] and a reasonable range of condensate particle sizes seems to reproduce the observations fairly well ( figure [ logglogt ] ) . if we accept the assumption that cold - trap processes are the dominant inhibitor of stratospheric temperature inversions in these giant planets , then more spectrally - resolved observations of planets within this parameter space could allow us to observationally constrain the maximum condensate size possible in a hot jupiter s dayside , atmosphere . this would provide us with constraints on the cloud growth processes and the bulk vertical mixing of the atmosphere . to properly assess this , however , we will also need to work on two- and three - dimensional atmosphere models . the current one - dimensional models are generally not capable of treating the local atmospheric variations that are presumable important in setting the average condensate size in an atmosphere . furthermore , while developing higher dimensional models would have an immediate impact here , we also note that it would also allow us to better treat the observed temperature gradients across planetary daysides and from day to night . pushing for higher dimensional models , and more detailed , spectrally - resolved , observations of hot jupiters atmospheres , will give us a much better window into the vertical temperature structure of these planets . these observations were made as a part of go program 13308 with the nasa / esa hubble space telescope at the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa , under the contract nas 5 - 26555 . t.g.b . , m.z . , and j.t.w . were partially supported by funding from the center for exoplanets and habitable worlds . the center for exoplanets and habitable worlds is supported by the pennsylvania state university , the eberly college of science , and the pennsylvania space grant consortium . a. t. was supported by the erc project exolights ( 617119 ) . m.z . and j.t.w . acknowledge nasa origins of solar systems grant nnx14ad22 g . this work has made use of nasa s astrophysics data system , the exoplanet orbit database and the exoplanet data explorer at exoplanets.org @xcite , the extrasolar planet encyclopedia at exoplanet.eu @xcite , the simbad database operated at cds , strasbourg , france @xcite , and the vizier catalog access tool , cds , strasbourg , france @xcite . lcl @xmath95&sqr . covariance amplitude & @xmath166 + @xmath97&sqr . covariance length - scale & @xmath167 + @xmath96&periodic covariance amplitude & @xmath168 + @xmath98&periodic covariance length - scale & @xmath169 + @xmath62&eclipse time ( @xmath71 ) & @xmath170 + @xmath171&log orbital period ( days ) & @xmath172 + @xmath64 & & @xmath74 + @xmath65 & & @xmath75 + @xmath76 & cosine of inclination & @xmath173 + @xmath78&radius of planet in stellar radii & @xmath174 + @xmath175&log semi - major axis in stellar radii & @xmath176 + @xmath69&eclipse depth ( ppm ) & @xmath1 + @xmath34&orbital period ( days ) & @xmath177 + @xmath80&semi - major axis in stellar radii & @xmath178 + @xmath179&inclination ( degrees ) & @xmath180 + @xmath181&impact parameter & @xmath182 + @xmath183&fwhm duration ( days ) & @xmath184 + @xmath185&ingress / egress duration ( days ) & @xmath186 + @xmath187&total duration ( days ) & @xmath188 + @xmath36&orbital eccentricity & @xmath189 + @xmath190&argument of periastron ( degrees ) & @xmath191 [ tab : broadbandresults ] cccccc @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 & @xmath197 + @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 + @xmath204 & @xmath205 & @xmath206 & @xmath207 & @xmath208 & @xmath209 + @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 + @xmath222 & @xmath223 & @xmath224 & @xmath225 & @xmath226 & @xmath227 + @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 & @xmath233 + @xmath234 & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath221 + @xmath239 & @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath244 + @xmath245 & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath250 + @xmath251 & @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 + @xmath257 & @xmath258 & @xmath259 & @xmath260 & @xmath238 & @xmath261 + @xmath262 & @xmath263 & @xmath264 & @xmath248 & @xmath265 & @xmath266 + @xmath267 & @xmath268 & @xmath269 & @xmath270 & @xmath271 & @xmath272 + @xmath273 & @xmath274 & @xmath275 & @xmath276 & @xmath277 & @xmath278 [ tab : spectralresults ] |
we discuss topological properties of the ground state of spatially homogeneous ensemble of fermions . there are several classes of topologically different fermionic vacua ; in each case the momentum space topology of the vacuum determines the low - energy ( infrared ) properties of the fermionic energy spectrum . among them there is class of the gapless systems which is characterized by the fermi - hypersurface , which is the topologically stable singularity . this class contains the conventional landau fermi - liquid and also the non - landau luttinger fermi - liquid . another important class of gapless systems is characterized by the topologically stable point nodes ( fermi points ) . superfluidhe - a and electroweak vacuum belong to this universality class . the fermionic quasiparticles ( particles ) in this class are chiral : close to the fermi points they are left - handed or right - handed massless relativistic particles . since the spectrum becomes relativistic at low energy , the symmetry of the system is enhanced in the low - energy edge . the low - energy dynamics acquires local invariance , lorentz invariance and general covariance , which become better and better when the energy decreases . interaction of the fermions near the fermi point leads to collective bosonic modes , which look like effective gauge and gravitational fields . since the vacuum of superfluidhe - a and electroweak vacuum are topologically similar , we can usehe - a for simulation of many phenomena in high energy physics , including axial anomaly . he - a textures induce a nontrivial effective metrics of the space , where the free quasiparticles move along geodesics . withhe - a one can simulate event horizons , hawking radiation , rotating vacuum , conical space , etc . [ @twocolumnfalse ] | cond-mat9902171 | the physical vacuum is a complicated condensed matter @xcite . at the moment we know only the low - energy properties of this substance , i.e. the properties at energies much smaller than the planck energy scale , @xmath1 , where @xmath2 is newton s constant . we know that at the low - energy edge our vacuum has many different symmetries : @xmath3 and @xmath4 gauge symmetries , lorentz invariance , general coordinate invariance , and dicrete cpt symmetry . with increasing energy more elements of symmetry are added : @xmath5 symmetry of weak interactions , probably the gut symmetry and even supersymmetry . when the temperature dicreases such symmetries become spontaneously broken . this is the traditional point of view , which is supported by the observation of the similar symmetry breaking in different ( many - body ) condensed matter systems : superfluids , superconductors , magnets , liquid and ordinary crystals . however there is another and actually opposite point of view : all symmetries known in the universe spontaneously ( but without any phase transition ) appear at the low - energy corner@xcite . they become more and more pronounced the lower the energy . the symmetries disappear at higher energies when the planck energy is approached . this conjecture is also supported by the condensed matter analogy . here we discuss the topological origin of such enhanced symmetry in the infrared limit . it is caused by the so - called topologically stable fermi points in the spectrum of the fermionic system . we show that if such points exist in quantum condensed matter , then at low enough temperature this system exhibits fully or partially the lorentz invariance , general covariance , and gauge invariance . moreover the collective modes which describe such a system at low @xmath6 all are represented by the chiral relativistic fermions , gauge fields , and gravity . all are the low - energy phenomena , which are absent at higher energies , where the condensed matter has only a very limited set of the global symmetries . from the second point of view some directions in physics look artificial . in particular , since the gravity exists only as the infrared phenomenon one should not quantize gravity : only low - energy gravitons can be quantized @xcite . the same concerns the multidimensional string theories , which also give rise to gravitation in the infrared limit . the fermi point mechanism does not require a high dimensionality for the space - time : the topologically stable fermi point is just a property of the conventional 3 + 1 dimensional space - time . the fermi point is a particular case of the topologically stable manifolds of zeroes in the fermionic spectrum . let us start with the simplest such manifold the fermi surface . the fermi surface appears in the noninteracting fermi gas , where the energy spectrum of fermions is @xmath7 and @xmath8 is the chemical potential . the fermi surface bounds the volume in the momentum space where the energy is negative , @xmath9 , and where the particle states are all occupied at @xmath10 . in this isotropic model the fermi surface is a sphere of radius @xmath11 . it is important that the fermi surface survives even if interactions between particles are introduced . such stability of fermi surface comes from the topological property of the feynman quantum mechanical propagator the one - particle green s function @xmath12 let us write the propagator for a given momentum @xmath13 and for the imaginary frequency , @xmath14 . the imaginary frequency is introduced to avoid the conventional singularity of the propagator `` on the mass shell '' , i.e. at @xmath15 . for noninteracting particles the propagator has the form @xmath16 obviously there is still a singularity : on the 2d hypersurface @xmath17 in the 4-dimensional space @xmath18 the propagator is not well defined . this singularity is stable , i.e. can not be eliminated by small perturbations . the reason is that the phase @xmath19 of the green s funtion @xmath20 changes by @xmath21 around the path embracing this 2d hypersurface in the 4d - space ( see fig . [ fermisurfaceasvortex ] ) . the phase winding number @xmath22 can not change continuously , that is why it is robust towards any perturbation . thus the singularity of the green s funtion on the 2d - surface in the momentum space is preserved , even when interactions between particles are introduced . exactly the same topological conservation of the winding number leads to the stability of the quantized vortex in superfluids and superconductors , the only difference being that , in the case of vortices , the phase winding occurs in the real space , instead of the momentum space . the complex order parameter @xmath23 changes by @xmath24 around the path embracing vortex line in 3d space or vortex sheet in 3 + 1 space . the connection between the topology in real space and the topology in momentum space is , in fact , even deeper ( see _ e.g. _ ref.@xcite ) . if the order parameter depends on space - time , the propagator in semiclassical aproximation depends both on 4-momentum and on space - time coordinates @xmath25 . the topology in the 4 + 4 dimensional space describes : ( i ) the momentum space topology in the homogeneous system ; ( ii ) the defects of the order parameter in real space ; and ( iii ) topology of the energy spectrum within the topological defects @xcite . in the more complicated cases , when the green s function is the matrix with spin and band indices , the phase of the green s function becomes meaningless . in this case one should use a general analytic expression for the integer topological invariant which is responsible for the stability of the fermi surface : @xmath26 here the integral is taken over an arbitrary contour @xmath27 in the momentum space @xmath28 , which encloses the fermi hypersurface ; and @xmath29 is the trace over the spin and band indices . the topological class of systems with fermi surface is rather broad . in particular it contains the conventional landau fermi - liquid , in which the propagator preserves the pole . close to the pole the propagator is @xmath30 evidently the residue @xmath31 does not change the topological invariant for the propagator , eq.([invariantforfs ] ) , wihch remains @xmath32 . this is essential for the landau theory of an interacting fermi liquid ; it confirms the assumption that in fermi liquids the spectrum of quasiparticles at low energy is similar to that of particles in a fermi gas . it is also important for the consideration of the bosonic collective modes of the landau fermi - liquid . the interaction between the fermions can not not change the topology of the fermionic spectrum , but it produces the effective field acting on a given particle by the other moving particles . this effective field can not destroy the fermi surface owing to its topological stability , but it can locally shift the position of the fermi surface . therefore a collective motion of the particles is seen by an individual quasiparticle as a dynamical mode of the fermi surface . these bosonic oscillative modes are known as the different harmonics of the zero sound . an example is shown in the upper part of fig.[collectivemodes ] . note that the fermi hypersurface exists for any spatial dimension . in the 2 + 1 dimension the fermi hypersurface is a line the vortex line in the 2 + 1 momentum space as is shown in fig . [ fermisurfaceasvortex](a ) . in the 1 + 1 dimension the green s function looses its pole , but nevertheless the fermi surface is still there @xcite . in 1 + 1 dimension the landau fermi liquid transforms to another states within the same topological class . example is provided by the luttinger liquid . close to the fermi surface the green s function for the luttinger liquid can be approximated as ( see @xcite ) @xmath33 where @xmath34 and @xmath35 correspond to fermi velocities of spinons and holons and @xmath36 . the above equation is not exact but reproduces the topology of the green s function in luttinger fermi liquid . if @xmath37 and @xmath38 , the singularity in the @xmath39 momentum space occurs on the fermi surface , i.e. at @xmath40 . the topological invariant in eq.([invariantfor fs ] ) remains the same @xmath32 , as for the conventional landau fermi - liquid . the difference from landau fermi liquid occurs only at real frequency @xmath41 : the quasiparticle pole is absent and one has the branch cut singularities instead of the mass shell , so that the quasiparticles are not well defined . the population of the particles has no jump on the fermi surface , but has a power - law singularity in the derivative @xcite . another example of the non - landau fermi liquid is the fermi liquid with exponential behavior of the residue @xcite . it also has the fermi surface with the same topological invariant , but the singularity at the fermi surface is exponentially weak . in this section we do not consider the bifurcations which lead to the overall reconstruction of the particle spectrum , which occur , say , during the transition to superconducting state . we assume that the temperature , though is small compared to the `` planck '' scale @xmath42 , is still above the superfluid transition : @xmath43 . but even in the superconducting region @xmath44 the winding number for the conventional green s function is preserved : the invariant in eq.([invariantfor fs ] ) is again @xmath32 . only instead of pole at @xmath45 in the green s function of the normal state , one has zero at @xmath45 in the superconducting state : @xmath46 where @xmath47 is the gap in the quasiparticle spectrum . this demonstrates the stability of the singularity in the green s function , though the fermi surface as the manifold of zeroes in the energy spectrum disappears in the superconductor . although the systems we have discussed contain fermionic and bosonic quantum fields , this is not the relativistic quantum field theory which we need for the simulation of quantum vacuum : there is no lorentz invariance and the oscillations of the fermi surface do not resemble the gauge field even remotely . the situation is somewhat better for superfluids and superconductors with fully gapped spectra ; examples , which provide useful analogies with dirac fermions and spontaneously broken symmetry in quantum fields , are conventional superconductors @xcite and superfluid @xmath0he - b @xcite . in @xmath0he - b the hamiltonian of free bogoliubov quasiparticles is the @xmath48 matrix : @xmath49 where the pauli @xmath50 matrices @xmath51 describe the conventional spin of fermions and @xmath50 matrices @xmath52 describe the bogoliubov isospin in the particle - hole space . the bogoliubov hamiltonian asymptotically approaches the dirac one in the low @xmath53 limit . this is however is not the low energy limit , which for the typical @xmath0he - b parameters occurs for @xmath53 close to @xmath54 . the low @xmath53 and low @xmath55 limits coincide at large negative @xmath56 , which is not the case in @xmath0he - b . nevertheless the @xmath0he - b also serves as a model system for simulations of many phenomena in particle physics and cosmology including experimental verification @xcite of the kibble mechanism describing formation of cosmic strings in the early universe @xcite . note that even for the fully gapped systems there can exist the momentum - space topological invariants , which characterize the vacuum states . typically this occurs in 2 + 1 systems , e.g. in the 2d electron system exhbiting quantum hall effect @xcite ; in thin film of @xmath0he - a ( see @xcite and sec.9 of ref.@xcite ) ; and in the 2d superconductors with broken time reversal symmetry @xcite . the quantum ( lifshitz ) transition between the states with different topological invariants occurs through the intermediate gapless regime @xcite . now we proceed to the topological class which most fully exhibits the fundamental properties needed for a realization of the relativistic quantum fields , analogous to those in particle physics and gravity . it is the class of systems , whose representatives are superfluid @xmath0he - a and the vacuum of relativistic left - handed and right - handed chiral fermions ( another example of this class in condensed matter has been discussed for gapless semiconductors @xcite ) . this class is characterized by points in the momentum space where the ( quasi)particle energy is zero . in particle physics the energy spectrum @xmath57 is characteristic of the massless neutrino ( or any other chiral lepton or quark in the standard model ) with @xmath58 being the speed of light . the energy of a neutrino is zero at point @xmath59 in the 3d momentum space . the hamiltonian for the neutrino the massless spin-1/2 particle is a @xmath50 matrix @xmath60 which is expressed in terms of the pauli spin matrices @xmath61 . the sign @xmath62 is for a right - handed particle and @xmath63 for a left - handed one : the spin of the particle is oriented along or opposite to its momentum , respectively . the bogoliubov matrix for the @xmath0he - a fermions is @xmath64 where @xmath65 and @xmath66 are real vectors , which in equilibrium or ground state are unit orthogonal vectors . the energy of the fermions @xmath67 is zero at 2 points , at @xmath68 . let us show that zeroes in the spectrum of the chiral fermions and in the spectrum of @xmath0he - a fermions are topologically stable and described by the same topological invariant . let us again consider the propagator of the particle @xmath69 on the imaginary frequency axis , @xmath14 . one can see that this propagator still has a singularity , which is now not on the surface but at the points in the 4d momentum space : @xmath70 and @xmath71 for chiral fermions and @xmath0he - a fermions respectively . an integer topological invariant which supports the stability of the point zeroes is expressed in terms of the propagator in the following way @xcite : @xmath72 where @xmath73 is the 3-dimensional surface embracing the point node in the 4-momentum space . for the chiral fermions in eq.([neutrino ] ) this invariant is @xmath74 , where the sign is determined by the chirality of the fermion . the meaning of this topological invariant can be easily visualized . let us consider the behavior of the particle spin @xmath75 as a function of its momentum @xmath13 in the 3d - space @xmath76 . for right - handed particles @xmath77 , while for left - handed ones @xmath78 . in both cases the spin distribution in the momentum space looks like a hedgehog ( see fig . [ fermisurfaceasvortex]b ) , whose spines are represented by spins : spines point outward for the right - handed particle and inward for the left - handed one . in the 3d - space the hedgehog is topologically stable . for the @xmath0he - a fermions in eq.([aphasefermions1 ] ) @xmath79 , where the sign is determined by the position of the node , @xmath80 . the topological invariant is twice larger because of the double degeneracy of the fermi point over the conventional spin of the @xmath0he atom . for each projection of spin one has @xmath81 . note that the bogoliubov spin @xmath82 in @xmath0he - a plays the same role as the conventional spin @xmath83 of chiral fermions in eq.([neutrino ] ) . on the other hand the conventional spin of the @xmath0he atom is responsible for the degeneracy , but not for chirality , and thus plays the part of the isospin ( see also sec.[degeneracy ] ) . this means that the origin of the spin responsible for the chirality of the ( quasi)particle is fully determined by the matrix structure of the fermi point . in this sense there is no principle difference between spin and isospin : changing the matrix structure one can convert isospin to spin , while the topological charge of the fermi point remains invariant . the high - temperature superconductors in cuprates most probably contain zeroes in their quasiparticle energy spectrum . the arpes experiments @xcite show that these are four lines in the 3d momentum space where the quasiparticle energy is zero or , equivalently , four point zeroes in the 2d cuo@xmath84 planes . the high - t superconductors thus belong to class of systems with fermi lines : the dimension @xmath85 of the manifold of zeroes is @xmath86 , which is intermediate between a fermi surface with @xmath87 and a fermi point with @xmath88 . the energy spectrum of quasiparticles near each of the 4 gap nodes can be written as @xmath89 the `` speeds '' of light @xmath90 and @xmath91 are the `` fundamental '' characteristics determined by the microscopic physics of the cuprates . this means that the system belongs to the same class as 2 + 1 qft with massless relativistic fermions . as in the other two classes of fermionic systems with the gapless quasiparticles , all low - energy ( infrared ) properties of cuprate superconductors are determined by zeroes . in particular , the density of the fermionic states is determined by the dimension of the zeroes : @xmath92 many low - temperature properties of these superconductors are obtained from a simple scaling arguments . for example , an external magnetic field @xmath93 has dimension of @xmath94 and thus of @xmath95 . at finite @xmath93 , the density of states is nonzero even at @xmath96 . substituting @xmath97 to eq.([dosgeneral ] ) one obtains @xmath98 and the following scaling law for the heat capacity : @xmath99 where @xmath100 is some function with the known asymptotes ( see @xcite ) . an experimental indication of such scaling with @xmath101 was reported for yba@xmath84cu@xmath102o@xmath103 in ref . @xcite . the lines of zeroes generally have no stability : there is no corresponding @xmath104 invariant , which can support the topological stability . the singular line in the momentum space from which the spines ( now the vector @xmath82 ) point outward ( see fig . [ fermisurfaceasvortex]c ) can be elmininated by the escape of the @xmath82-vector to a third dimension . this can be accomplished by an operation similar to the folding of an umbrella ( see fig . [ fermisurfaceasvortex]d ) . existence of the nodal lines can be prescribed , however , by the symmetry of the ground state . there are many nontrivial classes of superconductors , whose symmetry supports the existence of nodal lines in symmetric positions in the momentum space @xcite . the symmetry violating perturbations , such as impurities , an external magnetic field , etc . , destroy the lines of zeroes @xcite . one could expect different types of transformations of these lines of zeroes which depend on the perturbation . impurities , for example , can : ( i ) produce the gap in the fermionic spectrum @xcite ( see fig.[fermisurfaceasvortex]d ) , which corresponds to appearance of mass for the 2 + 1 relativistic fermions ; ( ii ) lead to the finite density of states @xcite , thus transforming the system to class ( 1 ) ; ( iii ) produce zeroes of fractional dimension , which means that the exponent in the density of states @xmath105 is non - integral @xcite and thus corresponds to a fractional @xmath85 of the manifold of zeroes ; and ( iv ) lead to localization @xcite . an open question is : can the quantum fluctuations do the same , in particular , can they change the effective dimension of the zeroes ? close to the fermi point @xmath106 in the 4d space one can expand the propagator in terms of the deviations from this fermi point , @xmath107 . if the fermi point is not degenerate , the general form of the propagator is @xmath108 here we returned back from the imaginary frequency axis to the real energy , so that @xmath109 instead of @xmath14 ; and @xmath110 . the quasiparticle spectrum @xmath111 is given by the poles of the propagator : @xmath112 thus in the vicinity of the fermi point the massless quasiparticles are described by the lorentzian metric @xmath113 . it is most important that this is the general form of the energy spectrum in the vicinity of any fermi point , even if the underlying fermi system is not lorentz invariant ; superfluid @xmath0he - a is an example . the fermionic spectrum necessarily becomes lorentz invariant near the fermi point , _ i.e._. if one applies this reasoning to our quantum vacuum , one may conclude that possibly the observed lorentz invariance of the physical laws is not a fundamental but a low - energy property of the vacuum which results from the existence of the topologically stable fermi points . let us consider the collective modes in such a system . the effective fields acting on a given particle due to interactions with other moving particles can not destroy the fermi point . that is why , under the inhomogeneous perturbation of the fermionic vacuum the general form of eqs.([generalpropagator]-[generalenergy ] ) is preserved . however the perturbations lead to a local shift in the position of the fermi point @xmath106 in momentum space and to a local change of the vierbein @xmath114 ( which in particular includes slopes of the energy spectrum ( see fig . [ collectivemodes ] ) . this means that the low - frequency collective modes in such fermi liquids are the propagating collective oscillations of the positions of the fermi point and of the slopes at the fermi point . the former is felt by the right- or the left - handed quasiparticles as the dynamical gauge ( electromagnetic ) field , because the main effect of the electromagnetic field @xmath115 is just the dynamical change in the position of zero in the energy spectrum : in the simplest case @xmath116 . the collective modes related to a local change of the vierbein @xmath114 correspond to the dynamical gravitational field . the quasiparticles feel the inverse tensor @xmath117 as the metric of the effective space in which they move along the geodesic curves @xmath118 therefore , the collective modes related to the slopes play the part of the gravity field ( see fig . [ collectivemodes ] ) . thus near the fermi point the quasiparticle is the chiral massless fermion moving in the effective dynamical electromagnetic and gravitational fields . another important property which results from the above equation is that the fermionic propagator in eq.([generalpropagator ] ) is gauge invariant and even obeys the general covariance near the fermi point . for example , the local phase transformation of the wave function of the fermion , @xmath119 can be compensated by the shift of the `` electromagnetic '' field @xmath120 . these attributes of the electromagnetic ( @xmath121 ) and gravitational ( @xmath113 ) fields also arise spontaneously as the low - energy phenomena . now let us discuss the dynamics of collective bosonic modes , @xmath121 and @xmath113 . since these are the effective fields their motion equations do not necessarily obey gauge inariance and general covariance . however , in some special cases such symmetries can arise in the low energy corner . what are the conditions for that ? the effective lagrangian for the collective modes is obtained by integrating over the vacuum fluctuations of the fermionic field . this principle was used by sakharov and zeldovich to obtain an effective gravity @xcite and effective electrodynamics @xcite , both arising from fluctuations of the fermionic vacuum . if the main contribution to the effective action comes from the vacuum fermions whose momenta @xmath13 are concentrated near the fermi point , _ i.e. _ where the fermionic spectrum is linear and thus obeys the `` lorentz invariance '' , the result of the integration is necessarily invariant under gauge transformation , @xmath120 , and has a covariant form . the obtained effective lagrangian then gives the maxwell equations for @xmath121 and the einstein equations for @xmath117 , so that the propagating bosonic collective modes do represent the gauge bosons and gravitons . thus two requirements must be fulfilled ( i ) the fermionic system has a fermi point and ( ii ) the main physics is concentrated near this fermi point . in this case the system acquires at low energy all the properties of the modern quantum field theory : chiral fermions , quantum gauge fields , and gravity . all these ingredients are actually low - energy ( infra - red ) phenomena . there is another important symmetry obeyed by massless relativistic weyl fermions , the conformal invariance the invariance under transformation @xmath122 . in the extreme limit when the vacuum fermions are dominatingly relativistic , the effective action for gravity must be conformly invariant . such gravity , the so - called weyl gravity , is a viable rival to einstein gravity in modern cosmology @xcite : the weyl gravity ( i ) can explain the galactic rotation curves without dark matter ; ( ii ) it reproduces the schwarzschild solution at small distances ; ( iii ) it can solve the cosmological constant problem , since the cosmological constant is forbidden if the conformal invariance is strongly obeyed ; etc . ( see @xcite ) . let us consider what happens in a practical realization of systems with fermi points in condensed matter in @xmath0he - a . close to the gap nodes , _ i.e. _ at energies @xmath123 , where @xmath47 is the maximal value of the gap in @xmath0he - a which plays the part of the planck energy , the quasiparticles do obey the relativistic equation @xmath124 here @xmath125 is the `` electric charge '' and simultaneously the chirality of the quasiparticles . let us consider the simplest situation , when the @xmath0he - a is in its vacuum manifold , which is characterized by two unit mutually orthogonal vectors @xmath126 and @xmath127 , and neglect the spin degeneracy of the fermi point . when we consider the low - energy collective modes , these vectors are slowly changing in space - time . in this situation the effective metric and effective electromagnetic field are given by : @xmath128 where @xmath129 is the superfluid velocity given by @xmath130 . from above equations it follows that the fields , which act on the `` relativistic '' quasiparticles as electromagnetic and gravitational fields , have very strange behavior . for example , the same texture of the @xmath131-vector is felt by quasiparticles as the effective magnetic field @xmath132 according to eq.([emfieldinaphase ] ) and simultaneously it enters the metric according to eq.([gravfieldinaphase ] ) . such field certainly can not be described by the maxwell and einstein equations together . actually the gravitational and electromagnetic variables coincide in @xmath0he - a only when we consider the vacuum manifold : outside of this manifold they split . @xmath0he - a , as any other fermionic system with fermi point , has enough number of collective modes to provide the analogs for the independent gravitational and electromagnetic fields . but some of these modes are massive in @xmath0he - a . for example the gravitational waves correspond to the modes , which are different from the oscillations of the @xmath131-vector . as distinct from the photons ( orbital waves propagating oscillations of the @xmath131-vector ) the gravitons are massive @xcite . all these troubles occur because in @xmath0he - a the main contribution to the effective action for bosonic fields come from the vacuum fermions at the `` planck '' energy scale , @xmath133 . these fermions are far from the fermi points and their spectrum is nonlinear . that is why in general the effective action is not symmetric . there are , however , exclusions , for example , the action for the @xmath131-field contains the term with the logarithmically divergent factor @xmath134 @xcite . it comes from the zero charge effect , experienced by the vacuum of the massless fermions for whom the @xmath131-field acts as electromagnetic field . due to its logarithmic divergence this term is dominanting at low frequency @xmath135 : the lower the frequency the larger is the contribution of the vacuum fermions from the vicinity of the fermi point and thus the more symmetric is the lagrangian for the @xmath136-field . as a result , in the very low - energy limit , when the non - logarithmic contributions can be neglected , the effective lagrangian for the @xmath121 becomes gauge invariant and even obeys the general covariance : @xmath137 where @xmath138 is the strength of the effective electromagnetic field @xmath121 from eq.([emfieldinaphase ] ) and @xmath113 is the effective gravitational field from eq.([gravfieldinaphase ] ) . in this regime the @xmath121 field does obey the maxwell equations in a curved space . on the other hand the `` einstein '' action for @xmath117 is highly contaminated by many noncovariant terms , which come from the integration over the `` nonrelativistic '' high energy degrees of freedom at `` planck '' scale . in this sense the @xmath0he - a , with its given physical parameters , is not a perfect model for quantum vacuum . to remove the polluting noncovariant terms , the integration must be spontaneously cut - off at energies much below the `` planck '' scale , @xmath139 , for example , due to strong quasiparticle relaxation . the main reason why @xmath0he - a is not a good substance , is that the fermi points of the left particles , i.e. at @xmath140 , and the fermi points of the right particles , i.e. at @xmath141 , are far from each orher in equilibrium . the `` perfect '' condensed matter would be such where all the fermi points are at the origin , at @xmath142 , as it happens in the standard model . however , inspite of the absence of general covariance even in the low - energy corner , many different properties of the physical vacuum with a fermi point , whose direct observation are still far from realization , can be simulated in @xmath0he - a . one of them is the chiral anomaly . the chiral anomaly is the phenomenon which allows the nucleation of the fermionic charge from the vacuum @xcite . such nucleation results from the spectral flow of the fermionic charge through the fermi point to high energy . since the flux in the momentum space is conserved , it can be equally calculated in the infrared or in the ultraviolet limits . in @xmath0he - a it is much easier to use the infrared regime , where the fermions obey all the `` relativistic '' symmetries . as a result one obtains the same anomaly equation , which has been derived by adler and by bell and jackiw for the relativistic systems . the rate of production of quasiparticle number @xmath143 from the vacuum in applied electric and magnetic fields is @xmath144 here @xmath145 and @xmath146 is the density of the right and left quasiparticles ; @xmath147 and @xmath148 are their charges ; and @xmath149 is the dual field strength . note that the above equation is fully `` relativistic '' . this equation has been verified in @xmath0he - a experiments @xcite , where the `` magnetic '' @xmath132 and `` electric '' @xmath150 fields have been simulated by the space and time dependent @xmath136-texture . in particle physics the only evidence of axial anomaly is related to the decay of the neutral pion @xmath151 , although the anomaly is much used in different cosmological scenaria explaining an excess of matter over antimatter in the universe ( see review @xcite ) . in @xmath0he - a the fermi point ( say , at the north pole ) is doubly degenerate owing to the ordinary spin @xmath83 of the @xmath0he atom . this means that in equilibtium the two zeroes , each with the topological invariant @xmath152 , are at the same point in momentum space . let us find out what can be the consequences of the fermi point degeneracy . it is clear that the collective motion can split the fermi points : positions of the two points can oscillate separately . moreover , since the propagator is now the @xmath48 matrix there can be the cross terms . if we neglect the degrees of freedom related to the vierbein then the collective variables of the system with the doubly degenerate fermi point enter the fermionic propagator as @xmath153 the new effective field @xmath154 acts on the chiral quasiparticles as @xmath5 gauge field . thus in this effective field theory the ordinary spin of the @xmath0he atoms plays the part of the weak isospin @xcite . the `` weak '' field @xmath154 is also dynamical and in the leading logarithmic order obeys the maxwell ( actually yang - mills ) equations . this implies that the higher symmetry groups of our vacuum can in principle arise as a consequence of the fermi point degeneracy . for example , the 4-fold degeneracy of the fermi point ( which implies that there are , say , 4 left - handed fermionic species , or 2 left - handed + 2 right - handed ) can produce the @xmath155 gauge group . in particle physics the collective modes related to the shift of the 4-momentum are discussed in terms of the `` generalized covariant derivative '' @xcite . in this theory the gauge fields , the higgs fields , and yukawa interactions , all are realized as shifts of positions of the degenerate fermi point , with degeneracy corresponding to different quarks and leptons . in the eq.([nonabelian fieldpropagator ] ) we did not take into account that dynamically the vierbein can also oscillate differently for each of the two elementary fermi points . as a result the number of the collective modes could increase even more . this is an interesting problem which must be investigated in detail . if the degenerate fermi point mechanism has really some connection to the dynamical origin of the non - abelian gauge fields , we must connect the degeneracy of the fermi point ( number of the fermionic species ) with the symmetry group of the gauge fields . naive approach leads to extremely high symmetry group . that is why there should be some factors which can restrict the number of the gauge and other bosons . for example there can be some special discrete symmetry between the fermions of the degenerate point , which restricts the number of massless bosonic collective modes . another source of the reduction of the number of the effective field has been found by chadha and nielsen @xcite . they considered the massless electrodynamics with different metric ( vierbein ) for the left - handed and right - handed fermions . in this model the lorentz invariance is violated . they found that the two metrics converge to a single one as the energy is lowered . thus in the low - energy corner the lorentz invariance becomes better and better , and at the same time the number of independent bosonic modes decreases . since the connections between qft in the standard model and in @xmath0he - a has been discussed in the earlier publications @xcite , we concentrate here on some problems related to gravitational analogy . as we have seen from eq.([gravfieldinaphase ] ) the gravitational field can be simulated in @xmath0he - a by the motion of the liquid with the superfluid velocity @xmath156 and by the @xmath131-texture @xcite . in this section we consider the situation when the @xmath131-vector is uniform and thus does not produce the effective gravitational field , so that the gravitational effects come only from the motion of the superfluid vacuu . the propagation of fermions in the moving liquid obeys the same equation as propagation of relativistic particles in the gravitational field . the same happens for propagating sound waves in normal fluids @xcite and phonons in superfluid @xmath157he . in the simplest case of the radial motion of superfluid @xmath157he , the effective metric is expressed in terms of the radial superfluid velocity @xmath158 as @xmath159 where @xmath58 now is a speed of sound in @xmath157he ( phonon velocity ) . for @xmath0he - a fermions the spherically symmetric metric occurs if @xmath131-field is radial . then it follows from eq.([gravfieldinaphase ] ) @xmath160 kinetic energy of superflow plays the part of the gravitational potential : @xmath161 . if one chooses the velocity field corresponding to the potential of the point body of mass @xmath162 @xmath163 one obtains the panlev - gullstrand form of schwarzschild geometry ( see e.g. ref.@xcite ) . here @xmath164 denotes the position of the event horizon , where the velocity reaches the `` speed of light '' @xmath58 . if the fluid moves towards the origin , the low - energy quasipartilcles are trapped behind the horizon , since their speed @xmath58 with respect to the fluid is less than the velocity of the fluid . such sonic black hole was first suggested by unruh for ordinary liquid @xcite . however since all the known normal liquids are classical , the most interesting quantum effects related to the horizon can not be simulated in such flow . also the geometry is such that it can not be realized : in such radial flow inward the liquid is accumulated at the origin , so that this sonic black hole can not be stationary . in the other scenario a horizon appears in moving solitons , if the velocity of the soliton exceeds the local `` speed of light '' @xcite . this scenario has the same drawback : in finite system the motion of the soliton can not be supported for a long time . in a draining bathtub geometry suggested in ref.@xcite the fluid motion can be made constant in time . however the friction of the liquid , which moves through the drain , is the main source of dissipation . the superfluidity of the liquid does not help much in this situation . horizon does not appear since the `` superluminal '' motion with respect to the boundaries of the drain tube is not allowed : the superflow becomes unstable and superfluidity collapses ( see @xcite ) . let us suggest a scenario , in which this collapse is avoided . the superfluid motion becomes quasi - stationary and exhibits the event horizon ; the life time of the flow state is determined by hawking radiation . the stationary black hole can be realized in the following geometry , which is the development of the bathtub geometry of ref.@xcite ( see fig . [ 2dbh](a ) ) . the superfluid @xmath0he - a film is moving towards the center of the disk , where it escapes to the third dimension due to the orifice ( hole ) . if the thickness of the film is constant , the flow velocity increases towards the center as @xmath165 and at @xmath166 reaches the speed of light ( now @xmath167 denotes the radial coordinate in the cylindrical system ) . outside the orifice the motion of the liquid is two dimensional and the effective metric for the low - energy bogoliubov quasiparticles is @xmath168 here we took into account that the @xmath169 vector in the film is fixed along the normal to the film , so that the `` speed of light '' for quasiparticles propagating along the film @xmath170cm / sec is much smaller than the fermi velocity @xmath171 which corresponds to the `` speed of light '' for quasiparticles propagating along the normal to the film . note that @xmath58 is much smaller than the speed of sound in @xmath0he - a , that is why the motion of fluid has no effect on the density of the liquid . the important element of the construction in fig . [ 2dbh ] is that the moving superfluid @xmath0he - a film is placed on the top of the superfluid @xmath157he film . this is made to avoid the interaction of the @xmath0he - a film with the solid substrate . the superfluid @xmath157he film effectively screens the interaction and thus prevents the collapse of the superfluid flow of @xmath0he - a with `` superluminal '' velocity : since the interactions with walls is removed the local observer moving with the superfluid velocity can not detect the relative motion with respect to the wall . the motion of the superfluid @xmath0he - a with respect to superfluid @xmath157he film is not dangerous : the superfluid @xmath157he is not excited even if @xmath0he - a moves with its superluminal velocity : @xmath58 for @xmath0he - a is much smaller than the landau velocity for radiation of quasiparticles in superfluid @xmath157he , which is about @xmath172 m / sec . in this consideration we neglected the radiation of surface waves , assuming that the thickness of @xmath157he film is small enough . finally one can close the superflow by introducing the toroidal geometry in fig . [ 2dbh](b ) , so that the superfluid condensate can circulate . in this case in addition to the black hole horizon the white hole horizon appears on the path where the superfluid @xmath0he - a flows out from the orifice . since the extrinsic mechanism of the friction of @xmath0he - a film the scattering of quasiparticles on the roughness of substrate is abandoned , we can consider now intrinsic mechanisms of dissipation . the most interesting one is the hawking radiation related to existence of a horizon . let us consider the simplest case of the 2d motion along the film in the bathtub geometry of fig . [ 2dbh](a ) . this can be easily generalized to the motion in the torus geometry . there are two important reference frames : ( i ) the frame of the observer , who is locally comoving with the superfluid vacuum . in this frame the local superfluid velocity is zero , @xmath173 , so that the energy spectrum of the bogoliubov fermions in the place of the observer is ( here we assume a pure 2d motion along the film ) @xmath174 in this geometry , in which the superflow velocity is confined in the plane of the film , the speed @xmath58 coincides with the landau critical velocity of the superfluid vacuum , @xmath175 . the vacuum as determined by the comoving observer is shown on fig . [ vacua](a ) : fermions occupy the negative energy levels in the dirac sea ( the states with the minus sign in eq.([energycomovingframe ] ) . it is the counterpart of the minkowski vacuum , which is however determined only locally : the comoving frame can not be determined globally . moreover for the comowing observer , whose velocity changes with time , the whole velocity field @xmath176 of the superflow is time dependent . this does not allow to determine the energy correctly . \(ii ) the energy can be well defined in the laboratory frame ( the rest frame ) . in this frame the system is stationary , though is not static : the effective metric does not depend on time , so that the energy is conserved , but this metric contains the mixed component @xmath177 . the energy in the rest frame is obtained from the local energy in the comoving frame by the doppler shift : @xmath178 in case of the radial superflow @xmath179 one has @xmath180 figs . [ vacua](b - c ) show how the `` minkowski '' vacuum of the comoving frame is seen by the rest observer ( note that the velocity is negative , @xmath181 ) . in the absense of horizon , or outside the horizon the local vacuum does not change : the states which are occupied ( empty ) in the minkowski vacuum remain occupied ( empty ) in the rest frame vacuum ( see fig . [ vacua](b ) ) . in the presense of horizon behind which the velocity of superflow exceeds the landau critical velocity the situation changes : behind the horizon the vacuum in the rest frame differs from that in the comoving frame . let us for simplicity consider the states with zero transverse momentum @xmath182 on the branch @xmath183 in the rest frame . if the system is in the minkowski vacuum state ( i.e. in the ground state as viewed by comoving observer ) , quasiparticles on this branch has reversed distribution in the rest frame : the negative energy states are empty , while the positive energy states are occupied ( see fig . [ vacua](c ) ) . for this branch the particle distribution corresponds to the negative temperature @xmath184 behind horizon . since the energy in the rest frame is a good quantum number , the fermions can tunnel across the horizon from the occupied levels to the empty ones with the same energy . thus if the system is initially in the minkowski vacuum in the comoving frame , the tunneling disturbs this vacuum state : pairs of excitations are created : the quasiparticle , say , is created outside the horizon while its partner the quasihole is created inside the horizon . this simulates the hawking radiation from the black hole . to estimate the tunneling rate in the semiclassical approximation , let us consider the classical trajectories @xmath185 of particles , say , with positive energy , @xmath186 , for the simplest case when the transverse momentum @xmath187 is zero fig . [ tunnelingtraj ] . the branch @xmath188 describes the incoming particles with @xmath189 which propagate through the horizon to the orifice ( or to the singularity at @xmath190 , if the orifice is infinitely small ) without any singularity at the horizon . the classical trajectories of these particles are @xmath191 the energies of these particles viewed by the comoving observer are also positive : @xmath192 . another branch @xmath183 in fig . [ tunnelingtraj ] contains two disconnected pieces describing the particle propagating from the horizon in two opposite directions : @xmath193 the eq.([tunnelingtrajectory2 ] ) describes the outgoing particles the particles propagating from the horizon to the exterior . the energy of these particles is positive in both frames , comoving and rest . the eq.([tunnelingtrajectory1 ] ) describes the propagation of particles from the horizon to the orifice ( or to the singularity ) . though for the rest frame observer the energy of these particles is positive , these particles , which live within the horizon , belong to the minkowski vacuum in the comoving frame . the classical trajectory in eqs.([tunnelingtrajectory2],[tunnelingtrajectory1 ] ) is thus disrupted at the horizon . there is however a quantum mechanical transition between the two pieces of the branch : the quantum tunneling . the tunneling amplitude can be found in semiclassical approximation by shifting the contour of integration to the complex plane : @xmath194 this means that the wave function of any particle in the minkowskii vacuum inside the horizon contains an exponentially small part describing the propagation from the horizon to infinity . this corresponds to the radiation from the minkowski vacuum in the presence of the event horizon . the exponential dependence of the probability on the quasiparticle energy @xmath195 suggests that this radiation looks as thermal . the corresponding temperature , the hawking temperature , is @xmath196 the radiation leads to the quantum friction : the linear momentum of the flow decreases with time . this occurs continuously until the superfluid minkowski vacuum between the horizons is completely exhausted and the superfluid state is violated . this leads to the phase slip event , after which the number @xmath197 of circulation quanta of superfluid velocity trapped by the torus is reduced . this process will repeatedly continue until the two horizons merge . the above construction in fig . [ 2dbh](b ) allows us ( at least in principle ) to obtain the event horizon in the quasi - stationary regime , when the main source of nonstationarity is the dissipation coming from the hawking radiation . as for the practical realization , there are , of course , many technical problems to be solved . on the other hand , if the black hole analog can in principle exist in condensed matter as the quasi - stationary object , its prototype the real black hole can also exist , at least in principle ( though it is not so easy to find the scenario of how this object can be obtained from the gravitational collapse of matter @xcite ) . it appears that in some range of the energies of hawking - radiated particles , close to the `` planck '' scale , where the fermionic spectrum becomes `` nonrelativistic '' , the hawking radiation does not exist any more . the particles which are radiated in the fully relativistic case , will be now andreev scattered back to the black hole . thus both partners ( particle and hole ) of the hawking radiation will remain within the horizon , so that the particle creation in high gravity field will disturb the quantum vacuum inside the horizon without any radiation outside . in principle the pair creation inside the horizon can be more important for the dissipation in the superfluid 3he film under discussion than the hawking radiation . the body moving in the vacuum with linear acceleration @xmath198 is believed to radiate the thermal spectrum with the unruh temperature @xmath199 @xcite . the comoving observer sees the vacuum as a thermal bath with @xmath200 , so that the matter of the body gets heated to @xmath201 ( see references in @xcite ) . linear motion at constant proper acceleration ( hyperbolic motion ) leads to velocity arbitrarily close to the speed of light . on the other hand uniform circular motion features constant centripetal acceleration while being free of the above mentioned pathology ( see the latest references in @xcite ) . the latter motion is stationary in the rotating frame , which is thus a convenient frame for study of the radiation and thermalization effects for uniformly rotating body . zeldovich @xcite was the first who predicted that the rotating body ( say , dielectric cylinder ) amplifies those electromagnetic modes which satisfy the condition @xmath202 here @xmath135 is the frequency of the mode , @xmath203 is its azimuthal quantum number , and @xmath204 is the angular velocity of the rotating cylinder . this amplification of the incoming radiation is referred to as superradiance @xcite . the other aspect of this phenomenon is that due to quantum effects , the cylinder rotating in quantum vacuum spontaneously emits the electromagnetic modes satisfying eq.([zeldovichcondition ] ) @xcite . the same occurs for any rotating body , including the rotating black hole @xcite , if the above condition is satisfied . distinct from the linearly accelerated body , the radiation by a rotating body does not look thermal . also , the rotating observer does not see the minkowski vacuum as a thermal bath . this means that the matter of the body , though excited by interaction with the quantum fluctuations of the minkowski vacuum , does not necessarily acquire an intrinsic temperature depending only on the angular velocity of rotation . moreover the vacuum of the rotating frame is not well defined because of the ergoregion , which exists at the distance @xmath205 from the axis of rotation . the problems related to the response of the quantum system in its ground state to rotation@xcite , such as radiation by the object rotating in vacuum @xcite and the vacuum instability caused by the existense of ergoregion @xcite , etc . , can be simulated in superfluids , where the superfluid ground state plays the part of the quantum vacuum . the quantum friction due to spontaneous emission of phonons in superfluid @xmath157he and bogoliubov fermions in superfluid @xmath0he - b has been discussed in @xcite . here we extend an analysis to the radiation of quasiparticles in @xmath0he - a . let us consider a cylinder of radius @xmath206 rotating with angular velocity @xmath204 in the ( infinite ) superfluid liquid . in this situation there are again two important reference frames . one of them is the laboratory frame . the energy of quasiparticles is not well determined in the laboratory frame , because , if the rotating body is not the perfect axysymmetric cylinder , it produces the time dependent perturbations of the liquid . however at distances far enough from the surface of the cylinder the influence of the rotating cylinder on the liquid can be neglected . in this approximation the superfluid vacuum is in rest with respect to the laboratory frame . this means that the superfluid velocity @xmath173 in the laboratory frame . the quasiparticle energy in this frame is that as in the comoving frame of previous section . it is @xmath207 if we consider superfluid @xmath157he , or the eq.([energycomovingframe ] ) , if we consider a pure 2d motion of @xmath0he - a with @xmath208 . such energy spectrum corresponds to the effective minkowski metric of flat space @xmath209 here @xmath210 for isotropic @xmath157he and @xmath211 for anisotropic @xmath0he - a . in the 3d case of @xmath0he - a there can be another effective metric far from the body . it can be caused by the normal boundary condition on the @xmath212-vector , which prefers the radial orientation of @xmath212-vector . in the case of the radial distribution , @xmath213 , the effective metric for the quasiparticles moving outside the cylinder follows from eq.([gravfieldinaphase ] ) : @xmath214 such an effective space is conical : the space outside the cylinder is flat , but the proper length of the circumference of radius @xmath167 around the cylinder is not equal to @xmath215 . in the relativistic theories such conical metric can arise outside the local cosmic strings , where there is an angle deficit . in our case the length of the circumference is @xmath216 , which is much larger than @xmath215 : this effective space exhibits a `` negative angle deficit '' ( for details see ref.@xcite ) . since the system far outside the rotating cylinder is not disturbed by the rotation of the body , the system will remain to be in or close to the minkowski vacuum ( or vacuum in the conical space in the case of radial @xmath217-vector ) as viewed in the laboratory frame . however this does not hold in the region ajacent to the cylinder , where the superfluid velocity field is disturbed and time - dependent . also the energy in the laboratory frame is not well determined because of the time - dependent perturbations . the energy is well determined in the frame corotating with the cylinder . in the corotating frame the cylinder is at rest , and thus the perturbations caused by rotation are stationary . the metric in the coorotating frame is simplest far outside the rotating body , where the superfluid velocity in the corotating frame is @xmath218 . substituting this @xmath129 into eq.([gravfieldinaphase ] ) one obtains the interval @xmath219 , which determines the propagation of quasiparticles in the corotating frame : @xmath220 here again @xmath210 for phonons in isotropic @xmath157he and @xmath211 for fermionic quasiparticles in anisotropic @xmath0he - a with @xmath208 . these metrics correspond to the interval in the rotating frame discussed in relativistic theories . it is convinient to write the quasiparticle spectrum in this frame in two different approximations . in classical description one has @xmath221 . in the other description , an azimuthal motion of quasiparticles is quantized in terms of the angular momentum @xmath222 , while the radial is still treated in the quasiclassical approximation . in this case the energy spectrum of phonons in the corotating frame is @xmath223 and the energy spectrum of the bogoliubov fermions @xmath224 the radius @xmath205 , where @xmath225 in the eq.([intervalrotation ] ) , marks the position of the ergoplane . in the ergoregion , i.e. at @xmath226 , the energy of quasiparticle in eq.([rotbogolonspectrum ] ) can be negative for any @xmath227 . this means that in the ergoregion the minkowski vacuum is not the vacuum for the corotating observer . situation is similar to the case discussed in previous section . however the relevant superfluid velocity is azimuthal now instead of radial . as a result there is an ergoplane instead of a horizon . from the point of view of the corotating observer , the minkowski vacuum is unstable towards the filling of the negative energy states in the ergoregion , which means the radiation of the quasiparticles from the rotating cylinder . however we need a real process , which leads to such radiation . this radiation can be caused only by the interaction between the superfluid minkowski vacuum and the rotating object . let us consider the slow rotations @xmath228 . in this case the linear velocity of the cylinder at the surface of the cylinder , @xmath229 , is much smaller than the landau critical velocity for nucleation of quasiparticles , @xmath230 . thus quasiparticles can not be nucleated near the surface of cylinder . the ergoregion , where @xmath231 and quasiparticles can be nucleated , is far from the cylinder , @xmath232 . the interaction with the cylinder , which produces the matrix element for the radiation , is very small . in this situation the most effective mechanism of the quasiparticle radiation is the tunneling of quasiparticles from the liquid ajacent to the surface of the rotating body , which plays the part of the rotating detector , to the ergoregion . the simplest rotating detector is the fermionic system , which is rigidly connected to the rotating body . in superfluids the rotating body can be effectively substituted by the rigidly rotating cluster of quantized vortices ( see fig . [ ergoregion](a ) ) . such clusters , that experience the solid - body rotation , are experimentally investigated in both phases of superfluid @xmath0he ( see e.g. @xcite ) . the vortex cluster rotating in the infinite superfluid liquid represents the vacuum in the corotating frame . however this vacuum state is not extended to the exterior of the cluster , where one has the analogue of minkowski vacuum . that is why such state is not equilibrium , but it is quasistationary metastable state which can live for a macroscopically long time . at @xmath10 the dominating mechanism of the relaxation of angular velocity @xmath204 of the cluster is the process of the radiation of quasiparticles . [ ergoregion](b ) shows the distribution of superfluid velocity @xmath233 in the laboratory frame . within the vortex cluster , i.e. at @xmath234 , the superfluid velocity , being averaged over the vortices , follows the velocity of the solid body rotation of the cluster : i.e. @xmath235 in the laboratory frame and thus @xmath236 in the frame corotating with the cluster . outside the cluster the superfluid velocity decays as @xmath237 , where @xmath238 is number of vortices in the cluster and @xmath239 is superfluid circulation around individual vortex . within the cluster one has @xmath236 in the corotating frame , so one can expect that the spectrum of quasiparticles , which live within the cluster , is the same as the energy spectrum in the frame comoving with the superfluid velocity , @xmath240 . in other words this spectrum has no @xmath241 shift of the energy levels , as distinct from quasiparticle spectrum far from the body , eqs.([quasiclassicalphononspectrum],[rotbogolonspectrum ] ) , which is measured in the same corotating frame . this is not exactly true , since the equation @xmath236 does not imply that @xmath173 locally : it essentially depends on the position in the vortex lattice . the main feature is nevertheless preserved , when one considers the spectrum of quasiparticles that live in the vortex core . their energy spectrum in the corotating state does not depend on rotation velocity @xmath204 : @xmath242 , where @xmath243 is the generalized angular momentum ; @xmath244 is the so called minigap , which depends on the core structure . this branch crosses zero energy level as a function of @xmath243 , if one considers this quantum number as continuous , and thus represents the fermion zero mode ( see left part of fig . [ ergoregion](c ) ) . the energy spectrum of such chiral fermions , that live in the vortex core , has been first calculated by caroli , de gennes and matricon for the abrikosov vortex in @xmath245-wave superconductors @xcite ; on the relation between the number of fermion zero modes and the winding number of the vortex see @xcite . for us it is important that for vortices in @xmath0he the quantum number @xmath243 is integer and thus one has the states in the detector with zero energy in the corotating frame . the radiation of bogoliubov quasiparticles can be considered as the process in which the particle from the zero energy state in the detector , @xmath246 , tunnels to the scattering state at the ergoplane , where also its energy is @xmath247 ( fig . [ ergoregion](c ) ) . in the quasiclassical approximation the tunneling probability is @xmath248 , where at @xmath249 and @xmath228 : @xmath250 thus all the particles with @xmath251 are radiated , but the radiation probability is smaller for higher @xmath203 : @xmath252 here @xmath253 is energy ( frequency ) of the radiated quasiparticles in the laboratory frame . if @xmath58 is substituted by the speed of light , eq.([fermiontunnelingprobability ] ) is proportional to the superradiant amplification of the electromagnetic waves by rotating dielectric cylinder derived by zeldovich @xcite . since each radiated fermion carries the angular momentum @xmath203 , the vortex cluster rotating in superfluid vacuum at @xmath10 is loosing its angular momentum and thus experiences the quantum rotational friction . the radation also leads to excitation of the detector matter . in principle the radiation can occur without excitation of the detector vacuum , via direct interaction of the particles in the minkowski vacuum with the rotating body ( cluster ) , but the contribution of this process the quantum friction is smaller @xcite . the rotational friction experienced by the body rotating in superfluid vacuum at @xmath10 , is caused by the spontaneous quantum emission of the quasiparticles from the rotating object to the `` minkowski '' vacuum in the ergoregion . the emission is not thermal and depends on the details of the interaction of the radiation with the rotating body . in the quasiclassical approximation it is mainly determined by the tunneling exponent , which can be approximately characterized by the effective temperature @xmath254 . the vacuum friction of the rotating body can be observed only if the effective temperature exceeds the temperature of the bulk superfluid , @xmath255 . for the body rotating with @xmath256rad / s , @xmath6 must be below @xmath257k . however , high rotation velocity can be obtained for clusters . the cluster containg two vortices rotate around their center of mass with @xmath258 , where @xmath206 is the radius of the circular orbit . if the radius @xmath206 is of order of superfluid coherence length , the effective temperature can reach @xmath259k . the process discussed in this section occurs only if there is an ergoplane in the rotating frame . if the superfluid is contained in a finite external cylinder of radius @xmath260 , this process occurs only at high enough rotation velocity , @xmath261 , when the ergoplane is within the superfluid . on the instability of the ergoregion in quantum vacuum towards emission see also in ref.@xcite . if @xmath262 and ergoregion is not present , then the interaction between the coaxial cylinders via the vacuum fluctuations becomes the main mechanism for dissipation . this causes the dynamic casimir forces between the walls moving laterally ( see review @xcite ) . as in @xcite the nonideality of the cylinders , i.e. violation of the rotational symmetry by the body , is the necessary condition for quantum friction . in the above examples of the nontrivial space , the effective gravitational field acts as a fixed external field . the dynamics of this field has not been discussed here . in most cases the effective gravity field do not obey the einstein equations . this is the main drawback of superfluid @xmath0he - a : since the fermi points in @xmath0he - a are too far apart from each other , the dynamical equations for the gravitational field are clumsy . nevertheless the above textures allow us to simulate many phenomena related to quantum vacuum in the presence of the strong gravity field . this is because many properties of the quantum vacuum in curved space , which are determined by the geometry , do not depend on the dynamical origin of the geometry . for example , it is well known that the hawking radiation is a purely kinematic effect and occurs in any geometry , if it exhbits an event horizon @xcite . that is why the @xmath0he - a quantum vacuum is a right object for simulation of many aspects of physics of vacuum in a curved space . for example what is the effect of the breakdown of the lorentz invariance at higher energy on hawking radiation . the entropy of the black hole can be also investigated using the above model , since the microstates within the horizon are well determined and ( at least in principle ) are completely known in the whole energy range including the `` transplanckian '' region . all these are different aspects of the problem of stability of the vacuum in strong gravitational and other fields . the superfluids provide many examples of the instability of the superfluid vacuum , and thus allow us to investigate different mechanisms of relaxation of the physical vacuum the ether . |
we use the ultra - deep hudf09 and the deep ers data from the hst wfc3/ir camera , along with the wide area/irac data from goods - s to derive seds of star - forming galaxies from the rest - frame uv to the optical over a wide luminosity range ( to ) from to . the sample contains , , , and 36 prior galaxies . median stacking enables the first comprehensive study of very faint high - z galaxies at multiple redshifts ( e.g. , = 27.4 \pm 0.1 $ ] ab mag for the sources at ) . at our faint median - stacked seds reach to and are combined with recently published results at high luminosity that extend to . we use the observed seds and template fits to derive rest frame uv - to - optical colors at all redshifts and luminosities . we find that this color does not vary significantly with redshift at a fixed luminosity . the uv - to - optical color does show a weak trend with luminosity , becoming redder at higher luminosities . this is most likely due to dust . at we find blue colors-[4.5]\sim-0.3\,$]mag that are most likely due to rest - frame optical emission lines contributing to the flux in the irac filter bandpasses . the scatter across our derived seds remains substantial , but the results are most consistent with a lack of any evolution in the seds with redshift at a given luminosity . the similarity of the seds suggests a self - similar mode of evolution over a timespan from 0.7 gyr to 1.5 gyr that encompasses very substantial growth in the stellar mass density in the universe ( from to ) | 1110.6441 | as a result of the installation of the wfc3/ir camera on the hubble space telescope , it is now possible to obtain extremely deep , high resolution observations in the near infrared . these observations match the exquisite optical data provided by the acs camera over deep and ultra - deep fields like the goods and the hudf @xcite . the upgrade has enabled several teams to extend their searches for lyman break galaxies ( lbgs ) to very high redshifts ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in combination , these two cameras have allowed for high s / n measurements of the rest - frame uv light of high redshift galaxies over a wide wavelength baseline . as a result of these new observations , the uv lfs at @xmath17 are known to high accuracy . another direct result from these observations is the determination of the uv slope of the seds , generally characterized by the parameter @xmath18 ( @xmath19 , e.g. , @xcite ) . this slope has been shown to become steeper ( bluer colors ) both with decreasing luminosity and with increasing redshift ( @xcite , but see @xcite ) . for a mid - aged ( @xmath20myr ) stellar population , this slope generally depends most strongly on the dust reddening so this has been interpreted as an increase in the dust content of galaxies with cosmic time and as galaxies increase in luminosity @xcite . since the rest - frame uv light primarily probes young short - lived stellar populations ( mostly o and b stars ) , it is interesting to extend these observations to longer wavelengths and study the relative importance of more evolved stellar populations . this is very challenging though and is now only possible thanks to the incredible sensitivity of the irac camera on the @xmath0 space telescope . some of the first observations of rest - frame optical light from @xmath17 galaxies showed quite substantial uv - to - optical colors @xcite . under the usual assumption of smooth , exponentially declining star formation histories ( sfhs ) , these colors were interpreted as indicative of an evolved stellar population ( ages @xmath21 myr ) in combination with the younger populations probed by the uv - light . follow - up work on @xmath22 galaxies continued to reveal systems with moderate size uv - to - optical breaks ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , even possibly to @xmath23 @xcite . however interesting , these studies are only possible for the brightest @xmath24 galaxies , due to the extreme depths required . some alternative explanations have been proposed for the red uv - to - optical colors observed in @xmath25 galaxies . for example , for young star - froming galaxies , one might expect strong line emission . at the right redshift , these lines could contaminate the 3.6 and 4.5 @xmath26 irac fluxes , resulting in red uv - to - optical colors similar to those produced by the balmer break @xcite . for emission lines to be strong enough to dominate the rest - frame optical fluxes , extremely young ages ( @xmath27myr ) are required . such models have ages shorter than the dynamical times of these high@xmath28 sources and are also somewhat uncertain . it would be surprising if the majority of these galaxies , which are selected over a redshift range that spans @xmath29myr , had such young ages . even if rest - frame optical emission nebular emission lines do not dominate the measured irac fluxes , it is expected that they will contribute some of the flux measured in the mid - ir , but the exact contribution currently remains hard to establish at these high redshifts . it has been routine , then , to compare the colors of these high redshift galaxies with synthetic stellar population ( ssp ) models of relatively evolved galaxies with smooth sfhs to derive physical properties from the observations . by studying lbg samples at different redshifts in the aforementioned way , one intriguing result has emerged . at a given intrinsic luminosity , the specific star formation rate ( sfr ) of galaxies seems to remain remarkably constant with redshift ( @xcite ; slightly modified to account for the latest dust corrections by @xcite ) . this result has called the attention of theorists and has proved very hard to reproduce in simulations ( e.g. , * * ; * ? ? ? the assumptions used to derive these result vary somewhat from group to group . in this work , we present the basic observations that lead to such a result , namely the observed seds of galaxies at @xmath22 . in the present work we study the observational properties of the typical high redshift galaxy . we do this by splitting our large sample in both redshift and uv - luminosity bins and constructing mean seds . the large number of galaxies allows us to study the typical colors at high s / n but also extend our measurements to faint limits that are currently inaccessible on an individual basis with the current data . although the scatter about the typical properties is also highly interesting , we defer their study to a future work which relies on even deeper @xmath0 observations . in section 2 we present the data and in section 3 the sample selection and photometry . section 4 describes the stacking procedure and we discuss the stacked seds and colors in section 5 . a summary is presented in section 6 . throughout , we use a ( @xmath30 ) = ( @xmath31 , 0.3 , 0.7 ) cosmology when necessary and we quote all magnitudes in the ab system @xcite . lcccc[h ! ] ers & 270 / 524 & 77 / 123 & 20 / 32 & 15 / 18 + hudf & 137 / 205 & 41 / 55 & 38 / 51 & 21 + + total & 407 / 729 & 118 / 178 & 58 / 83 & 36 + [ fig : magdist ] in this paper , we make use of the large number of @xmath17 galaxies found in the ultra - deep hudf data and wide - area ers data to construct median seds for star - forming galaxies at @xmath22 . the deep optical , near - ir , and deep irac coverage over these fields allow us to create seds that cover both the rest - frame uv and the rest - frame optical down to very faint fluxes . in constructing our median seds , we make use of the deep optical and near - ir hst data from acs and wfc3/ir . over the ers field , both the acs optical ( @xmath32@xmath33@xmath34@xmath35 ) and the wfc3/ir ( @xmath36@xmath37@xmath38 ) images from hst reach depths of @xmath39 mag ( @xmath40 measured on 0.35-diameter apertures ) . in the hudf field , the acs optical data ( @xmath41 ) are @xmath42 mags deeper , and the wfc / ir data ( @xmath43 ) are 1.5 times deeper than the ers . the hudf was observed with the @xmath44 filter instead of the @xmath36 filter used in the ers . because these fields are located inside the goods - s field , they both have deep @xmath0/irac coverage that amounts to @xmath45hr over the ers and twice that for the hudf which was observed to the same depths but in two different epochs with the irac camera rotated by @xmath46 . for irac channel 1 images ( @xmath47 ) , each single epoch reaches depths of 27.8 mags , measured over 2.5-diameter apertures ( @xmath48 ) . for channel 2 ( @xmath49 ) this limit corresponds to 27.2 mags . the criteria used to identify high redshift star forming galaxies has already been presented in numerous previous works . for details on the selection procedure and discussion of sample contamination please see @xcite . in short , the selection criteria used to find the sources presented here is as follows : + @xmath50-dropouts : @xmath51\ ] ] @xmath52 @xmath53-dropouts : @xmath54~\vee~(v_{606}-i_{775}>2)\}\ ] ] @xmath55 @xmath56-dropouts : @xmath57 the combined samples contain a total of 729 sources at @xmath4 , 178 at @xmath58 , and 83 at @xmath59 ( see table [ tbl : samplesummary ] ) . we complement the @xmath60 , and 6 samples with the sample of 36 @xmath3 sources presented by @xcite . we have adopted the @xmath3 stacks presented in that work . at these depths , the @xmath0/irac images are fairly crowded due to the size and shape of its psf . as a consequence , standard aperture photometry of the sources generally results in fluxes that are contaminated by flux coming from unrelated sources that are nearby in the sky . we are able to de - blend the fluxes from the different sources by making use of the higher resolution information available through the hst images . our method has been explained in previous works by our group ( e.g * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) but we give a short description here . we start by registering the irac images to the higher resolution hst image that will be used as a light profile template . in this case we use the wfc3/ir @xmath38-band image as our template because it is the closest in wavelength and has high signal to noise . we use the sextractor software @xcite to create segmentation maps . we isolate all the sources in a small area around each galaxy in our sample using these segmentation maps . we then convolve them with a kernel derived from the psfs of the hst and the irac images . we fit for the normalization fluxes of all sources ( including the source we are trying to clean ) simultaneously . finally we remove the flux from the unrelated sources to obtain a clean image of the galaxies in our sample . later we use these cleaned images to produce median stacks . the procedure described above does not always produce acceptable results . the reasons why it can fail vary but the most common is the presence of a source that is too bright , extended , and close to our region of interest . in these cases the amount of flux subtracted off our source of interest is large and more uncertain and in most cases results in large residuals . we have carefully inspected the residual images ( all sources subtracted , including the one in our sample ) and the cleaned images ( only unrelated neighbors subtracted ) of each of the sources in the original sample and come up with a sub - sample of sources with clean irac photometry . this `` by eye '' inspection criterion generally results in good agreement with a criterion that selects only the images with the lowest @xmath61 residuals but allows us to also remove sources that , even though have moderate @xmath61 values , present mild to strong residuals located very close to the source . overall , this reduces the sample to 60% of the original ( see table [ tbl : samplesummary ] ) . since the criterion adopted here depends on the complexity of the unrelated sources , it is not expected that it should introduce any biases that would be relevant in the determination of the mean rest - frame optical colors ( see figure [ fig : magdist ] ) . it has been shown that in star forming galaxies the total sfr appears to correlate with stellar mass , the so called main sequence of star forming galaxies @xcite . this sequence has been observed at high - redshifts ( @xmath17 ) but contrary to what has been observed at @xmath62 , the normalization of the relation at fixed stellar mass does not seem to evolve over @xmath63 @xcite . the lack of deeper rest - frame optical from @xmath0/irac for these sources is one of the big limitations to study this sequence to fainter magnitudes . in terms of the bulk of the population , however , progress can be made by stacking large numbers of sources to study their mean properties . in this section we seek to construct the median seds of galaxies at @xmath17 . due to the nature of the dropout search and the filters available , our full sample is naturally divided in redshift bins centered at @xmath64 plus the @xmath65 sample from @xcite . in this section , we further split each redshift sub - sample based on their observed rest - frame uv - luminosity . the filters used as reference uv - luminosity are the @xmath34 filter for the @xmath4 sources ; the @xmath35 for the @xmath58 sources ; the @xmath36 for the @xmath59 sources in the ers and the @xmath44 for the @xmath59 sources in the hudf ; and finally , the @xmath37 filter the @xmath3 sources . these filters were chosen to be close to rest - frame 1500 at the different redshifts and make the splitting criterion fairly uniform across redshifts . these bins are presented in figure [ fig : magdist ] . in the case of the hst acs and wfc3 imaging , the fluxes utilized correspond to the mag_auto values as determined using the sextractor code . for each uv - luminosity bin , the median flux in each band was determined from the individual flux measurements of the sources in the bin that also have clean irac determinations . the uncertainty in the median was estimated through a bootstrap process . in each realization , @xmath66 fluxes are drawn with replacement from the @xmath66 values available and perturbed by their individual uncertainties . generally the individual hst flux determinations have very high signal to noise . there are however , systematic uncertainties in the absolute calibration of the different bands and instruments across the sed . these uncertainties amount to at least 5% which we adopt as our minimum allowed uncertainty in any given band . in the case of the irac imaging , most sources are too faint to be individually detected even in these deep @xmath67hr integration images . instead of using individual fluxes , we first median combine ( pixel by pixel ) the cleaned stamps of the sources within a bin that have clean residuals from our irac cleaning procedure . we perform standard aperture photometry on the stacked images using a circular aperture of 2.5-diameter and correct the fluxes to total assuming psf profiles . this amounts to a factor 1.8 in both the [ 3.6 ] and the [ 4.5 ] irac filters . the local background is estimated from the stacks , in a ring between 6and 10diameters around the stamp center , where the source is expected to be , and then subtracted off the image . to estimate the uncertainty in the irac stacks , we also create bootstrap realizations . we treat the two epochs of irac data over the hudf as independent stamps in the stack , thereby effectively weighting with exposure time , i.e. , @xmath68 . in each bootstrap draw , then , we draw with replacement @xmath69 such stamps , and estimate fluxes as described above . the image noise in each stacks is expected to decrease with the total equivalent exposure time approximately as @xmath70 ( @xmath71hr ) . the actual photometric uncertainties that we derive are always greater than this because they also include the intrinsic scatter in the fluxes of the sources been stacked together ( see appendix ) . stamps for all the irac median stacks along with the derived magnitudes and uncertainties are presented in figures [ fig : bstackstamp ] , [ fig : vstackstamp ] , and [ fig : istackstamp ] . table [ tbl : seds ] is a compilation of all the stacked photometry including the @xcite stacks and information on the number of sources that go in each stack . it can be seen in the figures that even for sources as faint as 27mag , the stacks show good detections in both irac bands . the sample has been split according to their @xmath72 magnitude ( approx . rest frame 1500 ) in bins of 1 mag . the stamps are 10on a side . the number of sources stacked along with the total equivalent integration time has been included to the left of the stamps . the magnitudes included in the stamps were measured using circular apertures of 2.5 in diameter and corrected to total assuming stellar profiles which amounts to @xmath73 mags in both the [ 3.6 ] and the [ 4.5 ] channels . the background was estimated from the stacks and subtracted off the images . to estimate the uncertainties on the fluxes measured we created 200 random realizations of the stack . in each realization , sources were drawn with replacement and then median stacked pixel by pixel.,scaledwidth=40.0% ] but for the @xmath58 sample . in this case the sample was split according to their @xmath35 magnitude to approximate rest - frame 1500.,scaledwidth=40.0% ] but for the @xmath59 sample . in this case the sample was split according to their @xmath36 magnitude in the case of the ers sources and @xmath44 for the hudf sources ( approximately rest - frame 1500).,scaledwidth=40.0% ] the observed colors of galaxies can give us important information about the stellar populations that compose them . usually , this is accomplished by comparing their colors with those of synthetic stellar population ( ssp ) models . in this work , we chiefly explore the observed uv - to - optical colors and what can be learned from them and from their dependence on luminosity and redshift . when models are invoked , we use the ( * ? ? ? * bc03 ) models and we assume a @xcite imf , constant sfh , and @xmath75 metallicity , as these quantities can not be constrained from our data . to derive best fits we use the sed fitting code fast @xcite . the rest - frame uv - to - optical color is of particular interest at high redshifts . several studies have detected significant rest - frame uv - to - optical colors ( @xmath76mag ) in @xmath17 galaxies @xcite . the usual interpretation is that these colors derive from large balmer breaks which are associated with evolved stellar populations with ages @xmath29myr . this is surprising given how young the universe is at the time these galaxies are being observed . the main limitation for field studies is the depth of the @xmath0/irac imaging available , which constrains these studies to the brightest lbgs . searches behind clusters provide an alternative to probe the fainter populations ( see , e.g. , @xcite , and the recent tentative determination of a 1.5 mag balmer break in a @xmath59 galaxy by @xcite ) but the number of sources amenable to such studies , remain low . we work around this limitation by stacking large numbers of sources which allows us to estimate the typical rest - frame uv - to - optical colors of faint @xmath17 sources . we start by presenting the results found at @xmath4 where the similar and complementary work of @xcite allows us to study this colors over an unprecedentedly large range of uv - luminosities . galaxies in 7 bins of uv - luminosity . the faintest seds ( blue solid symbols and lines ) correspond to the seds derived from our sample . this sample has a mean redshift of @xmath77 and was split in bins of observed @xmath78mag , approximately covering the uv - luminosity range @xmath79 . the brightest stacks ( gray open symbols and dashed lines ) correspond to the median seds presented in @xcite , for a sample of galaxies with mean redshift @xmath80 . these seds allow us to extend the luminosity range to brighter luminosities up to @xmath11 . in @xcite the sources were split in six bins according to their observed @xmath81-band luminosity . for clarity , we only show every other bin but all are considered in figure [ fig : leebetaandutov ] . the x - axis shows wavelength and the approximate reference filter in their study . best fit bc03 models are included for references ( the @xmath82 and @xmath83band fluxes were excluded from the fits because they are likely biased see text , section 5.1 ) . a trend of bluer uv - slopes and smaller balmer breaks at fainter luminosities is observed.,scaledwidth=45.0% ] our sample of @xmath84dropouts has a mean redshift of @xmath85 . we divide this sample based on their @xmath34 luminosity ( which approximately corresponds to rest - frame 1500 ) in four bins of @xmath86 mag . these are the same bins as shown in the histogram in figure [ fig : magdist ] . the bins at @xmath87 mag are too faint to be detected in @xmath0/irac and will not be included in the analysis . assuming the mean redshift of the sample , these bins correspond approximately to @xmath88 and @xmath89 , which covers the moderately bright to faint end of the population ( @xmath90 , @xcite ) . in a recent study , @xcite focus on the median seds of the brightest @xmath4 galaxies found in the ground based noao deep wide - field survey . their sample has a mean redshift @xmath91 , and their seds also sample the rest - frame uv and optical using a combination of optical filters : @xmath92 ( mosaic camera ) , near - ir filters : @xmath93 ( newfirm camera ) , and the @xmath0/irac mid - ir channels 1 through 4 m and 8.0 @xmath94 m respectively ) is available for our sample in the goods - s , this data is still @xmath76mag shallower than the channel 1 & 2 . it is , therefore , much harder to obtain reliable stacked photometry in those bands , especially considering that our sample is much smaller and intrinsically fainter than the @xcite sample . we have not included these bands in the median - stacked seds presented in this work . ] . similar to the present work , they divide their sample according to the @xmath81band luminosity , although with uneven bins . their sources are intrinsically brighter than in this work , corresponding to @xmath95 sources roughly covering the range @xmath96 . the combination of the @xcite sample with our data spans the uv - luminosity range : @xmath97 . this represents an unprecedented dataset to study the seds , and especially the uv - to - optical colors , of @xmath4 galaxies over a large range of luminosities . the seds from both works are presented in figure [ fig : leecomparisonseds ] . in this figure the @xmath38 is shown with open symbols since it is likely biased and should be ignored as is explained in detail later . an overall trend to bluer colors ( both in the rest - frame uv and the rest - frame uv - to - optical colors ) towards fainter luminosities is already apparent in this figure . we quantify the uv - to - optical colors by estimating the interpolated rest - frame @xmath98 colors for the median - stacked seds in both sets . figure [ fig : leebetaandutov ] shows a not very steep but systematic trend of redder @xmath99 colors as a function of increasing uv - luminosity . the interpolated @xmath98 colors were determined from the best fits to the full observed sed , where the fits correspond to a linear combination of a set of template seds . these seds correspond to the default template set from the photometric code eazy @xcite . this template set was constructed and calibrated empirically to determine accurate photometric redshifts of galaxies of various types , from very blue starbursts to red ellipticals and has been shown to work well up to @xmath100 @xcite . for the @xmath101 and @xmath102 filters we have assumed narrow ideal filters that correspond to step functions of width 100 , centered at 3500 and 5500 for the @xmath101 and @xmath102 filters respectively . the colors were calculated from the best fit combination of templates using the filters described above . to estimate the uncertainty in the @xmath99 estimates , we perturbed the photometry and the redshift of each median - stacked sed within the allowed uncertainty , and a new fit was obtained . a new set of colors was measured from this best fit and the procedure was repeated a large number of times to derive the confidence intervals . a similar procedure was used to derive rest - frame @xmath103 magnitudes for an ideal filter centered at 1500 and with a width of 100 . for the data - point at @xmath104 , only an upper limit can be obtained due to the limitations of the template set used here . the eazy default template set only includes seds with @xmath105 . the sed at @xmath104 is likely to have a @xmath99 color of @xmath106mag , so no reliable lower limit can be obtained . we choose to show a @xmath107 upper limit . other more sophisticated algorithms to determine interpolated rest - frame colors for an observed sed yield very similar results ( e.g. , the interrest code , @xcite , which implements the algorithm described in @xcite ) . however , such algorithms also depend on the template set resulting in similar limitations . it is important to note that , when deriving this color from best fits , we have ignored the @xmath108band in our stacks and the @xmath82 and @xmath109bands in the @xcite stacks . given the redshift distribution of the sample it is expected that for some fraction of the sources ( the ones at the lowest redshifts ) the @xmath110band fluxes measured will include some rest - frame optical light . as a result , the estimated median rest - frame uv flux , will be biased towards larger values . to attempt a correction for this would require us knowing the exact redshift distribution of the sample , but even then a correction would be highly uncertain , so we opt to remove these photometric points from all fits . in recent works , @xcite accurately determine the uv - slope of lbgs at @xmath111 as a function of uv - luminosity and redshift . for a star - forming stellar population , the uv - slope , parametrized by @xmath18 ( @xmath112 ) , is most sensitive to the dust content of the galaxy . in consequence , the authors argue that the dependence of @xmath18 on uv - luminosity can be fully explained by an increase in the amount of dust that these galaxies contain as a they become brighter . if this is the case , a trend of redder @xmath99 colors for uv - brighter sources like the one observed here is also expected , since this color is also affected by dust extinction . the bottom panel of figure [ fig : leebetaandutov ] shows the @xmath99 colors corrected for dust extinction . these were derived as in @xcite , from the mean @xmath18 vs. uv - luminosity trend and the classic @xcite relation that relates the dust content with the uv - slope @xmath18 according to : @xmath113 where @xmath114 is the extinction in magnitudes at 1600 . a @xcite extinction curve is used to estimate the extinction at other wavelengths . color as a function of @xmath103 for the @xmath4 sources . blue solid symbols and errorbars correspond to the colors derived for the new seds in this work . similarly , we have derived interpolated @xmath99 colors for the seds presented in @xcite , dark open symbols and errorbars . although not very steep , a systematic trend to bluer @xmath99 colors with decreasing uv - luminosity is observed . the best fit trend of @xmath115 is shown by the dashed line and the 95 @xmath98 colors were determined from best fits to the observed data based on the default template set from the code eazy @xcite . we have used ideal narrow band @xmath101 and @xmath102 filters that correspond to step functions of 100 width , centered at rest - frame 3500 and 5500 respectively . the upper limit ( @xmath107 ) , arises from the limitations of the template set which only includes @xmath116 mags . the color of this sed is likely very close to this limit . _ bottom : _ the same @xmath99 colors after being corrected for dust extinction . the corrections were derived from the median uv - slope @xmath18 trends from @xcite and the local @xcite relation between @xmath18 and dust extinction . the corrected colors of both samples show flat trends with luminosity but seem to be disjoint . it is possible that the dust corrections are not equally adequate over the entire luminosity range . another possible explanation is that the contamination rates are different between the two samples . finally , it should be noted that the uv - slope in the @xcite sample is only poorly sampled . small offsets in the photometric calibration could bias the best fit models , from which the colors are estimated , to redder colors.,scaledwidth=45.0% ] cccrrrrrrrrrr[h ! ] @xmath34 & + 25 & 27 / 55 & 0 / 4 & @xmath117 & @xmath118 & @xmath119 & @xmath120 & @xmath121 & & @xmath122 & @xmath123 & @xmath124 & @xmath125 + 26 & 87 / 16 & 12 / 17 & @xmath126 & @xmath127 & @xmath128 & @xmath129 & @xmath130 & @xmath131 & @xmath132 & @xmath133 & @xmath134 & @xmath135 + 27 & 107 / 2 & 25 / 34 & @xmath136 & @xmath137 & @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath142 & @xmath143 & @xmath144 & @xmath145 + 28 & 45 / 78 & 43 / 74 & @xmath146 & @xmath147 & @xmath148 & @xmath149 & @xmath150 & @xmath151 & @xmath152 & @xmath153 & @xmath154 & @xmath155 + + @xmath35 & + 25 & 4 / 9 & 0 / 0 & @xmath156 & @xmath157 & @xmath158 & @xmath159 & @xmath160 & & @xmath161 & @xmath162 & @xmath163 & @xmath164 + 26 & 18 / 28 & 2 / 3 & @xmath165 & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 + 27 & 43 / 61 & 8 / 9 & @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & @xmath184 + 28 & 11 / 24 & 9 / 13 & @xmath185 & @xmath186 & @xmath187 & @xmath188 & @xmath189 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 + + @xmath195 & + 26 & 3 / 6 & 3 / 3 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 + 27 & 11 / 17 & 2 / 3 & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + 28 & 4 / 4 & 6 / 13 & @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 + + @xmath38 & + 26 & & & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & & @xmath231 & @xmath232 & @xmath233 & @xmath234 + 27 & & & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 & & @xmath240 & @xmath241 & @xmath242 & @xmath243 + 28 & & & @xmath244 & @xmath245 & @xmath246 & @xmath247 & @xmath248 & & @xmath249 & @xmath250 & @xmath251 & @xmath252 + and -19 . as seen in figure [ fig : colormag ] , the uv - to - optical colors of the seds depend weakly on redshift . here we perform the exercise of combining all the photometry from seds at different redshifts into a single sed . at different redshifts , our set of hst / acs+wfc3/ir+@xmath0/irac filters probe different rest - frame wavelengths . the combined seds are plotted in terms of their magnitudes relative to @xmath103 , which is measured from the best power law fit to the uv continuum of each median sed . an offset magnitude of @xmath253 and @xmath254 is added to reflect the approximate intrinsic magnitude of the seds that are being combined.,scaledwidth=45.0% ] . blue symbols correspond to the @xmath255 $ ] colors of the @xmath4 seds ; green : @xmath58 ; red : @xmath59 ; and black : @xmath3 . despite the sizable scatter , a systematic trend to redder uv - to - optical colors with increasing luminosity is observed at all redshifts . at the same time , there is no clear indication of a systematic trend with redshift . the dashed line is a best fit to all the points ( all redshifts simultaneously ) : @xmath256 = -0.11(\pm0.07 ) \times m_{1500,ab}+const$ ] , and the gray area corresponds to the 95 figure [ fig : leebetaandutov ] in that this figure shows the directly observed colors rather than interpolated rest - frame colors . in fact it is remarkable that this observed color , which samples different wavelength regions of the sed at the different redshift , remains fairly constant . this is probably an indication that the rest - frame optical colors measured by irac are fairly flat . _ center : _ the interpolated rest frame @xmath98 colors ( section 5.1 ) . the best fit trend to all the data is shown by the brown dashed line . for comparison , the black dashed line shows the best fit to the @xmath4 data only , as shown in figure [ fig : leebetaandutov ] , this trend is slightly flatter . _ bottom : _ the @xmath99 colors after being corrected by dust extinction following a simple prescription based on the uv - slopes and the local @xcite relation . except for the @xmath3 seds ( which have a much steeper @xmath18 vs. uv - luminosity relation ) , the dust corrected colors of all seds seem to show a flat trend ( although the scatter is large ) . this indicates that a change in dust is enough to explain the change in colors , both in the uv and the uv - to - optical . in particular , no age dependence on luminosity seems to be required.,scaledwidth=45.0% ] as can be seen from the figure , the dust corrected @xmath98 color is essentially constant with uv - luminosity . this again suggests that the main origin for the correlation of @xmath18 with the uv - luminosity is an increasing dust content at brighter magnitudes . otherwise we would expect some residual trend in the dust corrected color . this would be the case for example , if a strong age trend existed with uv - luminosity . as will be shown later , a similar flattening of the corrected colors is observed at all redshifts in the @xmath257 range . although both are flat , the trends of the corrected colors with uv - luminosity of the two samples are disjoint . the reasons for this are unclear but there are a couple of reasons that may help explain it . first , it is unclear that the mean trend used to derive dust corrections applies over such a large range of luminosities . this can help to explain that the corrected colors of the brightest sources are bluer than those of the faintest ones . however , it does not explain the break between the two samples . second there are differences between the selection of the two samples , in particular , it is not clear that the fraction of contaminants is comparable between the two samples . a larger fraction of contaminants in the brighter ground - based sample of @xcite as compared to the our data , could bias the brighter seds to smaller @xmath98 colors , resulting in excessive corrections for dust , which are derived from the mean trend on the uv with uv - luminosity . it should also be noted that in the case of the @xcite sample , the uv - slope is not very well sampled , with basically the @xmath81 and @xmath258bands driving the uv - slope ( since the @xmath110band has been excluded ) . even small systematic offsets of @xmath259mag in the zero - point calibration between these two filters can bias the best fit models to redder slopes and in consequence , smaller uv - to - optical colors . cccc[h ! ] + @xmath260 & @xmath261 & @xmath262 & @xmath263 + @xmath264 & @xmath265 & @xmath266 & @xmath267 + @xmath268 & @xmath269 & @xmath270 & @xmath271 + @xmath272 & @xmath273 & @xmath274 & @xmath275 + + + @xmath276 & @xmath277 & @xmath278 & @xmath279 + @xmath280 & @xmath281 & @xmath282 & @xmath283 + @xmath284 & @xmath285 & @xmath286 & @xmath287 + @xmath288 & @xmath289 & @xmath290 & @xmath291 + + + @xmath292 & @xmath293 & @xmath294 & @xmath295 + @xmath296 & @xmath297 & @xmath298 & @xmath299 + @xmath300 & @xmath301 & @xmath302 & @xmath303 + + + @xmath304 & @xmath305 & @xmath306 & @xmath307 + @xmath308 & @xmath309 & @xmath310 & @xmath279 + @xmath311 & @xmath312 & @xmath313 & @xmath314 + we now extend our determination of the sed of high-@xmath74 galaxies to @xmath58 and @xmath59 , and combine our determinations with the stacked seds of sources at @xmath3 recently presented in @xcite . figure [ fig : seds ] shows the seds at @xmath315 . the expected mean redshifts of these samples correspond to @xmath316 and @xmath317 , with typical spread in redshift of @xmath318 . the samples at each redshift have been split in bins of 1mag , using as reference the observed magnitude in the band closest to 1500 ( table [ tbl : seds ] contains the seds and information about the samples in each bin ) . best fit bc03 models with csf and @xmath319 metallicity are overlaid for reference . as explained in the previous section , we have excluded from the fitting the bands that are potentially biased ( marked by the open symbols ) . as can be seen in this figure , all these seds are remarkably similar and show sizable uv - to - optical colors . the similarity of the seds can also be appreciated in figure [ fig : combseds ] , where the seds from the different redshift samples have been grouped according to their approximate @xmath103 luminosities . the seds in the three @xmath103 bins have been re - normalized according to their @xmath103 luminosity and combined to produce over - sampled seds , which take advantage of the fact that the different filters probe different rest - frame wavelengths at the different redshifts . the combined seds show the typical flat uv slopes , but also rather flat colors in the rest - frame optical , albeit with a relatively large scatter . this scatter can in part be a consequence of the normalization which is done in the far uv , which effectively acts as the pivot point . different emission lines that sometimes contribute flux in the irac bands , depending on the redshift , can also be contributing to this scatter . we first consider the implications if the sizable red @xmath320 $ ] colors that we observe are assumed to be exclusively measuring the stellar continuum in these high redshift galaxies . the red color would arise from large balmer breaks , and would be indicative of evolved stellar populations . as can be seen in figure [ fig : agemodels]a , the @xmath320 $ ] colors characteristic of these sources would then require fairly large ages of @xmath321 gyr if no dust is allowed in the models ( the model tracks in the figure are csf models ) . allowing some reddening by dust ( using in this case a @xcite law ) , reduces their ages considerably ( figure [ fig : agemodels]c ) . the amount of dust reddening is constrained however . based on the uv continuum slopes typically observed in these galaxies @xcite , and using the @xmath322 @xcite relation between this slope and the dust content , it has been shown that the reddening at @xmath17 is typically @xmath323mag . the reddening is likely to be even smaller at higher redshifts @xcite . with this added constraint , the ages derived from sed fitting are generally estimated to be @xmath324 myr ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in our seds at @xmath13 we also find that the @xmath325-[4.5])$ ] color is consistently blue , with typical values of @xmath326mags . this color is very hard to reproduce with a model with any star formation history that is based only on stellar continuum . the addition of dust only makes the comparison worse ( figure [ fig : agemodels]c , d ) . it should be noted , however , that the uncertainties in this color are considerable and that in most cases the observed color is formally consistent with the stellar continuum only models . it has been claimed @xcite that at some of these redshifts , particularly @xmath59 and 7 , the seds could be better fit by very young stellar population models ( @xmath327 myr old ) with moderate dust content and very strong nebular emission lines . these lines would dominate the rest - frame optical fluxes . as the galaxies are forming stars , it is likely that they have nebular emission lines , and so we have assessed whether emission lines could be influencing our fluxes and fits . the observed position of the most prominent optical nebular emission lines at each redshift is marked in figure [ fig : seds ] by the dashed vertical lines ( assuming the mean redshift of each sample ) . it can be seen that at the different redshifts involved , some of the rest - frame optical emission lines would lie within the irac [ 3.6 ] and [ 4.5 ] filters . @xcite showed recently that at @xmath328 there are sources that show emission lines with large enough equivalent widths to dominate the broadband fluxes in the optical . however , even for their sample of high equivalent width selected sources , such objects are not very common . for the more typical cases , the lines that they find in their sample would have moderate contributions to the irac broad bands , usually making them @xmath329mags brighter . at higher redshifts , @xmath330 , the work of @xcite , based on spectroscopic redshifts and irac broadband photometry , finds similar results . they find that the sources in this redshift range show an excess of flux in [ 3.6 ] over [ 4.5 ] , which the authors think can be explained by h@xmath331 emission . the majority of the sources in their sample have @xmath325-[4.5])$ ] colors that are consistent with an h@xmath331 contribution of @xmath329mags to the [ 3.6 ] filter , indicating moderate contributions from emission lines . we consider the effects of a similar ( moderate ) contribution of emission lines to the colors of a 280myr old csf model . for the purpose of this exercise we assume a h@xmath331 rest - frame equivalent width @xmath332 . at @xmath59 , this would increase the luminosity in the [ 4.5 ] filter by 0.23mag . we only consider the strongest lines in the optical and assume the strength ratios presented by @xcite for a @xmath333 metallicity system . this corresponds to @xmath334 , respectively . the open squares in figure [ fig : agemodels]a , b , correspond to the median colors for a model like the one described ( the median takes into account the expected redshift distribution of each sample ) . as can be seen in figure [ fig : agemodels]a , the main effect of the contribution from the emission lines is that it allows for significantly younger ages to reproduce the @xmath320 $ ] colors ( by roughly a factor of two ) . since this implies lower stellar continuum fluxes than previously assumed , this would also result in lower best fit stellar masses than previously estimated for galaxies at these redshifts ( by a similar factor ) . simultaneously , given the line ratios and the redshift distributions expected for these samples , a model like this can also help in reproducing the blue [ 3.6]-[4.5 ] colors of the seds at @xmath13 ( figure [ fig : agemodels]b ) . while formally models with just stellar continua can fit the data , given the current large uncertainties , such colors are hard to obtain with stellar continuum only models . it is more likely that some degree of contribution of emission lines provides a better solution . the assessment presented here is necessarily very simplistic , given the current uncertainties , and is only meant to show the possible effects of emission lines on the colors of these galaxies . better estimates of the contribution of these lines and its effects on the properties derived from sed fitting will depend on the actual strengths and ratios of the lines . moderate improvement can be achieved if more precise redshifts are known for these sources , such data does not exist at this time for such faint high redshift galaxies and so this remains an open question until improved spectroscopic capability becomes available . regarding the trend of redder uv - to - optical colors for brighter uv - luminosities presented in figure [ fig : leebetaandutov ] for the @xmath4 sample , it can also be observed for the other samples . in fact , within the uncertainties in the color determination , the @xmath320 $ ] vs @xmath103 trend is roughly the same at @xmath58 and 6 ( figure [ fig : colormag ] , _ top _ ) . this is particularly remarkable in view of the fact that this color is an observed color , not a rest - frame color , and in consequence is probing different wavelengths for the different samples . this shows again how similar and flat these seds are . if all the seds are considered simultaneously , a trend of @xmath256 = -0.12(\pm0.07 ) \times m_{1500,ab}+const$ ] , is found . for simple csf models , this trend implies a variation in stellar m / l ratio . @xcite find that at @xmath4 , this variation is of a factor @xmath335 between @xmath336 and @xmath337 . so , despite the steep uv - luminosity functions characteristic at @xmath17 , the contribution of the faintest sources to the total stellar mass density is more modest than their contribution to the star formation rate density . similar to the analysis performed at @xmath4 , we have derived rest - frame @xmath98 colors for the median seds at all redshifts . these colors are shown in figure [ fig : colormag ] . the trend with uv - luminosity has large scatter and low significance but is systematic in the sense that brighter sources are redder in @xmath99 . a fit to the rest - frame colors as a function of m@xmath338 of all redshifts simultaneously , results in @xmath339 , shown as the brown dashed line in figure [ fig : colormag ] ( _ center _ panel ) . this is slightly steeper than the @xmath340 slope found for the @xmath4 data only , also plotted for reference as the black dashed line in the same figure . the latter trend included the bright sample of @xcite and is consistent within the uncertainties with the fit to all the points . using the median trends determined by @xcite regarding the @xmath18 slopes vs. uv - luminosity for @xmath341 lbgs , we corrected the @xmath342 colors for dust extinction using the @xcite relation ( figure [ fig : colormag ] , _ bottom _ ) . these corrected colors show large scatter around @xmath343 but there is no indication of a residual trend with uv - luminosity except for the @xmath3 sample ( black symbols ) . this may indicate again that at @xmath344 , the evolution can be fully explained by a change in dust content and there is no strong evolution in the ages of these sources as a function of uv - luminosity . this is consistent with the report by @xcite based on stellar population modeling of slightly brighter lbgs in this redshift range . it is also consistent with the predictions from the hydrodynamical simulations of @xcite , which find no trend of age with uv - luminosity . at @xmath3 there seems to be a residual trend in @xmath345 vs. uv - luminosity in the sense that brighter sources have systematically bluer intrinsic @xmath98 colors . in principle , this could indicate that the brighter sources are actually younger than the fainter sources , which would be difficult to imagine . alternatively , it is possible that the simple dust corrections ( calibrated at @xmath322 ) are not adequate at these high redshifts , due to , for example , changes in the imf , metallicity , or escape fraction , all of which could alter the uv - slope @xmath18 for a given dust content . it should also be noted that the @xmath18 vs. uv - luminosity at @xmath3 is considerably steeper than at the other redshifts and so the corrections to the @xmath99 vs. uv - luminosity trend are more extreme . nevertheless , the uncertainties are sufficiently large that also the @xmath3 trend is fully consistent with being constant . determining the rest - frame uv - to - optical colors of individual uv - faint galaxies at @xmath17 is challenging . at such redshifts the rest - frame optical lies at @xmath346 microns in the mid - ir and so it is hard to measure at the extremely faint magnitudes typical of galaxies in the first 1.5 gyr . to gain insight into the typical spectral properties of @xmath347 galaxies , we have taken advantage of the large samples of @xmath347 sources found in the deep hst acs and wfc3/ir images of the ers and hudf fields . the hst data give us rest - frame uv fluxes . we then use @xmath0/irac [ 3.6 ] and [ 4.5 ] micron data from the deep goods survey of these fields to get rest - frame optical fluxes by stacking the irac images and determining median flux values . this allows us to determine the seds of these galaxies covering both the rest - frame uv and the optical . these seds represent a first comprehensive study of the rest - frame uv - to - optical properties of @xmath17 galaxies as a function of both redshift and uv - luminosity down to very faint limits . at @xmath17 we have also combined our faint seds with the @xmath10 stacks presented in @xcite . this allows us to assess the colors of @xmath17 sources over an unprecedentedly large range of luminosities . our main findings are as follow : * at @xmath17 the @xmath98 rest - frame color ( interpolated from the data using a fitted sed from a set of templates is of @xmath348 mags . there is a shallow but systematic trend of redder colors for brighter uv - luminosities over the range @xmath349 ( figure [ fig : leebetaandutov ] , section 5.1 ) . a linear fit to this trend is @xmath350m@xmath351 * the seds of star forming galaxies at @xmath352 are remarkably similar at all luminosities from @xmath353 , showing fairly flat rest - frame uv and rest - frame optical colors ( section 5.2 and figures [ fig : seds ] , [ fig : combseds ] and [ fig : colormag ] ) . at a given redshift , there are weak indications of a subtle trend for redder colors for brighter sources . a simple fit to the data at all redshifts simultaneously results in @xmath354m@xmath355 ( figure [ fig : colormag ] _ center _ , section 4.2 ) . * the uv - to - optical color , as measured by the observed @xmath320 $ ] , color remains fairly constant with redshift , despite the @xmath0/irac bands probing different wavelength regions of the sed ( figure [ fig : colormag ] _ top _ ) . this suggests that the optical colors are fairly flat at all redshifts @xmath22 , although there is significant scatter , which can be caused by optical emission line contamination to the irac filters . * the @xmath13 seds show consistently blue @xmath8-[4.5]\sim-0.3 $ ] colors . this is hard to reproduce with models that only include stellar continuum , although it is still formally possible given the current uncertainties . nonetheless , it is probably better explained with moderate flux contributions from optical nebular emission lines in the two irac bands . including some contribution from emission lines leads to lower best fit ages and lower stellar masses than previously estimated for galaxies at these redshifts . changes of a factor 2 could well result , but we would caution that detailed assessments are not yet possible with the current data . * we derive dust - corrected @xmath12 colors using a simple dust correction based on the median uv - slope @xmath18 vs. uv - luminosity relation from @xcite and the @xcite relation ( figure [ fig : colormag ] , _ bottom _ ) . the dust - corrected uv - to - optical colors at @xmath344 are @xmath356mags . they do not appear to depend on uv - luminosity . this may suggest that moderate changes in the dust content of galaxies can explain the color dependency observed both in the uv - slopes and the uv - to - optical colors . in particular , it seems that no age evolution is required to match this color dependency . the recent deep and ultra - deep hst acs and wfc3/ir imaging programs over the goods - s field have resulted in large samples of @xmath4 galaxies . taking advantage of these large samples , along with a careful stacking analysis , we present a first comprehensive study of the typical seds over the rest - frame uv and the rest - frame optical for faint star forming galaxies in the @xmath357 redshift range . the stacked seds allow us to study the uv - to - optical colors of high - z star forming galaxies down to very low luminosities . we find a weak trend to bluer color at fainter luminosities . interestingly , these stacked seds also show a remarkable similarity across redshift in a period of considerable stellar mass growth from about 0.8 gyr to 1.5 gyr , suggesting a smooth self - similar mode of evolution at @xmath358 . we acknowledge support from nasa grant hst - go-11563 , and nasa grant hst - go-11144 . this work is based on observations made with the hubble space telescope and with the spitzer space telescope . support for this work was provided by nasa through an award issued by jpl / caltech . p.o . acknowledges support provided by nasa through hubble fellowship grant hf-51278.01 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa , under contract nas 5- 26555 . 51 natexlab#1#1 , p. , & fritze - v . alvensleben , u. 2003 , , 401 , 1063 , h. , et al . 2011 , arxiv:1109.0639 , e. , & arnouts , s. 1996 , , 117 , 393 , r. j. , et al . 2009 , , 705 , 936 , r. j. , illingworth , g. d. , franx , m. , & ford , h. 2007 , , 670 , 928 , r. j. , et al . 2011 , arxiv:1109.0994 . 2010 , , submitted , arxiv:1006.4360 . 2010 , , 709 , l133 , l. d. , et al . 2008 , , 678 , 647 , g. b. , van dokkum , p. g. , & coppi , p. 2008 , , 686 , 1503 , g. , & charlot , s. 2003 , , 344 , 1000 , a. j. , et al . 2010 , , 409 , 855 , d. , armus , l. , bohlin , r. c. , kinney , a. l. , koornneef , j. , & storchi - bergmann , t. 2000 , , 533 , 682 , j. s. , mclure , r. j. , robertson , b. e. , ellis , r. s. , stark , d. p. , cirasuolo , m. , & de ravel , l. 2011 , arxiv:1102.5005 , r. , santos , m. r. , kneib , j .- p . , & kuijken , k. 2001 , , 560 , l119 , l. p. , bunker , a. j. , ellis , r. s. , lacy , m. , stanway , e. r. , stark , d. p. , & chiu , k. 2007 , , 374 , 910 , l. p. , bunker , a. j. , stanway , e. r. , lacy , m. , ellis , r. s. , & doherty , m. 2005 , , 364 , 443 , s. l. , papovich , c. , giavalisco , m. , reddy , n. a. , ferguson , h. c. , koekemoer , a. m. , & dickinson , m. 2010 , , 719 , 1250 , s. l. , et al . 2011 , arxiv:1110.3785 , k. , oppenheimer , b. d. , & dav , r. 2011 , , 410 , 1703 , m. , et al . 2004 , , 600 , l93 , v. , labb , i. , bouwens , r. j. , illingworth , g. , franx , m. , & kriek , m. 2011 , , 735 , l34 + , v. , labb , i. , bouwens , r. j. , illingworth , g. , franx , m. , kriek , m. , & brammer , g. b. 2010 , , 713 , 115 , s. , & silk , j. 2011 , , 410 , l42 , m. , van dokkum , p. g. , labb , i. , franx , m. , illingworth , g. d. , marchesini , d. , & quadri , r. f. 2009 , , 700 , 221 , i. , bouwens , r. , illingworth , g. d. , & franx , m. 2006 , , 649 , l67 , i. , et al . 2010 , , 716 , l103 . 2010 , , 708 , l26 , n. , et al . 2011 , , 531 , a74 + , k .- s . , et al . 2011 , , 733 , 99 , r. j. , dunlop , j. s. , cirasuolo , m. , koekemoer , a. m. , sabbi , e. , stark , d. p. , targett , t. a. , & ellis , r. s. 2010 , , 403 , 960 , r. j. , et al . 2011 , arxiv:1102.4881 , g. r. , heckman , t. m. , & calzetti , d. 1999 , , 521 , 64 , k. g. , et al . 2007 , , 660 , l43 , p. a. , et al . 2010 , , 709 , l16 . 2011 , arxiv:1105.2297 , j. b. , & gunn , j. e. 1983 , , 266 , 713 , j. , kneib , j .- p . , ebeling , h. , stark , d. p. , egami , e. , & fiedler , a. k. 2011 , , 414 , l31 , j. , stark , d. p. , ellis , r. s. , george , m. r. , egami , e. , kneib , j .- smith , g. p. 2008 , , 685 , 705 , g. , et al . 2003 , , 599 , 847 , e. e. 1955 , , 121 , 161 , d. , & de barros , s. 2009 , , 502 , 423 . 2010 , , 515 , a73 + , h. , chary , r .- r . , dickinson , m. , lin , l. , spinrad , h. , stern , d. , & yan , c .- h . 2011 , arxiv:1103.4124 , d. p. , ellis , r. s. , bunker , a. , bundy , k. , targett , t. , benson , a. , & lacy , m. 2009 , , 697 , 1493 , e. n. , et al . 2009 , , 183 , 295 , s. m. , neistein , e. , & dekel , a. 2011 , arxiv:1103.3011 , s. m. , bunker , a. j. , stanway , e. , lorenzoni , s. , & caruana , j. 2011 , , 1340 , h. , dickinson , m. , giavalisco , m. , stern , d. , eisenhardt , p. r. m. , & ferguson , h. c. 2006 , , 651 , 24 , h. , et al . 2005 , , 634 , 109 , h. , windhorst , r. a. , hathi , n. p. , cohen , s. h. , ryan , r. e. , oconnell , r. w. , & mccarthy , p. j. 2010 , research in astronomy and astrophysics , 10 , 867 to quantify the rest - frame optical properties of very faint high - redshift galaxies , we stack the irac images for a large number of faint galaxies , after removing the flux contribution from nearby neighbors ( section 4 ) . while we do not expect large systematics in the photometry of individual sources from our deblending procedure to remove neighboring sources , it is possible that modest systematics could arise when stacking a large number of sources . the role of this appendix is to determine how large such systematics might be . fortunately we are able to demonstrate that the fluxes that we measure in the stacked irac images are unbiased and that their significance is properly determined . we started by selecting 200 empty areas in the hudf field . we select them based on the segmentation maps derived from the high resolution hst images . we processed each of the selected empty areas as if they were the position of one of the real sources in our catalog , i.e. we ran them through the de - blending code to remove the flux from nearby un - associated sources . to determine the rms for a given @xmath359 number of stacked stamps , we randomly draw ( with replacement ) @xmath359 stamps from the 200 empty areas , median combine them and perform aperture photometry in the median stack in the same fashion as for our real stacks . we repeat the drawing 300 times and measure the rms of these 300 trials . as can be seen in the upper panel of figure [ fig : stacklimits ] , the limiting flux decreases as @xmath360 , as expected for background - limited noise . in the case of the stacks that contain real sources , the rms is larger than the image noise because it is also affected by the intrinsic variation in the distribution of fluxes that are being stacked . limiting flux ( in @xmath94jy ) as a function of the number of irac stamps stacked . all the stamps have the same integration time of @xmath361h . the estimates ( open symbols ) come from stacking large numbers of empty areas of the goods images . each empty area was cleaned of neighboring fluxes using the code described in the text . the noise decreases @xmath362 , with n being the number of empty regions stacked . the filled symbols are the actual rms determined for each of our stacked stamps : blue for the @xmath4 stamps , green @xmath58 , and red @xmath59 ( see figures [ fig : bstackstamp ] , [ fig : vstackstamp ] , and [ fig : istackstamp ] ) . these rms values are @xmath363 than the ones derived for empty regions . this is expected because the uncertainties in the stack include both the image noise and intrinsic variations in the population being stacked . _ lower : _ the flux of the median combined stacks of cleaned empty regions as a function of the number of stamps combined . as expected , this is very close to zero.,scaledwidth=80.0% ] |
this paper proposes a new semi analytic modelling of galaxy properties in the ir / submm wavelength range , which is explicitly set in a cosmological framework . we start from a description of the non dissipative and dissipative collapses of primordial perturbations , and add star formation , stellar evolution and feedback , as well as the absorption of starlight by dust and its re emission in the ir and submm . this type of approach has had some success in reproducing the _ optical _ properties of galaxies . we hereafter propose a simple extension to the ir / submm range . the growth of structures is followed according to the standard cold dark matter model . we assume that star formation proceeds either in a `` quiescent '' mode , e.g. as in disks , or in a `` burst '' mode with ten times shorter time scales . in order to reproduce the current data on the evolution of the comoving cosmic sfr and gas densities , we need to introduce a mass fraction involved in the `` burst '' mode strongly increasing with redshift , probably reflecting the increase of interaction and merging activity . we estimate the ir / submm luminosities of these `` mild starburst '' and `` luminous uv / ir galaxies '' , and we explore how much star formation could be hidden in heavily extinguished , `` ultraluminous ir galaxies '' by designing a family of evolutionary scenarios which are consistent with the current status of the `` cosmic constraints '' , as well as with the _ iras _ 60 m luminosity function and faint counts , but with different high ir luminosity densities . however , these scenarios generate a cosmic infrared background whose spectrum falls within the range of the isotropic ir component detected by puget _ et al . _ ( 1996 ) and revisited by guiderdoni _ et al . _ ( 1997 ) . we give predictions for the faint galaxy counts and redshift distributions at ir and submm wavelengths . the submm range is very sensitive to the details of the evolutionary scenarios . as a result , the on going and forthcoming observations with _ iso _ and scuba ( and later with _ sirtf _ , sofia , _ first _ and _ planck _ ) will put strong constraints on the evolution of galaxies at and beyond . cosmology galaxy formation galaxy evolution infrared submm . | astro-ph9710340 | this paper describes a new modelling of galaxy evolution in the infrared and submm wavelength ranges , which are now open to high redshift exploration by the on going observations with _ iso _ and scuba , and forthcoming facilities and experiments such as _ sirtf _ , sofia , _ first _ and _ planck_. as a matter of fact , our knowledge of the early epochs of galaxies has recently increased thanks to the richness and precision of the observational evidence obtained by uv / visible / nir surveys of high redshift objects ( lilly _ et al . _ 1995 ; ellis _ et al . _ 1996 ; cowie _ et al . _ 1996 ; steidel _ et al . _ 1996 ; williams _ et al . the pattern of galaxy evolution which emerges from these data can be summarized as follows : i ) faint galaxy counts show the presence of a large number of blue objects , well in excess of no evolution expectations ( williams _ et al . ii ) these blue objects are `` sub@xmath4 '' galaxies undergoing strong bursts of star formation ( lilly _ et al . _ 1996 ; ellis _ et al . iii ) the fraction of these blue objects with unclassified / peculiar morphologies showing signs of tidal interaction and merging ( abraham _ et al . _ 1996 ) increases from local samples to the hst high resolution observations of the medium deep survey ( griffiths _ et al . _ 1994 ) and hubble deep field ( williams _ et al . iv ) the global star formation rate ( hereafter sfr ) of the universe declined by a factor of about ten since redshift @xmath3 ( lilly _ et al . _ 1996 ; madau _ et al . _ 1996 ; sawicki _ et al . _ 1997 ; connolly _ et al . v ) finally , this high sfr seems to be correlated to the decrease of the cold gas comoving density associated with damped lyman@xmath5 systems between @xmath6 and @xmath7 ( storrie lombardi _ et al . these results nicely match a picture in which star formation in bursts triggered by interaction / merging consumes the gas content of galaxies as time goes on . it is common wisdom that such a qualitative scenario is expected within the paradigm of hierarchical growth of structures . the implementation of hierarchical galaxy formation in semi analytic models quantitatively substantiates this view and may further suggest that we have seen the bulk of star formation and understood the broad features of galaxy formation ( baugh _ et al . _ 1997 ) . however , it should be emphasized that this seemingly consistent view is entirely based on visible / nir observations which only probe the rest frame uv / visible light of high@xmath1 objects . the total amount of energy released by star formation should be estimated by summing up the uv / visible / nir light of stellar populations directly escaping from galaxies , _ and _ the part which has been absorbed by dust and re emitted in the ir / submm wavelength range . the corrections needed to account for optical extinction by dust are rather uncertain and could easily induce an upward revision of the high redshift sfr deduced from rest frame uv / visible observations by factors of a few . moreover , a significant fraction of star formation might be completely hidden in heavily extinguished galaxies which are missed by the above mentionned surveys . even for normal objects like spiral galaxies , the fraction of energy emitted in the ir can amount to 30 % of the optical , and this can increase to more than 95 % for ultraluminous starbursts . as a matter of fact , the ir / submm range might be more adequate than the optical for observing high redshift , starburst galaxies provided they are dusty . frame ir / submm spectrum of dust heated by a young stellar population peaks between 50 to 100 @xmath0 m , pointing out the submm range as particularly relevant for the search of primeval , high redshift galaxies . at this time , we know very little about the `` optically dark '' side of galaxy evolution . the _ iras _ satellite has given the first complete survey of fir galaxy properties , in four bands between 12 and 100 @xmath0 m . many studies have emphasized the wide variety of ir luminosities , from `` normal spirals '' to `` mild starbursts '' , and finally to the `` luminous ir galaxies '' ( hereafter ligs ) , mostly interacting systems , and the spectacular `` ultraluminous ir galaxies '' ( hereafter uligs ) , which are mergers ( sanders and mirabel 1996 ; clements _ et al . _ unfortunately , while we have learned from _ iras _ that about one third of the bolometric luminosity of the local universe is released in the ir / submm ( soifer and neugebauer 1991 ) , we know very little about galaxy evolution in this wavelength range . faint galaxy counts and redshift surveys down to flux densities @xmath8 mjy ( at 60 @xmath0 m ) do not probe deeper than @xmath9 ( ashby _ et al . _ 1996 ; clements _ et al . _ 1996a ) . these surveys seem to show a strong luminosity and/or density evolution of _ iras _ sources , but it is difficult to extrapolate this trend to higher redshifts on a firm ground . in spite of its observational limits , _ iras _ has also revealed the existence of _ iras _ 10214 + 4724 , a very peculiar , `` hyperluminous '' galaxy at @xmath10 ( rowan robinson _ et al . _ 1991a ) , though this object is likely affected by lensing ( eisenhardt _ et al . it is expected that the _ iso _ satellite will complete and detail this picture , in a broader wavelength range from a few @xmath0 m to 200 @xmath0 m . other projects , such as _ sirtf _ , will give access to better sensitivity and imaging capabilities . the properties of galaxies in the submm range are sensitive to the spectral characteristics of dust , especially its emissivity at large wavelengths which is not constrained by _ iras _ observations alone . with respect to the relative wealth of data in the fir , the submm emission of galaxies is poorly known . the observational literature gathers the submm fluxes of only a few tens galaxies which have been measured from ground based or aircraft borne instruments . these observations are difficult and some of the estimates of the amount of energy released in this range happen to be strongly discrepant ( e.g. chini _ et al . _ 1986 ; stark _ et al . _ 1989 ; eales _ et al . _ 1989 ; chini and krgel 1993 ) . however , the observational situation will soon evolve with the start of scuba operations ( and later , with sofia and especially _ first _ ) . about twenty counterparts of radiogalaxies and quasars have already been observed at submm / mm wavelengths ( see e.g. hughes _ et al . _ 1997 ; and references therein ) . although their evolutionary status is not fully understood , these objects give a first flavour of the future deep surveys which will be achieved by forthcoming instruments . theoretical modelling is needed to optimise observational strategies and also to help assess the point source contribution of the forthcoming cmb satellites map , and especially _ planck_. these experiments aim at mapping the anisotropies of the submm / mm sky on scales well below the degree . on these scales , the separation of the various foregrounds and backgrounds which superimpose to the fluctuations of the cmb is more difficult than for the 7@xmath11 field of view of the very successful _ cobe_/dmr experiment . in the case of _ planck _ , preliminary studies have shown that several thousands of galaxies should be detected at several wavelengths between 400 @xmath0 m and 1 cm . the epoch of galaxy formation can also be observed by its imprint on the background radiation which is produced by the line of sight accumulation of extragalactic sources . the search for the `` cosmic optical background '' ( hereafter cob ) currently gives only upper limits . nevertheless , the shallowing of the faint counts obtained in the hdf suggests that we are now close to convergence ( williams _ et al . thus the lower limit to the cob obtained by summing up the contributions of faint galaxies is likely close to the real value . at longer wavelengths , the dirbe instrument on _ cobe _ has given upper limits on the fir background between 2 and 300 @xmath0 m ( hauser 1995 ) , while puget _ ( 1996 ) have discovered an isotropic component in the _ cobe_/firas residuals between 200 @xmath0 m and 2 mm , which could be the long sought `` cosmic infrared background '' ( hereafter cib ) . the presence of this component is confirmed by a new analysis restricted to the cleanest regions of the sky , where the foreground galactic components are negligible ( guiderdoni _ et al . such a detection seems to yield the first `` post_iras _ '' constraint on the high@xmath1 evolution of galaxies in the ir / submm range , before the era of _ iso _ results . its level comparable to the above mentionned estimate of the cob suggests that a significant fraction of the energy of young stars is absorbed by dust and released in the ir / submm . the models which have been proposed to predict the faint counts and background radiation in the ir / submm can be classified as `` backward evolution '' and `` forward evolution '' , following the good review by lonsdale ( 1996 ) . in the first class of models , the ir / submm luminosity function undergoes luminosity and/or number evolution which are simply parameterized as power laws of @xmath12 ( e.g. weedman 1990 ; beichman and helou 1991 ; hacking and soifer 1991 ; oliver _ et al . _ 1992 ; treyer and silk 1993 ; blain and longair 1993 ; pearson and rowan robinson 1996 ) . these power laws are generally derived from fits of _ iras _ galaxy counts ( which do not probe deeper than @xmath13 ) . then they are extrapolated up to redshifts of a few units . unfortunately , various analyses of _ iras _ deep counts yield discrepant results at @xmath14 mjy , and the issue of the evolution is very controverted ( see e.g. bertin _ et al . _ 1997 , for a new analysis and discussion ) . in the second class of models , the photometric evolution in the ir / submm is computed by implementing some of the involved physical processes . for instance : chemical evolution which rules the amount of dust responsible for the ir / submm emission is modelled in wang ( 1991a , b ) , and eales and edmunds ( 1996a , b ) . in addition to chemical evolution , the photometric evolution of stellar populations which heat the dust is modelled by franceschini _ ( 1991 , 1994 ) and fall , charlot and pei ( 1996 ) . while the previous models assume a simple relation between the dust content and the heavy element abundance of the gas , dwek and varosi ( 1996 ) try to explicitly model the processes of dust formation and destruction . however , both classes of models assume that all galaxies form at the same redshift @xmath15 and that there is no number evolution . but the paradigm of the hierarchical growth of structures implies that there is no clear cut redshift @xmath15 since galaxy formation is a continuous process . only blain and longair ( 1993a , b ) proposed a formalism to compute the redshift range of galaxy formation , in addition to chemical evolution , a consistent approach to the early evolution of galaxies is particularly important for any attempt at predicting their submm properties . fig . 1 shows _ model _ spectra of a luminous ir galaxy as it would be observed if placed at different redshifts . there is a wavelength range , between @xmath16 @xmath0 m and @xmath17 mm , in which the distance effect is counterbalanced by the `` negative k correction '' due to the huge rest frame emission maximum at @xmath18 @xmath0 m . in this range , the apparent flux of galaxies depends weakly on redshift to the point that , evolution aside , a galaxy might be easier to detect at @xmath19 than at @xmath20 ! the observer frame submm fluxes , faint galaxy counts and diffuse background of unresolved galaxies are consequently very sensitive to the early stages of galaxy evolution . note that this particular wavelength range brackets the maximum of emission of the cmb . the dissipative and non dissipative processes ruling galaxy formation in dark matter haloes have been studied by various authors ( see especially white and rees 1978 ; schaeffer and silk 1985 ; evrard 1989 ; cole 1991 ; blanchard _ et al . the modelling of these processes , complemented by star formation , stellar evolution and stellar feedback to the interstellar medium , has been achieved at various levels of complexity , in the so called _ semi analytic _ approach which has been successfully applied to the prediction of the statistical properties of galaxies ( white and frenk 1991 ; lacey and silk 1991 ; lacey _ et al . _ 1993 , kauffmann _ et al . _ 1993 , 1994 ; cole _ et al . _ 1994 ; heyl _ et al . _ 1995 ; kauffmann 1995 , 1996 ; baugh _ et al . _ 1996a , b , 1997 ) . it turns out that , in spite of differences in the details of the models , these studies lead to conclusions in the uv , visible and ( stellar ) nir which are remarkably similar . none of these models has been applied so far to the prediction of the properties of galaxies in the ir and submm ranges . this is the main aim of this paper which proposes a simple version of the semi analytic approach and gives predictions of faint counts at wavelengths between 60 @xmath0 m and 1.4 mm . this first study also intends to identify some of the difficulties arising in such a modelling . 2 quickly reviews the main physical processes which have to be introduced in order to describe the formation of galaxies : the non dissipative collapse of perturbations , the dissipative collapse of the gas , star formation and evolution , and feedback . 3 addresses the peculiar issue of dust absorption and re emission in the ir and submm . 4 extracts useful information from the recent uv / visible deep surveys and uses the new constraint arising from the cib in order to generate a family of evolutionary scenarios . sec . 5 gives the ir / submm counts predicted from these scenarios . finally , sec . 6 briefly discusses these results , as well as the shortcomings of such models , and concludes . in a previous study , we showed that the cib can be disentangled with a family of evolutionary scenarios which predict steep submm counts , and that the observations with _ ( at 175 @xmath0 m ) will soon constrain the ir / submm evolution at @xmath3 and beyond ( guiderdoni _ et al . _ 1997 ) . here we present the details of the modelling and apply it to other observations , more specifically in scuba bands ( through the narrow submm atmospheric windows ) . a preliminary version of this work was presented in guiderdoni _ ( 1996 ) . a forthcoming paper ( hivon _ et al . _ 1997 ) will show simulations of the anisotropies of the diffuse submm background due to galaxies and will study the possibility of their detection , with current and forthcoming instruments . other papers in this series will try to overcome some of the shortcomings of this first study . the formation and evolution of a galaxy in its dark matter halo can be briefly sketched as follows : the initial perturbation , which is gravitationally dominated by non baryonic dark matter , grows and collapses . after the ( non - dissipative ) collapse , and subsequent violent relaxation , the halo virializes , through the formation of a mean potential well seen by all particles , which consequently share the same velocity distribution . as detailed in the appendix , we use the classical top hat model for spherically symmetric perturbations , and we compute the mass distribution of collapsed haloes from the peaks formalism ( bardeen _ et al . _ 1986 ) , as in lacey and silk ( 1991 ) and lacey _ et al . _ hereafter , we shall consider the scdm model with @xmath21 km s@xmath22 mpc@xmath22 , @xmath23 , @xmath24 , and @xmath25 as an illustrative case . we take a baryonic fraction @xmath26 , consistent with primordial nucleosynthesis . the redshift evolution of the number density and formation rate of collapsed haloes is plotted in fig . the baryonic gas cools in the potential wells of dark matter haloes , by a process identical to cooling flows observed at the centre of rich clusters . the cooling time at halo radius @xmath27 is @xmath28 with @xmath29 for ionized primordial gas . the cooling curve @xmath30 takes into account various cooling processes . we do not consider the metallicity dependence of @xmath30 . neglecting this dependence leads to an _ overestimate _ of cooling times , which are already very short . the equation @xmath31 defines a cooling radius @xmath32 as a function of redshift @xmath1 . at this redshift , only gas inside @xmath32 ( or @xmath33 if @xmath34 ) cools and is available for star formation . this cooling criterion introduces a high mass cut - off in the mass distribution of cold baryonic cores . at the low mass end , the cooling is so efficient that almost all the gas can cool , leading to a slope of the mass distribution of baryonic cores @xmath35 at redshift @xmath7 , which is close to the slope @xmath36 for the number density of collapsed haloes . this is the so called `` overcooling '' problem ( cole 1991 ; blanchard _ et al . _ 1992 ) . the final radius of the cold gas in rotational equilibrium is related to the initial radius by conservation of angular momentum ( fall and efstathiou 1980 ) . approximately , @xmath37 , with the dimensionless spin parameter @xmath38 ( barnes & efstathiou 1987 ; efstathiou _ et al . _ 1988 ; zurek _ et al . previous studies only used the mean value of @xmath39 in this formula . hereafter , we introduce the @xmath39 distribution from barnes and efstathiou ( 1987 ) model c04 ( their fig . according to a fit based on fig . 3 of fall and efstathiou ( 1980 ) , the exponential disk which forms from the dissipative collapse of the gas has a length scale @xmath40 and a radius including half the cold baryonic mass @xmath41 . a dynamical time scale in the disk like core is @xmath42 . it is important to note that only disks can form in this formalism . the formation of elliptical galaxies ( and of bulges of spiral galaxies ) has to be explained by the merging of disks . ( 1993 ) and cole _ ( 1994 ) showed that this merging process can easily explain the current fraction of ge among bright galaxies . locally , the sfr depends on numerous physical parameters . nevertheless , phenomenological studies seem to show that , on galaxy scales , the sfr per unit surface density is proportional to the total gas surface density ( neutral plus molecular ) divided by the dynamical time scale of the disk ( kennicutt 1989 , 1997 ) . so we shall hereafter assume that the star formation time scale @xmath43 is proportional to the dynamical time scale of the disk @xmath44 and we introduce a first efficiency factor @xmath45 . with @xmath46 , we take : @xmath47 fig . 3 shows the predicted @xmath43 histogram compared to the histogram of `` roberts times '' for a sample of 63 bright disk galaxies observed by kennicutt _ ( 1994 ) . the roberts time is defined as @xmath48 where @xmath49 is the gas mass in the @xmath50 phase measured from the 21 cm line , @xmath51 is the gas mass in the @xmath52 phase measured from the co line , and @xmath53 is the total star formation rate measured from the h@xmath5 line , under an assumption about the shape of the initial mass function ( hereafter this is salpeter imf ) . the interesting result is that the model correctly predicts the _ shape _ and _ width _ of the histogram . we emphasize that this agreement is due to both the range of halo densities ( scaling as @xmath54 ) _ and _ dimensionless spin parameters @xmath39 . without the @xmath39 scatter , the predicted distribution would be about three times too narrow . the average value of the observed histogram can be reproduced by taking @xmath55 for our scdm model . finally , for sake of simplicity , we use a salpeter imf with index @xmath56 . stars have masses @xmath57 . we also assume that the mass fraction blocked in dark objects with masses below 0.1 @xmath58 is negligible . the explosion of massive stars can expel gas from the galaxies and quench star formation , leading to a strong increase of the mass to luminosity ratios in small objects . observationally , hi holes and x ray superbubbles are good evidence that such galactic winds are present in galaxies . the stellar feedback is introduced in a similar way by most authors , following the original work by dekel and silk ( 1986 ) . by equating the gas binding energy to the thermal energy ejected by supernovae , one gets : @xmath59 where @xmath60 is the number of sne per unit mass of stars , depending on the imf . for our choice of imf , @xmath61 @xmath62 . the output mechanical energy of a sn is @xmath63 erg . the escape velocity at radius @xmath64 in a singular isothermal sphere truncated at radius @xmath33 is @xmath65 . the maximum of the @xmath39 distribution corresponds to @xmath66 , leading to @xmath67 . since much of this energy is subsequently radiated away , we insert a second efficiency factor @xmath68 . after some algebra , the mass fraction of stars @xmath69 forming before the triggering of the galactic wind at time @xmath70 is given by : @xmath71 with @xmath72 and @xmath73 km s@xmath22 for the salpeter imf . nevertheless , there is much uncertainty on these parameters because of the cooling of supernova remnants before wind triggering . for instance , numerical simulations seem to suggest that @xmath74 ( thornton _ et al . _ 1997 ) . cole _ ( 1994 ) introduced a fit based on sph simulations of galaxy formation in which most of the feedback effect is due to momentum exchange rather than to ism heating ( navarro and white 1993 ) . for a typical feedback parameter @xmath75 , the numerical simulations can be fitted with @xmath76 and @xmath77 km s@xmath22 . we shall hereafter take these values as our standard parameters . the situation is still complicated by the existence of _ non local _ feedback processes in addition to the local ones . efstathiou ( 1992 ) and blanchard _ et al . _ ( 1992 ) suggest that high@xmath1 re ionization of the igm could prevent cooling in haloes with circular velocities below @xmath78 to 50 km s@xmath22 , and possibly as high as @xmath79 in case of adiabatic collapse . so it is very likely that the overall quenching of dwarf formation depends on redshift . introducing this redshift dependence of dwarf formation partly alleviates the problem of the steep slope of the luminosity function ( see e.g. kauffmann _ et al . so the situation appears to be very complicated , in the absence of a global theory of feedback processes . in the following , we shall simply model the combination of local and non local feedbacks by introducing a simple @xmath80 dependence for @xmath81 . a model of spectrophotometric evolution is used to compute the age dependence of the gas content , the uv to nir spectra of the stellar populations , and the mass to luminosity ratios . the stars are placed on the zero age main sequence of the hr diagram according to the imf . the models use compilations of stellar evolutionary tracks taking into account the various stages of stellar evolution in order to compute at each time step the distribution of stellar populations in the hr diagram . this distribution is combined with a library of stellar spectra and gives the synthetic spectrum @xmath82 . at the end of their lifetimes , stars die and return a fraction of their mass to the ism . the model which is used here is described in guiderdoni and rocca volmerange ( 1987 , 1988 ) , and rocca volmerange and guiderdoni ( 1988 ) , and includes upgraded stellar tracks from schaller _ et al . _ ( 1992 ) and charbonnel _ et al . _ ( 1996 ) . panel _ a _ of fig . 4 shows the grid of models with various @xmath43 from 0.33 to 20 gyr , which are introduced in order to follow the relative gas content of the galaxy @xmath83 . the heavy elements synthetised by stars are injected into the interstellar medium . the metallicity of the gas is estimated from the instant recycling approximation @xmath84 with a yield @xmath85 depending on the choice of the imf . b _ shows the time variation of @xmath86 for the grid of models . the photometric properties can be computed after taking into account the intrinsic extinction ( see the following section ) . as an example , panel _ c _ gives the face on @xmath87 colours . before examining the modelling of the ir emission of galaxies , we would like to quickly comment on the slopes of the @xmath88band luminosity and gas mass functions obtained by this simple version of the semi analytic approach . in spite of the variety of sfr histories , and of the strong variation of the mass to luminosity ratios in time scales of a few gyr , the _ shape _ of the luminosity function in the absence of feedback processes is surprisingly similar to that of the baryonic mass function . for @xmath89 , our model with @xmath90 and the standard mass loss ( @xmath91 and @xmath77 km s@xmath22 ) gives @xmath92 , whereas loveday _ ( 1992 ) find @xmath93 . such an uncomfortable situation is a robust result of _ all _ the recent attempts to model galaxy formation and evolution in a semi analytic approach ( see kauffmann _ et al . _ 1994 ; cole _ et al . although it seems in disagreement with the nearby surveys ( but other surveys seem to suggest an increase of the slope at the faint end , e.g. marzke _ et al . _ 1994 ) , this steep slope is necessary to reproduce the deep redshift surveys ( down to @xmath94 ) and the faint galaxy counts ( down to @xmath95 ) . subtle selection effects due to surface brightness could explain the discrepancy between the nearby luminosity function and the high@xmath1 one ( mcgaugh 1994 ; lobo and guiderdoni 1997 ) . indeed , this type of selection effect can be suspected because the predicted gas mass function is in better agreement with the observational data . the predicted mass distribution of gas at redshift @xmath7 has a slope @xmath96 ( with the standard values or the parameters ) . smaller objects form earlier on an average , with higher densities , smaller @xmath44 , smaller @xmath43 , and , consequently , they are more affected by mass loss and their relative gas content is lower . it is possible to correct statistically the total gas mass function in order to predict an hi mass function and compare it to the observational slope @xmath97 determined by briggs and rao ( 1993 ) . in the sample of 63 disk galaxies gathered by kennicutt _ ( 1994 ) , there is no systematic trend for @xmath98 versus @xmath99 , @xmath100 , or the @xmath101 . we have made an histogram of @xmath102 , with values from @xmath103 to 0 , and we have distributed the predicted number density at @xmath99 into the range of @xmath49 values according to this histogram . the effect of this correction turns out to be rather small . thus there is a good agreement of the predictions with the observations . it is worthwhile to emphasize that the slope @xmath97 of the data depends on the volume corrections of the survey for low mass objects and is somewhat uncertain . [ cols= " < , > , > , > , < " , ] [ ta : scuba ] we have used a simple semi analytic model of galaxy formation to derive the ir / submm statistical properties of galaxies : luminosity function , faint galaxy counts , redshift distributions , and the diffuse background . the model shares many common features with previous semi analytic works focussing upon the optical properties of galaxies : ( i ) the collapse of the perturbations is described by the classical top hat model under the assumptions of homogeneity and sphericity ; in this simple version , we used the peaks formalism to compute the formation rate of haloes . ( ii ) the dissipative cooling and collapse is introduced , with the usual `` overcooling '' problem which can be partly suppressed by introducing feedback due to overall re ionization and galactic winds triggered by sne . ( iii ) star formation is proportional to the ratio of the gas content to the dynamical time scale of the galaxies ; and ( iv ) stellar evolution is explicitly implemented . in order to make specific predictions for the ir / submm wavelength range , this model includes the following assumptions : 1 . an average optical depth of the disks is implemented and scales as the relative gas content and gas metallicity ( estimated through the assumption of instant recycling approximation ) ; no evolution of the dust composition is considered ; the variation of the extinction curve with metallicity is based on an interpolation of the milky way and the magellanic clouds which gives an extinction proportional to @xmath104 ; 2 . a simple `` slab '' geometry is assumed where dust and stars are mixed with equal height scales ; 3 . finally , the dust emission spectrum scales with the bolometric ir luminosity , following the observational relationship of _ iras _ colours with luminosities . such a dependence is obtained by changing the abundance of three components : pah , very small grains and big grains ( and the temperature of the big grains ) . -1truecm we introduced two modes of star formation . in the `` quiescent '' mode , the distribution of the characteristic time scales @xmath43 for star formation peaks at 3 gyr , and ranges from 0.3 to 30 gyr . in the `` burst mode '' , the distribution of @xmath43 peaks at 0.30 gyr , and ranges from 0.03 to 3 gyr . the `` quiescent '' mode is unable to reproduce the rapid evolution of the cosmic sfr and gas densities . the steep decline of the sfr and gas content since @xmath6 can be accommodated by assuming an increasing fraction of mass involved in the `` burst '' mode . since one knows that the fraction of peculiar objects with signs of interaction / merging increases with the depth of the survey , we suspect that the bursting behaviour of the cfrs and hdf galaxies is due to interactions and we introduce a phenomenological @xmath1 dependence fitting the evolution of the pair rates . since ligs and uligs are respectively interacting and merging systems , it is reasonable to estimate that , reciprocally , cfrs and hdf galaxies with peculiar morphologies should emit in the ir . for these `` mild starbursts '' and `` luminous ir / uv galaxies '' , we assume a conservative range of ir to blue luminosity ratios @xmath105 computed from our standard assumptions on average optical thickness and geometry . finally , in order to reproduce the bright end of ir galaxies with @xmath106 , we also introduced a population of heavily extinguished galaxies with massive star formation , similar to uligs . with these ingredients , we designed a family of evolutionary scenarios which fit the local overall energy budget and its evolution to @xmath3 . these scenarios also fit the cob estimate computed from faint galaxy counts , and are consistent with the observational range for the cib . so we feel confident that we roughly reproduce the local energy budget as well as a plausible range for the integrated budget along the line of sight . the scenarios predict uv and ir emissions strongly differring at high @xmath1 . with the same sfr history , different uv fluxes can be obtained with different extinctions and imf . in some of the scenarios , the uv fluxes at @xmath107 fit the current measurements ( or lower values ) of madau _ et al . _ ( 1996 ) , but with ten times more sfr than deduced _ under the assumption of no extinction_. such a factor 10 , which includes the contribution of heavily extinguished uligs , is much larger than the factor 3 corresponding to the correction for extinction derived by pettini _ ( 1997 ) for the hdf galaxies with lyman continuum drop outs . this quantifies the strong warning that the deduction of the sfr density from uv fluxes misses the optically dark side of galaxy formation . this family of scenarios was obtained within the context of the same cosmological model , with similar prescriptions for dissipative and non dissipative collapses , star formation and feedback . the changes only affect the efficiency of star formation in a dynamical time scale ( the @xmath45 parameter ) , the imf ( salpeter or only massive stars ) , and the extinction ( average extinction with `` slab '' geometry , or strong , `` screen''like extinction ) . we saw the extreme sensitivity of the predictions to the parameter choice for this poorly constrained astrophysics . the scenarios are consistent with _ iras _ data . they give similar predictions for the _ iras _ 60 @xmath0 m luminosity function and faint counts , provided the local fraction of uligs is lower than a few % . they are in rough agreement with the redshift surveys of _ iras _ samples , in spite of some uncertainty in the data . by construction , the ir spectra also fit the correlation of ir colours with total ir luminosity in the bright galaxy sample . the selected scenarios are consistent with the @xmath108 range for the observed cib and they are used to `` disentangle '' the background into faint galaxy counts . this cib is the first `` post_iras _ '' observation which helps constraining the high@xmath1 evolution of galaxies in the ir / submm and the energy budget integrated along the line of sight , before _ iso _ results . in contrast with the counts at fir wavelengths which are rather degenerate , the submm counts are very sensitive to the details of the evolutionary scenarios . thus our study illustrates the importance of the submm range to constrain the evolution of the energy budget of galaxies . as emphasized by guiderdoni _ ( 1997 ) , our understanding of the formation and evolution of galaxies is so far entirely based on surveys in the uv / visible window . we are missing the part of star / galaxy formation hidden by dust . we need to open a second window in the ir / submm , which is being explored by _ iso _ and scuba , and will be extensively studied by forthcoming telescopes and satellites such as _ sirtf _ , sofia , _ first _ and _ planck_. there are clearly many approximations and shortcomings in our approach , which could be listed as a research programme for forthcoming papers . ( i ) we need to implement the astrophysical processes in a code which explicitly follows the merging history trees of haloes and galaxies . this code would have to include the various aspects of interaction between galaxies , and to explain why the interaction rate strongly declined since @xmath109 . it would also have to include the effect of interactions on the sfr . ( ii ) the modelling of the ir emission could take into account the evolution of dust properties with chemical abundances , as well as more sophisticated transfer . ( iii ) the global energy budget of galaxies has to be addressed by fitting the faint galaxy counts in the uv / visible and @xmath110 , as well as the forthcoming nir counts from _ iso _ , in particular at 15 @xmath0 m . first results are already given here , and suggest an amount of evolution which is consistent with our predictions . ( iv ) the strong evolution of the comoving sfr density has consequences on the gas and heavy element abundances . the heavy elements made by the high sfrs are to be found in the stars , the cold gas , or the hot gas of the intergalactic medium . if the sfr turned out to be much higher than determined , for instance , by madau _ ( 1996 ) , a significant fraction of the heavy elements should be present in the igm . the amount of elements synthetised in bursts would still be higher if the imf is top heavy , as in the simple model of uligs which is used here . according to mushotzky and loewenstein ( 1997 ) , the absence of evolution of the iron content of clusters at @xmath111 is a first evidence that the global metal production is 25 times higher than inferred from the uv drop out technique . despite the limitations listed above , this paper provides the first predictions of the ir luminosity function , ir / submm galaxy counts , redshift distributions , and diffuse background , obtained from a semi analytic model of galaxy formation and evolution which takes into account the main physical processes from the collapse of the density fluctuations to the absorption of uv and visible star light by dust and the re emission in the ir / submm range . the predictions of this new model are consistent with the dynamical process of continuous galaxy formation predicted by the hierarchical growth of structures in a scdm universe , and represent a significant progress with respect to previous phenomenological models designed to make predictions in the ir / submm . tables giving the faint counts and redshift distributions at various wavelengths between 15 @xmath0 m and 2 mm are available by anonymous ftp at * ftp.iap.fr * in the directory /pub / from_users / guider / fir , and on the web node of the institut dastrophysique de paris at * http://www.iap.fr / users / guider / fir.html*. * acknowledgements . * we are pleased to thank stphane charlot , dave clements , franois xavier dsert , david elbaz , michael fall , karl glazebrook and jean loup puget for their comments and suggestions . abraham , r.g . , tanvir , n.r . , santiago , b.x . , ellis , r.s . , glazebrook , k. , van den bergh , s. , 1996 , mnras , 279 , l47 andreani , p. , franceschini , a. 1992 , a & a , 260 , 89 andreani , p. , franceschini , a. , 1996 , mnras , 283 , 85 ashby , m.l.n . , hacking , p.b . , houck , j.r . , soifer , b.t . , weisstein , e.w . , 1996 , apj , 456 , 428 bardeen , j.m . , bond , j.r . , kaiser , n. , szalay , a.s . 1986 , ap.j . , 304 , 15 barnes , j. , efstathiou , g. 1987 , apj , 319 , 575 baugh , c.m . , cole , s. , frenk , c.s . , 1996a , mnras , 283 , l15 baugh , c.m . , cole , s. , frenk , c.s . , 1996b , mnras , 283 , 1361 baugh , c.m . , cole , s. , frenk , c.s . , lacey , c.g . , 1997 , astro - ph/9703111 beichman , c.a . , helou , g. 1991 , apjl , 370 , l1 bertin , e. , dennefeld , m. , moshir , m. 1997 , a&a , 323 , 685 blain , a.w . , longair , m.s . , 1993a , mnras , 264 , 509 blain , a.w . , longair , m.s . , 1993b , mnras , 265 , l21 blanchard , a. , valls gabaud , d. , mamon , g. , 1992 , a&a , 264 , 365 bond , j.r . , cole , s. , efstathiou , g. , kaiser , n. , 1991 . apj , 379 , 440 bond , j.r . , 1988 . in _ the early universe _ , unruh & semenoff ( eds . ) , p. 322 bower , r.g . mnras , 248 , 332 briggs , f.h . , rao , s. , 1993 . , apj 417 , 494 burkey , j.m . , keel , w.c . , windhorst , r.a . , franklin , b.e . , 1994 , apj , 429 , l13 carlberg , r.g . , 1990 , apj , 359 , l1 carlberg , r.g . , pritchet , c.j . , infante , l. , 1994 , apj , 435 , 540 carico , d.p . , keene , j. , soifer , b.t . , neugebauer , g. , 1992 . , pasp , 104 , 1086 chini , r. , kreysa , e. , kr " ugel , e. , mezger , p.g . , 1986 , a&a , 166 , l8 charbonnel , c. , meynet , g. , maeder , a. , schaerer , d. , 1996 , a&ass , 115 , 339 chini , r. , kr " ugel , e. , 1993 , a&a , 279 , 385 clements , d.l . , andreani , p. , chase , s.t . , 1993 . , mnras , 261 , 299 clements , d.l . , sutherland , w.j . , saunders , w. , efstathiou , g.p . , mcmahon r.g . , maddox , s. , rowan robinson , m. , 1996a , mnras , 279 , 459 clements , d.l . , sutherland , w.j . , mcmahon r.g . , saunders , w. , 1996b , mnras , 279 , 477 cole , s. , 1991 . , apj , 367 , 45 cole , s. , aragn salamanca , a. , frenk , c.s . , navarro , j.f . , zepf , s.e . , mnras , 271 , 781 connolly , a.j . , szalay , a.s . , dickinson , m. , subbarao , m.u . , brunner , r.j . , 1997 , astro - ph/9706255 cowie , l.l . , songaila , a. , hu , e.m . , cohen , j.d . , 1996 , aj , 112 , 839 dekel , a. , silk , j. , 1986 , apj , 303 , 39 dsert , f.x . , boulanger , f. , puget , j.l . 1990 , a&a , 237 , 215 disney , m. , davies , j. , philipps , s. , 1989 , mnras , 239 , 939 draine , b.t . , lee , h.m . , 1984 , apj , 285 , 89 dwek , e. , vrosi , f. , 1996 , in _ unveiling the cosmic infrared background _ , e. dwek ( ed ) , aip conference proceedings 348 eales , s.a . , williams , c.g . , duncan , w.d . , 1989 , apj , 339 , 859 eales , s.a . , edmunds , m.g . , 1996a , mnras , 280 , 1167 eales , s.a . , edmunds , m.g . , 1996b , astro - ph/9609120 efstathiou , g. , rees , m.j . 1988 , mnras , 230 , 5p efstathiou , g. , frenk , c.s . , white , s.d.m . , davis , m. 1988 , mnras , 235 , 715 efstathiou , g. , 1992 , mnras , 256 , 43p eisenhardt , p. , armus , l. , hogg , d.w . , soifer , b.t . , neugebauer , g. , werner , m.w.,,1996 , apj , 461 , 72 ellis , r.s . , colless , m. , broadhurst , t.j . , heyl , j.s . , glazebrook , k. , 1996 , mnras , 280 , 235 evrard , a.e . , 1989 , apj , 341 , 26 fall , s.m . , efstathiou , g. , 1980 , mnras , 193 , 189 fall , s.m . , charlot , s. , pei , y.c . , 1996 , apj , 464 , l43 franceschini , a. , toffolatti , l. , mazzei , p. , danese , l. , & de zotti , g. , 1991 , apjss , 89 , 285 franceschini , a. , mazzei , p. , de zotti , g. , danese , l. 1994 , apj , 427 , 140 franceschini , a. , andreani , p. 1995 , apj , 440 , l5 gautier , t. n. , iii , boulanger , f. , perault , m. , puget , j. l. , 1992 , aj , 103 , 1313 griffiths , r.f . , _ et al . _ , 1994 , apj , 437 , 67 guiderdoni , b. , 1987 , a&a , 172 , 27 guiderdoni , b. , rocca volmerange , b. 1987 , a&a , 186 , 1 guiderdoni , b. , rocca volmerange , b. 1988 , a&ass , 74 , 185 guiderdoni , b. , hivon , e. , bouchet , f.r . , maffei , b. , & gispert , r. 1996 , in _ unveiling the cosmic infrared background _ , e. dwek ( ed ) , aip conference proceedings 348 guiderdoni , b. , bouchet , f.r . , puget , j.l . , lagache , g. , hivon , e. , 1997 , _ submitted _ hacking , p. , houck , j.r . , 1987 , apjss , 63 , 311 hacking , p.b . , soifer , b.t . 1991 , apjl , 367 , l49 haehnelt , m.g . , rees , m.j . , 1993 , mnras , 263 , 168 hauser , m.g . 1995 , in _ proceedings of the iau symp . n@xmath112 168 , examining the big bang and diffuse background radiation _ , the hague , august 1994 helou , g. , soifer , b.t . , rowan robinson , m. , 1985 , apj , 298 , l7 heyl , j.s . , cole , s. , frenk , c.s . , navarro , j.f . , 1995 , mnras , 274 , 755 hivon , e. , guiderdoni , b. , bouchet , f. , 1997 , _ in preparation _ hughes , d.h . , dunlop , j.s . , rawlings , s. , 1997 , astro - ph/9705094 kauffmann , g.a.m . , 1995 , mnras , 274 , 161 kauffmann , g.a.m . , 1996 , mnras , 281 , 487 kauffmann , g.a.m . , white , s.d.m . , 1993 , mnras , 261 , 921 kauffmann , g.a.m . , white , s.d.m . , guiderdoni , b. , 1993 , mnras , 264 , 201 kauffmann , g.a.m . , guiderdoni , b. , white , s.d.m . , 1994 , mnras , 267 , 981 kauffmann , g.a.m . , charlot , s. , white , s.d.m . , 1996 , mnras , 283 , l117 kennicutt , r.c . , 1989 , apj , 344 , 685 kennicutt , r.c . , 1997 , in _ starbursts : triggers , nature and evolution _ , b. guiderdoni and a. kembhavi ( eds . ) , editions de physique /springer verlag kennicutt , r.c . , tamblyn , p. , congdon , c.w . , 1994 , apj , 435 , 22 lacey , c. , silk , j. , 1991 , apj , 381 , 14 lacey , c. , guiderdoni , b. , rocca volmerange , b. , silk , j. , 1993 , apj , 402 , 15 lanzetta , k.m . , wolfe , a.m. , turnshek , d.a . , 1995 , apj , 440 , 435 lilly , s.j . , tresse , l. , hammer , f. , crampton , d. , le fvre , o. , 1995 , apj , 455 , 108 lilly , s.j . , le fvre , o. , hammer , f. , crampton , d. , 1996 , apj , 460 , l1 lonsdale , c.j . , hacking , p.b . , conrow , t.p . robinson , m. , 1990 , apj , 358 , 60 lobo , c. , guiderdoni , b. , 1997 , _ in preparation _ lonsdale , c.j . 1996 , in _ unveiling the cosmic infrared background _ , e. dwek ( ed . ) , aip conference proceedings 348 loveday , j. , peterson , b.a . , efstathiou , g. , & maddox , s.j . apj , * 390 * , 79 madau , p. , ferguson , h.c . , dickinson , m.e . , giavalisco , m. , steidel , c.c . , fruchter , a. , 1996 , mnras , 283 , 1388 maffei , b. 1994 , phd dissertation , universit paris vii marzke , r.o . , huchra , j.p . , geller , m.j . , 1994 , apj , 428 , 43 mathis , j.s . , mezger , p.g . , panagia , n. , 1983 , a&a , 128 , 212 mcgaugh , s.s . 1994 , nat . , 367 , 538 mushotzky , r.f . , loewenstein , m. , 1997 , astro - ph/9702149 natarajan , p. , pettini , m. , 1997 , astro - ph/9709014 natta , a. , panagia , n. , 1984 , apj , 287 , 228 navarro , j.f . , white , s.d.m . , 1993 , mnras , 265 , 271 oliver , s.j . , rowan robinson , m. , saunders , w. , 1992 , mnras , 256 , 15p oliver , s.j . , goldschmidt , p. , franceschini , a. , serjeant , s.b.g . , efstathiou , a.n . , _ et al . _ , 1997 , mnras , 847 , 1 pearson , c. , rowan robinson , m. , 1996 , mnras , 283 , 174 pei , y.c . , fall , s.m . , 1995 , apj , 454 , 69 pettini , m. , steidel , c.s . , dickinson , m. , kellogg , m. , giavalisco , m. , adelberger , k.l . , 1997 , astro - ph/9707200 puget , j.l . , abergel , a. , boulanger , f. , bernard , j.p . , burton , w.b . , _ et al . _ , 1996 , a&a , 308 , l5 reach , w.t . , _ _ 1995 , astro - ph 9504056 rice , w. , lonsdale , c.j . , soifer , b.t . , neugebauer , g. , kopan , e.l . , lloyd , l.a . , de jong , t. , habing , h.j . , 1988 , apjss , 68 , 91 rigopoulou , d. , lawrence , a. , rowan robinson , m. , 1996 , mnras , 278 , 1049 rocca volmerange , b. , guiderdoni , b. , 1988 , a&ass , 75 , 93 rowan robinson , m. , 1992 , mnras , 258 , 787 rowan robinson , m. , broadhurst , t. , lawrence , a. , mcmahon , r.g . , lonsdale , c. , _ et al . _ , 1991a , nature , 351 , 719 rowan robinson , m. , saunders , w. , lawrence , a. , leech , k. , 1991b , mnras , 253 , 485 rush , b. , malkan , m. , spinoglio , l. , 1993 , apjs , 89 , 1 sanders , d.b . , mirabel , i.f . , 1996 , araa , 34 , 749 saunders , w. , rowan robinson , m. , lawrence , a. , efstathiou , g. , kaiser , n. , ellis , r.s . , frenk , c.s . , 1990 , mnras , 242 , 318 sawicki , m.j . , lin , h. , yee , h.k.c . , 1997 , aj , 113 , 1 schaeffer , r. , silk , j. , 1985 , apj , 292 , 319 schaller , g. , schaerer , d. , meynet , g. , maeder , a. , 1992 , a&ass , 96 , 269 smith , b.j . , kleinmann , s.g . , huchra , j.p . , low , f.j . , 1987 , apj , 318 , 161 soifer , b.t . , sanders , d.b . , madore , b.f . , neugebauer , g. , danielson , g.e . , _ et al . _ , 1987 , apj , 320 , 238 soifer , b.t . , neugebauer , g. , 1991 , aj , 101 , 354 stark , a.a . , davidson , j.a . , harper , d.a . , pernic , r. , loewenstein , r. , platt , s. , engargiola , g. , casem , s. , 1989 , apj , 337 , 650 steidel , c.c . , giavalisco , m. , pettini , m. , dickinson , m. , adelberger , k.l . , 1996 , apj , 462 , l17 storrie lombardi , l.j . , mcmahon , r.g . , irwin , m.j . , 1996 , mnras , 283 , l79 thornton , k. , gaudlitz , m. , janka , h.t . , steinmetz , m. , 1997 , astro - ph/9706175 wang , b. , 1991a , apj , 374 , 456 wang , b. , 1991b , apj , 374 , 465 white , s.d.m . , rees , m.j . , 1978 , mnras , 183 , 341 white , s.d.m . , frenk , c.s . , 1991 , apj , 379 , 52 williams , r.e . , _ et al . _ , 1996 , aj , 112 , 1335 zurek , w.h . , quinn , p.j . , salmon , j.k . , 1988 , apj , 330 , 519 zepf , s.e . , koo , d.c . , 1989 , apj , 337 , 34 the initial , linear perturbation is assumed to be spherical and homogeneous . this is the so called `` top hat '' model , which has the friendly property of being entirely defined by two parameters , for instance the size @xmath113 and density contrast @xmath114 which are the linearly extrapolated values at @xmath7 , or , equivalently , by the mass @xmath115 and collapse redshift @xmath116 . if @xmath117 is the current mass density of the universe , we have @xmath118 . the linearly extrapolated density contrast , and the ratio of the radius of maximal expansion @xmath119 to the linearly extrapolated size @xmath113 , can be respectively computed as functions of the redshift of collapse : @xmath120 , ~~~~ r_m / r = r_m / r [ 1+z_{coll } ] .\ ] ] for instance , if @xmath23 , @xmath121 with @xmath122 , and @xmath123 . after collapse and violent relaxation , a mean potential forms and the virialized halo ends as a singular , isothermal sphere truncated at `` virial radius '' @xmath124 . if we define a `` circular velocity '' as @xmath125 , the density profile of the relaxed halo at @xmath64 is : @xmath126 and the mass included within radius @xmath27 is simply @xmath127 . for instance , if @xmath23 , the virial radius ( in mpc ) and the circular velocity ( in km s@xmath22 ) are : @xmath128 @xmath129 we hereafter take @xmath130 . finally , the virial temperature of the baryonic gas is : @xmath131 @xmath132 is the mass of the proton , and the mean molecular weight @xmath133 corresponds to ionized gas with respective h and he mass fractions @xmath134 and @xmath135 . then we compute the mass distribution of collapsed haloes from the peaks formalism ( bardeen _ et al . _ 1986 ) , as in lacey and silk ( 1991 ) and lacey _ et al . _ we start from the power spectrum of linear fluctuations @xmath136 which is normalized by @xmath137 , the variance at a _ top hat _ smoothing radius @xmath138 mpc . the peaks formalism uses momenta of the power spectrum with gaussian smoothing on radius @xmath139 . the mass included within such a radius is @xmath140 . so the relation between the gaussian radius @xmath139 and the top hat radius @xmath113 including mass @xmath115 is simply @xmath141 . these momenta are computed from : @xmath142 where @xmath143 is the fourier transform of the gaussian filtering function : @xmath144 the ( comoving ) number density distribution of peaks @xmath145 in which @xmath146 is smoothed on the gaussian scale @xmath139 is : @xmath147 with : @xmath148 @xmath149 such as @xmath150 . the functions @xmath151 and @xmath152 are computed as : @xmath153 the constant @xmath154 and an analytic fit of the function @xmath155 is given by equ . ( 4.4 ) and ( 4.5 ) of bardeen _ et al . _ if @xmath156 , it is easy to check that @xmath157 does not depend on @xmath139 and that @xmath158 . for @xmath23 , @xmath159 , and the cdm power spectrum with @xmath160 , we have slow variations : @xmath161 and @xmath162 for @xmath163 mpc . at that stage , we do not have the number density distribution of peaks per bin of density contrast _ and _ radius . for that purpose , we have to introduce , as in lacey & silk ( 1991 ) , a suggestion by bond ( 1988 ) who interprets equ . ( a9 ) as the total number density of peaks @xmath164 at scale @xmath165 . by derivating equ . ( a8 ) , it is easy to get : @xmath166 we neglect the slow variation of @xmath157 with @xmath139 . this is the number density of _ all _ peaks . we now need to count the peaks over a certain threshold @xmath167 for collapse , and we get the formation rate of haloes : @xmath168 after a change of variables , we get the halo formation rate , since @xmath169/\sigma_0(r_g)$ ] : @xmath170 \over d ( 1+z_{coll } ) } .\end{aligned}\ ] ] in order to link the gaussian smoothing to the top hat formalism , we take @xmath141 and @xmath171 . the total number of collapsed peaks at redshift @xmath12 is obtained by integrating equ . ( a14 ) on all redshifts larger than @xmath12 . the number density continuously increases with time , at all masses , and contrarily to what is found from the press schechter formalism . the peaks formalism follows the collapse of _ all _ peaks , and counts high peaks as well as the broader , shallower peaks in which these high peaks are included . this so called `` cloud in cloud '' problem results in a possible overestimate of the total number of galaxies . as a matter of fact , the total number density of peaks at the low mass end of the mass function varies as @xmath172 , and the total mass density grows logarithmically with the lower mass cut off . anyhow , since other astrophysical phenomena will act to `` suppress '' galaxy formation in low mass haloes , we find that this slow divergence is not too serious a problem . finally , we do not explicitly include the merging of galaxies in merging haloes . in our crude modelling , new galaxies form at each generation of peak collapse . this model could be fitted to the description of `` active '' objects such as luminous galaxies seen in the ir . such a simplifying assumption is very similar to that of haehnelt and rees ( 1993 ) who modelled qso formation in each generation of halo formation . the issue of merging should be explicitly addressed by making monte carlo realizations of the halo merging history tree and monitoring the merging of galaxies in merged haloes , for instance with the dynamical friction time scale ( kauffmann _ et al _ 1993 ; cole _ et al . _ 1994 ; and following works ) . such studies show that the galaxy merging rate is relatively low , and that the resulting mass and luminosity functions do not differ significantly from those computed with cruder assumptions . more precisely , galaxy merging seems to be sufficient to make the 10 % fraction of giant galaxies which are elliptical , but does not produce a significant change in the slope of the luminosity function at faint magnitudes . the implementation of the astrophysics hereafter described into this type of code is clearly one of the next steps of our programme . nevertheless , it is worthwhile to note that only merging following the slow process of dynamical friction has been currently modelled . two body encounters and tidal interactions , which trigger starburst activity as it clearly appears from observational evidence , are not modelled yet in the current monte carlo codes . |
the constraint satisfaction problem ( csp ) and its quantified extensions , whether without ( qcsp ) or with disjunction ( qcsp ) , correspond naturally to the model checking problem for three increasingly stronger fragments of positive first - order logic . their complexity is often studied when parameterised by a fixed model , the so - called template . it is a natural question to ask when two templates are equivalent , or more generally when one `` contain '' another , in the sense that a satisfied instance of the first will be necessarily satisfied in the second . one can also ask for a smallest possible equivalent template : this is known as the core for csp . we recall and extend previous results on containment , equivalence and `` coreness '' for qcsp before initiating a preliminary study of cores for qcsp which we characterise for certain structures and which turns out to be more elusive . | 1204.5981 | we consider the following increasingly stronger fragments of first - order logic : 1 . primitive positive first - order ( @xmath1 ) 2 . positive horn ( @xmath2 ) 3 . positive equality - free first - order ( @xmath3 ) ; and , 4 . positive first - order logic ( @xmath4 ) the _ model checking problem _ for a logic @xmath5 takes as input a sentence of @xmath5 and a structure @xmath6 and asks whether @xmath6 models @xmath7 . the structure @xmath6 is often assumed to be a fixed parameter and called the template ; and , unless otherwise stated , we will assume implicitly that we work in this so - called _ non - uniform _ setting . for the above first three fragments , the model checking problem is better known as the _ constraint satisfaction problem _ csp@xmath8 , the _ quantified constraint satisfaction problem _ qcsp@xmath8 and its extension with disjunction which we shall denote by qcsp@xmath9 . much of the theoretical research into csps is in respect of a large complexity classification project it is conjectured that csp@xmath8 is always either in p or np - complete @xcite . dichotomy _ conjecture remains unsettled , although dichotomy is now known on substantial classes ( e.g. structures of size @xmath10 @xcite and smooth digraphs @xcite ) . various methods , combinatorial ( graph - theoretic ) , logical and universal - algebraic have been brought to bear on this classification project , with many remarkable consequences . a conjectured delineation for the dichotomy was given in the algebraic language in @xcite . complexity classifications for qcsps appear to be harder than for csps . just as csp@xmath8 is always in np , so qcsp@xmath8 is always in pspace . no overarching polychotomy has been conjectured for the complexities of qcsp@xmath8 , as @xmath6 ranges over finite structures , but the only known complexities are p , np - complete and pspace - complete ( see @xcite for some trichotomies ) . it seems plausible that these complexities are the only ones that can be so obtained . distinct templates may give rise to the same model - checking - problem or preserve acceptance , * ( @xmath5-*equivalence * ) for any sentence @xmath11 of @xmath5 , @xmath12 models @xmath13 @xmath14 @xmath6 models @xmath13 * ( @xmath5-*containment * ) for any sentence @xmath11 of @xmath5 , @xmath12 models @xmath13 @xmath15 @xmath6 models @xmath13 . we will see that containment and therefore equivalence is decidable , and often quite effectively so , for the four logics we have introduced . for example , when @xmath5 is @xmath1 , any two bipartite undirected graphs that have at least one edge are equivalent . moreover , there is a canonical _ minimal _ representative for each equivalence class , the so - called _ core_. for example , the core of the class of bipartite undirected graphs that have at least one edge is the graph @xmath16 that consists of a single edge . the core enjoys many benign properties and has greatly facilitated the classification project for csps ( which corresponds to the model - checking for @xmath1 ) : it is unique up to isomorphism and sits as an induced substructure in all templates in its equivalence class . a core may be defined as a structure all of whose endomorphisms are automorphisms . to review , therefore , it is well - known that two templates @xmath12 and @xmath6 are equivalent iff there are homomorphisms from @xmath17 to @xmath18 and from @xmath18 to @xmath17 , and in this case there is an ( up to isomorphism ) unique core @xmath19 equivalent to both @xmath12 and @xmath6 such that @xmath20 and @xmath21 . the situation for @xmath2 and qcsp is somewhat murkier . it is known that non - trivial @xmath12 and @xmath6 are equivalent iff there exist integers @xmath22 and @xmath23 and surjective homomorphisms from @xmath24 to @xmath18 and from @xmath25 to @xmath17 ( and one may give a bound on these exponents ) @xcite . however , the status and properties of `` core - ness '' for qcsp were hitherto unstudied . we * might * call a structure @xmath18 a _ q - core _ if there is no equivalent @xmath17 of strictly smaller cardinality . we will discover that * this * q - core is a more cumbersome beast than its cousin the core ; it need not be unique nor sit as an induced substructure of the templates in its class . however , in many cases we shall see that its behaviour is reasonable and that like the core it can be very useful in delineating complexity classifications . the erratic behaviour of q - cores sits in contrast not just to that of cores , but also that of the _ @xmath26-@xmath27-cores _ of @xcite , which are the canonical representatives of the equivalence classes associated with @xmath28 , and were instrumental in deriving a full complexity classification a tetrachotomy for qcsp@xmath0 in @xcite . like cores , they are unique and sit as induced substructures in all templates in their class . thus , primitive positive logic and positive equality - free logic behave genially in comparison to their wilder cousin positive horn . in fact this manifests on the algebraic side also polymorphisms and surjective hyper - endomorphisms are stable under composition , while surjective polymorphisms are not . continuing to add to our logics , in restoring equality , we might arrive at positive logic . two finite structures agree on all sentences of positive logic iff they are isomorphic so here every finite structure satisfies the ideal of `` core '' . when computing a / the smallest substructure with the same behaviour with respect to the four decreasingly weaker logics positive logic , positive equality - free , positive horn , and primitive positive we will obtain possibly decreasingly smaller structures . in the case of positive equality - free and primitive positive logic , as pointed out , these are unique up to isomorphism ; and for the @xmath26-@xmath27-core and the core , these will be induced substructures . * a * q - core will necessarily contain the core and be included in the u - x - core . this phenomenon is illustrated on table [ tab : different - cores ] and will serve as our running example . .different notions of `` core '' ( the circles represent self - loops ) . [ cols="^,^,^,^",options="header " , ] the paper is organised as follows . in section [ sec : csp ] , we recall folklore results on csp . in section [ sec : mylogic ] , we recall results on coreness and spell out containment for @xmath28 that were only implicit in @xcite . in section [ sec : qcsp ] , we move on to qcsp and recall results on the decidability of containment from @xcite together with new lower bounds before initiating a study of the notion of core for qcsp . unless otherwise stated , we consider structures over a fixed relational signature @xmath29 . we denote by @xmath30 the domain of a structure @xmath12 and for every relation symbol @xmath31 in @xmath29 of arity @xmath22 , we write @xmath32 for the interpretation of @xmath31 in @xmath12 , which is a @xmath22-ary relation that is @xmath33 . we write @xmath34 to denote the cardinality of the set @xmath30 . a homomorphism ( resp . , strong homomorphism ) from a structure @xmath12 to a structure @xmath6 is a function @xmath35 such that @xmath36 , if ( resp . , iff ) @xmath37 . we will occasionally consider signatures with constant symbols . we write @xmath38 for the interpretation of a constant symbol @xmath39 and homomorphisms are required to preserve constants as well , that is @xmath40 . containment for @xmath1 is a special case of conjunctive query containment from databases @xcite . we state and prove these results for pedagogical reasons , before moving to the case of @xmath28 . let us fix some notation first . given a sentence @xmath11 in @xmath1 , we denote by @xmath41 its _ canonical database _ , that is the structure with domain the variables of @xmath11 and whose tuples are precisely those that are atoms of @xmath11 . in the other direction , given a finite structure @xmath12 , we write @xmath42 for the so - called _ canonical conjunctive query _ of @xmath12 , the quantifier - free formula that is the conjunction of the positive facts of @xmath12 , where the variables @xmath43 correspond to the elements @xmath44 of @xmath12 .. ] it is well known that there is a homomorphism from @xmath41 to a structure @xmath12 if , and only if , @xmath45 . moreover , a winning strategy for @xmath46 in the ( hintikka ) @xmath47-game is precisely a homomorphism from @xmath41 to @xmath12 . note also that @xmath12 is isomorphic to the canonical database of @xmath48 . let @xmath12 and @xmath6 be two structures . the following are equivalent . for every sentence @xmath11 in @xmath1 , if @xmath49 then @xmath50 . the exists a homomorphism from @xmath12 to @xmath6 . @xmath51 . where @xmath42 denotes the _ canonical conjunctive query _ of @xmath12 . a homomorphism corresponds precisely to a winning strategy in the @xmath52-game and ( ii ) and ( iii ) are equivalent . clearly , ( i ) implies ( iii ) since @xmath53 . we now prove that ( ii ) implies ( i ) . let @xmath54 be a homomorphism from @xmath12 to @xmath6 . if @xmath49 , then there is a homomorphism @xmath55 from @xmath41 to @xmath12 . by composition , @xmath56 is a homomorphism from @xmath41 to @xmath6 . in other words , @xmath56 is a winning strategy witnessing that @xmath50 . it is well known that the core is unique up to isomorphism and that it is an induced substructure @xcite . it is usually defined via homomorphic equivalence , but because of the equivalence between ( i ) and ( ii ) in the above theorem , we may define the core as follows . * the * _ core _ @xmath6 of a structure @xmath12 is a minimal substructure of @xmath12 such that for every sentence @xmath11 in @xmath1 , @xmath49 if and only if @xmath50 . let @xmath12 and @xmath6 be two structures . the following are equivalent . for every sentence @xmath11 in @xmath1 , @xmath49 if and only if @xmath50 . there are homomorphisms from @xmath12 to @xmath6 and from @xmath6 to @xmath12 . the core of @xmath12 and the core of @xmath6 are isomorphic . as a preprocessing step , one could replace the template @xmath12 of a csp by its core @xmath6 ( see algorithm 6.1 in @xcite ) . however , the complexity of this preprocessing step would be of the same order of magnitude as solving a constraint satisfaction problem . this drawback , together with the uniform nature of the instance in constraints solvers , means that this preprocessing is not exploited in practice to the best of our knowledge . the notion of a core can be extended and adapted suitably to solve important questions related to data exchange and query rewriting in databases @xcite . it is also very useful as a simplifying assumption when classifying the complexity : with the algebraic approach , it allows to study only idempotent algebras @xcite . for @xmath28 , it is no longer the homomorphism that is the correct concept to transfer winning strategies . a _ surjective hypermorphism _ @xmath57 from a structure @xmath12 to a structure @xmath6 is a function from the domain @xmath30 of @xmath12 to the power set of the domain @xmath58 of @xmath6 that satisfies the following properties . * ( * total * ) for any @xmath59 in @xmath30 , @xmath60 . * ( * surjective * ) for any @xmath61 in @xmath58 , there exists @xmath59 in @xmath30 such that @xmath62 . * ( * preserving * ) if @xmath63 holds in @xmath12 then @xmath64 holds in @xmath6 , for all @xmath65 @xmath66 . a _ strategy _ for @xmath46 in the ( hintikka ) @xmath52-game , where @xmath67 , is a set of mappings @xmath68 with one mapping @xmath69 for each existentially quantified variable @xmath70 of @xmath11 . the mapping @xmath69 ranges over the domain @xmath30 of @xmath12 ; and , its domain is the set of functions from @xmath71 to @xmath30 , where @xmath71 denotes the universally quantified variables of @xmath13 preceding @xmath70 . we say that @xmath68 is _ winning _ if for any assignment @xmath72 of the universally quantified variables of @xmath11 to @xmath30 , when each existentially quantified variable @xmath70 is set according to @xmath69 applied to @xmath73 , then the quantifier - free part @xmath74 of @xmath11 is satisfied under this overall assignment @xmath54 . when @xmath74 is disjunction - free , this amounts to @xmath54 being a homomorphism from @xmath75 to @xmath12 . [ lemma : strategy : transfert : mylogic ] let @xmath12 and @xmath6 be two structures such that there is a surjective hypermorphism from @xmath12 to @xmath6 . then , for every sentence @xmath11 in @xmath28 , if @xmath49 then @xmath50 . let @xmath57 be a surjective hypermorphism from @xmath12 to @xmath6 and @xmath11 be a sentence of @xmath28 such that @xmath49 . for any element @xmath61 of @xmath6 , let @xmath76 . we fix an arbitrary linear order over @xmath30 and write @xmath77 to denote the smallest antecedent of @xmath61 in @xmath30 under @xmath57 . let @xmath68 be a winning strategy in the @xmath52-game . we construct a strategy @xmath78 in the @xmath79-game as follows . let @xmath80 be an assignment to the universal variables @xmath71 preceding an existential variable @xmath70 in @xmath11 , we select for @xmath81 an arbitrary element of @xmath82 where @xmath83 is an assignment such that for any universal variable @xmath84 preceding @xmath70 , we have @xmath85 . this strategy is well defined since @xmath57 is surjective ( which means that @xmath86 is well defined ) and total ( which means that @xmath87 ) . note moreover that using @xmath88 in the definition of @xmath86 means that a branch in the tree of the game on @xmath6 will correspond to a branch in the tree of the game on @xmath12 . it remains to prove that @xmath78 is winning . we will see that it follows from the fact that @xmath57 is preserving . assume first that @xmath11 is a sentence of @xmath2 . let @xmath75 be the canonical database of the quantifier - free part @xmath74 of @xmath11 . the winning condition of the @xmath79-game can be recast as a homomorphism from @xmath75 . composing with @xmath57 the homomorphism from @xmath75 to @xmath12 ( induced by the sequence of compatible assignments @xmath86 to the universal variables and the strategy @xmath68 ) , we get a surjective hypermorphism from @xmath75 to @xmath6 . the map from the domain of @xmath75 to @xmath6 induced by the sequence of assignments @xmath89 and the strategy @xmath78 is a range restriction of this surjective hypermorphism and is therefore a homomorphism ( we identify surjective hypermorphism to singletons with homomorphisms ) . when @xmath11 is not a sentence of @xmath2 , we write its quantifier - free part in disjunctive normal form as a disjunction of conjunctions - of - atoms @xmath90 . the winning condition can now be recast as a homomorphism from some @xmath91 . the above argument applies and the result follows . consider the structures @xmath92 and @xmath93 from table [ tab : different - cores ] . the map @xmath57 given by @xmath94 is a surjective hypermorphism from @xmath92 to @xmath93 . the map @xmath55 given by @xmath95 is a surjective hypermorphism from @xmath93 to @xmath92 . the two templates are equivalent w.r.t . @xmath28 . we extend the notion of canonical conjunctive query of a structure @xmath12 . given a tuple of ( not necessarily distinct ) elements @xmath96 , define the quantifier - free formula @xmath97 to be the conjunction of the positive facts of @xmath98 , where the variables @xmath99 correspond to the elements @xmath100 . that is , @xmath101 appears as an atom in @xmath102 iff @xmath103 holds in @xmath12 . when @xmath98 enumerates the elements of the structure @xmath12 , this definition coincides with the usual definition of canonical conjunctive query . note also that there is a strong homomorphism from the canonical database @xmath104 to @xmath12 given by the map @xmath105 . let @xmath12 be a structure and @xmath106 . let @xmath98 be an enumeration of the elements of @xmath12 . @xmath107 observe that @xmath108 . indeed , we may take as witness for the variables @xmath109 the corresponding enumeration @xmath110 of the elements of @xmath12 ; and , for any assignment @xmath111 to the universal variables @xmath112 , it is clear that @xmath113 holds . [ lemma : canonical : sentence : mylogic ] let @xmath12 and @xmath6 be two structures . if @xmath114 then there is a surjective hypermorphism from @xmath12 to @xmath6 . let @xmath115 be witnesses for @xmath43 . assume that an enumeration @xmath116 of the elements of @xmath6 is chosen for the universal variables @xmath117 . let @xmath111 be the witness s.t . @xmath118 . let @xmath57 be the map from the domain of @xmath12 to the power set of that of @xmath6 which is the union of the following two partial hyperoperations @xmath54 and @xmath55 ( _ i.e. _ @xmath119 for any element @xmath120 of @xmath12 ) , which guarantee totality and surjectivity , respectively . * ( * totality * ) @xmath121 * ( * surjectivity * ) @xmath122 . it remains to show that @xmath57 is preserving . this follows from @xmath123 . let @xmath31 be a @xmath22-ary relational symbol such that @xmath124 holds in @xmath12 . let @xmath125 . we will show that @xmath126 holds in @xmath6 . assume for clarity of the exposition and w.l.o.g . that from @xmath127 to @xmath128 the image is set according to @xmath54 and from @xmath129 to @xmath130 according to @xmath55 : i.e. for @xmath131 , @xmath132 and for @xmath133 , there is some @xmath134 such that @xmath135 and @xmath136 . by definition of @xmath137 the atom @xmath138 appears in @xmath139 . it follows from @xmath123 that @xmath126 holds in @xmath6 . [ theo : containment : mylogic ] let @xmath12 and @xmath6 be two structures . the following are equivalent . for every sentence @xmath11 in @xmath28 , if @xmath49 then @xmath50 . the exists a surjective hypermorphism from @xmath12 to @xmath6 . @xmath140 where @xmath141 is a canonical sentence of @xmath28 that is defined in terms of @xmath12 and @xmath142 and that is modelled by @xmath12 by construction . by construction @xmath143 , so ( i ) implies ( iii ) . by lemma [ lemma : strategy : transfert : mylogic ] , ( ii ) implies ( i ) . by lemma [ lemma : canonical : sentence : mylogic ] , ( iii ) implies ( i ) . let @xmath26 and @xmath27 be two subsets of @xmath30 and a surjective hypermorphism @xmath54 from @xmath12 to @xmath12 that satisfies @xmath144 and @xmath145 . let @xmath6 be the substructure of @xmath12 induced by @xmath146 . then @xmath57 and @xmath55 , the range and domain restriction of @xmath54 to @xmath58 , respectively , are surjective hypermorphisms between @xmath12 and @xmath6 witnessing that @xmath12 and @xmath6 satisfy the same sentence of @xmath28 . note that in particular @xmath54 induces a retraction of @xmath12 to a subset of @xmath27 ; and , dually a retraction of the complement structure of @xmath6 to a subset of @xmath26 . additional minimality conditions on @xmath26 , @xmath27 and @xmath147 ensure that @xmath6 is minimal . it is also _ unique _ up to isomorphism and within @xmath6 the set @xmath26 and @xmath27 are _ uniquely determined_. consequently , @xmath6 is called * the * @xmath26-@xmath27-core of @xmath12 ( for further details see @xcite ) and may be defined as follows . * the * @xmath26-@xmath27__core _ _ @xmath6 of a structure @xmath12 is a minimal substructure of @xmath12 such that for every sentence @xmath11 in @xmath28 , @xmath49 if and only if @xmath50 . the map @xmath148 is a surjective hypermorphism from @xmath92 to @xmath92 with @xmath149 and @xmath150 . the substructure induced by @xmath151 is @xmath93 . it can be checked that it is minimal . the @xmath26-@xmath27-core is just like the core an induced substructure . there is one important difference in that @xmath26-@xmath27-cores should be genuinely viewed as a minimal equivalent substructure induced by _ two _ sets . indeed , when evaluating a sentence of @xmath28 , we may assume w.l.o.g . that all @xmath152 variables range over @xmath26 and all @xmath46 variables range over @xmath27 . this is because for any play of @xmath152 , we may extract a winning strategy for @xmath46 that can even restrict herself to play only on @xmath27 ( * ? ? ? * lemma 5 ) . hence , as a _ preprocessing step _ , one could compute @xmath26 and @xmath27 and restrict the domain of each universal variable to @xmath26 and the domain of each universal variable to @xmath27 . _ the complexity of this processing step is no longer of the same magnitude _ and is in general much lower than solving a qcsp@xmath0 . thus , even when taking into account the uniform nature of the instance in a quantified constraints solver , this preprocessing step might be exploited in practice . this could turn out to be ineffective when there are few quantifier alternation ( as in bilevel programming ) , but should be of particular interest when the quantifier alternation increases . another interesting feature is that storing a winning strategy over @xmath26 and @xmath27 together with the surjective hypermorphism @xmath54 from @xmath12 to @xmath12 , allows to recover a winning strategy even when @xmath152 plays in an unrestricted manner . this provides a _ compression mechanism _ to store certificates . in primitive positive and positive horn logic , one normally considers equalities to be permitted . from the perspective of computational complexity of csp and qcsp , this distinction is unimportant as equalities may be propagated out by substitution . in the case of positive horn and qcsp , though , equality does allow the distinction of a trivial case that can not be recognised without it . the sentence @xmath153 is true exactly on structures of size one . the structures @xmath154 and @xmath155 , containing empty relations over one element and two elements , respectively , are therefore distinguishable in @xmath156 , but not in @xmath2 . since we disallow equalities , many results from this section apply only to _ non - trivial _ structures of size @xmath157 . note that equalities can not be substituted out from @xmath158 , thus it is substantially stronger than @xmath28 . for @xmath2 , the correct concept to transfer winning strategies is that of _ surjective homomorphism from a power_. recall first that the _ product _ @xmath159 of two structures @xmath12 and @xmath6 has domain @xmath160 and for a relation symbol @xmath31 , @xmath161 ; and , similarly for a constant symbol @xmath39 , @xmath162 . the _ @xmath163th power _ @xmath164 of @xmath12 is @xmath165 ( @xmath163 times ) . [ lemma : strategy : transfert : qcsplogic ] let @xmath12 and @xmath6 be two structures and @xmath166 such that there is a surjective homomorphism from @xmath164 to @xmath6 . then , for every sentence @xmath11 in @xmath2 , if @xmath49 then @xmath50 . for @xmath167 , the proof is similar to lemma [ lemma : strategy : transfert : mylogic ] . a projection from @xmath164 to @xmath12 is a surjective homomorphism . this means that for every sentence @xmath11 in @xmath2 , if @xmath168 then @xmath49 . for the converse , one can consider the `` product strategy '' which consists in projecting over each coordinate of @xmath164 and applying the strategy for @xmath12 . for further details see ( * ? ? ? * lemma 1&2 ) . [ ex : bip ] consider an undirected bipartite graphs with at least one edge @xmath169 and @xmath16 the graph that consists of a single edge . there is a surjective homomorphism from @xmath169 to @xmath16 . note also that @xmath170 ( where @xmath171 stands for disjoint union ) which we write as @xmath172 . thus , @xmath173 ( as @xmath174 distributes over @xmath171 ) . hence , if @xmath169 has no isolated element and @xmath163 edges there is a surjective homomorphism from @xmath175 to @xmath169 . this examples provides a lower bound for @xmath163 which we can improve . [ prop : lowerbound : powerepimorphism ] for any @xmath176 , there are structures @xmath12 and @xmath6 with @xmath177 and @xmath178 such that there is only a surjective homomorphism from @xmath179 to @xmath6 provided that @xmath180 . we consider a signature that consists of a binary symbol @xmath181 together with a monadic predicate @xmath31 . consider for @xmath12 an oriented cycle with @xmath163 vertices , for which @xmath31 holds for all but one . consider for @xmath6 an oriented cycle with @xmath163 vertices , for which @xmath31 does not hold , together with a self - loop on which @xmath31 holds . the square of @xmath12 will consists of @xmath177 oriented cycles with @xmath163 vertices : one cycle will be a copy of @xmath12 , all the other will be similar but with two vertices on which @xmath31 does not hold ( this is depicted on figure [ fig : power : cycles ] in the case @xmath182 : white vertices do not satisfy @xmath31 while black ones do ) . it is only for @xmath183 that we will get as an induced substructure of @xmath179 one copy of an oriented cycle on which @xmath31 does not hold as in @xmath6 . there is also a _ canonical @xmath2-sentence _ which turns out to be in @xmath184-form , that is with a quantifier prefix of the form @xmath185 . we consider temporarily structures with @xmath163 constants @xmath186 ; let @xmath187 in @xmath188 describe the position of these constants in a structure @xmath12 ; and , write @xmath189 for the corresponding structure with constants . we consider the canonical conjunctive query of the structure with constants @xmath190 , ( where @xmath191 denote the product ) , identifying the constants with some variables @xmath192 and using variables @xmath109 for the other elements . we turn this quantifier - free formula into a sentence by adding the prefix @xmath193 . keeping this in mind , we can also give the following direct definition , but it dilutes the intuition somewhat . [ def : canonical : pi2:sentence ] let @xmath12 be a structure and @xmath106 . let @xmath98 be an enumeration of the elements of @xmath194 . @xmath195 } \ldots \land w_m = v_{{\ensuremath{\mathbf{t}}},{\ensuremath{\mathbf{t}}}[m ] } .\ ] ] observe that we may propagate the equalities out of @xmath196 to obtain an equivalent sentence : e.g. we remove @xmath197}$ ] and replace every occurrence of @xmath198}$ ] by @xmath199 . observe also that @xmath200 . indeed , assume that @xmath201 is the assignment chosen for the universal variables @xmath112 . there is a natural projection from @xmath190 to @xmath189 which is a homomorphism . this homomorphism corresponds precisely to a winning strategy for the existential variables @xmath109 . [ theo : containment : qcsplogic ] let @xmath12 and @xmath6 be two non - trivial structures . the following are equivalent . for every sentence @xmath11 in @xmath2 , if @xmath49 then @xmath50 . there exists a surjective homomorphism from @xmath202 to @xmath6 , with @xmath203 . @xmath204 where @xmath205 is a canonical sentence of @xmath2 with quantifier prefix @xmath206 that is defined in terms of @xmath12 and modelled by @xmath12 by construction . \(ii ) implies ( i ) by lemma [ lemma : strategy : transfert : qcsplogic ] . ( i ) implies ( iii ) since @xmath12 models @xmath205 . ( iii ) implies ( ii ) by construction of @xmath205 . we may chose for the universal variables @xmath112 an enumeration of @xmath6 . the winning strategy on @xmath6 induces a surjective homomorphism from @xmath202 ( for further details see ( * ? ? ? * theorem 3 ) and comments on the following page ) . following our approach for the other logics , we now define a minimal representative as follows . * a * _ q - core _ @xmath6 of a structure @xmath12 is a minimal substructure of @xmath12 such that for every sentence @xmath11 in @xmath2 , @xmath49 if and only if @xmath50 . consider @xmath93 and @xmath207 from table [ tab : different - cores ] . the map @xmath208 is a surjective homomorphism from @xmath93 to @xmath207 . the square of @xmath207 is depicted on figure [ fig : squarea2 ] ; and , a surjective homomorphism from it to @xmath93 is depicted on figure [ fig : surjhom : squarea2:a3 ] . thus @xmath93 and @xmath207 are equivalent w.r.t . one can also check that @xmath207 is minimal and is therefore a q - core of @xmath93 , and _ a posteriori _ of @xmath92 . the behaviour of the q - core differs from its cousins the core and the @xmath26-@xmath27-core . the q - core of a @xmath210-element structure @xmath12 is not always an induced substructure of @xmath12 . consider the signature @xmath211 involving a binary relation @xmath181 and two unary relations @xmath31 and @xmath212 . let @xmath12 and @xmath6 be structures with domain @xmath213 with the following relations . @xmath214 since @xmath6 is a substructure of @xmath12 , we have @xmath215 . conversely , the square of @xmath216 contains an edge that has no vertex in the relation @xmath31 and @xmath212 , which ensures that @xmath217 ( see figure [ fig : qcores ] ) . observe also that no two - element structure @xmath218 , and _ a fortiori _ no two - element substructure of @xmath12 agrees with them on @xmath2 . indeed , if a structure @xmath218 agrees on @xmath2 with @xmath6 , it agrees also on @xmath1 . thus , the core of @xmath6 is also the core of @xmath218 and must appear as an induced substructure of @xmath218 . this core is the one - element substructure of @xmath6 induced by 1 . in order to have a surjective homomorphism from a power of @xmath218 to @xmath6 , this power must contain a non - loop , and so does @xmath218 . this non - loop must in @xmath218 be adjacent to another vertex ( this is a @xmath2-expressible property that holds in @xmath6 @xmath219 ) . the structure @xmath218 would therefore be a two element structure satisfying @xmath220 a power of @xmath218 would therefore be connected , which is not the case of @xmath6 , preventing the existence of any surjective homomorphism . we do not know whether the q - core of a structure is unique . we will explore in the following section q - cores over some special classes and show that it behaves well in these cases . a boolean structure @xmath18 has domain @xmath221 . the results of this section apply to arbitrary ( not necessarily finite ) signatures . we give the following lemma ultimately for illustrative purposes ( the gist of its proof will be reused several times ) . the _ ph - type _ @xmath222 of @xmath223 is the set of all formulae @xmath224 in one free variable @xmath70 from @xmath2 such that @xmath225 ( ph stands for positive horn ) . [ lem : switch - aut ] if @xmath18 is a boolean structure such that the ph - types @xmath226 and @xmath227 in @xmath18 coincide , then @xmath18 has an automorphism swapping @xmath228 and @xmath229 . suppose there is no such automorphism , then we may assume there exists a conjunction of atoms @xmath230 , involving only variables @xmath70 and @xmath84 , such that @xmath231 but @xmath232[l]{\ensuremath{\models}}/ \ } \theta(1,0)$ ] . now , @xmath233 iff @xmath234 , since @xmath228 and @xmath229 are of the same ph - type . if @xmath233 then @xmath235 but @xmath232[l]{\ensuremath{\models}}/ \ } \forall x \theta(x,1)$ ] ( contradiction ) . similarly , if @xmath232[l]{\ensuremath{\models}}/ \ } \theta(0,0)$ ] then @xmath236 but @xmath232[l]{\ensuremath{\models}}/ \ } \exists x \theta(1,x)$ ] ( contradiction ) . [ thm : boolean ] let @xmath17 and @xmath18 be boolean structures that are equivalent w.r.t . @xmath2 . then @xmath17 and @xmath18 are isomorphic . we consider the ph - types @xmath237 and @xmath238 of @xmath228 and @xmath229 in @xmath17 , respectively ( likewise with @xmath18 superscripts for @xmath18 ) . case i. @xmath239 . it follows that @xmath240 , since @xmath241 iff @xmath242 iff @xmath243 iff @xmath244 etc . in this case the function @xmath245 given by @xmath246 and @xmath247 is an isomorphism , as in the proof of lemma [ lem : switch - aut ] . ( of course , @xmath248 and @xmath249 is also an isomorphism . ) @xmath237 and @xmath238 are incomparable . let @xmath250 be in @xmath237 but not in @xmath238 ; and let @xmath251 be in @xmath238 but not in @xmath237 . let @xmath252 be the witness of @xmath253 in @xmath18 ; and let @xmath254 be the witness of @xmath255 in @xmath18 . since @xmath256[l]{\ensuremath{\models}}/ \ } \exists x \theta_0(x ) \wedge \theta_1(x)$ ] , @xmath232[l]{\ensuremath{\models}}/ \ } \exists x \theta_0(x ) \wedge \theta_1(x)$ ] and @xmath257 . we claim @xmath258 ( @xmath259 , @xmath260 ) is an isomorphism from @xmath17 to @xmath18 . if not , then we may assume there exists an atom ( maybe with variables identified ) @xmath230 such that @xmath261 but @xmath232[l]{\ensuremath{\models}}/ \ } \theta(z(0),z(1))$ ] . now , @xmath262 iff @xmath263 iff @xmath264 iff @xmath265 iff @xmath266 . the proof is completed as in that of lemma [ lem : switch - aut ] . @xmath267 but @xmath268 . let @xmath251 be in @xmath238 but not in @xmath237 . let @xmath254 be the witness of @xmath255 in @xmath18 . @xmath254 is unique since @xmath232[l]{\ensuremath{\models}}/ \ } \forall x \theta_1(x)$ ] . let @xmath252 be the other element of @xmath18 . we claim @xmath258 is an isomorphism from @xmath17 to @xmath18 . if not , then there exists a conjunction of atoms @xmath230 such that ( we lose the ) either @xmath261 but @xmath232[l]{\ensuremath{\models}}/ \ } \theta(z(0),z(1))$ ] , or @xmath231 but @xmath256[l]{\ensuremath{\models}}/ \ } \theta(z(0),z(1))$ ] . in fact , we can still deal with both cases at once , since @xmath269 iff @xmath270 iff @xmath271 and @xmath234 . the proof concludes as in lemma [ lem : switch - aut ] . in the extended logic @xmath156 , it follows that every structure of size at most @xmath272 satisfies the ideal of core . for @xmath2 we can only say the following . every boolean structure @xmath18 is either a q - core , or its q - core is the substructure induced by either of its elements . in particular , * the * q - core of @xmath18 is unique up to isomorphism and is an induced substructure of @xmath18 . if @xmath18 generates the same qcsp as a one - element structure @xmath17 with domain @xmath273 , then it is clear that the ph - types @xmath274 , @xmath275 and @xmath237 coincide . this gives uniqueness and induced substructure in this case . otherwise , the only possibility to violate the statement is for @xmath18 to have as q - core a non - induced substructure . but this is impossible by theorem [ thm : boolean ] . lemma [ lem : switch - aut ] does not extend to structures of size three . there exists @xmath276 of size three such that every element of @xmath276 has the same ph - type and yet @xmath276 has no non - trivial automorphism ( @xmath277 and @xmath278 ) . however , @xmath276 is not a q - core . let @xmath29 be a fixed relational signature that consists of @xmath279 unary relation symbols @xmath280 . a structure over such a signature is deemed _ unary_. let @xmath281 be a string of length @xmath279 over the alphabet @xmath282 . we write @xmath283 as an abbreviation for the quantifier - free formula @xmath284=1 } m_i(x)$ ] . each element @xmath59 of a unary structure @xmath12 corresponds to a word @xmath281 , which is the largest word bitwise such that @xmath285 . let @xmath286 be the bitwise @xmath287 of the words associated to each element . the unary structure @xmath12 satisfies the _ canonical universal sentence _ @xmath288 ( note that @xmath286 is also the largest word bitwise among such satisfied universal formulae ) . the q - core of a unary structure @xmath12 is the unique substructure of @xmath12 defined as follows . the q - core of @xmath12 is the core @xmath289 of @xmath12 if they share the same canonical universal sentence and the disjoint union of @xmath289 with a single element corresponding to @xmath286 where @xmath288 is the canonical universal sentence of @xmath12 . a positive horn sentence over a unary signature is logically equivalent to a conjunction of formulae that do not share any variables . each conjunct is either a universal formulae of the form @xmath290 or an existential formulae of the form @xmath291 . the core @xmath289 is a substructure of the q - core , so we need only enforce that the q - core and @xmath289 satisfy the same canonical universal sentence . this is achieved with the optional addition of an element corresponding to @xmath286 where @xmath292 is the canonical universal sentence of @xmath12 . we say that a ph - sentence is a _ proper ph - sentence _ , if it has at least one universally quantified variable @xmath293 that occurs in some atom in the quantifier - free part . generally , we will say that a sentence is _ proper _ if any variable @xmath293 occurs in some atom and we will be always working with such sentences unless otherwise stated ( otherwise , we would simply discard @xmath293 and consider the equivalent sentence with one less variable ) . we will say that a proper ph - sentence @xmath74 induced from a proper ph - sentence @xmath13 by the removal of some conjuncts is a _ proper subsentence _ of @xmath74 . let @xmath29 be a signature that consists of finitely many relation symbols @xmath294 of respective arity @xmath295 . we will consider the set of minimal proper ph - sentences w.r.t . @xmath29 , that is all formulae of the form @xmath296 where the tuple @xmath297 is a permutation of the variables @xmath298 where @xmath293 has been transposed with some other variable . there are @xmath299 such formulae . let @xmath12 be a @xmath29-structure . the following are equivalent . 1 . @xmath12 does not satisfy any proper ph - sentence . 2 . @xmath12 does not satisfy any of the @xmath300 minimal proper ph - sentences w.r.t . 3 . @xmath301 contains an isolated element . the first point implies trivially the second . we show the converse . note that any proper ph - sentence @xmath13 contains as a proper subinstance a sentence of the form @xmath302 , where the @xmath303 represent some arbitrary quantifiers . by assumption , the structure @xmath12 models @xmath304 thus , it follows that @xmath12 models the weaker sentence @xmath305 @xmath306 ( the same strategy for selecting a witness for @xmath307 will work ) and the even weaker sentence where some universal quantifiers are turned to existential ones , namely those for which @xmath308 is universal ( the strategy for these new existential variable can be chosen arbitrarily ) . so , @xmath12 does not model the negation of this last sentence which is @xmath309 , which is a subinstance of @xmath13 . by monotonicity , @xmath12 does not model @xmath13 either . we now prove that the second point implies the third . let @xmath310 be the @xmath311th minimal proper ph - sentence w.r.t . @xmath29 . let @xmath120 be a witness for the unique existential variable of @xmath312 that @xmath12 does not satisfy @xmath310 . it is a simple exercise to check that @xmath313 is an isolated element of @xmath301 . conversely , if @xmath314 is an isolated element of @xmath301 then @xmath315 is a witness that @xmath301 does not satisfy any minimal proper ph - sentence @xmath310 . consequently , there exists some @xmath316 witnessing that @xmath12 does not satisfy @xmath310 and we are done . in the case of directed graphs , the minimal proper ph - sentences are @xmath317 and @xmath318 . a directed graph which does not satisfy them will satisfy their negation @xmath319 and @xmath320 . a witness for the existential @xmath293 in the first sentence will be a _ source _ , and in the second sentence a _ sink _ , respectively so a directed graph has a source and a sink if , and only if , it does not satisfy any proper ph - sentence , if and only if , its square has an isolated element . * the * q - core of a structure @xmath12 that does not satisfy any proper ph - sentence is the unique substructure of @xmath12 may be found as follows . the q - core of @xmath12 is the core @xmath289 of @xmath12 , if @xmath321 contains an isolated element , and the disjoint union of @xmath289 and an isolated element , otherwise . by assumption @xmath12 does not satisfy any proper ph - sentence . consequently a minimal structure @xmath289 ( both w.r.t . domain size and number of tuples ) which satisfies the same ph - sentences as @xmath12 will satisfy the same pp - sentences as @xmath12 and none of the proper ph - sentences either . it follows that @xmath289 must contain the core of @xmath12 ( and can be no smaller ) . if this core satisfies no proper ph - sentence then we are done and by the previous theorem @xmath301 has an isolated element . otherwise , we must look for a structure that contains the core of @xmath12 and does not satisfy any proper universal sentence . adding tuples to @xmath12 can clearly not force this property by monotonicity of qcsp . thus , the minimal ( and unique ) such structure will be obtained by the addition of an isolated element . note that in this second case we have also a substructure of @xmath12 . it follows that checking whether a structure with an isolated element is a q - core is of the same complexity as checking whether it is a core . recall that the latter is known to be a co - np - complete decision problem ( the induced sub - structure that ought to be a core is given via an additional monadic predicate @xmath322 ) , see @xcite . we show that the former is also co - np - complete ( we also assume a monadic predicate as in this particular case the q - core is an induced substructure ) , by showing inter - reducibility of both problem . ( hardness ) let @xmath323 be the input to the core problem ( we assume that it is not trivial and that @xmath322 does not contain any isolated element ) . if @xmath301 has an isolated element ( it can be done in polynomial time as @xmath300 does not depend on @xmath324 ) then we reduce to @xmath325 , and to @xmath326 otherwise , where @xmath327 consists of the disjoint union of @xmath12 with an isolated element and @xmath328 is the union of @xmath322 with this new element . ( co - np - complete algorithm ) let @xmath329 be the input to the q - core problem such that @xmath327 has an isolated element . we check whether the alleged q - core ( the substructure of @xmath327 induced by @xmath330 ) elevated to the @xmath22th power has an isolated element . if it does not we answer no . if the alleged q - core has more than one isolated element we answer also no . otherwise , we remove at most one isolated element from @xmath328 to derive @xmath331 and reduce to the core question w.r.t . @xmath332 . graphs are relational structures with a single symmetric relation @xmath181 . we term them _ partially reflexive _ to emphasise that any vertex may or may not have a self - loop . tree _ may contain self - loops but no larger cycle @xmath333 for @xmath334 . pseudotree _ contains at most one cycle @xmath333 for @xmath334 . ( resp . , _ pseudoforest _ ) is the disjoint union of p.r . trees ( resp . , pseudotrees ) . since forests ( resp . , pseudoforests ) are closed under substructures , we can be assured that a q - core of a forest ( resp . , pseudoforest ) is a forest ( resp . , pseudoforest ) . it is clear from inspection that the q - core of forest ( resp . , pseudoforest , cycle ) is unique up to isomorphism , but we do not prove this as it does not shed any light on the general situation . the doubting reader may substitute `` a/ all '' for `` the '' in future references to q - cores in this section . the complexity classifications of @xcite were largely derived using the properties of equivalence w.r.t . @xmath2 . this will be the central justification for the following propositions . let @xmath335 and @xmath336 be the reflexive and irreflexive @xmath311-cliques , respectively . let @xmath337:=\{1,\ldots , n\}$ ] . for @xmath338 $ ] and @xmath339 , let @xmath340 $ ] be the @xmath311th entry of @xmath341 . for @xmath342 , let @xmath343 be the path with domain @xmath337 $ ] and edge set @xmath344=1 \}$ ] for a tree @xmath345 and vertex @xmath346 , let @xmath347 be the shortest distance in @xmath345 from @xmath348 to a looped vertex ( if @xmath345 is irreflexive , then @xmath347 is always infinite ) . let @xmath349 be the maximum of @xmath350 . a tree is _ loop - connected _ if the self - loops induce a connected subtree . a tree @xmath345 is _ quasi - loop - connected _ if either 1 . ) it is irreflexive , or 2 . ) there exists a connected reflexive subtree @xmath351 ( chosen to be * maximal * ) such that there is a walk of length @xmath349 from every vertex of @xmath345 to @xmath352 . it is not true that , if @xmath353 is a forest , then either @xmath353 admits a majority polymorphism , and qcsp@xmath354 is in nl , or qcsp@xmath354 is np - hard . however , the notion of q - core restores a clean delineation . let @xmath353 be a forest . then either the q - core of @xmath353 admits a majority polymorphism , and qcsp@xmath354 is in nl , or qcsp@xmath354 is np - hard . we assume that graphs have at least one edge ( otherwise the q - core is @xmath355 ) . irreflexive forests are a special case of bipartite graphs , which are all equivalent w.r.t . @xmath2 , their q - core being @xmath16 when they have no isolated vertex ( see example [ ex : bip ] ) and @xmath356 otherwise . we assume from now on that graphs have at least one edge and one self - loop . the one vertex case is @xmath357 . we assume larger graphs from now on . if the graph contains an isolated element then its q - core is @xmath358 . assume from now on that the graph does not have an isolated element . we deal with the disconnected case first . if the graph is reflexive , then its q - core is @xmath359 . otherwise , the graph is properly partially reflexive in the sense that it embeds both @xmath360 and @xmath355 . if the graph has an irreflexive component then its q - core is @xmath361 . if the graph has no irreflexive component , then its q - core is @xmath362 where @xmath363 is the longest walk from any vertex to a self - loop . to see this last case gives an equivalent qcsp , we may consider power surjective homomorphisms , together with the fact that the q - core must not satisfy @xmath364 . we now follow the classification of @xcite . if a forest contains more than one tree , then the q - core is among those formed from the disjoint union of exactly two ( including the possibility of duplication ) of @xmath355 , @xmath365 , @xmath366 , @xmath16 . each of these singularly admits a majority polymorphism , therefore so does any of their disjoint unions . we now move on to the connected case , it remains to consider trees @xmath367 . if @xmath367 is irreflexive , then its q - core is @xmath16 or @xmath355 , which admit majority polymorphisms . if @xmath367 is loop - connected , then it admits a majority polymorphism @xcite . if @xmath367 is quasi - loop - connected , then it is qcsp - equivalent to one of its subtrees that is loop - connected @xcite which will be its q - core . in all other cases qcsp@xmath368 is np - hard , and @xmath367 does not admit majority @xcite . a _ pseudotree _ is a graph that involves at most one cycle . a _ pseudoforest _ is the disjoint union of a collection of pseudotrees . let @xmath353 be an irreflexive pseudoforest . then either the q - core of @xmath353 admits a majority polymorphism , and qcsp@xmath354 is in nl , or qcsp@xmath354 is np - hard . we follow the classification of @xcite . if @xmath353 is bipartite , then its q - core is either @xmath16 , @xmath355 , @xmath356 ( see @xcite ) and this admits a majority polymorphism . otherwise its q - core contains an odd cycle , which does not admit a majority polymorphism , and qcsp@xmath354 is np - hard . we may use theorem [ theo : containment : mylogic ] to provides a first algorithm ( algorithm [ algo : naiveqcore ] ) . this does not appear very promising if we wish to use q - cores as a preprocessing step . we will propose and illustrate a general and less naive method to compute q - cores by computing @xmath26-@xmath27-core and cores first . another nice feature of cores and @xmath26-@xmath27-cores which implies their uniqueness is the following : any substructure @xmath218 of @xmath12 that agrees with it on @xmath1 ( respectively on @xmath28 ) will contain the core ( respectively the @xmath26-@xmath27-core ) . consequently , the core and the @xmath26-@xmath27-core may be computed in a _ greedy fashion_. assuming that the q - core would not satisfy this nice property , why should this concern the q - core ? well , we know that * any * q - core will lie somewhere between the @xmath26-@xmath27-core and the core that are induced substructures : this is a direct consequence of the inclusion of the corresponding fragments of first - order logic and their uniqueness . moreover , according to our current knowledge , checking for equivalence appears , at least on paper , much easier for @xmath28 than @xmath2 : compare the number of functions from @xmath30 to the power set of @xmath58 ( @xmath369 ) with the number of functions from @xmath370 to @xmath58 ( @xmath371 ) where @xmath22 could be as large as @xmath372 and can certainly be greater than @xmath373 ( see proposition [ prop : lowerbound : powerepimorphism ] ) . so it make sense to bound the naive search for q - cores . furthermore , we know that the @xmath26-@xmath27-core can be identified by specific surjective hypermorphisms that act as the identity on @xmath27 and contain the identity on @xmath26 @xcite which makes the search for the @xmath26-@xmath27-core somewhat easier than its definition suggest ( see algorithm [ algo : uxcore ] ) . observe also that @xmath27 must contain the core @xmath218 of the @xmath26-@xmath27-core @xmath6 , which is also the core of the original structure @xmath12 ( this is because @xmath54 induces a so - called retraction of @xmath12 to the substructure @xmath374 induced by @xmath27 ) . thus we may compute the core greedily from @xmath27 . next , we do a little bit better than using our naive algorithm , by interleaving steps where we find a substructure that is @xmath2-equivalent , with steps where we compute its @xmath26-@xmath27-core ( one can find a sequence of distinct substructures @xmath6 , @xmath375 and @xmath376 such that @xmath6 is a @xmath26-@xmath27-core , which is @xmath2-equivalent to @xmath375 , whose @xmath26-@xmath27-core @xmath376 is strictly smaller than @xmath6 , see example [ example : interleaving : uxcore : qretract ] ) . algorithm [ algo : boundedsearchqcore ] describes this proposed method informally when we want to compute * one * q - core ( of course , we would have no guarantee that we get the smallest q - core , unless the q - core can be also greedily computed , which holds for all cases we have studied so far ) . in algorithm [ algo : boundedsearchqcore ] , we have purposely not detailed line [ algo : line : q - retractioncheck ] . we could use the characterisation of @xmath2-containment via surjective homomorphism from a power of theorem [ theo : containment : qcsplogic ] as in algorithm [ algo : naiveqcore ] . alternatively , we can use a refined form of ( iii ) in this theorem and use the canonical sentences in @xmath184-form @xmath377 and @xmath378 , with @xmath379 and @xmath380 ( see definition [ def : canonical : pi2:sentence ] ) . the test would consists in checking that @xmath6 satisfies @xmath378 ( where we may relativise to universal variables to @xmath381 and existential variables to @xmath27 ) and @xmath375 satisfies @xmath377 . this is correct because we know that we may relativise every universal variable to @xmath26 within @xmath6 . thus , it suffices to consider @xmath184-sentences with at most @xmath382 universal variables . [ example : interleaving : uxcore : qretract ] we describe a run of algorithm [ algo : boundedsearchqcore ] on input @xmath383 . during the initialisation , we compute its @xmath26-@xmath27-core @xmath384 and discover that @xmath149 and @xmath385 . we compute @xmath386 , the core of the substructure induced by @xmath27 . note that @xmath387 is isomorphic to @xmath388 . next the algorithm guesses a substructure @xmath375 of @xmath6 that contains @xmath218 : _ e.g. _ it drops the self - loop around vertex @xmath210 to obtain a structure isomorphic to @xmath389 and checks successfully equivalence w.r.t . @xmath2 ( there is a surjective homomorphism from @xmath389 to @xmath388 ; and , conversely we can use the surjective homomorphism from @xmath390 to @xmath391 composed with that from the former to @xmath389 ) . next the algorithm computes the @xmath392-@xmath393-core @xmath376 of @xmath375 which is @xmath207 ( witnessed by @xmath394 , @xmath395 , @xmath396 ) and sets @xmath397 . the algorithm stops eventually and outputs @xmath207 as it is minimal . we have introduced a notion of q - core and demonstrated that it does not enjoy all of the properties of cores and @xmath26-@xmath27-core . in particular , there need not be a unique minimal element size in the equivalence class of structures agreeing on ph - sentences . however , we suspect that the notion of q - core we give is robust , in that the q - core of any structure @xmath18 is unique up to isomorphism ; and , that it sits inside any substructure of @xmath6 that satisfies the same sentence of @xmath2 , making it computable in a greedy fashion . thus , the nice behaviour of q - cores is almost restored , but `` induced substructure '' in the properties of core or @xmath26-@xmath27-core must be replaced by the weaker `` substructure '' . generalising the results about q - cores of structures with an isolated element to disconnected structures is already difficult . just as the ph - theory of structures with an isolated element is essentially determined by their pp - theory , so the ph - theory of disconnected structures is essentially determined by its @xmath398 fragment ( see @xcite ) . |
it is proved that the median eigenvalues of every connected bipartite graph of maximum degree at most three belong to the interval $ ] with a single exception of the heawood graph , whose median eigenvalues are . moreover , if is not isomorphic to the heawood graph , then a positive fraction of its median eigenvalues lie in the interval $ ] . this surprising result has been motivated by the problem about homo - lumo separation that arises in mathematical chemistry . 2010 mathematics subject classification : 05c50 | 1309.7395 | in a recent work , fowler and pisanski @xcite introduced the notion of the _ hl - index _ of a graph that is related to the homo - lumo separation studied in theoretical chemistry ( see also jakli , fowler , and pisanski @xcite ) . this is the gap between the highest occupied molecular orbital ( homo ) and lowest unoccupied molecular orbital ( lumo ) . in the hckel model @xcite , the energies of these orbitals are in linear relationship with eigenvalues of the corresponding molecular graph and can be expressed as follows . let @xmath0 be a graph of order @xmath3 , and let @xmath4 be the eigenvalues of its adjacency matrix . the eigenvalues occurring in the homo - lumo separation are the median eigenvalues @xmath5 and @xmath6 , where @xmath7 the _ hl - index _ @xmath8 of the graph @xmath0 is then defined as @xmath9 a simple unweighted graph @xmath0 is said to be _ subcubic _ if its maximum degree is at most @xmath10 . in @xcite it is proved that every subcubic graph @xmath0 satisfies @xmath11 and that if @xmath0 is bipartite , then @xmath12 . the following is the main result from @xcite . [ thm : cubic ] the median eigenvalues @xmath13 and @xmath14 of every subcubic graph @xmath0 are contained in the interval @xmath15 $ ] , i.e. , @xmath16 . this result is best possible since the heawood graph ( the bipartite incidence graph of points and lines of the fano plane ) has @xmath17 . the following conjecture was proposed in @xcite . [ conj : planar ] if @xmath0 is a planar subcubic graph , then @xmath18 . the conjecture has been verified for planar bipartite graphs in @xcite . in this paper we prove a surprising extension of @xcite and of conjecture [ conj : planar ] that holds for all bipartite subcubic graphs with a single exception of the heawood graph ( or disjoint union of copies of it ) . the following are our main results . [ thm : cubic bipartite ] let @xmath0 be a bipartite subcubic graph . if every connected component of @xmath0 is isomorphic to the heawood graph , then @xmath19 . in any other case , the median eigenvalues @xmath13 and @xmath14 are contained in the interval @xmath1 $ ] , i.e. , @xmath20 . theorem [ thm : cubic bipartite ] shows that the median eigenvalues @xmath21 and @xmath22 are small , but our proof can be tweaked to give much more a positive fraction of ( median ) eigenvalues lie in the interval @xmath1 $ ] . [ thm : cubic bipartite linearly many ] there is a constant @xmath23 such that for every bipartite subcubic graph @xmath0 , none of whose connected components is isomorphic to the heawood graph , all its eigenvalues @xmath24 , where @xmath25 , belong to the interval @xmath1 $ ] . let us first recall that eigenvalues of bipartite graphs are symmetric with respect to 0 , i.e. , if @xmath26 is an eigenvalue , then @xmath27 is an eigenvalue as well and has the same multiplicity as @xmath26 . this in particular implies that @xmath28 and that @xmath29 . therefore , it suffices to consider @xmath21 . let us next recall the eigenvalue interlacing theorem ( cf . , e.g. , @xcite ) that will be our main tool in the sequel . [ thm : interlacing ] let @xmath30 be a vertex set of cardinality @xmath31 , and let @xmath32 . then for every @xmath33 , we have @xmath34 if @xmath35 is a partition of the vertices of @xmath0 , we denote by @xmath36 the subgraph of @xmath0 induced on @xmath37 . in the sequel we will consider vertex partitions @xmath35 , but the two parts @xmath38 will play different roles . thus , we shall consider such a partition as an ordered pair @xmath39 . given a partition @xmath39 of @xmath40 , let @xmath41 be the smallest integer such that @xmath42 , and let @xmath43 . then we say that the partition @xmath39 is _ @xmath44-imbalanced _ , and we define the _ imbalance _ of the partition @xmath39 as @xmath45 [ lem : k - unbalanced ] suppose that @xmath39 is an @xmath44-imbalanced vertex partition of a subcubic graph @xmath0 . if @xmath46 , then @xmath47 . in particular , if @xmath48 , then @xmath49 . conditions of the lemma give that @xmath42 and @xmath50 if @xmath51 is even , then @xmath52 . if @xmath3 is odd , then @xmath53 . in each case , @xmath54 since @xmath36 is obtained from @xmath0 by deleting @xmath55 vertices and @xmath56 , the eigenvalue interlacing theorem shows that @xmath57 . let @xmath39 be a partition of @xmath40 . suppose that @xmath58 is a set of vertices in @xmath59 . we say that @xmath60 _ increases imbalance _ of @xmath39 if @xmath61 . the following result will be our main tool for finding imbalance - increasing vertex - sets . [ lem : increase imbalance ] suppose that @xmath39 is a partition of @xmath40 and @xmath58 . let @xmath62 be the union of those connected components of @xmath63 that contain vertices in @xmath60 . if @xmath64 , then @xmath60 increases imbalance of @xmath39 . let @xmath39 be @xmath44-imbalanced . let @xmath65 and @xmath66 . then @xmath67 . note that @xmath68 and @xmath69 have the same connected components except for those contained in @xmath62 . since @xmath70 and @xmath71 , we have that @xmath72 . thus @xmath73 is @xmath74-imbalanced , where @xmath75 . hence , @xmath76 . for @xmath77 , let @xmath78 denote the set of all vertices in @xmath79 that have a neighbor in @xmath60 . the statement of lemma [ lem : increase imbalance ] has a converse under a mild restriction on @xmath78 . [ lem : increase imbalance converse ] suppose that @xmath39 is a partition of @xmath40 and @xmath58 . if @xmath80 and every vertex in @xmath78 has all its neighbors in @xmath59 , then @xmath60 increases imbalance of @xmath39 if and only if @xmath64 , where @xmath81 . let us first observe that the independence of @xmath78 in @xmath68 implies that the subgraph @xmath62 that appears in lemma [ lem : increase imbalance ] is equal to @xmath82 . therefore , it remains to prove only the direction converse to the one of the previous lemma , i.e. , that the increased imbalance implies that @xmath64 . let us adopt the notation used in the proof of lemma [ lem : increase imbalance ] and let @xmath83 be the smallest integer such that @xmath84 . let us assume that @xmath85 . since @xmath86 , it suffices to see that @xmath87 . however , this is an easy observation since @xmath68 and @xmath69 have the same eigenvalues apart from the eigenvalues of @xmath62 in @xmath69 that are replaced by @xmath88 eigenvalues , all equal to 0 , in @xmath68 . suppose now that @xmath0 is a bipartite graph and that @xmath39 is its bipartition . a set @xmath89 of vertices of @xmath0 is _ ( _ @xmath37-thick _ ) in @xmath0 if every vertex in @xmath90 ( resp . @xmath91 ) has at most one neighbor in @xmath92 , every vertex in @xmath93 ( resp . @xmath94 ) has at most one neighbor in @xmath89 , and @xmath95 ( resp . @xmath96 ) . the set @xmath89 is _ thick _ if it is either @xmath59-thick or @xmath37-thick . [ lem : thick unbalanced ] if @xmath89 is an @xmath59-thick set of vertices in @xmath0 , then the set @xmath97 increases imbalance of the bipartition @xmath39 . consider the subgraph @xmath62 of @xmath63 consisting of those connected components that contain vertices in @xmath60 , and let @xmath98 . thickness condition implies that after removing vertices in @xmath99 from @xmath62 , we are left with a graph consisting of a matching and isolated vertices , so its eigenvalues are all in the interval @xmath1 $ ] . by the eigenvalue interlacing theorem , we conclude that @xmath100 , so @xmath60 increases imbalance of @xmath39 by lemma [ lem : increase imbalance converse ] . in this section we prove that for every vertex @xmath101 of a connected bipartite subcubic graph @xmath0 with bipartition @xmath39 , if @xmath0 is not isomorphic to the heawood graph , then a small neighborhood around @xmath101 contains a set @xmath60 that can be used to increase imbalance of the partition @xmath39 or @xmath102 . from now on we assume that @xmath0 is bipartite and @xmath39 is the bipartition of @xmath0 . given a graph @xmath0 and its vertex @xmath103 , we denote by @xmath104 the set of all vertices of @xmath0 whose distance from @xmath103 is at most @xmath105 . we will sometimes consider the set @xmath104 as the subgraph of @xmath0 induced on this vertex set . [ lem : improve imbalance ] suppose that @xmath39 is the bipartition of a bipartite subcubic graph @xmath0 , @xmath106 , and the connected component of @xmath0 containing @xmath101 is not isomorphic to the heawood graph . then @xmath107 contains a set @xmath60 of vertices such that either @xmath58 and @xmath60 increases imbalance of @xmath39 , or @xmath108 and @xmath60 increases imbalance of @xmath102 . before giving the proof of lemma [ lem : improve imbalance ] , let us show how the lemma implies our main results , theorems [ thm : cubic bipartite ] and [ thm : cubic bipartite linearly many ] . the proof of theorem [ thm : cubic bipartite ] follows from the proof of theorem [ thm : cubic bipartite linearly many ] given below by taking @xmath109 , where @xmath101 is an arbitrary vertex of @xmath0 . thus we only need to take care of theorem [ thm : cubic bipartite linearly many ] . let @xmath0 be a bipartite subcubic graph with no component isomorphic to the heawood graph . for each vertex @xmath103 of @xmath0 , we have @xmath110 . therefore , @xmath0 contains a set @xmath111 of vertices that are mutually at distance at least 38 and @xmath112 , where @xmath113 . let us consider the bipartition @xmath39 of @xmath0 , and let , for each @xmath114 , @xmath115 be the vertex set @xmath60 obtained by applying lemma [ lem : improve imbalance ] . let @xmath116 denote the number of vertices @xmath114 such that @xmath117 , and let @xmath118 be the number of cases where @xmath119 . note that @xmath120 . thus , @xmath121 let @xmath73 be the partition obtained from @xmath39 by removing from @xmath59 and adding into @xmath37 all sets @xmath115 ( @xmath114 ) for which @xmath117 . since every @xmath115 is contained in @xmath122 , all these sets are pairwise at distance at least 4 from each other . consequently , their graphs @xmath123 are pairwise disjoint and non - adjacent . hence , each of these sets increases imbalance of @xmath39 by at least 1 ( lemma [ lem : increase imbalance converse ] ) . therefore , @xmath124 similarly , adding to @xmath59 all sets @xmath115 ( @xmath114 ) for which @xmath119 , we obtain a partition @xmath125 such that @xmath126 finally , all three inequalities ( [ eq : pf1])([eq : pf3 ] ) imply that @xmath127 by symmetry , we may assume that @xmath128 . then ( [ eq : pf4 ] ) implies that @xmath129 , and lemma [ lem : k - unbalanced ] gives the claim of the theorem with @xmath130 . the family @xmath131 of graphs listed in the appendix ( figures [ fig : appendix1][fig : appendix3 ] ) has the following property : if @xmath132 and @xmath60 is the bipartite set of its vertices that are shown as full circles or full squares in figures [ fig : appendix1][fig : appendix3 ] , then @xmath133 . lemma [ lem : increase imbalance converse ] shows that the following is the common outcome for all of these graphs : [ cor : appendix ] suppose that @xmath134 is one of the graphs depicted in figures [ fig : appendix1][fig : appendix3 ] . let @xmath60 be the bipartite class of its vertices that are drawn as full circles or full squares . if a bipartite subcubic graph @xmath0 with bipartition @xmath39 contains @xmath135 as an induced subgraph , where @xmath58 , and every vertex in @xmath60 has all its neighbors in @xmath135 , then @xmath60 increases imbalance of the bipartition @xmath39 of @xmath0 . the proof is clear by observing that the component @xmath62 in @xmath63 containing @xmath60 is equal to @xmath135 and by the remarks given in the paragraph before the corollary . we shall need some new concepts . let @xmath136 be an integer . we say that vertices @xmath137 and @xmath138 of @xmath0 are _ @xmath31-adjacent _ in @xmath0 if there is a path of length @xmath31 joining them . if @xmath135 is a subgraph of @xmath0 , a _ @xmath31-chord _ of @xmath135 in @xmath0 is a path @xmath139 of length @xmath31 , such that @xmath140 . having such a @xmath31-chord @xmath141 , we say that @xmath101 and @xmath142 are _ @xmath31-adjacent outside _ the subgraph @xmath135 is _ @xmath31-induced _ in @xmath0 if it has no @xmath143-chords for @xmath144 . note that the special case when @xmath145 gives the usual notions of being adjacent , a chord , or an induced subgraph . when dealing with vertices of degree 2 in the proof of lemma [ lem : improve imbalance ] , we will need the following result . [ prop : path in the ball ] let @xmath0 be a bipartite graph of girth at least 6 , let @xmath106 , and let @xmath105 be a positive integer . if @xmath146 , then there exists a 2-induced path from @xmath137 to @xmath138 that is contained in @xmath147 . for @xmath148 , we will denote by @xmath149 the distance in @xmath0 from @xmath150 to @xmath103 . let @xmath151 and @xmath152 be shortest paths from @xmath137 and from @xmath138 to @xmath101 , respectively . choose these paths so that their intersection is a path @xmath62 , and @xmath62 is as long as possible . since @xmath0 has no cycles of length 4 , both paths @xmath151 and @xmath152 are 2-induced . let @xmath153 be the first vertex of intersection of @xmath151 and @xmath152 when traversing the paths in the direction towards @xmath101 . then @xmath62 is a path from @xmath153 to @xmath101 . the path @xmath154 consisting of the segments of @xmath151 from @xmath137 to @xmath153 and of @xmath152 from @xmath153 to @xmath138 has length @xmath155 . first , we claim that @xmath154 is an induced path . namely , the segments of the two paths are induced , so if there were a chord @xmath156 of @xmath154 , then @xmath157 and @xmath158 . since @xmath0 is bipartite , we have @xmath159 , and it is easy to see that this contradicts our choice of the paths with @xmath62 being longest possible since we could replace one of the paths by a path using the edge @xmath156 and thus increasing the intersection of the two paths . if @xmath160 is not 2-induced , let @xmath161 be a 2-chord , where @xmath157 and @xmath158 . let us choose the 2-chord such that @xmath162 is maximum possible . as before , the maximality of @xmath62 shows that @xmath163 . let @xmath164 be the path from @xmath137 to @xmath138 obtained from @xmath154 by using the 2-chord @xmath161 instead of the path from @xmath150 to @xmath103 in @xmath154 . it is clear from our choices that any chords or 2-chords of @xmath164 must use the vertex @xmath165 . however , since @xmath0 has no 4-cycles , any such chord or 2-chord would give a contradiction to the maximality property of @xmath62 . therefore , @xmath164 is 2-induced . the conclusion from the above paragraph is that either @xmath154 is 2-induced , or @xmath164 exists and is 2-induced . in each case we obtain the statement of the proposition . in the proof , we do not intend to optimize the distance from @xmath101 in which we are able to find a set that increases imbalance . our main aim is to keep the proof simple . suppose that @xmath0 and its vertex @xmath101 give a counterexample to the theorem . we may assume that @xmath0 is connected . as before , we assume that @xmath39 is the bipartition of @xmath0 . we shall proceed through a sequence of claims , concerning vertices in vicinity of @xmath101 . in each claim we will assume that the claim is false and then define certain vertex set @xmath60 ( where @xmath58 or @xmath108 ) . we let @xmath166 . observe that @xmath167 is the subgraph @xmath62 that appears in lemmas [ lem : increase imbalance ] and [ lem : increase imbalance converse ] about increasing imbalance . _ claim 1 : each vertex in @xmath107 has degree at least 2 . _ if @xmath103 has degree at most 1 , then @xmath168 increases imbalance since @xmath169 and @xmath170 . _ claim 2 : @xmath171 contains no 4-cycles . _ suppose that @xmath172 is a 4-cycle in @xmath171 . for @xmath173 , let @xmath174 be the neighbor of @xmath175 that is not in @xmath176 ( if @xmath177 , then we set @xmath178 ) . if @xmath179 , then let @xmath180 . we may assume that @xmath58 . clearly , @xmath181 is isomorphic to the graph @xmath182 depicted in figure [ fig : appendix1 ] ( or to an induced subgraph of @xmath182 if @xmath183 or @xmath184 ) . corollary [ cor : appendix ] shows that @xmath60 increases imbalance of @xmath39 . similarly , if @xmath185 , we may take @xmath186 and @xmath60 increases imbalance of @xmath102 . finally , assume that @xmath187 and @xmath188 . let us now consider the 4-cycle @xmath189 . again , if @xmath190 and @xmath191 have no common neighbor outside this cycle , we are done by taking @xmath192 . ( note that @xmath193 since @xmath194 . ) if they have such a common neighbor , this must be @xmath195 , and hence @xmath196 . in this case , @xmath197 increases imbalance . _ claim 3 : if @xmath198 contains a vertex of degree 2 , then @xmath171 contains precisely two vertices of degree 2 and they are adjacent to each other . _ let us first prove that @xmath171 contains at most two vertices of degree 2 , and if there are two , one of them is in @xmath59 and the other one is in @xmath37 . to see this , suppose that @xmath199 have degree 2 and @xmath200 . by proposition [ prop : path in the ball ] , there exists a 2-induced path @xmath141 from @xmath150 to @xmath103 in @xmath107 . let @xmath201 . since @xmath141 is 2-induced and @xmath202 have degree 2 , the graph @xmath167 is isomorphic to the graph @xmath203 shown in figure [ fig : appendix1 ] , where the horizontal path shown at the bottom of the drawing is @xmath141 and @xmath204 is its length . since @xmath205 , corollary [ cor : appendix ] completes the proof . suppose now that @xmath206 , where @xmath207 . suppose first that @xmath103 does not belong to a cycle of length 6 . let @xmath208 be the neighbors of @xmath103 , and let @xmath191 ( @xmath209 ) be a neighbor of @xmath210 ( @xmath211 ) that is different from @xmath103 . finally , let @xmath212 . let @xmath213 . we claim that @xmath62 is isomorphic to the graph @xmath214 depicted in figure [ fig : appendix1 ] ( since @xmath215 ) . to see this , we have to prove that vertices in @xmath216 are distinct and non - adjacent , apart from their adjacencies in @xmath217 . clearly , @xmath103 is at distance at most 3 from all vertices in @xmath62 . if two of them were adjacent or the same ( apart from adjacencies in @xmath217 ) , then we would obtain a cycle of length at most 7 containing @xmath103 . since @xmath0 is bipartite and @xmath103 does not belong to a cycle of length 4 or 6 , this is not possible . this proves the claim . now , we are done by corollary [ cor : appendix ] . finally , let @xmath218 be a 6-cycle containing @xmath103 . as shown above , we may assume that @xmath219 . by symmetry , we may also assume that @xmath220 . it suffices to prove that @xmath221 . suppose for a contradiction that @xmath222 . for @xmath223 , let @xmath174 be the neighbor of @xmath175 that is not in @xmath224 . in the preceding paragraph we proved that taking the set @xmath225 gives the outcome of the theorem unless @xmath191 and @xmath226 are 2-adjacent ( which in turn gave rise to the 6-cycle @xmath227 ) . the same argument can be repeated on the sets @xmath228 , @xmath229 , and @xmath230 . they show that @xmath195 is the common neighbor of @xmath191 and @xmath231 , @xmath232 is the common neighbor of @xmath190 and @xmath226 , and that @xmath190 and @xmath231 have a common neighbor @xmath233 . ( here we used the fact that there are no 4-cycles in @xmath171 and that all treated vertices are in @xmath171 . ) if the subgraph @xmath234 of @xmath0 induced on the vertex set @xmath235 is 2-induced , then it is thick and we are done by lemma [ lem : thick unbalanced ] . therefore , two of its vertices are 2-adjacent . the only pairs that could be 2-adjacent outside @xmath234 without creating a 4-cycle are @xmath236 and @xmath237 . both of these give isomorphic outcomes since @xmath238 and @xmath165 would have played the role of @xmath195 and @xmath232 if taking the 6-cycle @xmath239 instead of @xmath227 . thus , we may assume that @xmath240 is a 2-chord . let @xmath241 . observe that @xmath242 since a neighbor of @xmath103 belongs to @xmath243 . then @xmath166 is isomorphic to the graph @xmath244 shown in figure [ fig : appendix1 ] . thus , we obtain the outcome of the theorem by corollary [ cor : appendix ] . this completes the proof of claim 3 . _ claim 3a . if @xmath245 and @xmath246 are adjacent degree-2 vertices of @xmath0 , then they are contained in a 6-cycle @xmath247 and there are vertices @xmath248 such that @xmath249 is also a 6-cycle in @xmath0 . _ we shall use the notation used in the last part of the proof of claim 3 , except that we now assume that @xmath250 and @xmath251 both have degree 2 . considering the set @xmath252 , we conclude that @xmath190 is adjacent to @xmath232 . this gives rise to @xmath253 . since a neighbor of @xmath254 belongs to @xmath243 , we conclude that @xmath255 . an induced @xmath204-cycle @xmath256 in @xmath0 , in which either no two vertices in the set @xmath257 are 2-adjacent outside @xmath258 , or no two vertices in the set @xmath259 are 2-adjacent outside @xmath258 , is called a _ good @xmath204-cycle_. _ claim 4 . @xmath107 contains neither good @xmath260-cycles nor good @xmath261-cycles . _ suppose not . let @xmath60 and @xmath262 be the vertex sets of a good 8- or 12-cycle from the definition of good cycles . it follows that either @xmath167 or @xmath263 is isomorphic to the graph @xmath264 or @xmath265 in figure [ fig : appendix1 ] ( if some of the vertices on the cycle were of degree 2 , @xmath167 or @xmath263 could be subgraph of one of these missing some of the degree-1 vertices ) . by corollary [ cor : appendix ] , @xmath60 or @xmath262 increases imbalance of either @xmath39 or @xmath102 . this proves the claim . since @xmath171 has no 4-cycles , every 8-cycle in @xmath171 is induced . by using claim 4 , it is easy to see that every 8-cycle can be written as @xmath266 , where @xmath210 and @xmath267 are 2-adjacent , and @xmath191 and @xmath268 are 2-adjacent . we shall denote such a subgraph by @xmath269 ( see figure [ fig : h_0 ] ) , and we shall later prove that every copy of @xmath269 in a vicinity of @xmath101 is 2-induced in @xmath0 ( see claim 8) . prior to that , we need some further properties . _ claim 5 . every vertex in @xmath243 is contained in a cycle of length 6 . _ suppose @xmath103 is not contained in a 6-cycle . since @xmath270 , claims 3 and 3a imply claim 5 for vertices of degree 2 and their neighbors . therefore , we may assume that @xmath103 and its neighbors are of degree 3 . let @xmath60 be the set consisting of @xmath103 and the six vertices that are at distance 2 from @xmath103 . then @xmath167 is isomorphic to the tree @xmath271 shown in figure [ fig : appendix1 ] ( or its induced subgraph with some of the leaves missing ) . by corollary [ cor : appendix ] , @xmath60 increases imbalance of either @xmath39 or @xmath102 . we say that an edge of @xmath0 is _ internal _ if it is contained in a 6-cycle in @xmath0 , and is _ external _ , otherwise . claim 5 implies that the set of external edges in the vicinity of @xmath101 form a matching in @xmath0 . by the remarks stated after claim 4 , every 8-cycle in @xmath107 is contained in a copy of the graph @xmath269 , which can be written as the union of three 6-cycles . this implies that every edge in the 8-cycle is internal . thus , external edges can not belong to cycles of length less than 10 . this fact will be used repeatedly in the proof of the next claim which shows that every 6-cycle is incident with at most one external edge . _ claim 6 . every 6-cycle in @xmath272 is incident with at most one external edge . _ let @xmath273 be a 6-cycle with two or more external edges . for @xmath274 , let @xmath275 be the edge incident with @xmath175 that is not on the cycle @xmath258 . ( we set @xmath276 if @xmath177 , and we say that @xmath277 is _ internal _ in such a case . ) since @xmath278 , claim 5 shows that there is a 6-cycle @xmath279 through @xmath174 . if @xmath277 is external , the two 6-cycles @xmath258 and @xmath279 induce a subgraph @xmath280 of @xmath0 that has only the edge @xmath277 in addition to the two cycles . if there were other adjacencies between @xmath258 and @xmath279 , then @xmath277 would be contained in a cycle of length less than 10 ( and would also be contained in @xmath281 ) , which is excluded as argued above . the same argument shows that @xmath280 is 2-induced . we may assume that @xmath282 is external , and we shall distinguish three cases , depending on whether @xmath283 , @xmath284 , is another external edge . let @xmath285 be the subgraph of @xmath0 induced on @xmath286 . figure [ fig : external edges ] shows graphs @xmath286 for @xmath284 . let us first observe that all vertices shown in figure [ fig : external edges ] are distinct in each of the three cases since any identification would give a cycle of length at most 8 through @xmath282 . ( this is clear for @xmath287 ; similarly in @xmath288 and @xmath289 where we have to observe that possible identification of @xmath137 and @xmath290 in @xmath288 or identification of @xmath137 and @xmath291 in @xmath289 or cases symmetric to these would force another identification of a neighbor of @xmath137 and thus yielding a cycle of length at most 8 containing @xmath282 . ) let @xmath60 be the set of vertices of @xmath285 that are depicted as black vertices in figure [ fig : external edges ] . since @xmath292 , we have that @xmath293 . let us first suppose that two of the vertices in @xmath60 have a common neighbor outside @xmath60 . the only possibilities ( up to symmetries ) that do not yield a cycle of length at most 8 through @xmath282 are the following : 1 . @xmath290 and @xmath138 in @xmath287 : in this case , there is a good 12-cycle in @xmath171 using the 2-chord from @xmath290 to @xmath138 , the path from @xmath138 to @xmath294 that passes through @xmath137 , and the path from @xmath294 to @xmath290 . 2 . @xmath290 and @xmath137 in @xmath288 : this case gives a good 12-cycle in @xmath171 . @xmath290 and @xmath153 in @xmath288 : this case also gives a good 12-cycle ( by using the path @xmath295 through @xmath258 ) since any 2-adjacency of white vertices on that 12-cycle would yield a short cycle through @xmath282 or through @xmath296 . 4 . @xmath290 and @xmath138 or @xmath150 and @xmath138 in @xmath289 : both cases give a good 12-cycle . thus we may assume that there are no additional 2-adjacencies between vertices in @xmath60 . if @xmath285 is as shown in figure [ fig : external edges ] , i.e. , the subgraph shown is induced in @xmath0 , the subgraph @xmath167 is isomorphic to one of the graphs @xmath297 , @xmath298 , and @xmath299 , shown in figure [ fig : appendix2 ] in the appendix ( with possibly some degree-1 vertices missing if @xmath285 contains vertices whose degree in @xmath0 is 2 ) . by corollary [ cor : appendix ] , @xmath60 increases imbalance . thus , we may assume that @xmath285 is not induced . the only possibilities for additional edges ( up to symmetries ) that do not yield a cycle of length at most 8 through @xmath282 are the following : 1 . the edge @xmath300 in @xmath287 : in this case , @xmath167 is isomorphic to the graph @xmath301 from the appendix ( figure [ fig : appendix2 ] ) , and we are done by corollary [ cor : appendix ] . 2 . the edge @xmath302 in @xmath288 or the edge @xmath303 in @xmath289 : each of these cases gives rise to a good 12-cycle ( using the path through the vertex @xmath137 ) . 3 . the edge @xmath300 in @xmath289 : this case also gives a good 12-cycle . 4 . the edge @xmath304 in @xmath289 : in this case , @xmath167 is isomorphic to the graph @xmath301 shown in figure [ fig : appendix2 ] ( where @xmath258 can be any of the bottom two hexagons ) , and we are done by corollary [ cor : appendix ] . this completes the proof of claim 6 . _ claim 7 . let @xmath305 . if every vertex at distance at most 2 from @xmath103 has degree 3 , then @xmath0 contains an 8-cycle that has a vertex at distance at most 2 from @xmath103 . _ let @xmath273 be a 6-cycle that contains @xmath103 . by claim 3a , all vertices in @xmath258 and all vertices adjacent to @xmath258 have degree 3 . let @xmath275 be the edge incident with @xmath175 that is not contained in @xmath258 . by claim 6 , at least five of the edges @xmath306 are internal . each 6-cycle @xmath307 containing @xmath277 must use another internal edge @xmath283 . if @xmath308 or @xmath309 ( values modulo 6 ) , then @xmath310 contains an 8-cycle , where @xmath103 is either on the cycle or adjacent to it . suppose now that @xmath311 . we may assume that @xmath312 and @xmath313 and that @xmath314 . since any two vertices in @xmath310 lie on a common 6-cycle , at most one of these vertices is incident with an external edge . we may assume that this vertex , if it exits , is @xmath315 . note that all vertices in @xmath310 have degree 3 . define the edges @xmath316 that are incident with @xmath317 and @xmath315 , respectively . if a 6-cycle @xmath318 through @xmath294 uses any of the edges @xmath319 , then the union @xmath320 contains an 8-cycle that is at distance at most 2 from @xmath103 . by symmetry , a similar conclusion holds for all other internal edges leaving @xmath310 . thus , we may assume that the 6-cycle through @xmath294 returns through @xmath296 , the cycle through @xmath321 returns through @xmath322 , and the 6-cycle leaving @xmath317 returns through @xmath315 . denote these 6-cycles by @xmath323 , respectively . note that any two of these 6-cycles together with a 6-cycle in @xmath310 form a subgraph of @xmath0 that is isomorphic to the graph @xmath324 shown in figure [ fig : hexagons through internal edges ] . it is easy to see that if @xmath324 is not 2-induced , then it contains an 8-cycle that passes through one of the vertices marked @xmath137 and @xmath138 in figure [ fig : hexagons through internal edges ] . since @xmath137 and @xmath138 are at distance at most 2 from @xmath103 , we get the claim . thus , we may assume that each of these subgraphs is 2-induced , which implies that also the graph @xmath325 ( see figure [ fig : hexagons through internal edges ] ) consisting of @xmath310 together with @xmath323 is 2-induced in @xmath0 . let @xmath326 and note that @xmath327 . since @xmath325 is 2-induced in @xmath0 , the subgraph @xmath167 is isomorphic to the graph @xmath328 shown in figure [ fig : appendix3 ] , and we are done by applying corollary [ cor : appendix ] . the last case to consider is when every 6-cycle using one of the edges @xmath277 returns to @xmath258 either through @xmath329 or through @xmath330 . suppose first that a 6-cycle @xmath331 through @xmath282 and a 6-cycle @xmath332 through @xmath296 both return through @xmath294 . if the union @xmath333 is an @xmath59-thick ( respectively @xmath37-thick ) subgraph of @xmath0 , then either @xmath334 ( resp . @xmath335 ) gives an imbalance - increasing vertex set in @xmath171 ( see lemma [ lem : thick unbalanced ] ) . therefore , either two vertices in @xmath59 or two vertices in @xmath37 are 2-adjacent . the corresponding 2-chord gives rise to two 6-cycles sharing a path of length 2 or 3 , which is the case we have already treated above . even though one of the 6-cycles @xmath331 or @xmath332 may play the role of @xmath258 in this case , it is still true that a resulting 8-cycle is at distance at most 2 from @xmath103 . by symmetry , we may now assume that there are precisely three 6-cycles using internal edges @xmath336 . we may assume that the 6-cycles are @xmath337 , where @xmath338 uses the edges @xmath277 and if a vertex @xmath339 and a vertex @xmath340 are either the same , adjacent , or 2-adjacent in @xmath0 , then we obtain two 6-cycles sharing a path of length 2 or 3 , which is the case we have already treated above . in that case we get an imbalance - increasing set in @xmath107 or an 8-cycle at distance at most 2 from @xmath103 . since the same can be said for the other pairs of the 6-cycles @xmath338 , we conclude that the graph @xmath341 is 2-induced in @xmath0 . we set @xmath335 and observe that @xmath167 is isomorphic to the graph @xmath342 shown in figure [ fig:6 ] . again , corollary [ cor : appendix ] applies . this completes the proof of claim 7 . let @xmath269 and @xmath343 be the graphs depicted in figure [ fig : h_0 ] , and let @xmath344 , @xmath345 , and @xmath346 be the graphs in figure [ fig : h_2 and h_3 ] . observe that @xmath344 and @xmath345 are both isomorphic to the heawood graph with one edge removed and that their subgraphs induced on vertices @xmath347 are isomorphic to @xmath343 . _ claim 8 . let @xmath258 be an 8-cycle in @xmath348 . then @xmath258 is contained in a 2-induced subgraph isomorphic to the graph @xmath269 . _ we have already seen that @xmath258 must be contained in a subgraph @xmath135 of @xmath0 that is isomorphic to @xmath269 . this subgraph is induced in @xmath0 since any chord in @xmath269 gives rise to a cycle of length 4 . it remains to see that @xmath135 is 2-induced . we will use the notation provided in figures [ fig : h_0 ] and [ fig : h_2 and h_3 ] , so the vertices of @xmath135 and other graphs isomorphic to one of @xmath349 will be denoted by the integers @xmath350 as shown in the figures . , @xmath345 , and @xmath346,width=453 ] suppose that @xmath135 is not 2-induced . note that any 2-chord starting at vertices 11 or 12 would give rise to a 4-cycle in @xmath0 . thus , we may assume , by symmetry , that we have a 2-chord 3 - 10 - 7 . since this 2-chord is part of another copy of @xmath269 , the subgraph obtained from @xmath135 by adding the 2-chord is induced in @xmath0 . by lemma [ lem : thick unbalanced ] , this subgraph can not be thick , so there is a 2-chord joining two of the vertices 4,8,10,12 . the only two pairs not giving a 4-cycle are 4 - 8 and 10 - 12 . they give rise to isomorphic graphs , so we may assume the 2-chord is 4 - 9 - 8 . thus we have a subgraph @xmath351 isomorphic to the graph @xmath343 ( shown in figure [ fig : h_0 ] ) . this subgraph is induced in @xmath0 since any two vertices in different bipartite classes belong to a common copy of @xmath269 inside @xmath351 . suppose first that @xmath351 is not 2-induced . by symmetry between the possible 2-adjacent pairs 9 - 11 and 10 - 12 , we may assume that 10 - 13 - 12 is a 2-chord . by lemma [ lem : thick unbalanced ] , two of the vertices 9,11,13 are 2-adjacent . if these are 9 and 11 , we obtain a copy of @xmath344 . otherwise , by symmetry , we have a 2-chord 9 - 14 - 13 which yields a copy of @xmath345 . let @xmath137 and @xmath138 be the degree-2 vertices in the obtained subgraph . if they were adjacent , we would get the heawood graph , so both graphs are isomorphic to the heawood graph minus an edge , and thus we may assume henceforth that we have @xmath344 . by symmetry between @xmath137 and @xmath138 and by symmetry of the bipartite classes @xmath59 and @xmath37 , we may assume that @xmath352 and that @xmath353 . by claim 3 , @xmath137 and @xmath138 have degree 3 in @xmath0 and there are two edges @xmath354 and @xmath355 going out of @xmath344 in @xmath0 . since the distance in @xmath344 between @xmath137 and @xmath138 is five , both @xmath356 and @xmath357 are external edges . observe that either @xmath358 or @xmath359 belongs to @xmath360 . this is clear if @xmath361 . otherwise , one of these vertices is on a shortest path from @xmath101 to @xmath258 , giving the same outcome . we may assume that @xmath362 . let @xmath363 be a 6-cycle through the vertex @xmath358 , and let @xmath364 . then the subgraph @xmath167 of @xmath0 is isomorphic to the graph @xmath365 shown in figure [ fig : appendix1 ] . ( to check this , note that the 8-cycle in the figure is 2 - 3 - 10 - 13 - 12 - 5 - 6 - 11 - 2 , where the vertex 11 is adjacent to the vertex @xmath366 having degree 1 . the only non - obvious possibility for @xmath167 being different is that a vertex in @xmath307 would be equal to @xmath359 and hence adjacent to @xmath138 . however , in that case , @xmath307 would be incident with two external edges @xmath356 and @xmath357 , contradicting claim 6 . ) now , corollary [ cor : appendix ] shows that @xmath367 increases imbalance . from now on , we may assume that @xmath351 is 2-induced . for vertices @xmath368 , let @xmath277 be the edge in @xmath0 incident with vertex @xmath369 but not contained in @xmath351 . note that for any @xmath370 of different parities , vertices @xmath369 and @xmath371 lie on a common 6-cycle in @xmath351 . since @xmath372 , we have @xmath373 . by claim 6 , at most one of the edges @xmath374 is external . by symmetry , we may assume that @xmath375 is internal . since @xmath351 is 2-induced and the distance between 9 and 11 in @xmath351 is four , @xmath375 and @xmath376 can not lie on a common 6-cycle . thus , we may assume ( by symmetry between @xmath377 and @xmath378 ) that a 6-cycle @xmath307 through @xmath375 uses the edge @xmath377 . in that case , @xmath307 is the cycle 9 - 8 - 7 - 10 - 15 - 16 - 9 , where 15 and 16 are new vertices . if a 6-cycle through @xmath376 uses @xmath377 , then the cycle is 11 - 14 - 15 - 10 - 7 - 6 - 11 ( 14 being a new vertex ) . in this case we obtain a good 8-cycle 11 - 14 - 15 - 16 - 9 - 8 - 1 - 2 - 11 . similarly , if a 6-cycle through @xmath378 uses @xmath375 . since @xmath376 and @xmath378 are incident with the same 6-cycle , they can not be both external by claim 6 . the only possibility for a 6-cycle ( excluding previously treated cases ) is the 6-cycle 12 - 13 - 14 - 11 - 6 - 5 - 12 ( or 12 - 13 - 14 - 11 - 2 - 1 - 12 ) where 13 and 14 are new vertices . this gives us the graph @xmath346 shown in figure [ fig : h_2 and h_3 ] . the subgraph @xmath346 is induced in @xmath0 ( or we get a case treated above ) . if it is not 2-induced , we may assume that we have the 2-chord 13 - 17 - 15 , and in this case we obtain a good 8-cycle : 12 - 13 - 17 - 15 - 16 - 9 - 8 - 1 - 12 . therefore @xmath346 is 2-induced . now , we take @xmath379 . the subgraph @xmath167 is isomorphic to the graph @xmath380 in figure [ fig : appendix1 ] . corollary [ cor : appendix ] shows that @xmath60 increases imbalance . this exhausts all possibilities and completes the proof of claim 8 . we now define a vertex @xmath381 as follows . if every vertex in @xmath382 is of degree 3 in @xmath0 , then we take @xmath383 . otherwise , claim 3a shows that there exists a vertex @xmath384 such that all vertices at distance at most 2 from it have degree 3 . claim 7 shows that there exists an 8-cycle @xmath258 at distance at most 2 from @xmath381 . by claim 8 , @xmath258 is contained in a 2-induced subgraph @xmath269 of @xmath0 , where @xmath269 is depicted in figure [ fig : h_0 ] . this gives the next conclusion . _ claim 9 . there is a 2-induced subgraph of @xmath0 isomorphic to @xmath269 that contains a vertex in @xmath385 . _ we shall now fix the subgraph of claim 9 , call it @xmath269 and denote its vertices by integers as in the figure . _ claim 10 . all vertices in @xmath269 have degree 3 in @xmath0 , and for @xmath386 , there is no @xmath291-chord joining a vertex in @xmath387 with a vertex in @xmath388 . _ by claim 3 , vertices 11 and 12 can not be of degree 2 . if another vertex , which we may assume is the vertex 3 , is of degree 2 , then let @xmath389 . since @xmath269 is 2-induced , the corresponding subgraph @xmath167 is isomorphic to the graph @xmath390 from figure [ fig : appendix1 ] , and we are done by corollary [ cor : appendix ] . this proves the first part of the claim . suppose next that there is a @xmath291-chord joining vertex 3 with @xmath388 . by claim 8 , we have @xmath391 . now it is an easy task to verify that the @xmath291-chord together with a path in @xmath269 gives rise to a good 8-cycle in @xmath0 . this completes the proof of claim 10 . if @xmath369 is a vertex of degree 2 in @xmath269 , then we will denote by @xmath277 the edge in @xmath392 that is incident with the vertex @xmath369 . if there is a 6-cycle that contains two of such edges , @xmath277 and @xmath283 , then we say that @xmath277 and @xmath283 are _ coupled_. if the corresponding vertices @xmath369 and @xmath371 are at distance @xmath291 in @xmath269 , then there is a @xmath393-chord @xmath394 of @xmath269 containing @xmath277 and @xmath283 . since @xmath269 is 2-induced , we know that @xmath395 in such a case . _ claim 11 . if @xmath296 is coupled with @xmath378 , then the edge @xmath396 is internal . _ suppose that the path @xmath397 is 3 - 13 - 14 - 12 ( where 13 and 14 are new vertices ) and suppose that @xmath398 is an external edge . since @xmath269 has a vertex in @xmath385 , the vertex @xmath116 lies in @xmath399 . by claim 5 , there is a 6-cycle @xmath400 through @xmath116 . since the edge @xmath401 is external , it can not be contained in a cycle of length less than 10 ( see the comment stated before claim 6 ) . this implies that the cycle @xmath227 is disjoint from @xmath402 and its only vertex that is incident to @xmath402 is @xmath116 . moreover , none of its vertices is 2-adjacent to a vertex in @xmath402 , except possibly the vertex @xmath403 that could be 2-adjacent to the vertex 8 . let @xmath404 . then @xmath167 is isomorphic to the graph @xmath405 shown in figure [ fig : appendix1 ] . now the proof is complete by corollary [ cor : appendix ] . _ claim 12 . if @xmath296 is coupled with @xmath396 and @xmath406 , then the subgraph @xmath407 induced on @xmath408 is either equal to @xmath409 or is equal to @xmath409 together with precisely one of the edges @xmath410 or @xmath411 . _ by claim 10 and since @xmath0 has no 4-cycles , the only possible edges of @xmath407 in addition to the edges in @xmath409 are the two edges @xmath410 and @xmath411 . thus it remains to see that both of them can not be present . suppose , for a contradiction , that they are . by claim 6 and symmetry , we may assume that the edge @xmath412 leaving @xmath413 at the vertex @xmath116 is internal . it is easy to see that its supporting 6-cycle must return to @xmath409 through the vertex @xmath403 . let @xmath414 be the corresponding 3-chord . let @xmath415 . by using claim 10 it is easy to see that the corresponding subgraph @xmath167 is isomorphic to the graph @xmath416 in figure [ fig : appendix3 ] , and we are done by corollary [ cor : appendix ] . , @xmath417 , and @xmath418,width=245 ] _ claim 13 . if @xmath296 is coupled with @xmath396 , then @xmath419 is not coupled with @xmath420 . _ suppose for a contradiction that @xmath296 is coupled with @xmath396 and that @xmath419 is coupled with @xmath420 . let us consider the graph @xmath421 . by claims 10 and 12 , this subgraph is isomorphic to the graph shown in figure [ fig : h_6 ] , where each of the two dotted edges may or may not be present . if just one of these two edges is present , then we may assume that this is the edge incident with the vertex @xmath403 . by claim 10 , the vertices @xmath422 can not be 2-adjacent outside this subgraph , except possibly for @xmath423 and @xmath403 . let @xmath424 . then @xmath167 is either isomorphic to one of the graphs @xmath425 , @xmath426 , @xmath427 in figure [ fig : appendix3 ] , or to a graph obtained from one of these graphs by identifying the neighbors of vertices @xmath423 and @xmath403 . the latter possibility does not happen if the left dotted edge in figure [ fig : h_6 ] is present . by the assumption made above , we may thus assume that @xmath423 and @xmath403 are not 2-adjacent when at least one of the dotted edges is present ; thus the only additional case arising this way is the graph @xmath428 in figure [ fig : appendix3 ] . in either case , we are done by corollary [ cor : appendix ] . _ claim 14 . if @xmath296 is coupled with @xmath376 , then @xmath296 is coupled with @xmath396 . _ let @xmath429 be the 4-chord @xmath430 . since @xmath431 is not thick ( lemma [ lem : thick unbalanced ] ) , there is a 2-chord joining two vertices in the larger bipartite class of this graph . by using claim 10 it is easy to see that the only possibility is a 2-chord joining the vertex @xmath432 with the vertex 4 . this gives the claim . we are now ready to complete the proof . we may assume that @xmath296 is an internal edge , and we may assume that @xmath296 is not coupled with @xmath396 ( by using claim 13 and symmetry ) . by claim 14 , @xmath296 is not coupled with @xmath376 and by claim 10 , it is not coupled with @xmath419 or @xmath420 . the only possibility remaining is that it is coupled with @xmath378 . claim 11 implies that @xmath396 is an internal edge . the same arguments as used for @xmath296 show that @xmath396 is coupled with @xmath376 . now , claim 10 implies that @xmath419 and @xmath420 can not be coupled with @xmath376 or @xmath378 . therefore , @xmath419 is coupled with @xmath420 . take @xmath60 to be the bipartite vertex class of @xmath433 containing the vertices 1 , 3 , 5 , 7 . then @xmath167 is isomorphic to the graph @xmath434 in figure [ fig : appendix3 ] . corollary [ cor : appendix ] applies again , and the proof is complete . in the appendix we list a collection of graphs and their critical eigenvalues that were used to obtain balance - increasing vertex sets in the proof of our main theorem . claims ( a)(l ) were checked by computer . the only graph that needs the proof is @xmath203 . let @xmath459 be the horizontal path of length @xmath204 in @xmath203 , and let @xmath460 . the subgraph @xmath461 is a matching consisting of @xmath462 disjoint edges . the interlacing theorem shows that @xmath463 . this completes the proof . part of the proof of lemma [ lem : some graphs ] is based on computer computation . for a reader that may be skeptical about a proof relying on computer evidence , we show in the sequel how to obtain a self - contained proof . we will provide a sketch for a direct proof for all of the cases ( in addition to the proof for @xmath203 given above ) that will suffice to support corollary [ cor : appendix ] which is used throughout in section [ sect : improving imbalance ] . let @xmath0 be a graph in one of figs . [ fig : appendix1][fig : appendix3 ] . let @xmath60 be the set of vertices of @xmath0 that are drawn as full circles or full squares . our goal is to show that @xmath464 . observe that in each case , @xmath465 consists of isolated vertices , thus the interlacing theorem implies that @xmath466 . thus it suffices to provide evidence that there is an eigenvalue @xmath26 , where @xmath467 . for graphs @xmath468 , and @xmath271 we can confirm this by describing an eigenvector for eigenvalue @xmath469 . for @xmath470 , the eigenvector has value @xmath471 on vertices of degree 2 and value @xmath472 on other vertices , where each vertex of degree 1 and its neighbor have the same value , and the values @xmath473 and @xmath474 alternate around the cycle . for @xmath214 , the vector has values 1 on vertices of degree 1 and 3 , value @xmath474 on vertices of degree 2 that are adjacent to the degree-3 vertices , and value @xmath475 on the vertex in the middle . for @xmath271 , the eigenvector has value 3 at the top vertex , values @xmath476 on adjacent vertices , and value @xmath474 at all other vertices . in the case of @xmath436 we can confirm that @xmath477 by using the interlacing theorem when we remove the vertex of degree 3 and the vertex that is opposite to it on the 6-cycle . the vertices removed in this case and in the cases treated below are shown as squares . in the case of the graph @xmath244 , we remove the two vertices on the left and two vertices on the right of the drawing . the resulting subgraph has one component isomorphic to @xmath478 , which has @xmath479 . again , the interlacing theorem applies . the same happens for graphs @xmath390 ( where we remove the two square vertices on the left of the drawing ) and @xmath365 ( where we remove five vertices ) . for the graph @xmath380 we provide an eigenvector for eigenvalue @xmath476 in figure [ fig : appendix4](a ) . for @xmath480 we first remove the three square vertices and then find an eigenvector for eigenvalue 1 for the remaining subgraph ( cf . figure [ fig : appendix4](b ) ) . for the graphs @xmath297 and @xmath301 , we remove six square vertices , and are left with a copy of the graph @xmath214 as the non - trivial component . since @xmath481 , the interlacing theorem applies . for the graph @xmath298 , we can remove its six square vertices , being left with a copy of the graph @xmath436 as the only non - trivial component . again , the interlacing theorem applies . for the graph @xmath299 , we remove five square vertices . the resulting non - trivial component @xmath62 has @xmath482 and @xmath483 . the evidence for this is shown by three eigenvectors in figure [ fig : appendix4](d ) , where the unfilled values for the eigenvectors of 0 are assumed to be 0 . ( note that @xmath484 as well , but this evidence is not needed for our proof . ) finally , the graph @xmath342 has eigenvalue 1 ; its eigenvector is shown in figure [ fig : appendix4](c ) . it remains to treat the graphs in figure [ fig : appendix3 ] . for the graph @xmath445 , we remove the four square vertices . the remaining nontrivial component @xmath62 consists of two hexagons sharing an edge plus two additional edges . figure [ fig : appendix4](e ) shows two independent eigenvectors for eigenvalue 1 of @xmath62 which implies that @xmath485 , and the interlacing theorem can be applied . for the graph @xmath453 , we remove its six square vertices . the remaining nontrivial component @xmath62 consists of a path @xmath486 with added pendant edges at vertices @xmath487 . this graph has characteristic polynomial @xmath488 note that @xmath489 and that @xmath490 . basic calculus shows that @xmath491 and @xmath492 for @xmath493 . for the graph @xmath458 , the removal of two square vertices from two of the three central hexagons and removal of three square vertices from the third hexagon give a non - trivial component that is isomorphic to @xmath214 , and we are done as in some of the previous cases . the proof is also easy for @xmath426 , and @xmath427 . the removal of indicated seven square vertices leaves only one nontrivial component , which is isomorphic to @xmath478 , whose second eigenvalue is 1 . from @xmath428 we remove five square vertices , being left with a non - trivial component @xmath62 consisting of two adjacent hexagons plus two edges . figure [ fig : appendix4](f ) contains evidence that @xmath62 has eigenvalue 1 of multiplicity at least 2 , which implies that @xmath485 , and interlacing arguments apply . finally , in the case of @xmath425 , removal of four square vertices leaves one non - trivial component , @xmath62 , which is isomorphic to the graph obtained from the path of length 10 to which we add two pendant edges at each end . the characteristic polynomial @xmath494 of @xmath62 is easily computed : @xmath495 then @xmath496 and @xmath497 . thus , one of @xmath498 is equal to 1 . however , @xmath484 , so we have that @xmath499 , and interlacing can be used again . this exhausts all graphs in figures [ fig : appendix1][fig : appendix3 ] . the author is grateful to tom boothby and to krystal guo for an independent verification of eigenvalue computations for graphs that appear in the appendix . computations were done by using both maple 16 ( http://www.maplesoft.com/products/maple/ ) and sage ( system for algebra and geometry experimentation , http://www.sagemath.org/ ) . |
we study the propagation and localization of classical waves in one - dimensional disordered structures composed of alternating layers of left- and right - handed materials ( mixed stacks ) and compare them with structures composed of different layers of the same material ( homogeneous stacks ) . for weakly scattering layers , we have developed an effective analytical approach and have calculated the transmission length within a wide range of the input parameters . this enables us to describe , in a unified way , the localized and ballistic regimes as well as the crossover between them . when both refractive index and layer thickness of a mixed stack are random , the transmission length in the long - wave range of the localized regime exhibits a quadratic power wavelength dependence with different coefficients of proportionality for mixed and homogeneous stacks . moreover , the transmission length of a mixed stack differs from the reciprocal of the lyapunov exponent of the corresponding infinite stack . in both the ballistic regime of a mixed stack and in the near long - wave region of a homogeneous stack , the transmission length of a realization is a strongly fluctuating quantity . in the far long - wave part of the ballistic region , the homogeneous stack becomes effectively uniform and the transmission length fluctuations are weaker . the crossover region from the localization to the ballistic regime is relatively narrow for both mixed and homogeneous stacks . in mixed stacks with only refractive - index disorder , anderson localization at long wavelengths is substantially suppressed , with the localization length growing with wavelength much faster than for homogeneous stacks . the crossover region becomes essentially wider and transmission resonances appear only in much longer stacks . the effects of absorption on one - dimensional transport and localization have also been studied , both analytically and numerically . specifically , it is shown that the crossover region is particularly sensitive to losses , so that even small absorption noticeably suppresses frequency dependent oscillations in the transmission length . all theoretical predictions are in an excellent agreement with the results of numerical simulations . | 0912.0362 | metamaterials are artificial structures having negative refractive indices for some wavelengths @xcite . while natural materials having such properties are not known , it was the initial paper @xcite that sought to realize artificial metamaterials which triggered the rapidly increasing interest in this topic . over the past decade , the physical properties of these structures , and their possible applications in modern optics and microelectronics , have received considerable attention ( see e.g. refs @xcite ) . the reasons for such interest are their unique physical properties , their ability to overcome the diffraction limit @xcite , and their potential role in cloaking @xcite , the suppression of spontaneous emission rate @xcite , and the enhancement of quantum interference @xcite , etc . until recently , most studies considered only ideal systems and did not address the possible effects of disorder . however , real metamaterials are always disordered , at least , in part , due to fabrication errors . accordingly , the study of disordered metamaterials is not just an academic question but is also relevant to their application . the first step in this direction was made in ref . @xcite where it was shown that the presence of a single defect led to the appearance of a localized mode . a metamaterial with many point - like defects was considered in ref . @xcite where it was demonstrated that even weak microscopic disorder might lead to a substantial suppression of wave propagation through a metamaterial over a wide frequency range . the next steps in this direction focused on the study of localization in metamaterials . anderson localization is one of the most fundamental and fascinating phenomenon of the physics of disorder . predicted in the seminal paper @xcite for spin excitations , it was extended to the case of electrons and other one - particle excitations in solids , as well as to electromagnetic waves ( see , e.g. , refs . @xcite ) , becoming a paradigm of modern physics . anderson localization results from the interference of multiply scattered waves , manifesting itself in a most pronounced way in one - dimensional systems @xcite , in which all states become localized @xcite so that the envelope of each state decays exponentially away from a randomly located localization center @xcite . the rate of this decay is non - random and is called the lyapunov exponent , @xmath0 , the reciprocal of which determines the size of the area of localization . in a finite , but sufficiently long , disordered sample , the localization manifests itself in the fact that the frequency dependent transmission amplitude is ( typically ) an exponentially decreasing function of the sample size . the average of this decrement is a size - dependent quantity , whose inverse ( _ i.e. _ , reciprocal ) is termed the _ transmission length , _ @xmath1 . in the limit as the sample becomes of infinite length , the decrement tends to a constant non - random value . the reciprocal of this value determines another characteristic spatial scale of the localized regime , which is the _ localization length , _ it is commonly accepted in both the solid state physics and optical communities , that the inverse of the lyapunov exponent , @xmath3 and the localization length , @xmath2 are always equal . while this is true for media with a continuous spatial distribution of the random dielectric constant , in the case of randomly layered samples , the situation , as we show in this paper , is more complicated . in particular , the inverse of the lyapunov exponent by itself , calculated , for example , in ref . @xcite , does not provide comprehensive information about the transport properties of disordered media . furthermore , it is unlikely that it can be measured directly , at least in the optical regime . the first study of localization in metamaterials was presented in ref . @xcite where wave transmission through an alternating sequence of air layers and metamaterial layers of random thicknesses was studied . localized modes within the gap were observed and delocalized modes were revealed despite the one - dimensional nature of the model . a more general model of alternating sequences of right- and left - handed layers with random parameters was studied in ref . there , it has been shown that in mixed stacks ( m - stacks ) with fluctuating refractive indices , localization of low - frequency radiation was dramatically suppressed so that the localization length exceeded that for homogeneous stacks ( h - stacks ) , composed solely of right- or left - handed slabs , by many orders of magnitude and scaling as @xmath4 or even higher powers of wavelength ( in what follows we refer to this result as the @xmath5 anomaly ) , in contrast to the well - known dependence @xmath6 observed in h - stacks baluni . as noted in ref . @xcite , a possible physical explanation of this is the suppression of phase accumulation in m - stacks , related to the opposite signs of the phase and group velocities in left- and right - handed layers . scaling laws of the transmission through a similar mixed multilayered structure were studied in ref . there , it was shown that the spectrally averaged transmission in a frequency range around the fully transparent resonant mode decayed with the number of layers much more rapidly than in a homogeneous random slab . localization in a disordered multilayered structure comprising alternating random layers of two different left - handed materials was considered in ref . @xcite , where it was shown that within the propagation gap , the localization length was shorter than the decay length in the underlying periodic structure , and the opposite of that observed in the corresponding random structure of right - handed layers . in this paper , we study the wave transmission through disordered m- and h - stacks of a finite size composed of a weakly scattering right- and left - handed layers . in the framework of the weak scattering approximation ( wsa ) , we have developed a unified theoretical description of the transmission and localization lengths over a wide wavelength range , allowing us to explain the pronounced difference in the transmission properties of m- and h - stacks at long wavelengths . when both refractive index and layer thickness of the mixed stack are random , the transmission length in the long wavelength part of the localized regime exhibits a quadratic power law dependence on wavelength with different constants of proportionality for mixed and homogeneous stacks . moreover , in the localized regime , the transmission length of a mixed stack differs from the reciprocal of the lyapunov exponent of the corresponding infinite stack ( to the best of our knowledge , in all one - dimensional disordered systems studied till now these two quantities always coincide ) . both m- and h - stacks demonstrate a rather narrow crossover from the localized to the ballistic regime . the h - stack in the near ballistic region , and the m - stack in the ballistic region are weakly scattering disordered stacks , while in the far ballistic region , the h - stack transmits radiation as an effectively uniform medium . we also consider the effects of loss and show that absorption dominates the effects of disorder at very short and very long wavelengths . the crossover region is particularly sensitive to losses , so that even small absorption suppresses oscillations in the transmission length as a function of frequency . all of the theoretical results mentioned above are confirmed by , and are shown to be in excellent agreement with , the results of extensive numerical simulation . in m - stacks with only refractive index disorder , anderson localization and transmission resonances are effectively suppressed and the crossover region between the localized and ballistic regimes is orders of magnitude greater . a more detailed study of the @xmath5-anomaly shows that the genuine wavelength dependence of the transmission length is not described by any power law and rather is non - analytic in nature . in what follows , sec . [ sec : mod ] presents a detailed description of our model . section [ sec : anal ] is devoted to the analytical studies of the problem , while the results of numerical simulations and a discussion of these are presented in sec . [ sec : num ] . we consider a one - dimensional alternating m - stack , as shown in fig . it comprises disordered mixed left- ( @xmath7 ) and right- ( @xmath8 ) handed layers , which alternate over its length of @xmath9 layers , where @xmath9 is an even number . the thicknesses of each layer are independent random values with the same mean value @xmath10 . in what follows , all quantities with the dimension of length are measured in units of @xmath10 . in these units , for the thicknesses of a layer we can write @xmath11 where the fluctuations of the thickness , @xmath12 @xmath13 are zero - mean independent random numbers . we take the magnetic permeability for right - handed media to be @xmath14 and for metamaterials to be @xmath15 , while the dielectric permittivity is @xmath16 where the upper and lower signs respectively correspond to normal ( right - handed ) and metamaterial ( left - handed ) layers . the refractive index of each layer is then @xmath17 where all @xmath18 and absorption coefficients of the slabs , @xmath19 , are independent random variables . with this , the impedance of each layer relative to the background ( free space ) is @xmath20 with the same choice of the sign . we begin with the general case when both types of disorder ( in refractive index and in thickness ) are present . two particular cases , each with only one type of disorder , are rather different . in the absence of absorption , the m - stack with only thickness disorder is completely transparent , a consequence of @xmath21 . however , the case of only refractive - index disorder is intriguing because , as is shown below , such mixed stacks manifest a dramatic suppression of anderson localization in the long wave region @xcite . although localization in disordered h - stacks with right - handed layers has been studied by many authors @xcite , here we consider this problem in its most general form and show that the transmission properties of disordered h - stacks are qualitatively the same for stacks comprised of either solely left- or right - handed layers . we introduce the transmission length @xmath1 of a finite random configuration as @xmath22 where @xmath23 is the random transmission coefficient of a sample of the length @xmath24 as a consequence of the self - averaging of @xmath25 , @xmath26 this means that for a sufficiently long stack ( in the localized regime ) the transmission coefficient is exponentially small @xmath27 . in what follows , we consider stacks composed of layers with low dielectric contrasts , _ i.e _ , @xmath28 so that the fresnel reflection coefficients of each interface , and of each layer , are much smaller than @xmath29 here , a thin stack comprising a small number of layers is almost transparent . in this , the ballistic regime , the transmission length takes the form @xmath30 involving the average reflection coefficient @xcite , which is valid in the case of lossless structures . this follows directly from eq . ( [ ll-0 ] ) by virtue of the conservation relationship , @xmath31 . thus , in the ballistic regime , @xmath32 where the length @xmath33 in this equation is termed the _ ballistic length_. accordingly , in studies of the transport of the classical waves in one - dimensional random systems , the following spatial scales arise in a natural way : * @xmath1 the transmission length of a finite sample ( [ ll-0 ] ) , * @xmath2 the localization length ( [ ll-3 ] ) , and * @xmath33 ballistic length ( [ ll-31 ] ) . note that in the case of absorbing stacks ( @xmath34 ) , the right hand side of eq . ( [ ll-0 ] ) defines the attenuation length , @xmath35 , which incorporates the effects of both disorder and absorption . in what follows , we show that contrary to commonly accepted belief , the quantities @xmath36 , @xmath2 , and @xmath33 are not necessarily equal , and , under certain situations , can differ noticeably from each other . in this paper , we study mainly the transmission length defined above by eq . this quantity is very sensitive to the size of the system and therefore is best suited to the description of the transmission properties in both the localized and ballistic regimes . more precisely , the transmission length coincides either with the localization length or with the ballistic length , respectively in the cases of comparatively thick ( localized regime ) or comparatively thin ( ballistic regime ) stacks . that is , @xmath37 another argument supporting our choice of the transmission length as the subject of investigation is that it can be found directly by standard transmission experiments , while measurements of the lyapunov exponent call for a much more sophisticated arrangement . the theoretical analysis involves the calculation of the transmission coefficient using a recursive procedure . consider a stack which is a sequence of @xmath9 layers enumerated by index @xmath38 from @xmath39 at the rear of the stack through to @xmath40 at the front . the total transmission ( @xmath41 and reflection ( @xmath42 ) amplitude coefficients of the stack satisfy the recurrence relations @xmath43 for @xmath44 , in which both the input and output media are free space . in eqs ( [ rec1 ] ) and ( [ rec2 ] ) , the amplitude transmission ( @xmath45 ) and the reflection ( @xmath46 ) coefficients of a single layer are given by @xmath47 here , @xmath48 , @xmath49 and @xmath50 denotes the dimensionless free space wavelength . while the sign of the phase shift across each slab @xmath51 varies according to the handedness of the material , the fresnel interface coefficient @xmath52 given by @xmath53 depends only on the relative impedance of the layer @xmath54 , a quantity whose real part is positive , irrespective of the handedness of the material . equations ( [ rec1])-([ro ] ) are general and provide an _ exact _ description of the system and will be used later for direct numerical simulations of its transmission properties . it was mentioned previously that we consider the special case of weak scattering for which the reflection from a single layer is small . _ i.e. _ , @xmath55 . this occurs either for weak disorder , or for strong disorder provided that the wavelength is sufficiently long . the transmission length then follows from @xmath56 and requires the following first order approximations derived from eqs . ( [ rec1 ] ) and ( [ rec2 ] ) : @xmath57 in deriving eq . ( [ rec4 ] ) , we omit the first - order term @xmath58 since it contributes only to the second order of @xmath59 already after the first iteration . then , by summing up logarithmic terms ( [ rec3 ] ) , we obtain @xmath60 this equation enables us to derive a general expression for the transmission length @xmath61 which is valid in all regimes ( see the next section ) . however , the ballistic length @xmath33 , according to eq . ( [ ll-31 ] ) ( and the average reflection coefficient as well ) , is determined only by the total reflection amplitude @xmath62 in the case of lossless strutures . in the ballistic regime , this amplitude coefficient , to the necessary accuracy , is given by @xmath63 from here on , we assume that the random variables @xmath18 of left - handed or right - handed layers , @xmath64 , and @xmath65 are identically distributed according to the corresponding probability density functions . this enables us to express all of the required quantities via the transmission and reflection amplitudes of a single right - handed or left - handed layer , @xmath66 , @xmath67 , and also to calculate easily all of the necessary ensemble averages . the average of the first term in eq . ( [ lnt1 ] ) can be written as @xmath68 next , we split the second term of eq . ( [ lnt1 ] ) into two parts @xmath69 where @xmath70 comprising contributions to the depletion of the transmitted field due to two pass reflections respectively between slabs of different materials ( _ i.e. _ , of opposite handedness ) , and between slabs of the same material ( _ i.e. _ , of like handedness ) . averaging these expressions , we obtain @xmath71 , \\ \langle \mathcal{r}_{2}\rangle & = & \frac{a_{r}^{2}b_{l}+a_{r}^{2}b_{l}}{2n } \left [ \frac{n}{1-b^2}+\frac{2(b^{n}-1)}{(1-b^2)^2}\right ] , \label{rterm4}\end{aligned}\ ] ] where @xmath72 the resulting transmission length is determined by the equation @xmath73 in the lossless case ( @xmath74 ) , the first term on the right hand side of eq . ( [ translength-1 ] ) corresponds to the so - called single - scattering approximation , which implies that multi - pass reflections are neglected so that the total transmission coefficient is approximated by the product of the single layer transmission coefficients , _ i.e. _ , @xmath75 in the case of very long stacks ( _ i.e. _ , as the length @xmath76 ) , we can replace the arithmetic mean , @xmath77 , by its ensemble average @xmath78 on the other hand , in this limit the reciprocal of the transmission length coincides with the localization length . using the energy conservation law , @xmath79 and is proportional to the mean reflection coefficient of a single random layer @xcite . the corresponding modification of eq . ( [ translength-1 ] ) then reads @xmath80 here , the first term corresponds to the single - scattering approximation , while the next two terms take into account the interference of multiply scattered waves as well as the dependence of the transmission length on the stack size . note that eq . ( [ translength-2 ] ) is appropriate only for lossless structures . in the presence of absorption , ( [ translength-1 ] ) should be used instead . from this point on , we assume that the statistical properties of the right - handed and left - handed layers are identical . as a consequence of this symmetry , the following relations hold for any real - valued function @xmath81 in either the lossless or absorbing cases : @xmath82 therefore , @xmath83 and @xmath84 are the complex conjugates of @xmath85 and @xmath86 , and @xmath87 is real quantity , as are both averages @xmath88 and @xmath89 . accordingly , as a consequence of the left - right symmetry ( [ symmetry ] ) , the transmission length of a m - stack depends only on the properties of a single right - handed layer and may be expressed in terms of three averaged characteristics : @xmath90 , @xmath91 , and @xmath92 ( in which we omit the subscript @xmath93 ) . with these observations , the transmission length of a finite length m - stack may be cast in the form : @xmath94 where @xmath95 and @xmath96 are , as we will see below , the inverse localization and inverse ballistic lengths . the function @xmath97 is defined as @xmath98 , \label{f}\ ] ] and introduces a new characteristic length termed the _ crossover length _ @xmath99 which arises in the calculations in a natural way and , as will be demonstrated below , plays an important role in the theory of the transport and localization in one - dimensional random systems . equations ( [ finalg3])-([bal ] ) completely describe the transmission length of a mixed stack in the weak scattering approximation . obviously , the characteristic lengths @xmath100 , @xmath101 , and @xmath102 appearing in eq . ( [ finalg3 ] ) are functions of wavelength . using straightforward calculations , it may be shown that the first two always satisfy the inequality @xmath103 , while , as we will see , the crossover length is the shortest of the three , _ i.e. _ , @xmath104 in the long wavelength region . in the case of a fixed wavelength @xmath105 and a stack so short that @xmath106 , the expansion of the exponent in eq . ( [ f ] ) yields @xmath107 , in which case the transmission length approaches @xmath108 . correspondingly , for a sufficiently long stack , @xmath109 , @xmath110 and the transmission length assumes the value of @xmath100 . in summary , @xmath111with the transition between the two ranges of @xmath9 being determined by the crossover length . while in the lossless case , the ballistic regime occurs when the stack is much shorter than the crossover length ( @xmath112 ) , it is important to note that , in the localization regime , the opposite inequality is not sufficient and the necessary condition for localization is @xmath113 . in what follows , we consider samples of an intermediate length , _ i.e. _ , @xmath114 . for a m - stack of fixed size @xmath115 the parameter governing the transmission is the wavelength , and the conditions for the localized and ballistic regimes should be formulated in the wavelength domain . to do this , we introduce two characteristic wavelengths , @xmath116 and @xmath117 defined by the relations @xmath118 it can be shown that the long wavelength region , @xmath119 , corresponds to localization where the transmission length coincides with the localization length , while in the extremely long wavelength region , @xmath120 , the propagation is ballistic , with the transmission length given by the ballistic length @xmath33 . that is , @xmath121 when @xmath122 , there exists an intermediate range of wavelengths , @xmath123 , which will be discussed below . to better understand the physical meaning of the expressions ( [ loc ] ) and ( [ bal ] ) for the localization and ballistic lengths , we will consider ensembles of random configurations in which the fluctuations @xmath124 and @xmath64 are distributed uniformly over the intervals @xmath125 $ ] and @xmath126 $ ] respectively , with @xmath127 . the average quantities that arise in eqs . ( [ loc ] ) and ( [ bal ] ) are presented in the appendix . these formulae allow for calculations with an accuracy of order @xmath128 for arbitrary @xmath129 . for the sake of simplicity , we assume that the fluctuations of the refractive index and thickness are of the same order , _ i.e. _ , @xmath130 so that the dimensionless parameter @xmath131is of order of unity . we also neglect the contribution of terms of order higher than @xmath132 . the short wavelength asymptotic of the localization length is then @xmath133 in the long wavelength limit , we obtain the following asymptotic forms for the corresponding single layer averages : @xmath134substitution of these expansions into eq . ( [ bal ] ) yields the long wavelength asymptotic of the ballistic length @xmath135 this asymptotic can be calculated directly from eq . ( [ ll-31 ] ) with the average reflection coefficient , determined from eq . ( [ refl ] ) , being @xmath136 the same substitutions into this equation ( [ refl - mix ] ) give the average total reflection coefficient @xmath137 in the ballistic regime , the final term is negligibly small . this , together with eq . ( [ ll-31 ] ) , again results in the value of the ballistic length given in eq . ( [ bbal ] ) . substituting the long wavelength expansions ( [ expansions - r])([expansions - t ] ) into eqs . ( [ loc ] ) and ( [ l ] ) , we derive the following asymptotic forms for the localization length @xmath138 and the crossover length @xmath139 in seeking to compare the result ( [ lloc ] ) for the localization length with the corresponding long wave asymptotic of the reciprocal lyapunov exponent , it is important to note that these two quantities are defined in different ways . the localization length is defined by eq . ( [ ll-3 ] ) via the transmittivity on a realization , while the lyapunov exponent describes an exponential growth of the envelope of a currentless solution far from a given point in which the solution has a given value @xcite . we have calculated the long wave asymptotic of the m - stack lyapunov exponent using the well known transfer matrix approach , assuming the same statistical properties of the modulus of dielectric constant and the thickness of both left- and right - handed layers , @xmath140 in the case of rectangular distributions of the fluctuations of the dielectric constants and thicknesses , this result reduces to @xmath141(we neglected the small corrections proportional to @xmath142 ) . thus , the disordered m - stack in the long wavelength region presents a unique example of a one - dimensional disordered system in which the localization length differs from the reciprocal of the lyapunov exponent . the long wavelength asymptotic of the lyapunov exponent @xmath143 and the asymptotic of the localization length , @xmath144 , eq . ( [ lloc ] ) have rather clear physical meaning . indeed , in the limit @xmath145 , the propagating wave is insensitive to disorder since the disorder is effectively averaged over distances of the order of the wavelength . this means that anderson localization is absent and the lyapunov exponent @xmath146 vanishes . for large but finite wavelengths ( _ i.e. _ , small wavenumbers @xmath147 , the lyapunov exponent is small and , assuming that its dependence on the wavenumber is analytic , we can expand it in powers of @xmath148 . this expansion commences with a term of order @xmath149 since the lyapunov exponent is real and the wavenumber enters the field equations in the form @xmath150 . accordingly , in the long wavelength limit , @xmath151 and so @xmath144 . the behavior of the lyapunov exponent given by ( [ lyapunov-3 ] ) does not depend on the handness of the layers ; compare this equation with the corresponding long wave asymptotic obtained in refs . @xcite for homogeneous stacks composed of only positive or only negative layers . however , it crucially depends on the propagating character of the field within a given wavelength region . when this region is within the gap of the propagation spectrum of the effective ordered medium , the frequency dependence of the lyapunov exponent differs from @xmath152 . it takes place , for example , for a stack composed of single negative layers , where only one of two characteristics ( @xmath153 , @xmath154 ) is negative @xcite . the main contribution to the ballistic length ( [ bbal ] ) is due to the final term in the right hand side of eq . ( [ bal ] ) , which corresponds to the single - scattering approximation discussed at the end of the sec . subsubsec : gen . while the ballistic length follows from the single scattering approximation , the calculation of the localization length ( lloc ) is more complex and requires that interference due to the multiple scattering of waves must be taken into account . in the case under consideration ( @xmath130 and @xmath155 ) , the two characteristic wavelengths @xmath116 and @xmath156 ( [ wavelengths-1 ] ) for an m - stack of a fixed length size @xmath9 take the form @xmath157 and are of the same order of magnitude . this means that the transmission length @xmath1 coincides with the localization length ( [ lloc ] ) for @xmath113 ) , and with the ballistic length ( [ bbal ] ) in the case when @xmath158 . the crossover between the localized and ballistic regimes occurs for @xmath159 , with the two bounds being proportional to @xmath160 and thus growing as @xmath161 . for long stacks , the transmission length in the crossover region can be described with high accuracy by the general equations ( [ finalg3 ] ) and ( [ f ] ) , where the ballistic length @xmath101 , the localization length @xmath162 , and the crossover length @xmath102 are replaced by their asymptotic forms ( [ bbal])([crossover-3 ] ) . in the case of solely refractive index disorder ( @xmath163 when all thicknesses are set to unity , the ballistic length and the short wavelength asymptotic of the localization length coincide with the limiting values of their counterparts for an m - stack with both refractive index and thickness disorder ( [ bbal ] ) and ( [ shortwave ] ) . however , the corresponding limiting values of the long wavelength localization length and crossover lengths have nothing to do with the genuine behaviour of the transmission length ( see sec . [ subsubsec : suppress ] ) . this means that the weak scattering approximation fails to describe the long wavelength asymptotics of the transmission length in both the localization and crossover regions . this is discussed in greater detail below . ( [ rterm3])([translength-1 ] ) ( or eqs . ( [ finalg3])(l ) ) in the symmetric case ) completely determine the behaviour of the transmission length for a mixed stack composed of weakly scattering layers . although @xmath164 has been introduced as a crossover length , the entire region @xmath165 does not necessarily support localization . correspondingly , the ballistic regime may exist outside the region @xmath166 all of the results obtained under the symmetry assumption ( [ symmetry ] ) are qualitatively valid in the general case . indeed , the existence of the crossover ( [ crossover-2 ] ) is related to the exponential dependence in eq . ( [ f ] ) with a real and positive crossover length @xmath164 . when the assumption of symmetry no longer holds , this length takes a complex value . however , the quantity @xmath167 in eq . ( [ rterm4 ] ) satisfies ( by its definition in eq . ( [ ref - trans ] ) ) the evident inequality @xmath168 . therefore , the corresponding analogue of the function @xmath169 ( [ f ] ) preserves all necessary limiting properties ( see the next sec . [ subsec : hom ] ) . in this section , we consider a h - stack composed entirely of normal material layers noting that the behaviour of a h - stack of metamaterial ( left - handed ) layers alone is exactly the same , a result which may be obtained directly from eq . ( [ translength-1 ] ) by replacing each @xmath2 by @xmath93 , after which any reference to the index @xmath93 may be omitted . the transmission length of an h - stack is then @xmath170,\ ] ] where the inverse localization length @xmath2 is @xmath171 now we consider a h - stack composed of weakly scattering layers . to simplify the discussion , we consider only refractive index disorder ( _ i.e. _ , @xmath172 ) . in this case , the asymptotic behavior of the localization length in the short and long wavelength limits are @xmath173 the main contribution to the localization length is related to the first term in eq . ( [ lochom ] ) . thus , the localization length of the h - stack in the long wavelength region is described completely by the single scattering approximation and coincides with the ballistic length ( [ bbal ] ) of the m - stack . using the transfer matrix approach , we can also calculate the long wavelength asymptotic of the lyapunov exponent for a h - stack . it is described by the same equation ( [ lyapunov-3 ] ) as the asymptotic for the m - stack , thus coinciding with the asymptotic of the reciprocal lyapunov exponent . in recent work @xcite , this coincidence was established analytically in a wider spectral region . however , the numerical calculations intended to confirm this result are rather unconvincing . indeed , the numerically obtained plots demonstrate strong fluctuations ( of the same order as the mean value ) of the calculated quantity , while the genuine lyapunov exponent is non - random and should be smooth without any additional ensemble averaging mentioned by the authors . if we cast @xmath174 in the second term of eq . ( [ transhom ] ) in the form @xmath175 we see that the crossover length of the h - stack is @xmath176 with its long wavelength asymptotic , according to eq . ( [ expansions - t ] ) , being @xmath177 here , the crossover length is proportional to the wavelength , in stark contradistinction to the situation for m - stacks , in which the crossover length is proportional to @xmath178 . to consider this further , we define the characteristic wavelengths @xmath179 and @xmath180 by the expressions ( [ wavelengths-1 ] ) . for a h - stack , these lengths are @xmath181 evidently , the second characteristic wavelength is always much larger than the first @xmath182 . as a consequence , the long wavelength region , where the ballistic regime is realized , can be divided into two subregions . the near subregion ( moderately long wavelengths ) is bounded by the characteristic wavelengths @xmath183 the main contribution to the ballistic length in the near ballistic region , @xmath184 is due to the first term in eq . ( [ lochom ] ) . thus the ballistic length @xmath185 has the same wavelength dependence as the localization length @xmath100 ( [ asympt - h ] ) @xmath186 and is well described by the single scattering approximation ( class - picture-2 ) . for a h - stack , the transition from the localized to the ballistic regime at @xmath187 is not accompanied by any change of the wavelength dependence of the transmission length . this change can occur at much longer wavelengths @xmath188 in the far long wavelength subregion . to derive the ballistic length @xmath189 in the far long wavelength region , we may proceed in a similar manner to that outlined in the case of a m - stack and expand the exponent @xmath190 in eq . ( [ transhom ] ) . however , because @xmath191 in this expression is complex , the situation is more complicated than was the case for the m - stack . in particular , the first two terms of the expansion do not contribute to the ballistic length . taking account of the second order leads to the following expression for the ballistic length , @xmath192 in the far long wavelength region @xmath193 in the case of a relatively short h - stack @xmath194 the contribution of the first term in the right hand side of this equation dominates , and hence the transition from the near to the far subregions is not accompanied by any change in the analytical dependence on the wavelength . the ballistic length is thus described by the same wavelength dependence over the entire ballistic region @xmath195 for sufficiently long h - stacks @xmath196 in the far long wavelength region , the second term is dominant and so @xmath197 thus , the wavelength dependence of the ballistic length of a sufficiently long h - stack is @xmath198 the same result for the ballistic length also follows from eq . ( [ ll-31 ] ) with eq . ( [ refl ] ) , in the case of a homogeneous stack , yielding @xmath199 in the long wave limit this leads to @xmath200 the final term in eq . ( [ refl - hom ] ) differs from the final term in eq . ( [ refl - mix ] ) and , therefore , in contrast to the m - stack , its contribution to the total reflection coefficient can be of the order of , or larger than , that of the first term . this together with eq . ( [ ll-31 ] ) is equivalent to the result in eqs . ( [ bal - h-3 ] ) and ( [ bal - h-5 ] ) , and is applicable to short and long stacks respectively . the far long wavelength ballistic asymptotic ( [ bal - h-4 ] ) has a simple physical interpretation . indeed , in this subregion , the wavelength essentially exceeds the stack size and so we may consider the stack as a single weakly scattering uniform layer with an effective dielectric permittivity @xmath201 in this case , the ballistic length of the stack according to eq.([ll-31 ] ) is @xmath202 where @xmath203 is the long wavelength form of the reflection amplitude for a uniform right - handed stack of the length @xmath9 and constant dielectric permittivity @xmath204 ( with @xmath205 ) . in the case of a uniform left - handed stack , the value of the reflection coefficient should be @xmath206 in which the value of @xmath204 is now negative ( with @xmath207 ) . to calculate the effective parameters , we use eq.([refl ] ) . neglecting small fluctuations of the layer thickness , the single layer reflection amplitude in the long wavelength limit reads @xmath208 the total reflection amplitude for a stack of length @xmath9 is @xmath209 corresponding to the effective dielectric permittivity determined by the expression @xmath210for sufficiently long stacks , the right hand side of this equation can be replaced by the ensemble average @xmath211 and hence @xmath212 this expression is similar to that of the two - dimensional case for disordered photonic crystals reported in ref . the same considerations as above , but for a homogeneous stack composed entirely with disordered metamaterial slabs , also lead to an effective value of the permittivity , @xmath213 eq . ( [ effective-4 ] ) , together with eqs ( [ effective-1 ] ) , ( [ effective-2 ] ) , leads immediately to the far long wavelength ballistic length ( [ bal - h-4 ] ) . we emphasize that because of the effective uniformity of the h - stack in the far ballistic region , the transmission length on a single realization is a less fluctuating quantity . in contrast , transmission length for h - stacks fluctuates strongly in the near ballistic region , as indeed it does over the entire ballistic region for m - stacks . to characterize the entire crossover region between the two ballistic regimes ( [ bal - h-5 ] ) in greater detail , we return to the general formula ( [ transhom ] ) , in which we represent @xmath214 as @xmath215 . then , in accordance with eq . ( [ expansions - t ] ) , we can write @xmath216 as a consequence , instead of the result in eq . ( [ bal - h-2 ] ) , we obtain @xmath217 we note that , strictly speaking , the expansion ( [ brrrr ] ) is valid inside the interval @xmath218 the second term in eq . ( [ bal - h-6 ] ) represents standard oscillations of the reflection coefficient of a uniform slab of finite size . in the far long wavelength limit @xmath219 this equation coincides with ( [ bal - h-2 ] ) . thus , eq . ( [ bal - h-6 ] ) describes the ballistic length over practically the whole ballistic region @xmath220 moreover , taking into account that the long wavelength asymptotic of the localization length coincides with that of the near long wavelength ballistic length , we see that the right hand side of eq . ( [ bal - h-6 ] ) serves as an excellent interpolation formula for the reciprocal of the transmission length @xmath221 of a sufficiently long stack ( @xmath222 ) over the entire long wavelength region @xmath223 away from the transition regions , the transmission length can exhibit three types of long wavelength asymptotics described by the right hand sides of eqs . ( [ bbal ] ) , ( [ lloc ] ) , and ( [ bal - h-4 ] ) . the first ( [ bbal ] ) corresponds to the single scattering approximation where the inverse transmission length is proportional to the average reflection coefficient of a single random layer . in the absence of absorption , this characterizes both the localization and ballistic lengths . the second asymptotic form ( [ lloc ] ) takes into account the interference of multiply scattered waves and describes the localization length . the third asymptotic ( [ bal - h-4 ] ) corresponds to transmission through a uniform slab with an effective dielectric constant given by eq . ( [ effective-5 ] ) and is relevant only in the ballistic regime . in the case of a m - stack , the first two expressions ( [ bbal ] ) and ( [ lloc ] ) characterize the ballistic and localized regimes respectively , while the third asymptotic is never realized in m - stacks . in relatively short h - stacks ( [ bal - h-3 ] ) , the long wavelength behaviour of the transmission length is described by the same dependence ( [ bbal ] ) in both the localized and ballistic regimes . finally , in the case of long h - stacks ( [ bal - h-4 ] ) , the transmission length follows from eq . ( [ bbal ] ) in the localized and near ballistic regimes , while in the far ballistic region it is described by the right hand side of eq . ( [ bal - h-4 ] ) . these results predict the existence of a different wavelength dependence of the transmission length in different wavelength ranges . for m - stacks , the crossover between them occurs at the wavelength @xmath116 ( wavelengths-3 ) where the size of the stack is comparable with the localization length @xmath224 . for long h - stacks , the crossover occurs when the wavelength becomes comparable to the stack size @xmath225 . short h - stacks exhibit the same wavelength dependence of the transmission length in all long wavelength regions , _ i.e. _ , in both localized and ballistic regimes . the results of the numerical simulations presented below correspond to uniform distributions of the fluctuations @xmath226 @xmath227 , with widths of @xmath228 and @xmath129 , respectively , and with a constant value of the absorption @xmath229 in each layer . results are presented for ( a ) direct simulations based on the exact recurrence relations ( [ rec1 ] ) and ( [ rec2 ] ) ; ( b ) the weak scattering analysis for the transmission length based on eqs . ( [ finalg3 ] ) and ( transhom ) ; and ( c ) asymptotics for short ( [ shortwave ] ) and long wavelengths [ ( bbal ) , ( [ lloc ] ) , and ( [ bal - h-4 ] ) ] . for the mean reflectivity , we used the asymptotic forms ( [ totalrefl ] ) and ( [ bal - h-6 ] ) . in all cases , unless otherwise is mentioned , the ensemble averaging is taken over @xmath230 realizations . the results up to sec . [ subsec : losses ] are for lossless stacks ( @xmath74 ) only . we first consider stacks having refractive index and thickness disorder , with @xmath231 and @xmath232 . shown in fig . [ fig2 ] are transmission spectra for a m - stack of @xmath233 layers and a h - stack of length @xmath234 . there are two major differences between the results for these two types of samples : first , in the localized regime ( @xmath235 ) , the transmission length of the m - stack exceeds or coincides with that of the h - stack ; second , in the long wavelength region , the plot of the transmission length of the m - stack exhibits a pronounced bend , or kink , in the interval @xmath236 $ ] , while there is no such feature in the h - stack results . these two types of behaviour are discussed in more detail below . the weak scattering approximation ( wsa ) of eq . ( [ translength-1 ] ) is an excellent method by which to calculate the transmission length for m - stacks . this is seen in fig . [ fig2 ] where the curves obtained by numerical simulations and by the wsa are indistinguishable ( solid line ) . the characteristic wavelengths ( wavelengths-3 ) of this mixed stack are @xmath237 and @xmath238 . therefore , the transmission length describes the localization properties of a random sample in the region @xmath239 , whereas longer wavelengths , @xmath240 , correspond to the ballistic regime . the crossover from the localized to the ballistic regime demonstrates the kink - type behaviour that occurs within the region @xmath241 . the short and long wavelength behaviour of the transmission length is also in excellent agreement with the calculated asymptotics in both regimes . to analyze the long wavelength region ( @xmath242 ) more carefully , we plot in fig . [ fig3 ] the transmission lengths of m - stacks of three different sizes , @xmath243 in all cases , there is excellent agreement between the simulations and the wsa predictions . the characteristic wavelengths for @xmath234 are @xmath244 and @xmath245 , while for @xmath246 they are @xmath247 and @xmath248 , and we see that the observed crossover regions are bounded exactly by these characteristic wavelengths in all three cases . to confirm the ballistic nature of the transmission in the region @xmath249 , we plot in fig . [ fig4](a ) the logarithm of the mean value of the reflectance for the same three stack sizes as a function of the logarithm of the wavelength . in all cases , the plots exhibit a linear dependence @xmath250 in the ballistic regime , which is bounded from below by the crossover wavelength @xmath251 . the straight lines are calculated from eq . ( [ totalrefl ] ) and confirm that the reflection coefficient in the mixed stack is proportional to the stack length . within the localized region @xmath252 , the reflection coefficient is close to unity in all three cases . the behaviour of the transmission length is illustrated by the phase diagram in the @xmath253 plane shown in fig . [ fig4](b ) . the two slanted lines @xmath254 and @xmath255 separate the plane into three parts corresponding to the localization ( i ) , the crossover region ( ii ) and the ballistic region ( iii ) . the intersections of these lines with the horizontal lines @xmath234 , @xmath233 , and @xmath246 define the characteristic wavelengths @xmath256 and @xmath251 for the three stack sizes considered here . it is easy to see that these wavelengths , determined with the aid of the phase diagram , perfectly bound the crossover regions in both fig . [ fig3 ] and fig . [ fig4](a ) . until now , we have dealt only with the transmission length @xmath257 , which was defined through an average value . however , more detailed information can be obtained from the transmission length @xmath258 for a single realization , defined by the equation @xmath259 in the localized regime , for a sufficiently long ( _ i.e. _ , @xmath260 ) m - stack , the transmission length for a single realization @xmath261 is practically non - random and coincides with @xmath257 , while in the ballistic region it fluctuates . the data displayed in fig . [ fig5new ] ) enables us to estimate the difference between the transmission length @xmath257 ( solid line ) and the transmission length @xmath258 for a single randomly chosen realization ( dashed line ) , and the scale of the corresponding fluctuations . both curves are smooth , coincide in the localized region , and differ noticeably in the ballistic regime . the separate discrete points in fig . [ fig5new ] present the values of the transmission length @xmath258 calculated for different randomly chosen realizations . it is evident that fluctuations in the ballistic region become more pronounced with increasing wavelength . the absence of any kink in the h - stack transmission length spectrum in fig . [ fig2 ] follows from sec . [ subsec : hom ] in which it was shown that the crossover to the far ballistic regime occurs at the wavelength @xmath262 ( [ crossover - h-2 ] ) . for @xmath234 , this is of the order of @xmath263 and so the kink does not appear . in order to study the crossover , we plot in fig . [ fig5 ] the transmission lengths of h - stacks with @xmath234 and @xmath263 over the wavelength range extended up to @xmath264 . as for the m - stack , the simulation results for h - stacks can not be distinguished from those of the wsa ( [ transhom ] ) . the transition from the localized to the near ballistic regime occurs without any change in the analytical dependence of transmission length , in complete agreement with the results of sec . subsec : hom . the crossover from the near to the far ballistic regime is accompanied by a change in the analytical dependence that occurs at @xmath265 , which for these stacks is of the order of @xmath263 and @xmath266 respectively . the crossover is accompanied by prominent oscillations described by eq . ( [ bal - h-6 ] ) . finally , we note that the vertical displacement between the moderately long and extremely long wavelength ballistic asymptotes does not depend on wavelength , but grows with the size of the stack , according to the law @xmath267 which stems from eq . ( [ bal - h-5 ] ) . to study the ballistic transmission regime in the region @xmath268 more closely , we plot in the upper panel of fig . [ fig6](a ) ( on a logarithmic scale ) the mean value of the reflection coefficient for the same two stack sizes . that part of each plot which presents the ballistic propagation comprises two linear asymptotes of the form @xmath269 and the other in the far ballistic region @xmath270 . the difference between the values of these two constants corresponds precisely with the right hand side of eq . ( [ shift ] ) , with the far long wavelength asymptotic given by the second term in eq . ( [ bal - h-6 ] ) . within the localized regime @xmath271 the reflectance is almost unity in all three cases . by analogy with m - stacks , the behaviour of the transmission length in the @xmath272 plane is illustrated by the phase diagram displayed in the lower panel of fig . [ fig6 ] . two slanted lines @xmath254 and @xmath273 divide the plane into three parts corresponding to the localization regime ( i ) , the crossover regime ( ii ) and the ballistic regime ( iii ) . the wavelength , at which these lines intersect the horizontal lines corresponding to the stack lengths , @xmath234 , and @xmath274 , defines the characteristic wavelengths ( [ crossover - h-3 ] ) for each stack size . from fig . [ fig6](b ) , it follows that the separation between them grows with increasing size in accordance with eq . ( [ shift ] ) . it is easy to see that the wavelengths @xmath275 determined from the phase diagram perfectly bound the near ballistic region in both fig . [ fig5 ] and fig . [ fig6](a ) . we now consider the transmission length for a h - stack computed using a single realization . for extremely long stacks ( @xmath276 ) , the transmission length becomes practically non - random , not only within the localized region ( as is the case for m - stacks ) but also in the far ballistic region because of the self - averaging nature of the effective dielectric constant ( [ effective-4 ] ) . for less long stacks , however , @xmath258 also fluctuates in the far ballistic region . to demonstrate this , we have plotted in fig . [ fig8new ] the transmission length @xmath277 ( solid line ) and the transmission length @xmath258 for a single randomly chosen realization ( dashed line ) . it is evident that , in contrast to the results for the m - stack , the h - stack single realization transmission length in the near ballistic region is a complicated and irregular function , similar to the well known `` magneto fingerprints '' of magneto - conductance of a disordered sample in the weak localization regime @xcite . in the far ballistic region , these fluctuations are moderated since they vanish in the limit as @xmath278 . to support this statement , we display in fig . [ fig8new ] the set of separate discrete points , each of them presenting @xmath258 calculated for a different randomly chosen realization . in summary , we note that the results of sec . [ subsec : refthick ] show excellent agreement between the numerical simulations and the analytical predictions of the weak scattering approximation of sec . [ sec : anal ] . moreover , even for the case of an h - stack of length @xmath234 and strong disorder , @xmath279 and @xmath232 , the results of direct simulation and those of the wsa analysis coincide completely in the long wavelength region and differ by only a few percent in the short wave region where the scattering is certainly not weak . this is because the perturbation approach based on eqs . ( [ rec3 ] ) , ( [ rec4 ] ) is related not to the calculated quantities but to the equations they satisfy . here , we present results for stacks with only refractive index disorder ( rid ) , with @xmath231 . for h - stacks , the transmission length demonstrates qualitatively and quantitatively the same behaviour as was observed in the presence of both refractive index and thickness disorder . corresponding formulae for the transmission , localization , and ballistic lengths can be obtained from the general case by taking the limit as @xmath280 . in the case of m - stacks , however , the situation changes markedly . while in the short wavelength region and in the ballistic regime the numerical results are still in a excellent agreement with the predictions of the weak scattering approximation , the wsa fails in the long wavelength part of the localization regime . this discrepancy manifests itself also for short stacks , in which the localization regime in the long wave region is absent . in fig . [ fig8 ] we present the transmission length spectrum for a m - stack of @xmath234 layers , showing that , for @xmath281 , the numerical results for the transmission lengths of m - stacks differ by an order of magnitude from those predicted by the wsa analysis , and also those observed for the corresponding h - stacks . for longer stacks , the differences in the transmission length spectra exhibited by m - stacks and h - stacks becomes much more pronounced . in fig . fig9 , we plot transmission length spectra for different values of the m - stack length with @xmath246 ( using @xmath282 realizations ) , @xmath283 ( using @xmath282 realizations ) and @xmath284 layers ( using only a single realization ) , with the dashed - dotted straight line in fig . [ fig9 ] showing the long wavelength ballistic asymptote ( [ bbal ] ) . in the moderately long wavelength region corresponding to the localization regime , the transmission length @xmath285 coincides with the localization length @xmath286 . it exceeds the h - stack localization length by a few orders of magnitude and is characterized by a completely different wavelength dependence . this substantial suppression of localization was revealed in our earlier paper @xcite where the localization length @xmath1 was reported to be proportional to @xmath5 , in contrast to the classical @xmath178 dependence ( [ lloc ] ) that is observed for h - stacks , and which is also valid for m - stacks with both refractive index and thickness disorder . the reason for this difference is the lack of phase accumulation caused by the phase cancelation in alternating layers of equal thicknesses . to study this behaviour in more detail , we generate a least squares fit @xmath287 to the transmission length data . respectively , for @xmath288 , @xmath283 , and @xmath284 layers , the best fits are @xmath289 , @xmath290 and @xmath291 , indicating that the asymptotic form for the localization length differs from a pure power law , and perhaps is described by a non - analytic dependence . in the crossover part of the long - wave region where @xmath292 , the transmission length of m - stacks also differs essentially from that for h - stacks . moreover , the width of the crossover region on the @xmath293axis remains of the same order of magnitude , although on the @xmath294axis it grows for @xmath295 to four orders of magnitude , being much wider for longer stacks . in the long wavelength region corresponding to the ballistic regime , the transmission length of the rid m - stack coincides with that of rid h - stack . an important signature of the localization regime is the presence of transmission resonances ( see , for example , refs . lifshits , azbel , bliokh-2 ) , which appears in sufficiently long , open systems and which are a `` fingerprint '' of a given realization of disorder . these resonances are responsible for the difference between two quantities that characterize the transmission , namely @xmath296 and @xmath297 . the former reflects the properties of a typical realization , while the main contribution to the latter is generated by a small number of almost transparent realizations associated with the transmission resonances . the natural characteristic of the transmission resonances is the ratio of the two quantities mentioned above : @xmath298 in the absence of resonances , this value is close to unity , while in the localization regime @xmath299 . in particular , in the high - energy part of the spectrum of a disordered system with gaussian white - noise potential , this ratio takes the value @xmath300 @xcite . in fig . [ fig11 ] , we plot the ratio @xmath301 as a function of the wavelength for rid m- and h - stacks and for the corresponding stacks with thickness disorder . in all cases , the stack length is @xmath234 and it is evident that for the rid m - stack @xmath302 , _ i.e._. the stack length is too short for the localization regime to be realized . in other three cases , however , @xmath303 , which means that the localization takes place even in a comparatively short stack . thus , there are two ways in which to introduce transmission resonances . the first is to increase the length of the stack . [ fig12 ] displays the rid m - stack transmittance @xmath304 for a single realization as a function of @xmath105 for two lengths : @xmath233 ( solid line ) and @xmath305 ( dotted line ) . it is readily seen that while there are no resonances in the shorter stacks , they do appear for the longer sample . the second way to generate transmission resonances is to introduce thickness disorder . to demonstrate this , we plot in fig . fig13 the transmittance of a single m - stack with both thickness and refractive index disorder . it clearly shows that while the rid m - stack is too short to exhibit transmission resonances at @xmath306 , resonances do emerge at longer wavelengths for the m - stack with thickness disorder . here , we analyze the effect of thickness disorder on the @xmath307-anomaly that is the @xmath5 dependence of the transmission length . in fig . [ fig14 ] , we plot the transmission length @xmath1 for an m - stack with fixed refractive index disorder ( @xmath308 for various values of the thickness disorder . it is evident that the transmission length changes from @xmath4 to the classical dependence @xmath309 as the thickness disorder increases from @xmath310 ( top curve ) to @xmath311 ( bottom curve ) . in all cases , the number of layers @xmath312 is longer than the transmission length , guaranteeing that fig . [ fig14 ] represents the genuine localization length @xmath2 . we have also found that the anomalous dependence @xmath4 ( or with higher power ) is extremely sensitive to the alternation of left- and right - handed layers . to demonstrate this , we consider an m - stack of length @xmath313 in which each subsequent layer is chosen with equal probability to be either right- or left - handed . figure [ fig15 ] shows the transmission length spectrum for this case , which is almost the same for both the h - stacks and m - stacks . the only difference is a fairly modest suppression of localization , which occurs within the wavelength interval @xmath314 . this result confirms that it is the additional correlation between left - handed and right - handed layers in the alternating stack which is responsible for the suppression of localization . in this section , we study the transmission through layered media with absorption , which is characteristic of real metamaterials . in this case , the exponential decay of the field is due to both anderson localization and absorption @xcite , and in some limiting cases , it is possible to distinguish between these contributions . for an m - stack with weak fluctuations of the refractive index , weak thickness disorder and weak absorbtion , the wsa theory in the limits of short or long waves leads to the well - known formula @xmath315 where @xmath1 is the disorder - induced transmission length in the absence of absorption , and the absorption length is @xmath316 for short wavelengths , @xmath317 is a constant given by eqs ( [ shortwave ] ) , ( [ bbal ] ) and ( [ lloc ] ) , while for long wavelengths , in either the localized or ballistic regimes , its contribution is proportional to @xmath318 and thus is negligibly small in comparison with the contribution due to losses , which is always proportional to @xmath319 . accordingly , at both short and long wavelengths , the attenuation length coincides with the absorption length ( [ abs ] ) , with disorder contributing significantly to the attenuation length only in some intermediate wavelength region , provided that the absorption is sufficiently small . the results of the numerical calculations shown in fig . [ fig16 ] completely confirm the theoretical predictions presented above . for weak absorption @xmath320 , the direct simulation and wsa theory give exactly the same result ( solid curve in fig . [ fig16 ] ) . over a reasonably wide wavelength range , @xmath321 , disorder contributes significantly to the attenuation . for such a stack , the characteristic wavelengths are @xmath322 and @xmath323 , implying that the contribution of disorder is significant in all regions , from the short wavelength part of the localized regime to the long wavelength ballistic region . due to the losses , however , there are fewer oscillations evident in the transmission length spectrum than in the lossless case of fig . [ fig2 ] . for stronger absorption , @xmath324 , the wavelength range , over which disorder contributes to the attenuation is reduced , as well as the relative value of the contribution itself . the agreement between the numerical simulations and the wsa calculations is reasonable . we have studied the transmission and localization of classical waves in one - dimensional disordered structures composed of alternating layers of left- and right - handed materials ( m - stacks ) and have compared this to the transport in homogeneous structures composed of different layers of the same material ( h - stacks ) . for weakly scattering layers and general disorder , where both refractive index and thickness of each layer is random , we have developed an effective analytical approach , which has enabled us to calculate the transmission length over a wide range of input parameters and to describe transmission through m- and h - stacks in a unified way . all theoretical predictions are in excellent agreement with the results of direct numerical simulations . there are remarkable distinctions between the transmission and localization properties of the two types of stacks . when both types of disorder ( refractive index and layer thickness ) are present , the transmission length of a h - stack in the localized regime coincides with the reciprocal of the lyapunov exponent , while for m - stacks these two quantities differ by a numerical prefactor . this is quite surprising and , to the best of our knowledge , it is the first time that a system where such a difference exists has been discovered . it is shown that the stacks of m- and h - type manifest quite different behaviour of the transmission length as a function of wavelength . this difference is most pronounced in the long wavelength region . in the localized regime , stacks of both types are strongly disordered and reflect the incident wave almost entirely . in the ballistic regime , where stacks are almost transparent , the transport properties of h - stacks and m - stacks are markedly different . h - stacks , over the moderately long wavelength ballistic region , and m - stacks , over the entire long wavelength region , are weakly scattering disordered structures . in the extremely long wavelength ballistic region , the h - stack becomes effectively uniform a regime which is absent for m - stacks because of the intrinsic non - uniformity caused by the alternating nature of the structure . the crossover regions between different regimes are comparatively narrow . the transmission length for a single realization is non - random in the localized regime for both types of stacks . it fluctuates strongly over the entire ballistic region for m - stacks and in the near ballistic regime for h - stacks . in the far long wave region for h - stacks , the fluctuations apparent in the transmission length are moderate and decrease with increasing stack length . in the case of m - stacks , the transition from the localized to the ballistic regime is accompanied by a change in the wavelength dependence of the transmission length . in contrast , for h - stacks , the corresponding change occurs in the vicinity of the transition from the near to the far ballistic regime . again , the crossover regions between the different regimes are comparatively narrow . in m - stacks with only refractive - index disorder , anderson localization is substantially suppressed , and the localization length grows with increasing wavelength much faster than the classical square law dependence . the crossover region becomes significantly wider , and transmission resonances occur in much longer stacks than in the corresponding h - stacks . the effects of absorption on the one - dimensional transport and localization have also been studied , both analytically and numerically . in particular , it has been shown that the crossover region is particularly sensitive to losses , so that even small absorption noticeably suppresses the oscillations of the transmission length in the frequency domain . this work was supported by the australian research council through the discovery and centres of excellence programs , and also by the israel science foundation ( grant # 944/05 ) . we thank l. pastur , k. bliokh and yu . bliokh for useful discussions . we also acknowledge the provision of computing facilities through nci ( national computational infrastructure ) and intersect in australia . + if the fluctuations @xmath18 and @xmath64 are uniformly distributed in the intervals @xmath125 $ ] and @xmath325 $ ] respectively , and absorption is the same in all slabs ( @xmath326 ) , the transmission , localization , and ballistic lengths given by eqs . ( [ finalg3])-([l ] ) can be calculated explicitly in the weak scattering approximation . the results are : |
in this article we discuss an implementation of renormalization group ideas to spin foam models , where there is no a priori length scale with which to define the flow . in the context of the continuum limit of these models , we show how the notion of cylindrical consistency of path integral measures gives a natural analogue of wilson s rg flow equations for background - independent systems . we discuss the conditions for the continuum measures to be diffeomorphism - invariant , and consider both exact and approximate examples . | 1407.7746 | the application of renormalization techniques to quantum gravity theories has always been problematic . this is due to the inherent tension between the notion of scale - dependent physics of the renormalization group on the one side , and the role of background - independence of general relativity on the other side . the renormalization group is a tool for dealing with a phenomenon that is inherent to many physical systems with a very large number of degrees of freedom , which are nonlinear , i.e. exhibit nontrivial self - interaction . to describe such systems , it is convenient to not consider all degrees of freedom on an equal footing , but rather to order them according to some hierarchy . this way one considers only a finite and manageable amount of the system at a time . this hierarchy often runs along a length scale , introducing a notion of `` smaller '' and `` larger '' degrees of freedom . often the dynamics on different scales are related to each other , in that collective effects of smaller degrees of freedom combine to larger ones . the non - linearity of the interaction between the small degrees of freedom influences the dynamics of the system on larger scales . this leads to a scale - dependence of the effective actions for the system , and this `` running of coupling constants '' is a key concept in wilson s approach to the renormalization group @xcite . on the other hand , one of the key concepts of general relativity is its general covariance , which makes the theory background - independent . in particular , diffeomorphisms of the space - time manifold act as gauge symmetries . arguably , the corresponding quantum theory should possess the same symmetries . this makes an easy distinction of `` small '' and `` large '' degrees of freedom difficult , because there is no external geometry to measure any length with . rather , because geometry itself is dynamical , spatial dimensions become observables , rather than parameters . the length scales themselves become degrees of freedom , making the applicability of usual renormalization techniques to quantum gravity theories difficult . + in this article , we will consider the tentatively background - independent spin foam model approach to quantum gravity , which has been very successful in describing a path integral for quantum gravity in terms of quantized geometry @xcite , @xcite , see also @xcite for a review . a crucial ingredient to spin foam models is an abstract discretization of space - time . we will show how working with _ embedded _ , rather than abstract , discretizations provides not only a natural notion of continuum limit , it also results in a conceptually clear analogue of `` scale '' in these models , as well as a mathematically well - defined notion of renormalization group flow for the background - independent context . [ [ the - plan - of - the - paper ] ] the plan of the paper : + + + + + + + + + + + + + + + + + + + + + + in section [ sec : holonomysfm ] we will discuss an approach to spin foam models based on the kkl - formalism @xcite , which works with embedded , rather than abstract , discretizations of space - time . as a result , there is a natural notion of continuum limit in the sense of a projective family . in order to construct the continuum path integral measure , the partial measures need to satisfy a strong condition , called _ cylindrical consistency_. we show how these conditions are precisely the analogue of wilson s rg flow equation in the background - independent context , and discuss the physical interpretation in terms of configurations and measurements . in section [ sec : diffinvariance ] we will demonstrate that the condition for the continuum path integral measures to be invariant under space - time diffeomorphisms , can be posed naturally within this context , and show how this is a very strong condition on the partial measures defined at each scale . in section [ sec : example ] we will discuss an easy example , where the rg flow can be solved completely and exactly , and demonstrate how the condition of diffeomorphism - invariance severely restricts the set of solutions , most of which spontaneously break background - independence . in section [ sec : approximations ] we will discuss several suggestions for how to derive approximate solutions in more complicated cases , and give an example with a quartic interaction term , which can be treated with these methods . while appendix [ app : properties ] provides mathematical details for some of the constructions , in appendix [ app : canonical ] the connection to refined algebraic quantization techniques is discussed , to make contact to the canonical framework in the case that space - time has a boundary . in this section we review the definition of ( holonomy ) spin foam models @xcite . these are generalizations of lattice gauge theories defined on arbitrary , irregular lattices . they encompass candidate models for quantum gravity such as the barrett - crane @xcite and the eprl - fk model @xcite , as well as versions of group- and tensor field theories @xcite . spin - net models , which are lower - dimensional analogues of spin foam models @xcite , are very close on a technical level , so that many results are transferrable @xcite . [ [ complexes ] ] 2-complexes : + + + + + + + + + + + + throughout the article , we will work in the semi - analytical category , which allows us to circumvent certain technical inconveniences which arise when working with smooth structures , while being more `` local '' than using analytic functions . so all maps , manifolds and connections are supposed to be semi - analytic @xcite in what follows . + to begin with we fix a manifold @xmath0 , which is supposed to be the model for space - time on which the fields are defined . for the time being @xmath0 is supposed to be closed . the case of compact @xmath0 with nonempty boundary will be treated in appendix [ app : canonical ] . the quantum theory that is developed then describes the fluctuation of connections of some type on @xmath0 . the key concept of the approach will be that of a @xmath1-complex . by a @xmath1-complex we mean an abstract complex @xmath2 embedded in @xmath0 , where all 1- and 2-cells ( `` edges '' and `` faces '' ) are oriented . wlog all 0-cells ( `` vertices '' ) are negatively oriented . for an edge @xmath3 and a face @xmath4 which contains @xmath5 as part of its boundary , define by @xmath6=\pm1 $ ] , depending on whether the orientation induced on @xmath5 by the given one on @xmath7 coincides or is opposite to the given one on @xmath5 . similarly @xmath8=\pm 1 $ ] whenever @xmath5 is incoming / outgoing w.r.t . the vertex @xmath9 . we note that the set of all such embedded @xmath1-complexes possesses an ordering relation : we write @xmath10 if and only if every cell of every dimension in @xmath2 is composed ( disregarding orientation ) of cells of the appropriate dimension in @xmath11 . note that this depends crucially on the embedding of @xmath2 , @xmath11 in @xmath0 . in particular , every vertex in @xmath2 is necessarily a vertex in @xmath11 . [ [ configuration - spaces ] ] configuration spaces : + + + + + + + + + + + + + + + + + + + + + fix a compact lie group @xmath12 . for each such @xmath1-complex @xmath2 we define the quantum configuration space @xmath13 to be a direct product of as many copies of @xmath12 as there are pairs @xmath14 such that the edge @xmath5 is part of the boundary of the face @xmath7 . the set of all these pairs will be denoted as @xmath15 , so we have @xmath16 note that this results in not one link variable , representing the parallel transport as an element of the gauge group @xmath12 , along the edge @xmath5 , but in as many link variables @xmath17 as there are faces meeting at @xmath5 . for the special cases of bf or gauge theories these will all set to be equal . for the general case , the way that these different parallel transports are allowed to depend on the face is precisely the imposition of the simplicity constraints . for any pair of two complexes @xmath18 one defines the map @xmath19 via @xmath20}.\end{aligned}\ ] ] here @xmath21=\pm1 $ ] is the relative orientation of @xmath22 and @xmath5 . the @xmath23 constitute a coarse graining procedure , and determine how the ( microscopic ) degrees of freedom in @xmath24 combine to the the ( macroscopic ) degrees of freedom in @xmath25 . note how in our convention the order of the product is from the right to the left , as befits parallel transports . note also how , for @xmath26 one has that @xmath27 with this data it is possible to define the projective limit of the family @xmath28 , to be @xmath29 the space @xmath30 represents the quantum continuum connections for the gauge group @xmath12 . note that the limit does not constitute of performing any limiting procedure in the analytical sense . rather , a continuum connection @xmath31 is given by the collection of all its partial representatives @xmath32 , which satisfy the consistency conditions @xmath33 , whenever @xmath34 . note that there is an associated projection @xmath35 for each @xmath2 , given by @xmath36 satisfying @xmath37 . see appendix [ app : properties ] for more information about @xmath30 . [ [ path - integrals - on - overlinemathcala ] ] path integrals on @xmath30 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + there is a well - understood way to construct ( regular borel ) measures @xmath38 on @xmath30 , which have the interpretation of path - integral measures in quantum theory , or probability measures in statistical physics . first we note that , given a regular borel measure @xmath38 on @xmath30 , then we get a family of measures @xmath39 on the @xmath13 via push - forward with the map ( [ eq : projectionmap ] ) , i.e. @xmath40 clearly , these measures satisfy the so - called cylindrical consistency condition @xmath41 a collection of measures @xmath42 satisfying ( [ eq : cylindricalconsistency ] ) is also called a promeasure . we are in the lucky situation that the converse is also true : given a promeasure , then this comes from a unique measure @xmath38 on @xmath30 . see appendix [ app : properties ] for details . + [ [ observables ] ] observables : + + + + + + + + + + + + in this context , it is straightforward to define what observables should be : they should correspond to continuous functions on @xmath30 . the sup - norm provides a @xmath43-topology for that space , and a dense subset is given by _ continuous cylindrical functions_. these are complex - valued functions @xmath44 on @xmath30 such that there is a @xmath1-complex @xmath2 and a continuous function @xmath45 such that @xmath46 such an observable is also called _ cylindrical over @xmath2_. the expectation values of cylindrical functions are then the mathematical realizations of observables in the path integral @xmath47 it should be noted that , whenever @xmath44 is cylindrical over @xmath2 , then it is automatically cylindrical over all @xmath48 , because of ( [ eq : projectionconsistency ] ) : @xmath49 this fact will play a crucial role later , when we consider diffeomorphism - invariant measures @xmath38 . [ [ physical - dictionary - and - renormalization ] ] physical dictionary and renormalization : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in this section we would like to recap the physical interpretation of the mathematical concepts introduced in the last section . in particular we would like to demonstrate how they form the natural framework for renormalization in a background - independent context , similarly to the concepts described in @xcite . the space of quantum continuum connections @xmath30 contains the full degrees of freedom of the theory , and thusly is quite large in particular it is no ( not even an infinite - dimensional ) manifold . however , each @xmath1-complex @xmath2 provides a cut - off , in the sense that is specifies a finite amount of degrees of freedom of the theory , the @xmath17 with @xmath50 . the corresponding configuration space @xmath13 is a finite - dimensional manifold , and hence much more accessible than @xmath30 . it can be regarded as a `` finite - dimensional slice through @xmath30 '' . the ordering relation @xmath34 provides a hierarchy between the degrees of freedom . in particular , the coarse graining maps @xmath23 provide the information on how microscopic degrees of freedom , associated to the `` finer '' @xmath1-complex @xmath11 , combine to macroscopic ones , which are associated to the `` coarser '' @xmath1-complex @xmath2 . in this sense the @xmath1-complexes @xmath2 are the natural analogue of the notion of `` scale '' in the background - independent context . + as we have already stated , the measure @xmath38 provides the full path integral measure for the continuum . on the other side , the @xmath51 correspond to the effective path integral measures for the degrees of freedom in @xmath13 . in this sense , the continuum theory is given by the collection of all effective theories at all scales . the cylindrical consistency conditions ( [ eq : cylindricalconsistency ] ) then give a relation between the effective integration measures on @xmath2 and @xmath11 , providing a background - independent version of wilson s renormalization group flow . to see this in a familiar context , consider standard lattice gauge theory as an example . given , say , a regular @xmath52-torus with a background being a flat metric having volume @xmath53 , consider the set of all regular hypercubic lattices , centered at a common point , having lattice length @xmath54 for @xmath55 . to the lattice with lattice length @xmath56 associate the @xmath1-complex @xmath57 consisting of its edges and faces consider theories of the form @xmath58 the function @xmath59 is called an _ action _ , and it is an @xmath56-dependent function of the link variables @xmath60 . more specifically , it can depend on several parameters @xmath61 , called _ coupling constants _ which in turn are @xmath56-dependent . condition ( [ eq : cylindricalconsistency ] ) then becomes @xmath62 which can be readily seen to be precisely the rg flow equation for effective lattice actions , in terms of their coupling constants . + in the general case , when there is no background metric , one has to keep the collection of all @xmath1-complexes @xmath2 , together with the information of how they are embedded into each other . in this case , the condition for there to be a continuum measure @xmath38 is still given by the cylindrical consistency conditions ( [ eq : cylindricalconsistency ] ) , but instead of an easy set of equations one now generically has an uncountable set of equations for an uncountable set of measures . so , while the rg flow in the background - dependent case runs along a single parameter ( the length scale @xmath56 , figure [ fig : figure02 ] ) , in the background - independent context the rg flow runs along the partially ordered set of all @xmath1-complexes @xmath2 ( see figure [ fig : figure03 ] ) . note an important difference to the approach in @xcite : the flow described here is not invertible , because during the coarse graining process information about microscopic degrees of freedom is lost . in @xcite , the flow is defined to go in both directions , leading to the notion of a renormalization groupoid . we will see an example for this general case in section [ sec : example ] , and a discussion for how to simplify this complicated collection of equations in section [ sec : approximations ] . the group of space - time diffeomorphisms is a gauge symmetry of general relativity . it is natural to demand that it should also appear as a symmetry of the path integral measure . while one can achieve this on a formal level in the linearized regime @xcite , in general it is not known how to construct a non - perturbative path integral measure with full diffeomorphism symmetry . the first example of a rigorous diffeomorphism - invariant measure on the space of all connections ( i.e. not just flat ones ) is the construction in @xcite , which led to the remarkable success of loop quantum gravity . the measure in question , however , is not the path integral measure for covariant quantum gravity , but rather the kinematical measure for canonical quantum gravity . consequently , it is invariant under diffeomorphisms of the cauchy surface only , and contains no knowledge about the dynamics . still , key elements of its construction play a central role in our approach as well . in four space - time dimensions , the tension between continuum diffeomorphism symmetry on the one side , and the discretization in some quantum gravity approaches on the other side is deep @xcite . incorrect implementation of diffeomorphism symmetry on the discrete level can lead to the emergence of spurious degrees of freedom , as one would expect from a breaking of classical gauge symmetry in the quantum theory @xcite . still , from lower - dimensional models there are hints that at the ir fixed point diffeomorphism symmetry can emerge @xcite , and that one can even construct the discrete theories with full continuum diffeomorphism symmetry , leading to the quantization of discretizations of the correct continuum degrees of freedom @xcite , although this might be highly nontrivial in four space - time dimensions @xcite . it should be noted that succeeding in this would be equivalent to constructing a nontrivial four - dimensional manifold invariant @xcite . for these not many nontrivial examples are known , in particular not if the invariant should not also be topological . + in our framework , it is easy to define what is meant by diffeomorphism - invariant path - integral measures , and in the following we will see that the condition for diffeomorphism - invariance will pose severe restrictions on the partial measures @xmath51 . for this , the fact that we have kept the information about how each @xmath1-complex @xmath2 is embedded in the manifold @xmath0 , is a necessary requirement . denote the group of semi - analytic diffeomorphisms by @xmath63 , then the action of this group on @xmath0 can be pulled back to an action on the space of @xmath1-complexes . so e.g. for each edge @xmath5 in @xmath2 , there is a corresponding edge @xmath64 in @xmath65 , and so on . in the following we assume that the family of @xmath1-complexes considered is large enough , so that for each @xmath2 and each @xmath66 , the @xmath1-complex @xmath65 is also a member of that family . if that is the case , @xmath66 induces a lie group isomorphism between @xmath13 and @xmath67 . it is not hard to see that this map respects the ordering relation ( [ eq : orderingrelation ] ) , i.e. if @xmath34 , then @xmath68 . consequently , the action of @xmath69 can be pulled back to @xmath30 : if @xmath70 is given by ( [ eq : definitioncontinuumlimit ] ) , then @xmath71 furthermore , given a collection of cylindrically consistent measures @xmath42 , the measures @xmath72 are also cylindrically consistent , and hence define a measure on @xmath30 . of course , this measure coincides with the push - forward @xmath73 , given by the action @xmath74 on @xmath30 + in this context it is a natural question to ask whether a measure @xmath38 is invariant under the action of @xmath63 , i.e. whether or not the equation @xmath75 holds . for this , take some observable @xmath44 cylindrical over a @xmath1-complex @xmath2 , i.e. @xmath76 with @xmath45 . also , for some semi - analytic diffeomorphism @xmath69 , there exists , by definition , a @xmath1-complex @xmath11 which is finer than both @xmath2 and @xmath65 , i.e. @xmath77 . then the two observables @xmath78 \mathcal{o}_2\;&:=&\;\big(\phi_*\mathcal{o}_\gamma\big)\,\pi_{\gamma'\phi(\gamma)}\end{aligned}\ ] ] are both cylindrical over @xmath11 , and from @xmath75 and ( [ eq : expectationvalueobservable ] ) it is easy to deduce that they should have the same expectation value , i.e. @xmath79 it is important to note that this is a strong restriction on @xmath80 . in particular , the observables supported on two sub - complexes of @xmath11 which can be connected by a semi - analytic diffeomorphism on @xmath0 , need to have the same expectation values , when evaluated with the measure @xmath11 . for this it is _ not _ necessary that the whole diffeomorphism maps @xmath11 onto itself , only that it maps the two sub - complexes @xmath2 and @xmath65 onto each other . the converse is also true : let @xmath38 be a measure on @xmath30 with the following property : for each @xmath2 , and each @xmath81 such that there exists @xmath82 such that @xmath83 , one has that @xmath84 for each @xmath85 cylindrical over @xmath86 , and @xmath87 . then @xmath75 for all @xmath88 . in other words , diffeomorphism - invariance can be checked separately for each @xmath51 . it should be noted , however , that in order to do so , it is not sufficient to know just the combinatorial data of @xmath2 . one also needs the information about how @xmath2 embeds into @xmath0 . in the following we consider a very simple example , which can be treated completely and analytically . for more interesting systems we do not expect that the rg equations can be solved in full generality . let @xmath0 be a @xmath1-dimensional , connected , closed , oriented , semi - analytic manifold . the family of @xmath1-complexes we consider consists of all semi - analytic cellular decompositions of @xmath0 , i.e. @xmath1-complexes that completely `` fill out '' @xmath0 . we want to consider @xmath89 gauge theory on ( a trivial line bundle over ) @xmath0 , so to a @xmath1-complex @xmath2 with @xmath90 edges we associate the partial configuration space @xmath91 in this case , the coarse - graining maps @xmath92 are given by @xmath93}\end{aligned}\ ] ] where @xmath94=\pm 1 $ ] , depending on whether the orientations of @xmath95 agree or disagree . we formulate the rg equations considering the family of measures @xmath96 where the product ranges over all @xmath97 faces of @xmath2 , @xmath98 and @xmath99 . as customary , we denote by @xmath100 the heat kernel ( on @xmath89 ) @xmath101 at time @xmath102 , and by @xmath103 the ordered product of @xmath104 of edges @xmath5 around the face @xmath7 . the observables in this case are simply given by continuous functions on @xmath30 , a dense set of which is generated by the so - called _ charge - network - functions _ on a graph @xmath105 . @xmath106 with @xmath107 . it is an elementary calculation to show that the expectation value of ( [ eq : exampleobservable ] ) with the measure ( [ eq : examplemeasure ] ) is given by @xmath108n_{f_1}+[e , f_2]n_{f_2},0}\end{aligned}\ ] ] where for an edge @xmath5 we denote by @xmath109 the two faces that meet at @xmath5 . using the orientability of @xmath0 , one can see that ( [ eq : exampleexpectationvalue ] ) vanishes if @xmath110 is not gauge - invariant , i.e. if at some vertex @xmath9 the condition @xmath111}\;\stackrel{!}{=}\;0\end{aligned}\ ] ] is violated . if ( [ eq : eyamplegaugeinvariance ] ) holds , on the other hand , then ( [ eq : exampleexpectationvalue ] ) can be evaluated by computing just one sum over @xmath112 . using an elementary property of the heat kernel , it is easy to show that the rg equations @xmath113 for a pair @xmath34 of @xmath1-complex are given by @xmath114 \theta_f\;=&\;\sum_{f'\subset f}[f',f]\theta_{f ' } \end{aligned}\label{eq : examplergequations}\end{aligned}\ ] ] where @xmath115=\pm 1 $ ] , depending on whether the orientations of @xmath116 agree or disagree . + there is quite a simple set of solutions to ( [ eq : examplergequations ] ) : choose a semi - analytic riemannian metric @xmath117 and a semi - analytic @xmath1-form @xmath118 , and define @xmath119 \theta_f\;&:=\;\int_f\theta \end{aligned}\label{eq : examplergsolution}\end{aligned}\ ] ] it is straightforward to see that ( [ eq : examplergsolution ] ) satisfy ( [ eq : examplergequations ] ) , defining a measure @xmath120 on @xmath30 . there are two interesting limiting cases of the solutions ( [ eq : examplergsolution ] ) . one corresponds to formally taking @xmath121 , i.e. all @xmath122 , while keeping @xmath123 fixed . the result exists as a measure on @xmath30 and can be easily identified with the ashtekar - lewandowski measure @xmath124 @xcite , also called _ high temperature fixed point_. another limiting solution is comprised of the limit @xmath125 , i.e. of taking all @xmath126 . it can be shown that the limit measure @xmath127 exists as a measure on @xmath30 . in this limit we arrive at a rigorous version of the expression @xmath128-\theta\big)\ ; \mathcal{o}[a].\end{aligned}\ ] ] in the case of @xmath129 , this becomes simply the measure of @xmath89 bf theory on @xmath0 . in the case of @xmath12 being finite instead of @xmath130 , and @xmath129 , the state sum @xmath131 in ( [ eq : examplelimitmeasure ] ) is finite and becomes the well - known dijkgraaf - witten invariant of @xmath1-dimensional manifolds @xcite . [ [ diffeomorphism - invariance ] ] diffeomorphism - invariance : + + + + + + + + + + + + + + + + + + + + + + + + + + it can be easily shown that for a semi - analytic diffeomoprhism @xmath69 one has that @xmath132 so generically the continuum measures @xmath120 on @xmath30 are not diffeomorphism - invariant . in particular , the condition for diff - invariance imposes further restrictions , additional to the rg flow equations ( [ eq : examplergequations ] ) . to see this , we consider a @xmath1-complex @xmath86 and a diffeomorphism @xmath69 which is different from the identity map only in a small neighborhood of an interior point of an edge @xmath5 , moving the edge slightly ( see figure [ fig : figure06 ] ) . denote by @xmath2 the obvious @xmath1-complex finer than @xmath86 and @xmath133 , and consider a charge - network function @xmath110 on @xmath86 . from figure [ fig : figure06 ] it can be seen that the condition @xmath134 for all @xmath135 leads to @xmath136 a_{f_1}\;&=\;a_{f_5 } & a_{f_2}\;&=\;a_{f_6}+a_{f_7}\\[5pt ] a_{f_3}\;&=\;a_{f_5 } + a_{f_6 } & a_{f_4}\;&=\;a_{f_7 } \end{aligned}\end{aligned}\ ] ] and similar equations for the @xmath137 . since these equations have to hold for every edge , relations for all @xmath138 ( and all @xmath137 ) for all @xmath2 can be deduced . it s not hard to see that the only solutions for this are either ( formally ) @xmath139 for all @xmath7 , or , if @xmath138 is finite , then @xmath140 , hence @xmath141 these two solutions correspond ( formally ) to the measures @xmath124 and @xmath142 . indeed , it is not difficult to check that both the ahtekar - lewandowski measure @xmath124 and the bf - theory measure @xmath142 are invariant under diffeomoprhisms . we see that imposing diffeomorphism - invariance has drastically reduced the allowed solutions , by introducing new equations , additionally to the rg flow equations . we have seen that there are uncountably many rg flow equations given by the cylindrical consistency conditions ( [ eq : cylindricalconsistency ] ) . in the easy example from section [ sec : example ] it was possible to write them in a closed form for a specific set of measures ( [ eq : examplemeasure ] ) , and to guess some very obvious solutions . this was due to the easy nature of the heat kernel on @xmath89 in 2d , and will certainly not be possible in more interesting scenarios . as in almost all cases of physical theories , one needs to make approximations to be able to perform even numerical calculations . in the following we will present three different ways of approximating the rg flow equations ( [ eq : cylindricalconsistency ] ) , making them more accessible to analytical and numerical investigation . afterwards we will apply all of them in order to treat a spin foam model with quartic interaction term . [ [ restricting - the-2-complexes ] ] 1 . ) restricting the @xmath1-complexes + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + one reason for the complexity of the full set of the rg equations is that the set of @xmath1-complexes is not only uncountable , they can also be completely irregular , making bookkeeping difficult . so instead of taking on all @xmath1-complexes at the same time , one could restrict the set under consideration , simplifying the analysis immensely . one example would certainly be considering a sequence of regular lattices which are sublattices of each other , as in the regular hypercubic calculations in lattice gauge theory . the difference is that there would be no lattice length attached to it , rather , the lattice length would be one observable among many . in this case it might be possible to retrieve standard rg flow equations in the sense of `` evolving constants '' which lie at the heart of the relational framework for dirac observables @xcite . although a severe restriction will almost certainly violate the condition at the beginning of section [ sec : diffinvariance ] , which was necessary to define the action of the diffeomorphisms , one can still look for diffeomorphism - invariant measures in this case , because the condition for diffeo - invariance can be pulled back to each single @xmath1-complex @xmath2 , even e.g. regular lattices , via ( [ eq : conditiondiffinvariance ] ) . [ [ restricting - the - observables ] ] 2 . ) restricting the observables + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + given @xmath34 , then the condition @xmath143 can be rephrased as @xmath144 instead of demanding ( [ eq : cylindricalconsistencyalternative ] ) to hold for all observables @xmath145 , one can select a small subset of them , and require ( [ eq : cylindricalconsistencyalternative ] ) to hold only for those . by considering only a few important observables , the problem can be simplified considerably , as we will see in an example later . [ [ restricting - the - parameter - space ] ] 3 . ) restricting the parameter space + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + one of the most important approximations , is restricting the space of measures @xmath42 one formulates equation ( [ eq : cylindricalconsistency ] ) in . this can be done by making an ansatz for the measure , e.g. in terms of an action function @xmath146 an example for this would be e.g. the ansatz ( [ eq : examplemeasure ] ) we were using in the example . even more dramatically would be the restriction to actually finitely many parameters , so that the rg flow equations would be equations for a finite set of coupling constants depending on @xmath2 , i.e. @xmath147 . the standard rg equations for lattice gauge theory certainly fall into this category . in the example in section [ sec : example ] we were able to solve the rg equations ( [ eq : cylindricalconsistency ] ) within this ansatz . in most interesting cases this will not be possible . one of the reasons for that is that any non - topological exact solution to the rg flow equations will necessarily generate non - local couplings @xcite , with the 2d heat kernel being one of the very few exceptions to this rule . in particular , any ansatz which factorizes over the constituents of the @xmath1-complexes , as is the case with e.g. all present spin foam models , will not be an exact solution to the rg flow equations , because it will not include non - local couplings . however , one can still look for approximate solutions in a realm in which non - local couplings are small , and restrict the flow to the space of measures parametrized as ( [ eq : measureansatz ] ) , by demanding that the error @xmath148 is small , with @xmath149 if one chooses @xmath150 to be the supremum norm on @xmath151 , then ( [ eq : approximatergflowequation ] ) is just the minimization of the operator norm of the difference of linear functionals corresponding to the measures @xmath152 and @xmath153 , respectively , via the riesz representation theorem . by varying this norm one can weigh the errors on some observables to be more severe than on others , introducing more freedom in the procedure . note that this approximation is akin to the one employed in the functional renormalization approach to quantum gravity ( @xcite and references therein ) , where the space of allowed actions is kept fixed as well . of course , this approximation might lead to solutions which are very far away from an actual solution , if the space of measures is restricted too strongly . just as in the frg approach , though , one can make a truncation to few parameters , and hope to find nontrivial , i.e. interacting , uv fixed points of the action . afterwards , one can enlarge the parameter space and check whether the characteristic fixed point is stable . this has led to quite some success in that approach , so that one can hope that similar procedures can also work for spin foam models . it should also be noted that , due to the specific nature of this approximation , one can suppress non - local couplings by specifically excluding them from the ansatz for the partial measures @xmath51 . this is reminiscent of the migdal - kadanoff approximation within lattice gauge theory @xcite , and the highest eigenvalue approximation within tensor network renormalization . it would be very interesting to compare these methods with the one presented here , and we hope to come back to this point in future work . [ [ example ] ] example : + + + + + + + + in what follows we will utilize all three of the above approximation methods to compute the rg flow of a simple @xmath154 model with gauge group @xmath89 and quartic interaction term . consider a single face @xmath97 diffeomorphic to the closed disc , bounded by the single edge @xmath90 , forming the @xmath1-complex @xmath2 , and two different subdivisions @xmath155 of @xmath2 into @xmath156 and , respectively , into @xmath157 faces . for @xmath158 we make a similar ansatz to ( [ eq : examplemeasure ] ) , with @xmath159 constant for all faces and @xmath160 , including a quartic term @xmath161 with the heat kernel @xmath162 ( [ eq : heatkernelu(1 ) ] ) , @xmath163 normalizing the measure to @xmath164 , and @xmath165\,\phi_e,\end{aligned}\ ] ] with all @xmath166 . thus we have two coupling constants , @xmath167 and @xmath168 , on @xmath86 . a similar measure can be defined on @xmath169 . we relate the two sets of coupling constants by comparing the expectation values of the observables on @xmath2 . there is one gauge - invariant observable for each @xmath170 on @xmath2 , given by @xmath171 a straightforward calculation reveals that @xmath172 it is not hard to show that ( [ eq : approximateevalue ] ) is symmetric under @xmath173 , so we only consider @xmath174 . in order to explore the uv limit , we consider large @xmath156 and @xmath175 , with @xmath176 , and demand that the difference of expectation values of @xmath177 are minimal , i.e. @xmath178 with @xmath179 . given @xmath180 , condition ( [ eq.approximatecondition ] ) determines @xmath181 . in the range of @xmath182 , @xmath183 . ] in the range of @xmath182 , @xmath183 . ] in the range of @xmath184 , @xmath185 . ] in the range of @xmath184 , @xmath185 . ] within a numerical treatment , we considered several cases of @xmath186 , finding the rg flows being relatively stable under change of the @xmath186 . a depiction of the phase diagram is shown in figure [ fig : figure07 ] , while in figures [ fig : figure08 ] , [ fig : figure09 ] , [ fig : figure10 ] , [ fig : figure11 ] some numerical flows with @xmath187 are presented . note that , due to the compactness of @xmath89 , the quartic coupling @xmath188 can be negative . + the analysis shows quite unsurprisingly that there is the @xmath189-theory fixed point @xmath190 , corresponding to the gaussian theory , with @xmath56 being a relevant , @xmath188 being an irrelevant coupling . despite several different uv scenarios , all trajectories flow to the high temperature fixed point in the ir . the fact that the flow remains virtually unchanged upon varying the @xmath186 suggests that one can arrive at good approximations for the flow by considering only few observables . in this article , we have discussed the basic framework for background - independent renormalization in spin foam models . by carefully taking care of the embeddings of @xmath1-complexes in the space - time manifold , a clear notion of continuum limit arises , motivated by the construction of regular borel measures on infinite - dimensional spaces . the embedded @xmath1-complexes take on the meaning of scale , in the sense that they provide a cut - off for degrees of freedom , since they determine a finite - dimensional portion of infinite - dimensional state space . also , since embedded @xmath1-complexes form a directed , partially ordered set , there is a natural hierarchy among the degrees of freedom , which does not use any notion of length scale . the partial measures @xmath51 defined on each scale @xmath2 define the effective dynamics at that scale . the condition of cylindrical consistency ensures that the effective actions at different scales are compatible , and is the key ingredient to ensure the existence of the continuum measure @xmath38 . the cylindrical consistency conditions are the background - independent analogue of the renormalization group flow equations . indeed , if there is a background metric to associate lengths to the scales , the cylindrical consistency conditions indeed turn into wilson s rg flow equations for effective actions . it is important to note that the continuum theory does not consist of a limit in the usual sense . rather , the continuum theory is equivalent to the collection of all effective theories at all scales . also , because one keeps track of embeddings , there is a clear action of space - time diffeomorphisms on the continuum theory . this allows for a natural notion of diffeomorphism - invariance of the path - integral measure . from the examples we also have seen that , although the rg flow equations itself are background - independent , the solutions not necessarily are . rather , some solutions spontaneously break diffeomorphism - symmetry by introducing a background solution . as expected , the group of diffeomorphisms then maps these solutions into each other , transforming the background structure accordingly . for a solution to be background - independent , its effective theories need to satisfy very strong conditions , as given by ( [ eq : conditiondiffinvariance ] ) . this is advantageous for two reasons : firstly , it allows to check for diffeomoprhism - invariance without knowing the solution in full generality ( which will most likely not be possible in the case of quantum general relativity ) . secondly , it strengthens the hope that the condition of diffeomorphism - invariance will severely restrict the rg flow , so that the non - renormalizability that plagues the perturbative ansatz might not affect the full , background - independent theory . this is certainly a point worthwhile to investigate further . [ [ open - questions ] ] open questions : + + + + + + + + + + + + + + + one of the immediate limitations of the framework presented here is the condition of compactness of the gauge group @xmath12 . this only allows for the treatment of quantum gravity with a riemannian signature of the metric . although there exist finite expressions for the vertex amplitude for the lorentzian signature case , in which the gauge group is @xmath191 @xcite , the full path integral measure for one @xmath1-complex , does not , at this point , exist . in particular , it is not known how to make the sum over all intertwiners and spins finite , since for noncompact groups , there is no heat kernel regularization , and most likely no regularization at all , which does not break local gauge - invariance . although there are finite versions of the full amplitude involving quantum deformations of @xmath192 @xcite , these do not fall into the framework presented in this article , due to the non - commutativity of the observables in those cases . also , they are restricted to a positive cosmological constant . note that the problem of defining a finite amplitude for lorentzian signature is notorious for most other background - independent approaches to quantum gravity as well . additionally , the construction of spaces of generalized connections @xmath30 for non - compact groups is an unsolved problem until today . also , cylindrical consistency alone will not guarantee the existence of the continuum measure , as the conditions for kolmogorov s theorem are not automatically satisfied , as in the compact case @xcite . even if all these problems might be overcome , it should be noted that , due to the fact that in the lorentzian eprl model all edges are fixed to be of a certain signature , in most cases time - like , such a model seems to essentially fix the causal structure from the outset . one would expect that the resulting theory only lets the conformal factor fluctuate , which would seem to miss the point of the geometric principles underlying the construction of the eprl amplitude . all of these points , as well as making contact with the implementation of rg flow in other background - independent approaches to quantum gravity @xcite , should provide interesting research projects for the future . the author thanks dr . bianca dittrich and dr . astrid eichhorn for an invitation to perimeter institute , where the content of this article was presented as a talk . this work was funded by the project ba 4966/1 - 1 of the german research foundation ( dfg ) . in this section we go over the mathematical details of the construction of @xmath30 in more detail . recall that @xmath0 is a semi - analytic , closed manifold . a two - complex @xmath2 is a finite collection of cells , i.e. vertices , edges and faces , where each of those is , respectively , a @xmath193- , @xmath164- or @xmath1-dimensional , embedded semi - analytic manifold with the topology of , respectively , a point , a compact interval , or the bounded disc @xmath194 . we demand the usual rules for cell complexes , i.e. for every cell , its boundary is a collection of lower - dimensional cells , and for any two cells one of the three following cases is true : they do not intersect in @xmath0 , their intersection is a common subcell , or one is completely contained in the other s boundary . two @xmath1-complexes @xmath2 and @xmath11 satisfy the relation @xmath34 , if and only if , for each @xmath52-cell @xmath195 of @xmath2 ( with @xmath196 or @xmath1 ) , there is a collection of @xmath52-cells @xmath197 in @xmath11 such that @xmath198 because we work in the semi - analytic category , for two @xmath1-complexes @xmath155 there is always a @xmath1-complex @xmath11 finer than both , i.e. @xmath199 . its construction is straightforward by subdividing both @xmath1-complexes along their intersection , which are again semi - analytic submanifolds .- complexes there is a technically very involved way around this @xcite , while whether a similar generalization of our framework to smooth , say cw , @xmath1-complexes exists or not is not clear at this point . so using the semi - analytic category for the moment is a technically convenient restriction , which might be overcome in the future . ] note that with this relation , the set of @xmath1-complexes becomes a directed , partially ordered set . an orientation @xmath200 of @xmath2 denotes an individual orientation for each edge @xmath5 and face @xmath7 in @xmath2 ( vertices always implicitly carry the negative orientation ) . denote the set of orientations @xmath201 , and the set of all pairs of edges @xmath5 and faces @xmath7 such that @xmath202 as @xmath15 , and let @xmath12 be a compact lie group , then the configuration space @xmath13 is the set of all maps @xmath203 such that @xmath204 , depending on whether the orientations of the edge @xmath5 in @xmath200 and @xmath205 agree or disagree . of course , upon a choice of orientation of @xmath2 , @xmath13 becomes naturally isomorphic to @xmath206 , an isomorphism that we will use implicitly in what follows , as well as the rest of the article .- complexes do not , technically , form a partially ordered set . this is because differently oriented @xmath2 , @xmath207 satisfy @xmath208 and @xmath209 , but @xmath210 . nevertheless , in the rest of the article we will ignore this technicality , and assume that each @xmath2 carries an orientation . ] for @xmath34 , define the coarse graining map @xmath19 ( for a chosen orientation on both @xmath1-complexes ) as in ( [ eq : coarsegrainingmap ] ) to be @xmath211},\end{aligned}\ ] ] where @xmath21=\pm 1 $ ] depending on whether the orientation on @xmath22 agrees or disagrees with the one induced by the orientation of @xmath5 . the ordered product is the product of group elements @xmath212 going through all @xmath95 , starting at the beginning of @xmath5 until the end , according to its orientation . it is an easy task to check that this definition is in fact independent of the choice of orientation on either @xmath1-complex , making @xmath23 well - defined , and satisfies @xmath213 . + following @xcite , we define the projective limit @xmath30 to be the subset of the tychonoff product @xmath214 satisfying cylindrical consistency , i.e. @xmath215 where @xmath216 is the surjective projection onto the @xmath2th factor . because all @xmath217 are continuous in the tychonoff topology , and all @xmath23 are continuous as well , @xmath30 is a closed subset of @xmath218 , and because the latter is compact , so is the former , when we choose the subset topology , which we will do in what follows . the continuous functions on @xmath30 contain the set of cylindrical functions @xmath219 as dense subset . for @xmath34 one has the maps @xmath220 , defined by @xmath221 . note that the embedding maps @xmath222 can be continued to isometries on the hilbert spaces @xmath223 via @xmath224 the collection @xmath225 has a directed limit in the category of hilbert spaces , coinciding with @xmath226 , where @xmath124 is the ashtekar - lewandowski measure . + consider a collection @xmath51 of regular borel measures on @xmath13 , i.e. for each @xmath2 there is a positive linear map @xmath227 satisfying @xmath228 let furthermore the measures be normalized : @xmath229 , and satisfy @xmath230 then there is a positive linear map @xmath231 defined by @xmath232 for @xmath44 being cylindrical over @xmath2 . due to ( [ eq : appendixcylindricalconsistency02 ] ) , @xmath233 is well - defined . clearly , @xmath233 is bounded by @xmath164 , so that it can be continued to @xmath234 , so that there is a regular borel measure @xmath38 on @xmath30 . it is an elementary exercise to prove that the haar measures @xmath235 on each @xmath13 satisfy the cylindrical consistency conditions ( [ eq : appendixcylindricalconsistency02 ] ) . the resulting measure is precisely the ashtekar - lewandowski measure @xmath124 . the framework presented in this article needs only to be slightly amended in order to include manifolds with boundary . so let us assume that @xmath0 is compact with closed boundary @xmath236 . for @xmath1-complexes @xmath2 we allow only those which satisfy @xmath237 , i.e. a face may intersect with the boundary only in some of its boundary edges . thus @xmath238 forms a graph . the ordering relation between @xmath1-complexes is equivalent to the case in which there is no boundary : @xmath34 if and only if every cell in @xmath2 can be composed of cells in @xmath11 . note that , because @xmath2 and @xmath11 are both embedded in @xmath0 , this implies @xmath241 to be a subgraph of @xmath242 . for this we similarly write @xmath243 . for @xmath34 we define the coarse graining map @xmath19 to be @xmath244}\quad\textrm{for all } e\subset f\\[5pt]\label{eq : appendixcoarsegraining2 } \pi_{\gamma'\gamma}\big(\{h_{e'f'},g_{e'}\}\big)_{e}\;:=\;\overleftarrow{\prod_{e'\subset e}}g_{e}^{[e , e ' ] } \quad\textrm{for all } e\subset\partial\gamma.\end{aligned}\ ] ] which is completely analogous to ( [ eq : cylindricalconsistency ] ) . notice , however , that fubini s theorem allows us to integrate out all @xmath17 , without integrating out the @xmath247 , leaving us , for each @xmath2 , with a positive linear map @xmath248 , by @xmath249\;:=\;\int_{g^{\partial e}}\left(\int_{g^{e\ltimes f}}d\mu_{\gamma}(h_{ef } , g_e)\right)\,\psi(g_e).\end{aligned}\ ] ] * the linear map @xmath252 does _ not _ depend on @xmath2 , only on @xmath238 . to see this just note that for each @xmath86 and @xmath169 with @xmath253 , there is a @xmath11 with @xmath199 and @xmath254 . + we can therefore also write @xmath255 to indicate that it just depends on the boundary graph @xmath256 . * if there are two boundary graphs @xmath257 with number of edges @xmath258 , then one has @xmath259 where the embedding map @xmath260 is the pull back of the projection maps @xmath261 on the boundary ( [ eq : appendixcoarsegraining2 ] ) , i.e. @xmath262 . with the boundary graphs @xmath263 and the respective configuration spaces @xmath264 , forming a directed set as well , one can form the boundary configuration space @xmath265 , which is a standard construction in loop quantum gravity @xcite , going along the same lines as the construction of @xmath30 in chapter [ sec : holonomysfm ] . assume we have a manifold @xmath268 with @xmath269 a compact , connected @xmath270-manifold , interpreted as `` space '' . then @xmath271 defines a regular borel measure on @xmath265 , hence a linear functional on a dense subset of the kinematical boundary hilbert space @xmath272 , the states of which have an interpretation as quantized @xmath270-geometries @xcite . thus , @xmath271 defines a generalized boundary state , which plays the role of everett s `` wave function of the universe '' @xcite . + another case is that of a manifold @xmath0 with topology @xmath273 $ ] , @xmath274 again playing the role of `` space '' as a cauchy hypersurface , as in figure [ fig : figure12 ] . then the boundary is @xmath275 , and @xmath271 is a positive linear form on @xmath276 , which can in turn be used to define a sesquilinear form on a dense subset of the kinematical boundary hilbert space via @xmath277\end{aligned}\ ] ] with @xmath278 . here the `` complex conjugate '' of a state is the one obtained by reversing the orientation of all edges on the boundary ( but not of the vertices ) . this defines a rigging map @xmath279 in the sense of refined algebraic quantization @xcite , and if the continuum path integral measure is such that @xmath280=\overline{\eta[\overline{\phi}\otimes \psi]}$ ] , then this defines a physical inner product . note that this is automatically the case if @xmath38 is invariant under all diffeomorphisms of @xmath0 , since these include the ones switching the two boundaries . if should be noted that if @xmath38 is invariant under @xmath281 , then @xmath271 is invariant _ only _ under those diffeomorphisms of the boundary which can be extended to a diffeomorphism on all of @xmath0 . in particular , the physical inner product might not necessarily be invariant under large diffeomorphisms of @xmath274 , only under those which are in the connected component of the identity . this in particular suggests that the elements of @xmath282 should act as unitary operators on the kinematical hilbert space , but not as gauge transformations ( see e.g the discussion in @xcite ) . we should note that it is strictly necessary in this approach to allow for several faces to meet in one boundary edge , in order to have the group @xmath281 act on the space of @xmath1-complexes _ and _ the @xmath1-complexes forming a partially ordered set . so allowing for more than one - valent edges on the boundary is crucial to determine the action of the diffeomorphism group on the bulk and the boundary hilbert space the correct way . for those two - complexes @xmath2 in which all boundary edges @xmath5 are part of the boundary of precisely one face @xmath7 , the condition of gauge - invariance , which is a sensible requirement for the measures @xmath51 , but not strictly necessary , will severely restrict the interplay between @xmath247 and @xmath17 . in most spin foam models , these two will actually be set equal , i.e. the measure will necessarily contain a term of the form @xmath283 , see e.g. @xcite . also , one might want to construct the measures @xmath51 in such a way , that some kind of cobordism functoriality as in topological field theories should hold , i.e. for manifolds @xmath284 , @xmath285 glued together along some common boundary to satisfy @xmath286 it should be noted that an equation like ( [ eq : appendixfunctoriality ] ) will only hold in this strong form for cases in which the group @xmath12 is finite and hence the boundary hilbert space associated to one graph finite dimensional . indeed , then one has arrived at the axioms for a tqft in the sense of atiyah @xcite , which have the finite - dimensionality of their physical hilbert spaces as an a posteriori consequence . in general , @xmath271 will be a sesquilinear form rather than an operator , for which an equation of the type of ( [ eq : appendixfunctoriality ] ) can not even be formulated . j. w. barrett and l. crane , j. math . * 39 * ( 1998 ) 3296 [ gr - qc/9709028 ] . j. engle , e. livine , r. pereira and c. rovelli , nucl . b * 799 * ( 2008 ) 136 [ arxiv:0711.0146 [ gr - qc ] ] . l. freidel and k. krasnov , class . * 25 * ( 2008 ) 125018 [ arxiv:0708.1595 [ gr - qc ] ] . j. engle , m. han and t. thiemann , class . * 27 * ( 2010 ) 245014 [ arxiv:0911.3433 [ gr - qc ] ] . a. baratin and d. oriti , phys . d * 85 * ( 2012 ) 044003 [ arxiv:1111.5842 [ hep - th ] ] . e. r. livine and s. speziale , phys . d * 76 * ( 2007 ) 084028 [ arxiv:0705.0674 [ gr - qc ] ] . l. freidel and s. speziale , phys . d * 82 * ( 2010 ) 084040 [ arxiv:1001.2748 [ gr - qc ] ] . w. kaminski , m. kisielowski and j. lewandowski , class . * 27 * ( 2010 ) 095006 [ erratum - ibid . * 29 * ( 2012 ) 049502 ] [ arxiv:0909.0939 [ gr - qc ] ] . b. bahr , f. hellmann , w. kaminski , m. kisielowski and j. lewandowski , class . quant . * 28 * ( 2011 ) 105003 [ arxiv:1010.4787 [ gr - qc ] ] . b. bahr , b. dittrich , f. hellmann and w. kaminski , phys . d * 87 * ( 2013 ) 044048 [ arxiv:1208.3388 [ gr - qc ] ] . d. oriti , in * oriti , d. ( ed . ) : approaches to quantum gravity * 310 - 331 [ gr - qc/0607032 ] . r. gurau and v. rivasseau , europhys . * 95 * ( 2011 ) 50004 [ arxiv:1101.4182 [ gr - qc ] ] . b. dittrich , new j. phys . * 14 * ( 2012 ) 123004 [ arxiv:1205.6127 [ gr - qc ] ] . b. dittrich , m. martn - benito and e. schnetter , new j. phys . * 15 * ( 2013 ) 103004 [ arxiv:1306.2987 [ gr - qc ] ] . b. dittrich , m. martin - benito and s. steinhaus , arxiv:1312.0905 [ gr - qc ] . r. oeckl , nucl . b * 657 * ( 2003 ) 107 [ gr - qc/0212047 ] . e. manrique , r. oeckl , a. weber and j. a. zapata , class . * 23 * ( 2006 ) 3393 [ hep - th/0511222 ] . c. rovelli and m. smerlak , class . * 29 * ( 2012 ) 055004 [ arxiv:1010.5437 [ gr - qc ] ] . b. dittrich , arxiv:0810.3594 [ gr - qc ] . b. bahr and b. dittrich , class . * 26 * ( 2009 ) 225011 [ arxiv:0905.1670 [ gr - qc ] ] . b. bahr , r. gambini and j. pullin , sigma * 8 * ( 2012 ) 002 [ arxiv:1111.1879 [ gr - qc ] ] . b. bahr and b. dittrich , phys . d * 80 * ( 2009 ) 124030 [ arxiv:0907.4323 [ gr - qc ] ] . b. bahr and b. dittrich , arxiv:0909.5688 [ gr - qc ] . b. bahr , b. dittrich and s. steinhaus , phys . d * 83 * ( 2011 ) 105026 [ arxiv:1101.4775 [ gr - qc ] ] . c. rovelli , arxiv:1107.2310 [ hep - lat ] . b. dittrich and s. steinhaus , phys . d * 85 * ( 2012 ) 044032 [ arxiv:1110.6866 [ gr - qc ] ] . b. dittrich , pos qgqgs * 2011 * ( 2011 ) 012 [ arxiv:1201.3840 [ gr - qc ] ] . j. w. barrett , j. math . phys . * 36 * ( 1995 ) 6161 [ gr - qc/9506070 ] . h. pfeiffer , gr - qc/0404088 . b. dittrich , w. kaminski and s. steinhaus , arxiv:1404.5288 [ gr - qc ] . m. reuter and f. saueressig , new j. phys . * 14 * ( 2012 ) 055022 [ arxiv:1202.2274 [ hep - th ] ] . j. engle and r. pereira , phys . d * 79 * ( 2009 ) 084034 [ arxiv:0805.4696 [ gr - qc ] ] . w. kaminski , arxiv:1010.5384 [ gr - qc ] . w. j. fairbairn and c. meusburger , pos qgqgs * 2011 * ( 2011 ) 017 [ arxiv:1112.2511 [ gr - qc ] ] . m. han , j. math . * 52 * ( 2011 ) 072501 [ arxiv:1012.4216 [ gr - qc ] ] . y. yamasaki `` measures on infinite dimensional spaces '' , world scientific publishing 1985 j. b. hartle and s. w. hawking , phys . d * 28 * ( 1983 ) 2960 . d. marolf , gr - qc/9508015 . d. giulini and d. marolf , class . * 16 * ( 1999 ) 2479 [ gr - qc/9812024 ] . d. giulini , in * fauser , b. ( ed . ) : quantum gravity * 161 - 201 [ math - ph/0606066 ] . b. bahr , j. phys . conf . ser . * 360 * ( 2012 ) 012042 [ arxiv:1112.3567 [ gr - qc ] ] . m. atiyah , inst . hautes etudes sci . * 68 * ( 1989 ) 175 . |
let be a real polynomial of degree , and be an oval contained in the level set . we study complete abelian integrals of the form where are real and is a maximal open interval on which a continuous family of ovals exists . we show that the-dimensional real vector space of these integrals is not chebyshev in general : for any , there are hyperelliptic hamiltonians and continuous families of ovals , , such that the abelian integral can have at least - 1 $ ] zeros in . our main result is theorem [ main ] in which we show that when , exceptional families of ovals exist , such that the corresponding vector space is still chebyshev . | math0211386 | take real polynomials @xmath18 $ ] and let @xmath19 , @xmath13 , be a continuous family of ovals . for sufficiently small @xmath20 and generic @xmath21 the limit cycles of the perturbed plane hamiltonian system @xmath22 which tend to certain ovals from the continuous family when @xmath23 , are in one - to - one correspondence with the zeros of the complete abelian integral @xmath24 for this reason the problem of finding the zeros of @xmath14 in terms of the degrees of @xmath25 was called by arnold @xcite the _ weakened 16th hilbert problem _ " ( compare to hilbert @xcite , see also arnold @xcite ) . note that the level sets @xmath26 will contain in general several continuous families of ovals which need be considered separately . it follows from the varchenko - khovanskii theorem that the number of the zeros of @xmath14 is bounded by a constant @xmath27 , uniformly in all @xmath25 , such that @xmath28 , @xmath29 , @xmath30 . in the so called hyperelliptic case " ( @xmath2 ) it was proved by novikov and yakovenko@xcite that there exists an algorithm producing a function @xmath31 such that @xmath32 , which is given by a tower function ( an iterated exponent ) of height at least five . their proof is based on the analytic properties of a suitable picard - fuchs system satisfied by abelian integrals ( including the magnitude of the coefficients of the system and some restrictions on its monodromy group ) . the progress in solving the weakened 16th hilbert problem ( finding the _ exact _ number of the zeros of @xmath14 ) concerned so far the elliptic case " only ( the complex algebraic curve @xmath4 is of genus at most one ) . it was proved for instance that in several cases , the vector space @xmath33 of abelian integrals of degree @xmath34 polynomials along the ovals of @xmath11 , obeys the so called _ chebyshev property _ ( the number of the zeros of each integral is smaller than the dimension of the vector space @xmath33 ) , see @xcite . in this relation arnold asked in ( * ? ? ? * the 7th problem ) whether the @xmath9-dimensional vector space of abelian integrals @xmath35 where @xmath36>1 , $ ] is chebyshev . in an attempt to solve this problem givental @xcite obtained a non - oscillation theorem for lagrangian planes of the picard - fuchs system satisfied by the abelian integrals @xmath37 . he used the fact that every picard - fuchs system has a hamiltonian ( or poisson ) structure . he failed , however , to produce bounds for the zeros of the abelian integrals . in the present paper we try to explore the algebro - geometrical properties of the abelian integrals , without using the picard - fuchs system . our main result is theorem [ main ] in which we show that when @xmath17 and @xmath38 , exceptional families of ovals @xmath8 , @xmath39 , exist such that every abelian integral of the form @xmath40 has at most one simple zero on the interval @xmath16 . at a first sight this seems to be an easy observation ( equivalent to the claim that @xmath41 is monotonous on @xmath16 ) . this is , however , the first result of such a kind for non - elliptic curves @xmath4 . the proof uses the riemann bilinear relations on differentials of the first kind together with the fact that a jacobian variety with its polarization can not be a direct product of principally polarized abelian varieties . our arguments can be adapted to other situations ( for instance when the degree of @xmath0 is @xmath42 ) , but we shall not do this here . we give also a negative answer ( proposition [ nocheb ] ) to the initial question posed by arnold : it turns out that there exist abelian integrals of the form ( [ question ] ) with exactly @xmath15 - 1 $ ] zeros in a neighborhood of the origin . in this section we prove that in general , the integral ( [ question ] ) does not belong to a chebyshev space . consider first the particular case when @xmath43 . let @xmath44 where @xmath45 , @xmath46 , @xmath47 , and denote by @xmath48 a continuous family of cycles vanishing at the origin as @xmath49 tends to zero . [ cyclicity ] the abelian integral @xmath50 can have a zero at the origin of multiplicity at most two . moreover , for fixed @xmath51 , @xmath52 , there exist a sufficiently small @xmath53 and @xmath54 , @xmath55 in a small neighborhood of zero , such that @xmath56 has exactly two simple zeros in the interval @xmath57 . the above shows that the real vector space generated by the functions @xmath58 , @xmath59 is not chebyshev . recall that a real vector space @xmath60 of functions defined on some interval @xmath61 is said to be chebyshev , provided that each @xmath62 has at most @xmath63 zeros ( counted with multiplicity ) on @xmath61 , and chebyshev with accuracy @xmath64 , if each @xmath62 has at most @xmath65 zeros there . * proof of proposition [ cyclicity ] . * for a small @xmath66 , denote @xmath67 . then an easy calculation yields the inverse transformation @xmath68 therefore , for small positive @xmath49 , @xmath69 \varphi'(x)}{y}dx}\\[5 mm ] { \displaystyle = \int_{y^2+x^2=t } \frac{\alpha_0[1+o(x^2 ) ] - \alpha_1[\frac32a_1x^2 + \frac52(\frac{21}{8}a_1 ^ 3-\frac72a_1a_2+a_3)x^4 + o(x^6)]}{y}dx}\\[5 mm ] = 2\pi\alpha_0[1+o(t)]-2\pi\alpha_1[\frac34a_1t+ \frac{15}{16}(\frac{21}{8}a_1 ^ 3- \frac72a_1a_2+a_3)t^2 + o(t^3 ) ] . \end{array}\ ] ] when @xmath70 , the function @xmath56 has a double zero at the origin since @xmath52 . taking @xmath71 , @xmath72 and @xmath73 , one finishes the proof of the proposition . @xmath74 clearly , the above approach could be applied to hyperelliptic hamiltonians of any degree that have a center . denote by @xmath75 the set of hamiltonians @xmath76 , where @xmath77 , @xmath78 . denote by @xmath79\geq 2 $ ] the genus of the hyperelliptic curve . given @xmath80 , let @xmath81 , @xmath82 , be an oval from the continuous family of ovals surrounding the center at the origin . consider the vector space @xmath83 formed by the abelian integrals @xmath84,\quad deg\ , g < g.\ ] ] let us point out that @xmath85 in general , but @xmath86 $ ] provided that @xmath11 contains no odd degree monomials . in the proposition below , we give a partial answer to the question raised by v. arnoldd in @xcite . [ nocheb ] there exist @xmath87 and a polynomial @xmath88 such that the corresponding integral @xmath89 has exactly @xmath15 - 1 $ ] small positive simple zeroes . thus , such a space @xmath83 could be chebyshev with accuracy at least @xmath90 $ ] in @xmath91 . given an integer @xmath92 and @xmath87 , denote @xmath93 the proof of the above proposition is based on the following lemma . [ int ] the following asymptotic expansions hold for small positive @xmath49 : _ ( i)_@xmath94,\;\;\mbox{\it where } \;\ ; c_k=2\pi\frac{(2k-1)!!}{(2k)!!},$ ] _ ( ii)_@xmath95,$ ] therefore the coefficients in the last expansion belong to the ideal generated by the odd - numbered parameters @xmath96 . * take as above @xmath97 , then the inverse transformation @xmath98 for small @xmath99 reads @xmath100 , where the coefficients @xmath101 are polynomials . it is easy to see that if one attaches a weight @xmath102 to the coefficient @xmath103 , then @xmath104 is a polynomial of weight @xmath105 . indeed , directly from the definition of @xmath106 it follows that @xmath107 applying this identity to @xmath108 we get @xmath109 . to calculate the coefficient at @xmath103 in @xmath110 , we take @xmath111 , and easily obtain that it equals @xmath112 . therefore , one deduces that @xmath113 where @xmath114 belongs to the polynomial ideal generated by @xmath115 . in case ( i ) , one immediately obtains that for small positive @xmath49 , @xmath116 to calculate the integral in case ( ii ) , we write @xmath117 in the form @xmath118 @xmath119\right)\ ] ] @xmath120\right)\ ] ] @xmath121 + \phi(x)\ ] ] where @xmath122 is a series in @xmath99 containing only even order powers . therefore , @xmath123 @xmath124.\;\;\box\ ] ] * proof of proposition [ nocheb ] . * denote @xmath125 $ ] and take in ( [ g ] ) @xmath126 then by lemma [ int ] , @xmath127+{\textstyle\frac12 } \sum_{j=1}^g a_{2j-1}(2m+2j-1)t^{m+j-1}[c_{m+j-1}+o(t)].\ ] ] now , freeze the even - numbered coefficients in @xmath11 and then choose @xmath128 and @xmath129 so that @xmath130 , @xmath131 , @xmath132 and @xmath133 such a choice guarantees that @xmath56 will have exactly @xmath134 - 1 $ ] zeroes in a certain small interval @xmath135 . @xmath74 this result raises the problem to describe the hyperelliptic hamiltonians @xmath11 and the continuous families of ovals @xmath8 for which the space of integrals ( [ question ] ) obeys the chebyshev property . in the rest of the paper , we concentrate our efforts on the simplest case when @xmath38 . we begin with some preliminaries . we consider the hyperelliptic hamiltonian @xmath136 where @xmath137 $ ] and @xmath38 . assume that a certain level set @xmath4 contains an oval ( compact closed smooth curve without critical points ) . as the poincar index of the respective hamiltonian vector field @xmath138 is zero , then @xmath138 has at least two different real critical points . without any loss of generality , we can place all the four critical points at @xmath139 , @xmath140 , @xmath141 and @xmath142 , where either @xmath143 ( the real case ) or @xmath144 ( the complex case ) . finally , we can assume that the coefficient at @xmath145 is positive . then after rescaling of the @xmath146 variable , we come to the normal form @xmath147 and the hamiltonian flow @xmath138 takes the form @xmath148 until the end of the paper , we will use this normal form of @xmath11 only . clearly , the origin is a hyperbolic saddle if @xmath149 , and @xmath142 is a nondegenerate center provided that @xmath150 . in addition , when @xmath151 in the real case , @xmath140 is a center and @xmath152 is a saddle . denote by @xmath153 and @xmath154 the corresponding critical levels of @xmath11 . one has @xmath155 h_\lambda=-\frac{\lambda^3}{60}(3\lambda^2 - 5\lambda\mu-5\lambda+10\mu),\\[2 mm ] h_\mu=-\frac{\mu^3}{60}(3\mu^2 - 5\lambda\mu-5\mu+10\lambda ) . \end{array}\ ] ] if @xmath151 , we have @xmath156 , @xmath157 , @xmath158 . the bifurcation diagram @xmath159 in the real case consists of the boundary of the triangle @xmath160 and the curve @xmath161 having an equation @xmath162 see fig . [ bdrc ] . there are two open components in @xmath163 giving the two generic cases . namely , for parameters @xmath164 above the curve @xmath165 , @xmath138 has three period annuli , and below this curve @xmath138 has just two period annuli . there are also five codimension - one and four codimension - two degenerate ( nongeneric ) cases corresponding to the parameters on the bifurcation diagram @xmath159 . in the complex case , there is a unique period annulus around @xmath142 which terminates at the saddle - loop through @xmath139 and the phase portraits for all @xmath166 are topologically equivalent . we will denote the period annuli ( continuous families of ovals ) around @xmath142 , @xmath167 and the eight loop by @xmath168 , @xmath169 and @xmath170 , respectively . below we describe the dynkin diagram of the real polynomial @xmath171 when the critical levels are distinct . recall that it is a graph defined in the same way as the dynkin diagram ( or d - diagram ) of a germ of an analytical function with an isolated singularity , see @xcite . the vertices of the diagram are in one - to - one correspondence with the vanishing cycles of the polynomial . two vertices are connected by an edge if the intersection number of the cycles is not zero . to an edge one associates also a label equal to the intersection number , but it will be omitted here because is of no importance for us . the vanishing cycles vanish " along suitable paths , and hence the dynkin diagram depends also on the family of paths . the precise definition in our situation is as follows . consider first the real case @xmath151 . there are five possible distributions of the critical values as follows : @xmath172 for definiteness , assume that @xmath173 . denote @xmath174 and let @xmath175 \rightarrow \{h\in \c : im\,h \geq 0\}\ ] ] be continuous paths connecting some fixed regular value @xmath176 to @xmath177 , @xmath154 , @xmath178 , @xmath179 respectively , as shown on fig . [ realdynk ] . the dynkin diagram of @xmath11 with respect to the above paths is a graph with four vertices corresponding to the four families of cycles vanishing along the paths @xmath180 , @xmath181 , @xmath182 , @xmath183 , which we denote by @xmath184 , @xmath185 , @xmath186 , @xmath187 respectively . two vertices are connected by an edge if the intersection number of the cycles for @xmath188 is not zero ( in which case it is @xmath189 and may be supposed equal to one ) . the computation of the dynkin diagram of a real polynomial with real critical points only is well known , see for instance @xcite . it yields the diagram shown on fig . [ realdynk ] . we note that the computation depends only on the distribution of the four critical points @xmath151 and not on the distribution of the respective critical values ( hence , the dynkin diagram is the same for all possible arrangements of the critical values with corresponding paths @xmath180 , @xmath181 , @xmath182 , @xmath183 ) . consider now the complex case when @xmath190 . the method we use is a deformation of @xmath11 to a polynomial having real critical points . denote @xmath191 and let @xmath192 \rightarrow { \cal d}\ ] ] be continuous paths connecting some fixed regular value @xmath176 to @xmath177 , @xmath154 , @xmath178 , @xmath179 respectively , as shown on fig . [ dynkin1 ] . the dynkin diagram related to these paths is defined as above . consider in the complex plane the hyperbola @xmath193 defined by the equation @xmath194 and drawn on fig . [ imconj ] . ( 30,15)(0,0 ) ( 0,5 ) ( 15,15)(0,0 ) ( 0,0 ) ( 0,0.5)@xmath154 ( 3,0 ) ( 3,0.5)@xmath179 ( 6,0 ) ( 6,0.5)@xmath178 ( 9,0 ) ( 9,0.5)@xmath177 ( 6,0 ) ( 0,0)(0,5)(6,6 ) ( 3,0)(3,2)(6,6 ) ( 6,0)(5,3)(6,6 ) ( 9,0)(7.5,2.5)(6,6 ) ( 6,6 ) ( 6.5,6)@xmath195 ( 20,5 ) ( 15,15)(0,0 ) ( 0,0)(1,0)3 ( 0,0 ) ( 3,0)(1,0)3 ( 3,0 ) ( 6,0)(1,0)3 ( 6,0 ) ( 9,0 ) ( 0,.5)@xmath196 ( 2.5,.5)@xmath197 ( 5.5,.5)@xmath198 ( 8.5,.5)@xmath199 ( 7,0 ) ( i ) ( 22,0 ) ( ii ) suppose for definiteness that @xmath200 lies inside its left branch . first , by continuous deformation of @xmath201 such that in the course of the deformation @xmath202 remains in the left branch of ( [ hyperbola ] ) , we may achieve that @xmath203 . this , combined with @xmath204 already implies that * the intersection number of @xmath205 and @xmath206 is equal to @xmath189 ; * the intersection number of @xmath207 , @xmath208 with @xmath209 is zero ; * the intersection number of @xmath207 and @xmath208 is equal to @xmath189 ; * the intersection number of either @xmath207 or @xmath210 with @xmath206 is equal to @xmath189 . finally , we may use the real structure of the complex algebraic curve @xmath211 for @xmath212 . this structure is defined by an antiholomorphic involution @xmath213 on the curve which is the complex conjugation . it induces on its hand an involution @xmath214 of @xmath215 , @xmath216 . it is easy to see that for @xmath217 we have @xmath218 this combined with the identity @xmath219 shows that @xmath220 this implies the dynkin diagram as shown on fig . [ dynkin1](ii ) ) . the remaining cases shown there are studied in a similar way . ( 30,15)(0,0 ) ( 0,5 ) ( 15,15)(0,0 ) ( 0,0 ) ( 0,0.5)@xmath154 ( 3,0 ) ( 3,0.5)@xmath177 ( 1.5,1.5 ) ( 1.5,2)@xmath178 ( 1.5,-1.5 ) ( 2,-1.5)@xmath179 ( 3,0 ) ( 0,0)(0,5)(6,6 ) ( 3,0)(3,2)(6,6 ) ( 1.5,1.5)(4,5)(6,6 ) ( 1.5,-1.5)(3,3.5)(6,6 ) ( 6,6 ) ( 6.5,6)@xmath195 ( 20,5 ) ( 15,15)(0,0 ) ( 0,0)(1,0)3 ( 0,0 ) ( 3,0)(1,1)2.1213 ( 3,0 ) ( 3,0)(1,-1)2.1213 ( 5.1213,2.1213 ) ( 5.1213,-2.1213 ) ( 2.5,0.5)@xmath199 ( 0,0.5)@xmath196 ( 5.6213,2.1213)@xmath197 ( 5.6213,-2.1213)@xmath198 ( 5.1213,-2.1213)(0,1)4.2426 ( 7,0 ) ( i ) ( 22,0 ) ( ii ) ( 30,15)(0,0 ) ( 3,7 ) ( 15,15)(0,0 ) ( 0,0)(1,0)3 ( 0,0 ) ( 3,0)(1,1)2.1213 ( 3,0 ) ( 3,0)(1,-1)2.1213 ( 5.1213,2.1213 ) ( 5.1213,-2.1213 ) ( 2.5,0.5)@xmath196 ( 0,0.5)@xmath199 ( 5.6213,2.1213)@xmath197 ( 5.6213,-2.1213)@xmath198 ( 5.1213,-2.1213)(0,1)4.2426 ( 20,5 ) ( 15,15)(0,0 ) ( 0,0)(1,0)4 ( 0,0 ) ( 4,0)(0,1)4 ( 4,0 ) ( 0,4)(1,0)4 ( 0,0)(0,1)4 ( 4,4 ) ( 0,4 ) ( -.5,- 0.8)@xmath199 ( 4.3,-.8)@xmath221 ( -.5,4.5)@xmath198 ( 4.3,4.5)@xmath196 ( 0,0)(1,1)4 ( 7,0 ) ( iii ) ( 22,0 ) ( iv ) we finish this section by introducing a property of the continuous families of ovals which is dependent on the structure of the related dynkin diagram . suppose that the real polynomial @xmath44 has distinct critical values . let @xmath222 be a continuous family of ovals defined on a maximal open interval @xmath223 , where for @xmath224 the oval degenerates to a point @xmath225 which is a center and for @xmath226 the oval becomes a homoclinic loop of the hamiltonian system @xmath227 . the family @xmath228 represents a continuous family of cycles vanishing at the center @xmath225 . to formulate our main result we shall need the following [ excd ] _ it is said that @xmath228 is _ an exceptional family of ovals _ , provided that for every polynomial one - form @xmath229 the abelian integral @xmath230 allows an analytic continuation in the sector @xmath231 for some strictly positive @xmath20 . the corresponding family of vanishing cycles represented by @xmath232 is called _ an exceptional family of vanishing cycles . _ _ the above definition has in fact a geometric nature . indeed , @xmath14 has an analytic continuation in the sector @xmath233 if and only if @xmath232 has an appropriate intersection number with each of the other families of vanishing cycles . for instance , if @xmath200 , @xmath234 are real and @xmath235 , then the family of cycles @xmath185 is exceptional , because it has zero intersection number with @xmath236 . on the contrary , the family @xmath237 is not exceptional , because it has a non - zero intersection number with @xmath236 . the remaining possible distributions of the critical values , as well as the case @xmath238 , are studied in the same way . we summarize all this in the following [ exceptional ] suppose that the hamiltonian @xmath171 is taken in a normal form @xmath239 and has four distinct critical values . then : @xmath240 in the real case @xmath241 , the continuous family of ovals @xmath168 surrounding the center @xmath142 is exceptional , and the continuous family of ovals @xmath169 surrounding the center @xmath167 is not exceptional . @xmath242 in the complex case @xmath243 , the continuous family of ovals @xmath168 is exceptional if and only if @xmath200 lies inside the left branch of the hyperbola @xmath193 @xmath244see @xmath245 and _ fig . assume that the level curve @xmath4 contains an oval @xmath3 and consider the integral ( oriented along with the vector field ( [ system ] ) ) @xmath246 which is the derivative of the area inside @xmath3 and hence determines the period of the periodic orbit lying on @xmath3 . denote by @xmath247 and by @xmath248 the minimal and the maximal values of @xmath66 for @xmath249 . these are among the real solutions of the equation @xmath250 . then one can express @xmath251 in the form @xmath252 consider first the case @xmath151 and assume that @xmath3 belongs to either of the continuous families of ovals @xmath169 , @xmath168 surrounding centers @xmath167 and @xmath142 , respectively . the corresponding families of cycles vanish respectively at @xmath253 and @xmath254 . denote this value in both cases by @xmath255 and assume that the real ovals are defined in @xmath256 . finally , denote by @xmath257 the @xmath66-coordinate of the corresponding center . we keep the same notation in the complex case and in the cases when either @xmath258 or @xmath259 or @xmath260 . then @xmath257 is the @xmath66-coordinate of the unique nondegenerate center ( as long as @xmath261 ) . the integral @xmath251 has an analytic continuation for @xmath262 . we wish to prove that @xmath263 in @xmath264 . the precise statement is as follows . [ nozero ] assume that the parameters @xmath265 in @xmath239 are either real or lie inside the hyperbola @xmath193 . then the integral @xmath251 related to any period annulus around a nondegenerate center of @xmath266 has an analytic continuation in @xmath267 and takes positive values there . * proof . * the claim that @xmath251 has an analytic continuation in the interval @xmath267 is obvious in the complex case ( @xmath268 ) . in the real case this follows from the dynkin diagram shown on fig . [ realdynk ] . for instance , if @xmath235 as on fig . [ realdynk ] , and if @xmath269 , @xmath270 , then the claim is obvious . if @xmath269 , @xmath271 , then the claim is true again because the intersection number of @xmath272 and @xmath273 is equal to zero . the remaining four configurations of critical values and period annuli are examined in the same way ( with the help of fig . [ realdynk ] ) . note , however , that the integral @xmath251 corresponding to the period annulus @xmath170 ( in this case @xmath274 ) has no analytic continuation on the real axis outside @xmath16 . this is so because the singularities in both @xmath275 and @xmath276 are unavoidable . the same conclusion holds for some of the degenerate cases as well . next we shall prove the positivity of the period integral @xmath251 on @xmath267 . we first choose a proper formula to present @xmath251 when @xmath277 . take @xmath278 , @xmath262 and consider the polynomial @xmath279 . by taylor s formula , @xmath280\\[2 mm ] \equiv&\!\!\!h - q(x , y)-iyr(x , y)\end{array}\ ] ] given @xmath262 , consider the equation @xmath281 and denote by @xmath282 , @xmath283 , the two branches of the algebraic function defined by @xmath284 which satisfy @xmath285 . these functions are unique since @xmath286 for @xmath287 . for @xmath288 , one has @xmath289 and @xmath290 where @xmath291 are as above . for @xmath277 , one has @xmath292 . for a similarity in notation , we put @xmath293 and @xmath294 where @xmath295 . denote by @xmath296 the curve in the complex plane connecting the points @xmath247 , @xmath248 along @xmath297 , if @xmath277 , and along the real line @xmath298 , if @xmath288 . then the integral @xmath251 is expressed as @xmath299 let @xmath277 . we will establish below that @xmath300 can be used as a local coordinate on @xmath296 . thus , @xmath301dy$ ] where @xmath302 . denote for short @xmath303 and write @xmath304 . replacing @xmath146 with @xmath305 in the first integral , we obtain @xmath306 below we study the curve @xmath307 and establish that on this curve the real - valued function @xmath308 is positive for all @xmath277 . recall that @xmath309 has an equation @xmath310 . we will study how this algebraic curve changes when varying the parameters @xmath265 . it turns out that a local analysis can be used to determine the global behavior of @xmath309 . to begin with , we observe that the poincar index of any sufficiently big circle @xmath311 subject to the vector field @xmath312 is @xmath313 . the critical points of @xmath314 are determined from @xmath315 , namely \(i ) @xmath298 , @xmath316 ; \(ii ) @xmath317 , @xmath318 . it is easy to verify that the hessian @xmath319 is negative at any ( real ) nondegenerate critical point @xmath320 . indeed , it equals @xmath321 and @xmath322 , respectively in cases ( i ) and ( ii ) . hence , @xmath314 has in general three nondegenerate saddles and when two of the saddles collide , the resulting degenerate critical point has only hyperbolic sectors around . in particular , ( ii ) has no solutions for @xmath151 , since all the three critical points are then given by ( i ) . also , in this case the curve @xmath309 is free of critical points because the value of @xmath314 at the saddle @xmath323 is @xmath324 . in general , one can prove that @xmath309 has no critical points outside the real line . indeed , taking a critical point @xmath325 with @xmath326 , one has @xmath327 unless @xmath328 or @xmath329 ( which may happen in the complex case ) . however , by ( ii ) , @xmath330 is a critical point if and only if @xmath200 lies on the hyperbola @xmath193 , which is not our case . we have proved that the only critical points @xmath309 may have are @xmath139 , @xmath141 and @xmath142 provided that @xmath258 , @xmath331 , @xmath260 , respectively . as a consequence , the branches of the level curve @xmath309 do not intersect outside the real line and all they do escape to infinity ( in the directions of @xmath66 and @xmath146 altogether , as shown by the asymptotics at infinity ) . this is because any closed compact curve necessarily surrounds a critical point having an index @xmath332 and @xmath333 has no such points . assume first that @xmath151 . then the four branches of @xmath309 have no common points at all . this implies that the branches through @xmath139 and @xmath167 go to @xmath334 in the @xmath66-direction while the branches through @xmath141 and @xmath142 go to @xmath335 , see fig . [ curve ] . when running any of the branches , the @xmath146 variable changes in a monotone way . ( otherwise , there would exist a horizontal line @xmath336 having at least 6 intersections with @xmath309 which is impossible . ) finally , denote by @xmath337 any of @xmath338 . the local equation of @xmath309 near the point @xmath323 is @xmath339(x-\xi)+o((x-\xi)^2)$ ] where the coefficient is negative for @xmath340 , positive for @xmath341 and changes sign if @xmath342 . this latter is because @xmath343 and @xmath344 change sign in @xmath345 . on the @xmath346 curve where @xmath347 vanishes , the local equation of @xmath309 becomes @xmath348 for @xmath342 . summing up , we obtain the form of the curve @xmath309 as given in fig . [ curve ] . the arrows indicate the direction of the hamiltonian vector field @xmath349 . it is easy to see that @xmath350 does not change sign when @xmath351 . as @xmath352 , it suffices to show that @xmath353 . we first note that @xmath354 on the branch through @xmath355 we consider . then , by cauchy - riemann equations , @xmath356 and we obtain on @xmath357 that @xmath358 this implies that @xmath359 for @xmath351 and hence by ( [ y ] ) , @xmath360 for @xmath262 . the same analysis is applicable when @xmath258 or @xmath259 or @xmath260 , the unique difference is that on fig . [ curve ] the corresponding branches touch each other at @xmath298 . let us turn now to the complex case . apart of the real case , the branches through @xmath139 and @xmath142 ( let us denote them by @xmath361 and @xmath362 ) may bifurcate significantly . this is because they are not necessarily monotone with respect to @xmath146 and may escape to either @xmath334 or @xmath335 , in the @xmath66-direction . the complex plane is divided by the hyperbola @xmath193 into three parts . in the outer domain @xmath363 , @xmath361 and @xmath362 behave as in fig . [ curve ] , but are not necessarily monotone in @xmath146 . inside the left branch of the hyperbola , @xmath361 and @xmath362 look like the branches through @xmath200 and @xmath364 on fig . [ curve ] , respectively . and inside the right branch , @xmath361 and @xmath362 behave as the branches through @xmath365 and @xmath234 on fig . [ curve ] , respectively . all this is easily verified by a simple deformation argument , with a starting point on the appropriate side of triangle @xmath345 . to avoid complications , we regard only the parameters @xmath200 , @xmath234 in the complex plane which belong to the interior of @xmath193 ( which is the case to be considered in our main theorem ) . taking @xmath366 , then @xmath367 . we are going to prove that @xmath362 is monotone with respect to @xmath146 in this domain . if so , one obtains just as in the real case that @xmath368 and the proof is complete . consider first the interior of the left branch of @xmath193 . then @xmath362 lies in the half - plane @xmath369 . to establish monotonicity , we rewrite @xmath370 as a polynomial in @xmath371 . one obtains @xmath372 assume that @xmath373 . as the first two coefficients of the polynomial are positive , then according to descartes rule of signs , ( [ descartes ] ) can have three positive roots only if the coefficient at @xmath374 is negative and the coefficient at @xmath49 is positive . however , these two conditions contradict each other . therefore each line @xmath336 can intersect @xmath362 in at most two points ( counting multiplicity ) and hence in a single point because the number of intersections should be an odd number . now take @xmath200 in the interior of the right branch of @xmath193 and denote by @xmath320 the right - most point of @xmath362 in the half - plane @xmath375 . thus , @xmath376 and @xmath377 . as above , one obtains with @xmath378 that @xmath379 as @xmath380 if @xmath381 and @xmath382 otherwise , the coefficient at @xmath374 is negative for @xmath383 . then by descartes rule , ( [ descart1 ] ) has only two positive roots and therefore the part of @xmath362 above the line @xmath384 is monotone . it remains to consider the part below it ( if @xmath385 . in this domain ( [ descart1 ] ) can have two complex roots and descartes rule does not yield the required result . we can proceed as follows . assume that the part of @xmath362 in the upper half - plane lying below @xmath384 is not monotone . denote by @xmath386 the first local maximum on @xmath362 ( starting from @xmath142 ) , thus @xmath387 and @xmath388 . consider also the branch @xmath389 of @xmath309 which is placed in the upper half - plane to the right of the line @xmath390 ( and above the line @xmath391 ) . an easy calculation yields the equations @xmath392 for the tangential lines of @xmath309 at infinity . as their common point is different from @xmath386 , there would always exist a tangential line @xmath393 to @xmath389 at a finite point , which goes through @xmath386 . however , then @xmath393 would intersect @xmath389 in at least two points and @xmath362 in at least three points , which is impossible because @xmath309 is a quartic curve . the monotonicity of @xmath362 is established . @xmath74 [ nozero1 ] assume that @xmath3 is an exceptional family of ovals according to definition @xmath394 . then the period integral @xmath395 has an analytic continuation in the complex domain @xmath396 and does not vanish there . * proof . * the proof that @xmath251 has an analytic continuation in @xmath397 follows immediately from the definition of an exceptional family of ovals . to count the zeros of the analytic function @xmath251 in @xmath397 , we apply the argument principle to the complex domain @xmath398 obtained from @xmath397 by removing the discs @xmath399 , @xmath400 , where @xmath401 is a small enough constant . for this purpose we evaluate the increase of the argument of @xmath402 when @xmath403 runs the boundary of @xmath398 in a positive direction . we consider first the real case ( @xmath404 ) . the picard - lefschetz formula implies that in a suitable complex neighborhood of @xmath405 it holds @xmath406 where @xmath407 are locally holomorphic and , as it is easily checked , @xmath408 . the behavior of @xmath402 for @xmath409 sufficiently big follows from the picard - lefschetz formula too , but the argument is more delicate . let @xmath393 be a simple closed path which makes one turn in a positive direction around all the critical values of @xmath11 . it induces a monodromy map @xmath410 it is shown for instance in @xcite that @xmath411 coincides with the operator of classical monodromy of the singularity @xmath412 ( which is the highest weighted homogeneous part of @xmath11 ) . the characteristic polynomial @xmath413 of @xmath414 is now easily computed to be equal to @xmath415 ( see @xcite ) . this shows that in a neighborhood of @xmath416 on the projective sphere the integral @xmath251 is a meromorphic function of @xmath417 . the substitution @xmath418 defines an isotopy of the regular fibers of @xmath11 and the regular fibers of its weighted homogeneous part @xmath412 @xcite . from this fact we deduce that @xmath419 for a further use we note that @xmath420 . indeed , the one - form @xmath421 is not cohomologous to zero on the regular fibers of the polynomial @xmath412 ( this follows from ( * ? ? ? * theorem 1.1 ) ) . on the other hand , the form of the characteristic polynomial of @xmath411 shows that it is irreducible over the field of rational numbers . therefore the span of @xmath422 , @xmath423 generates the whole homology group of the fiber . this implies that if @xmath424 , then the restriction of @xmath421 on the regular fibers of @xmath425 defines the zero cohomology class which is a contradiction . the above consideration shows that the decrease of the argument of @xmath402 along @xmath426 is close to @xmath427 and the increase of the argument of @xmath402 along @xmath428 is close to zero . we claim further that the imaginary part of @xmath402 along @xmath429 does not vanish . namely , denote the two determinations of @xmath402 on @xmath429 by @xmath430 . similarly , denote by @xmath431 the two determinations of the continuous family @xmath3 on @xmath429 . since @xmath432 is a real analytic function along @xmath433 , we have @xmath434 and hence @xmath435 . on the other hand @xmath3 is an exceptional family of ovals . therefore the picard - lefschetz formula implies that for @xmath436 , @xmath437 where @xmath438 is a continuous family of cycles vanishing at @xmath439 , with intersection number @xmath440 . finally , we note that after substituting @xmath146 by @xmath441 , the family of cycles @xmath438 corresponds to a family of ovals surrounding a center . moreover , in the real case the full bifurcation diagram ( consisting of the boundary of @xmath345 and the parts of the curves @xmath442 , @xmath443 , @xmath444 inside ) images onto itself . for later use we also mention that in the complex case , the interior of the hyperbola @xmath193 images onto itself . all this holds because the corresponding hamiltonians when taken in a normal form ( [ normal ] ) are related by the formula @xmath445 hence , it follows from proposition [ nozero ] that @xmath446 does not vanish along @xmath447 . therefore if @xmath448 , then @xmath449 and hence the increase of the argument of @xmath450 along @xmath451 is at most @xmath452 . summing up the above information we conclude that the increase of the argument of @xmath402 along the boundary of @xmath398 is strictly less than @xmath453 . by the argument principle , @xmath402 does not vanish in @xmath398 , and hence in @xmath397 . the complex case when @xmath454 is studied in the same way . suppose that the real polynomial @xmath44 has distinct critical values . let @xmath222 be a continuous family of ovals defined on a maximal open interval @xmath223 , where for @xmath224 the oval degenerates to a point @xmath225 which is a center and for @xmath226 the oval becomes a homoclinic loop of the hamiltonian system @xmath227 . denote @xmath455 . the following theorem is our main result . [ main ] let @xmath232 , @xmath456 be an exceptional family of ovals of @xmath239 . then the real vector space @xmath457 of abelian integrals @xmath458 is chebyshev in @xmath397 and hence in @xmath16 . * proof . * let @xmath3 where @xmath459 be an exceptional family of ovals . according to proposition [ nozero1 ] , it suffices to show that the analytic function @xmath460 has at most one zero in the domain @xmath397 . to prove this claim we apply the argument principle to @xmath461 in the complex domain @xmath398 obtained from @xmath397 by removing the discs @xmath399 , @xmath400 , where @xmath401 is a small enough constant . as in the proof of proposition [ nozero1 ] , we may check that : * if @xmath462 , then @xmath463 is meromorphic with respect to @xmath464 and @xmath465 * if @xmath466 , then @xmath467 where @xmath468 and @xmath469 are locally holomorphic . moreover , it is easily verified that @xmath470 where @xmath471 is the saddle point corresponding to the critical value @xmath472 . therefore the increase of the argument of @xmath461 along @xmath426 is close to @xmath473 and the increase of the argument of @xmath461 along @xmath474 is close to zero or strictly negative . we claim further that * the imaginary part of @xmath461 along @xmath451 does not vanish . if this claim were proved then summing up the above information we would get that the argument of @xmath461 increases along the boundary of @xmath398 by at most @xmath475 . thus the function @xmath461 can have at most one zero in @xmath398 and hence in @xmath476 . it remains to prove the above claim . recall that , since @xmath232 is an exceptional family of ovals , then the two determinations of @xmath461 over @xmath477 , namely @xmath478 , are analytic along @xmath477 . the function @xmath461 is real analytic along @xmath267 and hence @xmath479 denote by @xmath480 the two determinations of @xmath232 , @xmath481 on @xmath477 , and let @xmath482 it was shown in the proof of proposition [ nozero1 ] that @xmath483 then we obtain for each @xmath484 @xmath485 \int_{\delta^{- } ( h ) } \frac{x dx}{y } & \int_{\delta^{- } ( h ) } \frac { dx}{y } \end{array } \right ) } { |\int_{\delta^{\pm } ( h ) } \frac { dx}{y}|^{2}}.\ ] ] therefore we have to prove that the analytic function @xmath486 \int_{\delta^{- } ( h ) } \frac{x dx}{y } & \int_{\delta^{- } ( h ) } \frac { dx}{y } \end{array } \right),\quad h \in ( h_s,\infty)\ ] ] does not vanish . suppose first that @xmath487 is a critical value of @xmath11 . the affine curve @xmath488 is singular of arithmetic genus two . let @xmath489 be the corresponding completed and normalized riemann surface . as the geometric genus of @xmath489 is one and the intersection number @xmath490 equals @xmath332 , then @xmath491 and @xmath492 form a canonical homology basis of @xmath489 . we have @xmath493 where @xmath494 for @xmath495 . therefore , @xmath496 induces a holomorphic one - form on @xmath489 and @xmath421 induces a meromorphic one - form of the third kind on @xmath489 with simple poles at @xmath497 , where @xmath497 are the two distinct pre - images of the singular point @xmath498 under the canonical projection @xmath499 . the reciprocity law for differentials of the first and third kind @xcite applied to @xmath496 , @xmath421 on the elliptic curve @xmath489 gives @xmath500\\ & = & \frac{\pm 2 \pi \sqrt{-1}}{\sqrt{(x_0-x_1)(x_0-x_2)(x_0-x_3 ) } } \int_{p^-}^{p^+}\frac{dx}{\sqrt{(x - x_1)(x - x_2)(x - x_3 ) } } \end{aligned}\ ] ] where the path of integration is on the riemann surface @xmath501 of the affine elliptic curve @xmath502 and @xmath503 . the function @xmath504 is an uniformizing variable on the elliptic curve @xmath501 which shows that @xmath505 as @xmath506 , then @xmath507 . suppose now that @xmath484 is a regular value of @xmath11 . the intersection number @xmath490 equals @xmath332 and it is easy to check that we may choose a canonical homology basis @xmath508 of the lattice @xmath509 such that @xmath510 , @xmath511 . further , let @xmath512 be a normalized basis of holomorphic differentials on @xmath513 . the corresponding period lattice @xmath514 reads @xmath515 where @xmath516 we get that @xmath517 and hence @xmath518 let @xmath519 be the lattice generated by the columns of @xmath520 . the riemann bilinear relations @xcite on @xmath520 imply @xmath521 therefore if @xmath522 , then the jacobian variety @xmath523 is a direct product of the two elliptic curves @xmath524 it is well known that this is impossible ( e.g. the remark on page 49 in @xcite ) which completes the proof of theorem [ main ] . for convenience of the reader , we present a proof of the last claim . consider the riemann theta function @xmath525 and the corresponding theta divisor @xmath526 where @xmath527 by the riemann theorem , the divisor @xmath528 is isomorphic to the curve @xmath529 and in particular is irreducible . if @xmath530 , then @xmath531 and @xmath532 factorizes into a product of two elliptic theta functions @xmath533 therefore the riemann theta divisor @xmath528 is a product of the two elliptic curves ( [ curves ] ) which is a contradiction . @xmath74 we finish this paper with some general hypotheses concerning the zeros of the function @xmath534 . in the previous section we demonstrated that this linear combination belongs to a chebyshev space , provided that the continuous family of ovals @xmath8 is exceptional . in the non - exceptional case a big part of our proof still holds , provided that @xmath200 and @xmath234 are real . this fact combined with the exact local result obtained in proposition [ cyclicity ] suggests the following more general statement . * conjecture 1 . * assume that the polynomial @xmath0 has distinct real critical values . then the corresponding vector space of abelian integrals @xmath535 is chebyshev with accuracy at most one in @xmath536 . one may also ask whether the upper bound ( two ) for the number of the zeros near the center always holds in the whole interval @xmath16 and next , whether the bounds for the real zeros when @xmath13 and for the complex zeros when @xmath403 belongs to a certain complex extension of @xmath16 do coincide . it turns out that the answers of all these questions are negative . in particular , we establish below that when @xmath0 has complex critical values , then there are three zeros in @xmath16 for appropriate @xmath537 . one can try to predict the possible number of real zeros by studying the behavior of the ratio @xmath538 near the ends of the interval @xmath16 . for reader s convenience , we first recall some formulas . let us take a continuous function @xmath539 and consider the integral @xmath540 where @xmath541 for the families @xmath168 , @xmath169 and @xmath542 for the family @xmath170 . using picard - lefschetz formula or direct calculations of integrals , one can obtain the following asymptotical expansions in a neighborhood of the critical levels @xmath405 corresponding to a saddle or a cuspidal point @xmath543 : \(i ) for the saddle - loop and eight - loop cases , @xmath544 where @xmath545 is a specific positive constant independent on @xmath546 and @xmath547 is a continuous function in a neighborhood of @xmath405 . we are not going to write up the exact expression of @xmath545 since do not need it here . ( in fact , these formulas hold in a much more general context . ) applying the above expansions to @xmath461 , we get @xmath551 for the heteroclinic loop case and @xmath552 for all the remaining cases . moreover , what concerns the saddle - loops , it is evident that @xmath553 in @xmath16 and the sign of the expression depends on the position of the saddle with respect to the family of ovals . consider now the case of @xmath170 where the situation is more complicated . for @xmath276 , we obtain from ( [ normal ] ) that @xmath554 where @xmath555 and @xmath556 , @xmath557 are the left - most and the right - most points of the eight - loop ( or the cuspidal loop , as a limit case ) . by ( i)(iii ) , the function @xmath558 is continuous in a neighborhood of @xmath178 , which yields immediately that @xmath559 clearly , @xmath560 when @xmath561 , @xmath562 when @xmath563 and @xmath564 , respectively . as @xmath565 is equivalent to @xmath566 , @xmath567 is equivalent to @xmath331 and @xmath568 is equivalent to @xmath260 , this means that the curve @xmath569 connects the points @xmath139 and @xmath570 and lies in the part of @xmath345 above @xmath165 . surprisingly , although expressed in terms of incomplete elliptic integrals , it turns out that @xmath571 is a part of an algebraic curve of degree 6 . this can be verified by transforming the integrals to a standard form . from the explicit algebraic equation thus obtained it is easily seen that @xmath571 is a simple and connected curve ( we omit the details ) . the curve @xmath571 is an element of the bifurcation diagram because for @xmath403 close to @xmath178 , the function @xmath572 will change the sign when crossing @xmath571 in the parameter space . among other , the above analysis implies that for @xmath331 , the function @xmath573 can have three zeros in @xmath574 . hence , the same fact remains true for @xmath168 and the respective @xmath575 in the complex case , at least for parameters @xmath366 with @xmath576 and @xmath577 small . based on the behavior of @xmath578 near the other endpoint which corresponds to a center ( see below ) , we conjecture that the domain in the parameter space @xmath537 where the equation @xmath579 can have three real zeros , is surrounded by a curve on which they join into a triple zero . looking at the bifurcation diagram again , let us denote by @xmath580 the curve @xmath581 @xmath582 further , let us denote by @xmath583 the ( hypothetical ) curve @xmath584 we believe that @xmath583 is a simple closed curve in the complex plane going through @xmath139 and @xmath142 . on @xmath583 , a bifurcation of a triple real zero into a simple one occurs . next , in the real case , denote by @xmath585 the part of @xmath345 placed between the curves @xmath580 and @xmath165 , by @xmath586 the part of @xmath345 placed between the curve @xmath571 and the line @xmath331 and finally , in the complex case , denote by @xmath587 the unit disk inside @xmath580 and by @xmath588 the interior of @xmath583 ( see figures [ imconj ] and [ reconj ] ) . ( 25,15)(5,0 ) ( 2,8 ) ( 5,0)(1,0)20 ( 12,-7)(0,1)15 ( 9,-6,)(14.96,0)(9,6 ) ( 4,0)(15,-6,)(9.04,0)(15,6 ) ( 12,0 ) ( 0,0)(0,6)(2,6 ) ( 0,0)(0,-6)(2,-6 ) ( 2,6)(4,6)(4,0 ) ( 2,-6)(4,-6)(4,0 ) ( 16,0 ) ( 0,0)(0,2)(2,2 ) ( 0,0)(0,-2)(2,-2 ) ( 2,2)(4,2)(4,0 ) ( 2,-2)(4,-2)(4,0 ) ( 25.5,-.3)@xmath589 ( 13.5,.3)@xmath588 ( 17,.3)@xmath587 ( 11,-.8)@xmath365 ( 16.5,-.8)@xmath364 ( 20.5,-.8)@xmath590 ( 15,-6)@xmath583 ( 19.5,-2)@xmath580 ( 19.7,6)@xmath193 ( 8,6)@xmath193 ( 11,8.5)@xmath591 ( 0,0)(10,4.5)(20,6 ) ( 0,0)(9,6)(20,10 ) ( 0,0)(15,10)(20,20 ) ( 0,0)(1,1)20 ( 0,-.6)@xmath365 ( 20,-.6)@xmath364 ( 22,.6)@xmath200 ( 0.6,19.6)@xmath234 ( 10,3.2)@xmath580 ( 16,7)@xmath585 ( 14,8.1)@xmath165 ( 17,14.6)@xmath571 ( 10,9.1)@xmath586 [ twocycles ] the vector spaces @xmath592 , @xmath593 , @xmath594 of abelian integrals @xmath534 corresponding to the continuous families of ovals @xmath169 for @xmath595 , @xmath170 for @xmath596 , and @xmath168 for @xmath597 , are not chebyshev in the respective intervals @xmath16 . * proof . * the assertion follows from local analysis . consider first @xmath169 . then @xmath598 and @xmath599 , @xmath600 . by the proof of proposition [ cyclicity ] , we obtain that @xmath601 is equivalent to @xmath602 . this coefficient is easily calculated from the normal form ( [ normal ] ) and the result is @xmath603 which proves the claim . the proof in the case of @xmath168 is similar . the assertion for @xmath170 follows from the definition of @xmath571 . @xmath74 * conjecture 2 . * ( i ) . the vector spaces @xmath592 , @xmath593 are chebyshev with accuracy one in @xmath16 for the cases considered in proposition [ twocycles ] , and they are chebyshev in @xmath16 for all the remaining cases . the vector space @xmath594 is chebyshev with accuracy one in @xmath16 for @xmath604 , it is chebyshev with accuracy two in @xmath16 for @xmath605 and it is chebyshev in @xmath16 for the remaining cases . the next problem in difficulty should be to study the space of abelian integrals of the first kind related to genus three hyperelliptic curves . we hope that this problem can be settled by combining the methods of the present paper with deformation arguments as in @xcite . properly understood , this might lead to the solution of the problem for arbitrary genus @xmath9 . 99 n. acampo , le groupe de monodromie du dploiement des singularits isoles de courbes planes . i. _ math . * 213 * ( 1975 ) , 132 . e. arbarello , m. cornalba , p.a . griffiths , j. harris , _ geometry of algebraic curves _ , vol . 1 , springer , new york , 1985 . v.i . arnold , s.m . gusein - zade , a.n . varchenko , _ singularities of differentiable maps _ , vols . 1 and 2 , monographs in mathematics , birkhuser , boston , 1985 and 1988 . arnold , yu.s . ilyashenko , ordinary differential equations , in : _ dynamical systems _ i , encyclopaedia of math . , vol . 1 , springer , berlin , 1988 . v.i . arnold , _ geometrical methods in the theory of ordinary differential equations _ , springer , new york , 1988 . arnold , sur quelques problmes de la thorie des systmes dynamiques , _ topological methods in nonlinear analysis _ , * 4 * ( 1994 ) , 209225 . arnold , some unsolved problems in the theory of differential equations and mathematical physics , _ russian math . surveys _ * 44 * ( 1989 ) , no . 4 , 157171 . v.i . arnold , ten problems , in : _ theory of singularities and its applications _ , soviet math . _ * 1 * , pp . 18 , amer . providence , 1990 . e. brieskorn , die monodromie der isolierten singularitten von hyperflchen , _ manuscripta math . _ * 2 * ( 1970 ) , 103161 . a. dimca , _ singularities and topology of hypersurfaces _ , springer , berlin , 1992 . a. gasull , w. li , j. llibre , zh . zhang , chebyshev property of complete elliptic integrals and its application to abelian integrals , _ pacific j. math . _ * 202 * ( 2002 ) , no . 2 , 341362 . l. gavrilov , isochronicity of plane polynomial hamiltonian systems , _ nonlinearity _ * 10 * ( 1997 ) , no . 2 , 433448 . l. gavrilov , petrov modules and zeros of abelian integrals , _ bull . _ * 122 * ( 1998 ) , no . 8 , . l. gavrilov , abelian integrals related to morse polynomials and perturbations of plane hamiltonian vector fields , _ ann . * 49 * ( 1999 ) , no . 2 , 611652 . l. gavrilov , nonoscillation of elliptic integrals related to cubic polynomials with symmetry of order three , _ bull . london math . * 30 * ( 1998 ) , no . 2 , 267273 . l. gavrilov , the infinitesimal 16th hilbert problem in the quadratic case , _ invent . * 143 * ( 2001 ) , 449497 . givental , sturm s theorem for hyperelliptic integrals , _ leningrad math . j. _ * 1 * ( 1990 ) , 11571163 . p.a . griffiths , j. harris , _ principles of algebraic geometry _ , pure and appl . john wiley and sons , new york , 1978 . d. hilbert , mathematische probleme , _ gesammelte abhandlungen _ iii , springer - verlag , berlin ( 1935 ) , pp . d. novikov and s. yakovenko , tangential hilbert problem for perturbations of hyperelliptic hamiltonian systems , _ electron . soc . _ * 5 * ( 1999 ) , 5565 ( electronic ) . petrov , number of zeros of complete elliptic integrals , _ funct.anal.appl . _ * 18 * ( 1984 ) , 7273 ; * 20 * ( 1986 ) , 3740 ; * 21 * ( 1987 ) , 8788 ; * 22 * ( 1988 ) , 3740 ; * 23 * ( 1989 ) , 8889 ; * 24 * ( 1990 ) , 4550 . pontryagin , on dynamic systems close to hamiltonian systems , _ zh . eksp . * 4 * ( 1934 ) , 234238 ( russian ) . r. roussarie , _ bifurcation of planar vector fields and hilbert s sixteenth problem _ , progress in mathematics , vol . 164 , birkhuser , basel , 1998 . |
we review millimeter interferometric phase variations caused by variations in the precipitable water vapor content of the troposphere , and we discuss techniques proposed to correct for these variations . we present observations with the very large array at 22 ghz and 43 ghz designed to test these techniques . we find that both the fast switching and paired array calibration techniques are effective at reducing tropospheric phase noise for radio interferometers . in both cases , the residual rms phase fluctuations after correction are _ independent of baseline length for b b . _ these techniques allow for diffraction limited imaging of faint sources on arbitrarily long baselines at mm wavelengths . we consider the technique of tropospheric phase correction using a measurement of the precipitable water vapor content of the troposphere via a radiometric measurement of the brightness temperature of the atmosphere . required sensitivities range from 20 mk at 90 ghz to 1 k at 185 ghz for the mma , and 120 mk for the vla at 22 ghz . the minimum gain stability requirement is 200 at 185 ghz at the mma assuming that the astronomical receivers are used for radiometry . this increases to 2000 for an uncooled system . the stability requirement is 450 for the cooled system at the vla at 22 ghz . to perform absolute radiometric phase corrections also requires knowledge of the tropospheric parameters and models to an accuracy of a few percent . it may be possible to perform an ` empirically calibrated ' radiometric phase correction , in which the relationship between fluctuations in brightness temperature differences with fluctuations in interferometric phases is calibrated by observing a strong celestial calibrator at regular intervals . a number of questions remain concerning this technique , including : ( i ) over what time scale and distance will this technique allow for radiometric phase corrections when switching between the source and the calibrator ? and ( ii ) how often will calibration of the t - relationship be required ? | astro-ph9904248 | the most important difference between mm and cm interferometry is the effect of the troposphere at mm wavelengths . the optical depth of the troposphere becomes significant below 1 cm , leading to increased system temperatures due to atmospheric emission , and increased demands on gain calibration due to variable opacity ( yun et al . , 1998 , kutner and ulich 1981 ) . even more dramatic is the effect of the troposphere on interferometer phases . variations in the tropospheric water vapor column density lead to variations in electronic pathlength , and hence variations in interferometric phase . this can cause loss of amplitude of the cross correlations , or ` visibilities ' , over the integration time ( ` coherence ' ) , reduced spatial resolution ( ` seeing ' ) , and pointing errors ( ` anomalous refraction ' ) . until recently , mm interferometric arrays have been restricted to maximum baselines of a few hundred meters . diffraction limited resolution images at mm wavelengths can be made on such baselines , since the tropospheric phase fluctuations on such baselines at good observatory sites are typically one radian or less under good weather conditions . however , many existing mm arrays are expanding to baselines of 1 km or more , and the planned large millimeter array facility at the chajnantor site in chile will have baselines as long at 10 km . even at this premium site , phase fluctuations due to the troposphere will be larger than 1 radian at 230 ghz on these baselines , except under the best weather conditions . hence , tropospheric seeing will preclude diffraction limited resolution imaging at frequencies of 230 ghz for arrays larger than about 1 km , even at the best possible sites , if no corrections are made for tropospheric phase noise . at higher frequencies , the maximum baselines which would permit uncorrected observations would be even shorter . in this paper we review the theory of tropospheric phase noise in mm interferometry , along with examples showing tropospheric induced phase fluctuations and their effect on images made with interferometric arrays . we then consider three techniques for reducing tropospheric phase noise : ( i ) fast switching phase calibration , ( ii ) paired array phase calibration , and ( iii ) radiometric phase calibration . the first two techniques entail using celestial calibration sources near the target source , with either a calibration cycle time fast enough to ` stop ' the troposphere ( fast switching ) , or by using some of the antennas as a tropospheric ` calibration array ' ( paired array ) . we present extensive observational data using the very large array ( vla ) in socorro , nm , usa , at 22 ghz and 43 ghz designed to test the efficacy of these first two techniques on baselines longer than a few km . the radiometric phase correction technique entails real - time estimation of the precipitable water vapor content along each antenna s line of sight through the troposphere via a radiometric measurement of the brightness temperature of the atmosphere above each antenna . a number of issues are addressed , including : ( i ) the required radiometric sensitivity as a function of frequency and site quality , ( ii ) the constraints on ancillary data , such as atmospheric data and models , in order to perform an absolute radiometric phase correction , and ( iii ) the limitations to making radiometric phase corrections by calibrating the relationship between brightness temperature fluctuations and interferometric phase using celestial sources . the troposphere is the lowest layer of the atmosphere , extending from the ground to the stratosphere at an elevation of 7 km to 10 km . the temperature decreases with altitude in this layer , clouds form , and convection can be significant ( garratt 1992 ) . the troposphere is composed predominantly of n@xmath4 , o@xmath4 , plus trace gases such as water vapor , n@xmath4o , and co@xmath4 , plus particulates such as liquid water and dust in clouds . the troposphere becomes increasingly opaque with increasing frequency , mostly due to absorption by o@xmath4 and h@xmath4o . figure 1 shows models of the atmospheric transmission at cm and mm wavelengths for the vla site at 2150 m altitude , and the planned millimeter array ( mma ) site in chile at 4600 m altitude ( holdaway and pardo 1997 , liebe 1989 ) . the plot shows a series of strong absorption lines including the water lines at 22 ghz and 183 ghz , and the o@xmath4 lines at 60 ghz and 118 ghz , plus a systematic decrease in the transmission with increasing frequency between the lines . this ` pseudo - continuum ' opacity is due to the sum of the pressure broadened line wings of a multitude of sub - mm and ir lines of water vapor . the plot for the mma site at chajnantor in chile includes the typical value for the column density of precipitable water vapor , w@xmath5 = 1 mm , while the water vapor column for the vla site is assumed to be w@xmath5 = 4 mm , where precipitable water vapor ( pwv ) = the depth of the water vapor if converted to the liquid phase . figure 2 shows the relative contributions from water vapor and dry air ( o@xmath4 plus other trace gases ) for the vla site . below 130 ghz , both o@xmath4 and h@xmath4o contribute significantly to the optical depth . above 130 ghz , h@xmath4o dominates the optical depth . the troposphere has a non - unit refractive index , @xmath6 . the refractive index is defined by the phase change experienced by an electromagnetic wave , @xmath7 , propagating over a physical distance , d : @xmath8 or , in terms of ` electrical pathlength ' , l@xmath9 : @xmath10 the refractive index of air is non - dispersive ( ie , independent of frequency ) except near the strong resonant water and o@xmath4 lines , and is typically given as a difference with respect to vacuum ( @xmath11 @xmath12 1 ) , in parts per million , @xmath13 , as ( waters 1976 ) : @xmath14 the index of refraction of air is typically separated into the dry air component , @xmath15 , and the water vapor component , @xmath16 . these terms behave as ( waters 1976 , bean and dutton 1968 ) : @xmath17 where @xmath18 is the total mass density in gm @xmath19 , and @xmath20 is the water vapor mass density . the inverse dependence on temperature for @xmath16 is due to the increased effect of collisions on the ( mis-)alignment of the permanent electric dipole moments of the water molecules with increasing temperature ( waters 1976 , bean and dutton 1968 ) . a detailed derivation of these relationships can be found in thompson , moran , and swenson ( 1986 ) . for water vapor alone , it can be shown that @xmath20 = @xmath21 . using the equations above then leads to the relationship between the electrical pathlength , l@xmath9 , and the precipitable water vapor column , w : @xmath22 or : @xmath23 for t@xmath24 @xmath25 270 k. this relation between electrical pathlength and precipitable water vapor column has been verified experimentally for a number of atmospheric conditions ( hogg , guiraud , and decker 1981 ) . variations in precipitable water vapor lead to variations in the effective electrical path length , corresponding to variations in the phase of an electromagnetic wave propagating through the troposphere ( tatarskii 1978 ) . such variations are seen as ` phase noise ' by radio interferometers . since the troposphere is non - dispersive , the phase contribution by a given amount of water vapor increases linearly with frequency ( except in the vicinity of the strong water lines ) . hence , tropospheric phase variations are most prominent for mm and sub - mm interferometers , and can be the limiting factor for the coherence time and spatial resolution of mm interferometers ( hinder and ryle 1971 , lay 1997 , wright 1996 ) . the standard model for tropospheric phase fluctuations involves variations in the water vapor column density in a turbulent layer in the troposphere with a mean height , h@xmath26 , and a vertical extent , w , which moves at some velocity , v@xmath27 . this model includes the ` taylor hypothesis ' , or ` frozen screen approximation ' , which states that : ` if the turbulent intensity is low and the turbulence is approximately stationary and homogeneous , then the turbulent field is unchanged over the atmospheric boundary layer time scales of interest and advected with the mean wind ' ( taylor 1938 , garratt 1992 ) . under this assumption one can relate temporal and spatial phase fluctuations with a simple eulerian transformation between baseline length , b , and v@xmath27 : b = v@xmath27@xmath28time . in the following sections we adopt a value of v@xmath27 = 10 m s@xmath29 . this process is shown schematically in figure 3 . a demonstration of tropospheric phase fluctuations is shown in figure 4 for observations made with the vla at 22 ghz . the vla is an aperture synthesis array comprised of 27 parabolic antennas of 25 m diameter , operating at frequencies from 75 mhz to 50 ghz ( napier , thompson , and ekers 1983 ) . the antennas are situated in a ` y ' pattern , along three arms situated north , southwest , and southeast . the maximum physical baseline in the largest configuration is 33 km . the complex cross correlations of the electric fields measured at each antenna are calculated between all pairs of antennas ( ` interferometers ' ) as a function of time . the fourier transform of these measurements of the spatial coherence function of the electric field then gives the sky brightness distribution ( clark 1998 ) . for these observations , two subarrays were employed , one observing the celestial calibrator 0423 + 418 , and the second observing the calibrator 0432 + 416 . the sub - arrays were ` inter - laced ' , meaning that every second antenna along each arm of the array observed a given source . antenna - based phase and amplitude solutions were derived from the data in each sub - array using self - calibration with an averaging time of 30 sec ( cornwell 1998 ) . figure 4 shows the antenna - based phase solutions from two pairs of neighboring antennas along the southwest arm . the antennas at stations w16 and w4 were observing 0423 + 418 while the antennas at w18 and w6 were observing 0432 + 416 . for adjacent pairs of antennas ( w16-w18 and w6-w4 ) the temporal variations in the phase track each other closely . this close relationship for phase variations between neighboring antennas in the two different subarrays is the signature that the phase variations are primarily tropospheric in origin , and are correlated on relevant timescales and baseline lengths . an important aspect of tropospheric phase fluctuations arising from the taylor hypothesis is the relationship between the amplitude of the fluctuations and the time scale : large amplitude fluctuations occur over long periods and are partially correlated between antennas , while small amplitude fluctuations occur over short periods and are uncorrelated between antennas , depending on the baseline length . this effect can be seen in figure 4 by the fact that the antennas at the outer stations ( w14 and w16 ) show larger amplitude fluctuations relative to the inner stations ( w4 and w6 ) . this occurs because the antenna - based phase solutions for each subarray are referenced to antennas at the center of the array , such that the reference antennas are within about 250 m of w4 and w6 , but are separated from w16 and w14 by about 2000 m. an example of what occurs in the image plane due to tropospheric phase fluctuations is shown in figure 5 . observations were made of the celestial calibrator 2007 + 404 at 22 ghz with the vla at a resolution of 0.1@xmath30 ( maximum baseline = 30 km ) for a period of 1 hour . the data were self - calibrated using a long solution averaging time of 30 minutes , ie . just a mean phase was removed from each half of the data . ` snap - shot ' images were then made from one minute of data at the beginning and end of the observation ( upper left and upper right frames in figure 5 , respectively ) . two important trends are apparent in these two frames . first , notice the positive - negative side - lobe pairs straddling the peak , indicative of antenna - based phase errors ( ekers 1998 ) . these image artifacts are due to phase fluctuations which arise in small scale water vapor structures in the troposphere that are not correlated between antennas . second , notice that the peak in each image has shifted from the true source position . this position shift is due to phase fluctuations which arise in large scale water vapor structures in the troposphere that are correlated between antennas , ie . a phase gradient across the array . these two frames are analogous to optical ` speckle ' images , although the timescales in the radio are much longer than in the optical due to the larger spatial scales for the turbulence . the lower left frame shows the image of 2007 + 404 made using the full hour of data , but with a self - calibration averaging time of 30 minutes . the source appears extended in this image . the lower right frame shows the same image after self - calibration with an averaging time of 30 seconds , and in this case the source is unresolved , ie . the source size is equivalent , within the noise , to the interferometric synthesized beam . the lower left image has a peak surface brightness of 1.0 jy beam@xmath29 , an off - source rms noise level of 47 mjy beam@xmath29 , and a total flux density of 1.5 jy . the lower right image has a peak surface brightness of 1.6 jy beam@xmath29 , an off - source rms noise level of 5 mjy beam@xmath29 , and a total flux density of 1.6 jy . not correcting for tropospheric phase noise in the lower left frame has : ( i ) increased the off - source noise in the image , ( ii ) decreased the ` coherence ' ( ie . lowered the peak surface brightness ) , and ( iii ) degraded the resolution ( ` seeing ' ) . the important lesson from figure 5 is that , while tropospheric phase errors can be quantified in terms of an antenna - based phase error , the errors are partially correlated between antennas on certain spatial and temporal scales , leading to positional shifts of sources as well as the standard positive - negative side - lobe pairs . also , it is important to keep in mind that short baselines only ` sample ' the power in the phase screen on scales of order the baseline length . tropospheric phase fluctuations are usually characterized by the spatial phase structure function , @xmath31 , @xmath32 where b is the distance between two antennas , @xmath33 is the atmospheric phase measured at one antenna and @xmath34 is the atmospheric phase measured at the second antenna , and the brackets represent an ensemble average . usually in radio astronomy the ensemble average is replaced by a time average on one particular baseline . an interferometric array will sample the phase structure function at several baselines . for a single interferometer , the taylor hypothesis ( which asserts that temporal phase fluctuations are equivalent to spatial phase fluctuations ) permits us to measure temporal phase fluctuations on a single baseline and translate these into the equivalent spatial phase structure function . in the following discussion we consider the square root of the phase structure function ( the ` root phase structure function ' ) , which corresponds to the rms phase variations as a function of baseline length : @xmath35 . kolmogorov turbulence theory ( coulman 1990 ) predicts a function of the form : @xmath36 where b is in km , and @xmath37 is in mm . a typical value of k = 100 for the mma site in chajnantor under good weather conditions , and k = 300 is typical for the vla site ( carilli , holdaway , and sowinski 1996 , sramek 1990 ) . kolmogorov turbulence theory predicts @xmath38 = @xmath39 for baselines longer than the width of the turbulent layer , w , and @xmath38 = @xmath40 for baselines shorter than w ( coulman 1990 ) . the change in power - law index at b @xmath41 w is due to the finite vertical extent of the turbulent layer . for baselines shorter than w the full 3-dimensionality of the turbulence is involved ( thick - screen ) , while for longer baselines a 2-dimensional approximation applies ( thin - screen ) . turbulence theory also predicts an ` outer - scale ' , l@xmath5 , beyond which the rms phase variations should not increase with baseline length ( ie . @xmath38 = 0 ) . this scale corresponds to the largest coherent structures , or maximum correlation length , for water vapor fluctuations in the troposphere , presumably set by external boundary conditions . recent observations with the vla support kolmogorov theory for tropospheric phase fluctuations . figure 6 shows the root phase structure function made using the bna configuration of the vla . this configuration has good baseline coverage ranging from 200 m to 20 km , hence sampling all three hypothesized ranges in the structure function . observations were made at 22 ghz of the vla calibration source 0748 + 240 . the total observing time was 90 min , corresponding to a tropospheric travel distance of 54 km , using v@xmath27 = 10 m s@xmath29 ( see section 6.2 ) . the open circles show the nominal tropospheric root phase structure function over the full 90 min time range . the solid squares are the rms phases after subtracting ( in quadrature ) a constant electronic noise term of 10@xmath42 , as derived from the data by requiring the best power - law on short baselines . the 10@xmath42 noise term is consistent with previous measurements at the vla indicating electronic phase noise increasing with frequency as 0.5@xmath42 per ghz ( carilli and holdaway 1996 ) . the three regimes of the structure function as predicted by kolmogorov theory are verified in figure 6 . on short baselines ( b @xmath43 1.2 km ) the measured power - law index is [email protected] and the predicted value is 0.83 . on intermediate baselines ( 1.2 @xmath43 b @xmath43 6 km ) the measured index is 0.41@xmath44 0.03 and the predicted value is 0.33 . on long baselines ( b @xmath45 6 km ) the measured index is [email protected] and the predicted value is zero . the implication is that the vertical extent of the turbulent layer is : w @xmath25 1 km , and that the outer scale of the turbulence is : l@xmath5 @xmath25 6 km . the increase in the scatter of the rms phases for baselines longer than 6 km may be due to an anisotropic outer scale ( carilli and holdaway 1997 ) . in practice , single baseline site testing interferometers tend to see values of @xmath38 which form a continuous distribution between the theoretical thin and thick layer values of @xmath39 and @xmath40 ( holdaway et al . , 1995 ) . the distribution cuts off fairly sharply at these two theoretical values . such a distribution could be due to the presence of both a thin turbulent layer associated with the ground or an inversion layer and a thick turbulent layer . by changing the relative weight of each of these two layers with their theoretical power law exponents , the resulting phase structure function will be very close to a power law with an exponent between the two theoretical values . alternatively , the intermediate power law values could just reflect the transition between thin and thick turbulence . tropospheric phase noise leads to a number of adverse effects on interferometric observations at mm wavelengths . first is the loss of coherence of a measured visibility on a given baseline over a given averaging time due to phase variations . for a given visibility , v = v@xmath5e@xmath46 , the effect on the measured amplitude due to phase noise in a given averaging in time is : @xmath47 assuming gaussian random phase fluctuations with an rms variation of @xmath48 over the averaging time ( thompson , moran , and swenson 1986 ) . for example , for @xmath48 = 1 rad , the coherence is : @xmath49 = 0.60 , meaning the observed visibility amplitude is reduced by 40@xmath50 from the true value . a second effect of tropospheric phase fluctuations is to limit the spatial resolution of an observation in a manner analogous to optical seeing , where optical seeing is due to thermal fluctuations rather than water vapor fluctuations . since interferometric phase corresponds to the measurement of the position of a point source ( perley 1998 ) , it is clear that phase variations due to the troposphere will lead to positional variations of a source , and hence ` smear - out ' a point source image over time ( figure 5 ) . the magnitude of tropospheric seeing can be calculated by considering the coherence as a function of baseline length . since the coherence decreases for longer baselines given an averaging time long compared to the array crossing time , the observed visibility amplitude decreases with increasing baseline length , as would occur if the source were resolved by the array . using equation 3 for the root phase structure , and equation 4 for the coherence , the visibility amplitude as a function of baseline length becomes : @xmath51 ^ 2)\ ] ] note that the exponent must be in radians , so k@xmath52 = k @xmath28 @xmath53 . the baseline length corresponding to the half - power point of the visibility curve , b@xmath54 , then becomes : @xmath55 for example , at 230 ghz using the typical value of @xmath38 = 5/6 , and a typical value for k@xmath52 at the mma site of 1.7 , the value of b@xmath54 = 0.9 km . this means that the resolution of the array is limited by tropospheric seeing to : @xmath56 @xmath25 @xmath57 @xmath25 0.3@xmath30 at 230 ghz . for average weather conditions , tropospheric seeing precludes diffraction limited resolution imaging for arrays larger than about 1 km at the mma site in chile , if no corrections are made for tropospheric phase noise . a rigorous treatment of tropospheric seeing , with predicted source sizes under various assumptions about the turbulence , can be found in thompson , moran , and swenson ( 1986 ) . two important points need to be remembered when considering tropospheric seeing . first is that the root phase structure function flattens dramatically on baselines longer than @xmath25 1 km , such that the tropospheric seeing degrades slowly with longer baselines . and second , there is an explicit connection between the coherence loss and the seeing : on the short baselines , where the phase errors are smaller , there is less coherence loss , and the correct flux density is measured even for a long averaging time . on the longer baselines , the phase errors are larger , causing decorrelation of the visibilities , either within an integration or across many integrations , thereby fictitiously resolving the source . it is the selective loss of coherence on the long baselines which determines the seeing . this phenomenon can be seen in the lower left frame of figure 5 , in which the peak surface brightness is only 60@xmath50 of the expected peak , but the total flux density averaged over the ` seeing disk ' is 94@xmath50 of the true value , ie . the shortest baselines see the total flux density of the source even for long averaging times . a final problem arising from tropospheric phase variations is ` anomalous refraction ' , or tropospheric induced pointing errors ( holdaway 1997 , butler 1997 , holdaway and woody 1998 ) . this effect corresponds to tropospheric seeing on the scale of the antenna itself . phase gradients across the antenna change the apparent position of the source on a time scale @xmath25 @xmath58 @xmath25 1 second , for an antenna with a diameter @xmath59 m. a straight - forward application of snells law shows that the effect in arc - seconds should decrease with antenna diameter as @xmath60 , or about @xmath61 for median @xmath38 of 0.6 . the decrease in the pointing error with dish diameter is due to the fact that the value of @xmath38 in the root phase structure function is less than unity , and hence the angle of the ` wedge ' of water vapor across the antenna becomes shallower with increasing antenna size . however , in terms of fractional beam size , the effect becomes worse with antenna size as @xmath62 . for the 10 m mma antennas the expected magnitude of the effect at an elevation of 50@xmath63 is @xmath25 0.6@xmath30 , which is about 50@xmath50 of the pointing error budget for the antennas . an important point to keep in mind is that while tropospheric phase variations can be quantified in terms of a baseline - length dependent structure function , the errors are fundamentally antenna - based , and hence can be corrected by antenna - based calibration schemes , such as self - calibration or fast switching calibration . a straight forward method of reducing phase errors due to the troposphere is self - calibration ( cornwell 1998 ) . self - calibration removes the baseline - dependent term in the root structure function , @xmath64(b ) , leaving the residual tropospheric phase noise dictated by the ` effective baseline ' : b@xmath1 = @xmath65 = half the distance the troposphere moves during the self - calibration averaging time , t@xmath66 . the factor of two arises from the fact that the mean calibration applies to the middle of the solution interval . the taylor hypothesis dictates a relationship between temporal and spatial fluctuations such that the longer baselines will not sample the full power in the root phase structure function if the calibration cycle time is shorter than the baseline crossing time for the troposphere . of course , we would like to make @xmath67 as short as possible , but for a target source of some given brightness , we are limited in that we must detect the source in t@xmath66 on each baseline with sufficient signal - to - noise ratio ( snr @xmath25 2 for arrays with large numbers of antennas ; cornwell 1998 ) to be able to solve for the phase . hence , there will be sources which are so weak that they can not be detected in a time short enough to track the atmospheric phase fluctuations . for the mma at 230 ghz , self - calibration should be possible on fairly weak continuum sources ( of order 10 mjy ) , with fairly short integration times ( @xmath25 30 sec ) , leading to residual rms phase errors @xmath43 20@xmath63 . for the completed vla 43 ghz system the limit is 50 mjy sources with 30 second averaging times with residual rms phase variations of 10@xmath63 . self - calibration is not possible for weaker continuum sources , or for weak spectral line sources , or in the case where absolute positions are required . in these cases other methods must be employed to ` stop ' tropospheric phase variations . another method for reducing tropospheric phase variations is ` fast switching ' ( fs ) phase calibration . this method is simply normal phase calibration using celestial calibration sources close to the target source ( fomalont and perley 1998 ) , only with a calibration cycle time , t@xmath68 , short enough to reduce tropospheric phase variations to an acceptable level ( holdaway 1992 , holdaway and owen 1995 , carilli and holdaway 1996 , 1997 ) . the expected residual phase fluctuations after fs calibration can be derived from the root phase structure function ( equation 3 ) , assuming an ` effective baseline length ' , b@xmath1 , given by : @xmath69 where @xmath70 = wind speed , and d = the physical distance in the troposphere between the calibrator and source . the fs technique will be effective for calibration cycle times shorter than the baseline crossing time of the troposphere = @xmath71 . moreover , a significant gain is made when b@xmath1 @xmath72 1 km , thereby allowing for corrections to be made on the steep part of the root phase structure function ( figure 6 ) , implying a timescale of 200 seconds or less for effective fs corrections . as with self - calibration , the calibrator source must be detected with sufficient snr on time scales short enough to track the atmospheric phase fluctuations . the effectiveness of fs phase calibration is shown in figure 7 for 22 ghz data from the vla on baselines ranging from 100 m to 20 km . the solid squares show the nominal tropospheric root phase structure function averaged over 90 minutes ( figure 6 ) . the open circles are the rms phases of the visibilities after applying antenna based phase solutions averaged over 300 seconds . the stars are the rms phases of the visibilities after applying antenna based phase solutions averaged over 20 seconds . the residual root structure function using a 300 second calibration cycle parallels the nominal tropospheric root structure function out to a baseline length of 1500 m , beyond which the root structure function saturates at a constant rms phase value of 20@xmath42 . the implied wind velocity is then : v@xmath27 = @xmath73 = 10 m s@xmath29 . using a 20 second calibration cycle reduces b@xmath1 to only 100 m , which is shorter than the shortest baseline of the array , and the saturation rms is 5@xmath42 . the important point is that , after applying standard phase calibration techniques on timescales short compared the array crossing time of the troposphere , the resulting rms phase fluctuations are _ independent of baseline length for b @xmath0 b@xmath1 . _ the fs technique allows for diffraction limited imaging of faint sources on arbitrarily long baselines . note that this conclusion should also apply to very long baseline interferometry ( vlbi ) , although the problems are accentuated due to the fact that antennas can be observing at very different elevations at any given time , and that vlbi observations typically employ low elevation observations in order to maximize mutual visibility times for widely separated antennas ( see the discussion of the vlbi phase referencing technique by beasley and conway 1995 ) . an important question to address when considering fs phase calibration is : are there enough calibrators in the sky in order to take advantage of a switching time as short as 40 seconds ? this depends on the slew rate and settling time of the telescope , the set - up time of the electronics , the sensitivity of the array , and the sky surface density of celestial calibrators . holdaway , owen , and rupen ( 1994 ) estimated the calibrator source counts at 90 ghz . they measured the 90 ghz flux densities of 367 flat spectrum quasars known from centimeter wavelength surveys , thereby determining the distribution of the spectral index between 5 ghz and 90 ghz , which was statistically independent of source flux density . by applying this spectral index distribution to well understood 5 ghz flat spectrum source counts , they were able to estimate the integral source counts for appropriate phase calibrators at 90 ghz . they estimate that the integral source counts over the whole sky between 0.1 to 1.0 jy is about 170 @xmath74 , where s@xmath75 is the 90 ghz source flux density in jy . then , the typical distance to the nearest appropriate calibrator at 90 ghz will be about @xmath76 . for calibrators with s@xmath77 @xmath45 50 mjy , and assuming a slew rate of 2 deg s@xmath29 , the 40 element mma should be able to employ fs phase calibration on most sources with total cycle times @xmath43 20 sec , leading to residual rms phase fluctuations @xmath43 20@xmath63 at 230 ghz , and on - source duty cycles @xmath25 75@xmath50 . one important practical problem is the lack of all - sky surveys at high frequency from which to generate calibrator source lists . however , the mma will be both agile and sensitive enough to survey a few square degrees around the target source in a few minutes prior to the observations , thereby permitting the determination of the optimal calibrator . a third method for reducing tropospheric phase noise for faint sources is paired antenna , or paired array , calibration . paired array ( pa ) calibration involves phase calibration of a ` target ' array of antennas using a separate ` calibration ' array , where the target array is observing continuously a weak source of scientific interest while the calibration array is observing a nearby calibrator source ( holdaway 1992 , counselman et al . , 1974 , asaki et al . , 1996 , 1998 , drashkik and finkelstein 1979 ) . in its simplest form pa calibration implies applying the phase solutions from a calibration array antenna to the nearest target array antenna at each integration time . an improvement can be made by interpolating the solutions from a number of nearby calibration array antennas to a given target array antenna at each integration time . ultimately , the discrete measurements in space and time of the phases from the calibration array could be incorporated into a physical model for the troposphere to solve for , and remove , the effects of the tropospheric phase screen on the target source as a function of time and space using some intelligent method of data interpolation , such as forward projection using a physical model for the troposphere and kalman filtering of the spatial time series ( zheng 1985 ) . for simple pairs of antennas , the residual phase error can be derived from the root phase structure function with : @xmath78 where d is the same as in equation 10 , and @xmath79b is the baseline length between the calibration antenna and the target source antenna . figure 4 shows an observation for which pa calibration was implemented . again , observations were made at 22 ghz using two ` inter - laced ' subarrays observing two close calibrators , 0432 + 416 and 0423 + 416 . notice how the phase variations for adjacent antennas in the different subarrays track each other closely . this correlation between phase variations from neighboring antennas in different subarrays observing different sources implies that the tropospheric phase variations can be corrected using pa calibration . in figure 4 the temporal variations for neighboring antennas track each other well , but the mean phase over the observing time range is different between antennas . this phase off - set is due to the electronics and/or optics at each antenna , and should be slowly varying in time . before interpolating phase solutions from the calibration array to the target array one must first determine , and remove , the electronic phase off - sets . this can be done by observing a celestial calibrator every 30 min or so . a demonstration of this process is shown in figure 8 . the upper frame in figure 8 shows a random phase distribution along the west arm at a given time before the mean electronic phase is removed , while the lower frame in figure 8 shows a smooth phase gradient along the arm after correction for the electronic phase term , indicating a tropospheric phase ` wedge ' down the array arm at this time . a quantitative measure of the effects of pa calibration can be seen in the root phase structure function plotted in figure 9 . the open triangles show the root phase structure function for the given observing day , as determined from the data with only the mean phase calibration ( 30 min averaging ) applied . this function can be fit by a power - law in rms phase versus baseline length with index 0.65 . the rms magnitude of 35@xmath63 on a baseline of 1000 m is somewhat higher than the expected value of about 25@xmath63 on a typical summer evening at the vla ( carilli etal . the stars in figure 9 correspond to the ` noise floor ' for the phase measurements , as determined by calculating the root structure function from self - calibrated data with an averaging time of 30 seconds . the solid squares in figure 9 show the root structure function for the data with pa calibration applied . the pa process in this case entailed interpolating phase solutions from neighboring antennas observing the ` calibration source ' 0423 + 418 , to the antennas observing the ` target source ' 0432 + 416 . the residual rms phase values are about 10@xmath63 on short baselines , and increase very slowly with baseline length . these data indicate a significant improvement in rms phase fluctuations after application of pa calibration for baselines longer than about 300 m. the increased noise floor for the pa calibrated data relative to self - calibration indicates residual short - timescale phase differences which do not replicate between the target and calibration arrays . this noise floor is a combination of ` jitter ' in the electronic phase contribution , and residual tropospheric phase noise as determined by b@xmath1 above . note that the residual noise floor increases slowly with baseline length . this is due to the logarithmically increasing separation between vla antennas along the arm . the brightness temperature of atmospheric emission , t@xmath80 , can be measured using a radiometer , and is given by the radiometry equation ( dicke et al . , 1946 ) : @xmath81 where t@xmath24 is the physical temperature of the atmosphere , and @xmath82 is the optical depth , which depends on , among other things , the precipitable water vapor content ( pwv ) of the troposphere . by measuring fluctuations in atmospheric brightness temperature with a radiometer , one can infer the fluctuations in the column density of water vapor of the troposphere ( barrett and chung 1962 , staguhn et al . , 1998 , staelin 1966 , westwater and guiraud 1980 , rosenkranz 1989 , bagri 1994 , sutton and hueckstadt 1997 , lay 1998 ) . the relationship between electrical pathlength and water vapor column ( equation 1 ) can then be used to derive the variable contribution from water vapor to the interferometric phase ( elgered 1993 ) . this technique has been used with varying degrees of success at connected - element mm interferometers ( welch 1994 , woody and marvel 1998 , bremer et al . , 1997 ) , and in geodetic vlbi experiments ( elgered et al . 1991 ) . we assume that the atmospheric opacity can be divided into three parts : @xmath83 where : ( i ) a@xmath77 is the optical depth per mm of pwv as a function of frequency , ( ii ) w@xmath5 is the temporally stable ( mean ) value for pwv of the troposphere , ( iii ) b@xmath84 is the total optical depth due to dry air as a function of frequency ( also assumed to be temporally stable ) , and ( iv ) w@xmath85 is the time variable component of the pwv of the troposphere . it is this time variable component which causes the tropospheric phase ` noise ' for an interferometer . in effect , we assume a constant mean optical depth : @xmath86 , with a fluctuating term due to changes in pwv : @xmath87 , and that @xmath88 @xmath89 @xmath90 . inserting equation 8 into equation 7 , and making the reasonable assumption that a@xmath84@xmath28w@xmath91 , leads to : @xmath92 ~~ + ~~ t_{\rm atm}\times e^{-\tau_{\circ}}\times[a_{\nu}\times w_{\rm rms } ~ + ~ { { ( a_{\nu}\times w_{\rm rms})^2}\over{2 } } ~ + ~ ... ] \ ] ] the first term on the right - hand side of equation 9 represents the mean , non - varying t@xmath93 of the troposphere . the second term represents the fluctuating component due to variations in pwv , which we define as : @xmath94\ ] ] at first inspection , it would appear that equation 10 applies to fluctuations in a turbulent layer at the top of the troposphere , since the fluctuating component is fully attenuated ( ie . multiplied by @xmath95 ) . however , for a turbulent layer at lower altitudes there is the additional term of attenuation of the atmosphere above the turbulent layer by the turbulence . it can be shown that the terms exactly cancel for an isobaric , isothermal atmosphere , in which case equation 10 is _ independent _ of the height of the turbulence . absolute radiometric phase correction entails measuring variations in brightness temperature with a radiometer , inverting equation 10 to derive the variation in pwv , and then using equation 1 to derive the variation in electronic phase along a given line of sight . as benchmark numbers for the mma we set the requirement that we need to measure changes in tropospheric induced phase above a given antenna to an accuracy of @xmath96 at 230 ghz at the zenith , or @xmath48 = 18@xmath63 . this requirement inserted into equation 10 then yields a required accuracy of : w@xmath85 = 0.01 mm . this value of w@xmath85 then sets the required sensitivity , t@xmath2 , of the radiometers as a function of frequency through equation 10 . for the vla we set the @xmath96 requirement at 43 ghz , leading to : w@xmath85 = 0.05 mm . in its purest form , the inversion of equation 10 requires : ( i ) a sensitive , absolutely calibrated radiometer , ( ii ) accurate models for the run of temperature and pressure as a function of height in the atmosphere , and ( iii ) an accurate value for the height of the pwv fluctuations . figure 10 shows the required sensitivity of the radiometer , t@xmath2 , given the benchmark numbers for w@xmath85 for the vla and the mma and using equation 10 . it is important to keep in mind that lower numbers on this plot imply that more sensitive radiometry is required in order to measure the benchmark value of w@xmath85 . the required t@xmath2 values generally increase with increasing frequency due to the increase in a@xmath77 , with a local maximum at the 22 ghz water line , and minima at the strong o@xmath4 lines ( 59.2 ghz and 118.8 ghz ) . the strong water line at 183.3 ghz shows a ` double peak ' profile , with a local minimum in t@xmath2 at the frequency corresponding to the peak of the line . this behavior is due to the product : a@xmath77 @xmath28 e@xmath97 in equation 10 . the value of a@xmath77 peaks at the line frequency , but this is off - set by the high total optical depth at the line peak . this effect is dramatic for the vla case , where the required t@xmath2 at the 183 ghz line peak is very low . in this section we consider making an absolute correction to the electronic phase at a given antenna using an accurate , absolutely calibrated measurement of t@xmath93 , and accurate measurements of tropospheric parameters ( temperature and pressure as a function of height , and the scale height of the pwv fluctuations ) . we consider requirements on the gain stability , sensitivity , and on atmospheric data , given the benchmark values of w@xmath85 and using equation 10 to relate w@xmath85 and t@xmath2 . we consider the requirements at a number of frequencies for the mma site , including : ( i ) the water lines at 22.2 ghz and 183.3 ghz , ( ii ) the half power of the water line at 185.5 ghz , and ( iii ) two continuum bands at 90 ghz and 230 ghz . for the vla we only consider the 22.2 ghz line . the results are summarized in table 1 . row 1 shows the optical depth per mm pwv , a@xmath84 , at the different frequencies for the model atmospheres discussed in section 4 , while row 2 shows the total optical depth , @xmath82 , for the models . row 3 shows the required t@xmath2 values as derived from equation 10 . it is important to keep in mind that these values are simply the expected change in t@xmath93 given a change in w of 0.01 mm for the mma and 0.05 mm for the vla , for a single radiometer looking at the zenith . all subsequent calculations depend on these basic t@xmath2 values . the values range from 19 mk at 90 ghz , to 920 mk at 185.5 ghz , at the mma site , and 120 mk for the vla site at 22 ghz . we first consider sensitivity and gain stability . row 4 lists approximate numbers for expected receiver temperatures , t@xmath98 , in the case of cooled systems ( eg . using the astronomical receivers for radiometry ) . row 5 lists the contribution to the system temperature from the atmosphere , t@xmath99 , and row 6 lists the expected total system temperature , t@xmath100 ( sum of row 4 and 5 ) . row 7 lists the rms sensitivity of the radiometers , t@xmath85 , assuming 1000 mhz bandwidth , one polarization , and a 1 sec integration time . in all cases the expected sensitivities of the radiometers are well below the required t@xmath2 values , indicating that sensitivity should not be a limiting factor for these systems . row 8 lists the required gain stability of the system , defined as the ratio of total system temperature to t@xmath2 : @xmath101gain @xmath12 @xmath102 . values range from 210 for the 185.5 ghz measurement to 5800 for the 90 ghz measurement at the mma , and 450 for the vla site at 22 ghz . rows 9 and 10 list total system temperatures and expected rms sensitivities in the case of uncooled radiometers . we adopt a constant total system temperature of t@xmath100 = 2000 k , but the other parameters remain the same ( bandwidth , etc ... ) . the radiometer sensitivity is then 63 mk in 1 second . this sensitivity is adequate to reach the benchmark t@xmath2 values in row 3 , although at 230 ghz the sensitivity value is within a factor two of the required t@xmath2 . the required gain stabilities in this case are listed in row 11 . the requirement becomes severe at 230 ghz ( @xmath101gain = 15000 ) . we next consider the requirements on atmospheric data , beginning with t@xmath24 . the dependence of t@xmath2 on t@xmath24 comes in explicitly in equation 10 through the first multiplier , and implicitly through the effect of t@xmath24 on @xmath88 . for simplicity , we consider only the explicit dependence , which will lead to an under - estimate of the expected errors by at most a factor @xmath25 2 adequate for the purposes of this document ( sutton and hueckstaedt 1997 ) . under this simplifying assumption the required accuracy , @xmath101t@xmath24 , becomes : @xmath103 } } ~ k\ ] ] the values @xmath101t@xmath24 are listed in row 12 . values in parentheses are the percentage accuracy in terms of the ground atmospheric temperature . values are typically of order 1 k , or a few tenths of a percent of the mean . a related requirement is the accuracy of the gradient in temperature : @xmath101@xmath104 @xmath25 @xmath105 , in the case of a turbulent layer at h@xmath26 = 2 km , and assuming a very accurate measurement of t@xmath24 on the ground and a very accurate measurement of h@xmath26 . these values are listed in row 13 . the accuracy requirements range from 0.25 k km@xmath29 to 1.5 k km@xmath29 , or roughly 10@xmath50 of the mean gradient . similarly , we can consider the required accuracy of the measurement of the height of the troposphere , @xmath101h@xmath26 @xmath25 @xmath106 , assuming a perfect measurement of the ground temperature and temperature gradient . these values are listed in row 14 . values are typically a few tenths of a km , or roughly 10@xmath50 of h@xmath26 . we consider the requirements on atmospheric pressure given the t@xmath2 requirements . the relationship between t@xmath93 and w is affected by atmospheric pressure through the change in the pressure broadened line shapes . an increase in pressure will transfer power from the line peak into the line wings , thereby flattening the overall profile . the expected changes in optical depth ( or brightness temperature ) as a function of frequency have been quantified by sutton and hueckstaedt ( 1997 ) , and their coefficients relating changes in pressure with changes in optical depth are listed in row 17 . note the change in sign of the coefficient on the line peaks versus off - line frequencies . sutton and hueckstaedt point out that , since the integrated power in the line is conserved , there are ` hinge points ' in the line profiles where pressure changes have very little effect on t@xmath93 , ie . for an increase in pressure at fixed total pwv the wings of the line get broader while peak gets lower . these hinge points are close to the half power points in t@xmath93 of the lines . rows 15 and 16 list the requirements on the accuracy of p@xmath24 , and on the value of h@xmath26 . the values of @xmath101p@xmath24 are derived from the equation : @xmath101p@xmath24 = @xmath107 , where @xmath108 is the coefficient listed in row 17 . we find that the value of p@xmath24 needs to be known to about 1@xmath50 , and the height of the turbulent layer needs to be known to a few percent . the exception is at the hinge point of the line ( @xmath25 185.5 ghz ) , where the optical depth is nearly independent of p@xmath24 . there are a few potential difficulties with absolute radiometric phase corrections which we have not considered . first is the question of how to make a proper measurement of the ` ground temperature ' ? it is possible , and perhaps likely , that the expected linear temperature gradient of the troposphere displays a significant perturbation close to the ground . the method for making the ` correct ' ground temperature measurement remains an important issue to address in the context of absolute radiometric phase correction . second , we have only considered a simple model in which the pwv fluctuations occur in a narrow layer at some height h@xmath26 , which presumably remains constant over time . if the fluctuations are distributed over a large range of altitude then one needs to know the height of the dominant fluctuation at _ each time _ to convert t@xmath93 into electrical pathlength . and when fluctuations at different altitudes contribute at the same time , this conversion becomes problematical . again , the required accuracies for the height of the fluctuations are given in rows 14 and 16 in table 1 . a possible solution to this problem is to find a linear combination of channels for which the effective conversion factor is insensitive to altitude under a range of conditions , ie . hinge points generalized to a multi - channel approach ( lay 1998 , staguhn et al . , 1998 ) . and third , the shape of the pass band of the radiometer needs to be known very accurately in order to obtain absolute t@xmath93 measurements . a final uncertainty involved in making absolute radiometric phase corrections are errors in the theoretical atmospheric models relating w and t@xmath93 ( elgered 1993 ) . sutton and hueckstaedt ( 1997 ) point out that model errors are by far the dominant uncertainties when considering absolute radiometric phase correction , and they have calculated a number of models with different line shapes and different empirically determined water vapor continuum ` fudge - factors ' . row 18 in table 1 lists the approximate differences between the various models at various frequencies . models can differ by up to 3 mm in pwv , corresponding to 19.5 mm in electrical pathlength , or 30@xmath109 rad in electronic phase at 230 ghz . the differences are most pronounced in the continuum bands , but are only negligible close to the peak of the strong 183 ghz line . this is an area of very active research , and it may be that this uncertainty is greatly reduced in the near future ( rosenkranz 1998 ) . given the status of current atmospheric models , radiometric phase correction then requires some form of empirical calibration of the water vapor continuum contribution in order to relate t@xmath93 to w. the exception may be a measurement close to the peak of the 183 ghz line , but in this case saturation becomes a problem . perhaps most importantly , if the calibrated continuum term is due to incorrect line shapes , it will depend on both t@xmath24 and p@xmath24 , in which case the continuum term may require frequent calibration . overall , absolute radiometric phase correction requires : ( i ) systems that are sensitive ( 19 @xmath43 t@xmath85 @xmath43 920 mk ) , and stable over long timescales ( 200 @xmath43 @xmath101gain @xmath43 15000 ) , and ( ii ) knowledge of the tropospheric parameters , such as t@xmath24 , p@xmath24 , and h@xmath26 , to a few percent or less . and even if such accurate measurements are available , fundamental uncertainties in the atmospheric models relating t@xmath93 and pwv may require empirical calibration of the t@xmath2 - w@xmath85 relationship at regular intervals . lay ( 1998 ) has recently presented an interesting radiometric phase correction method using multifrequency measurements of the 183 ghz water line profile . his method is insensitive to atmospheric parameters , since it relies on using the line profile , and in particular , the hinge points of the lines . this method may allow for an absolute radiometric phase correction to be made without great uncertainties due to the atmospheric models . many of the uncertainties in table 1 arise from the fact that we are demanding an absolute phase correction at each antenna based on the measured t@xmath93 plus ancillary data ( t@xmath24 , p@xmath24 , h@xmath26 , ... ) , using a theoretical model of the atmosphere to relate t@xmath93 to w. this sets very stringent demands on the absolute calibration , on the accuracy of the ancillary data , and on the accuracy of the theoretical model atmosphere . the current atmospheric models under - predict w by large factors in the continuum bands , thereby requiring calibration of ( possibly time dependent ) water vapor continuum ` fudge - factors ' . one way to avoid some of these problems is to calibrate the relationship between fluctuations in t@xmath2 with fluctuations in antenna - based phase , @xmath48 , by observing a strong celestial calibrator at regular intervals . this empirically calibrated phase correction method would circumvent dependence on ancillary data and model errors ( woody and marvel 1998 ) , and mitigate long term gain stability problems in the electronics . this technique can be thought of as calibrating the ` gain ' of both the atmosphere and the electronics , in terms of relating t@xmath2 to @xmath48 . in its simplest form , empirically calibrated radiometric phase correction would be used only to increase the coherence time on source . no attempt would be made to connect the phase of a celestial calibrator with that of the target source using radiometry , and hence the absolute phase on the target source would still be obtained from the calibration source . such a process is being implemented at the owen valley radio observatory ( woody and marvel 1998 ) . in this case the absolute phase is obtained from the first accurate phase measurement on the celestial calibrator , while the subsequent time series of phase measurements on the calibrator are then used to derive the t@xmath2 to @xmath48 relationship . this process results in additional phase uncertainty in a manner analogous to fast switching phase calibration ( holdaway and owen 1995 ) . the residual error is set by the distance between the calibrator and source , and the time required to obtain the first accurate record : t@xmath110 ( the slew time + the integration time required for the first accurate phase measurement ) . in this case : @xmath111 where d is the physical distance in the troposphere set by the angular separation of the calibrator and the source , and b@xmath112 is the ` effective baseline ' to be inserted into equation 3 in order to estimate the residual uncertainty in the absolute phase . for example , assuming t@xmath113 = 10 sec , and the calibrator - source separation = 2@xmath63 , leads to b@xmath112 = 170 m , or @xmath48 = 20@xmath63 at 230 ghz at the mma site . note that the temporal character of this ` phase noise ' is unusual in that the short timescale ( t @xmath114 t@xmath68 ) variations are removed by radiometry , while the long timescale variations ( t @xmath89 t@xmath68 ) are removed by celestial source calibration ( lay 1997 ) . an example of such a relatively calibrated radiometric phase correction is shown in figure 11 , using data from the vla at 22 ghz . observations were made of the celestial calibrator 0319 + 415 ( 3c 84 ) . in the upper frame , the dash line shows the interferometric phase time series measured between antennas 5 and 9 , corresponding to a baseline length of about 3 km . the solid line shows the predicted phase time series derived by differencing measurements of the 22 ghz system temperature at each antenna . a single scale factor relating phase fluctuations and fluctuations in system temperature differences was derived from all the data , by requiring a minimum residual rms scatter in the phase fluctuations after applying radiometric phase correction . a constant off - set was also applied to each data set . note the clear correlation between measured interferometric phase variations , and the phase variations predicted by radiometry . the middle frame shows the residual phase variations after radiometric correction . the rms variations in the raw phase time series before correction are 32@xmath63 . after applying the radiometric correction , the rms phase variations are reduced to 17@xmath63 . the lower frame shows the residual phase variations after radiometric correction , but now making a correction for the first and second half of the data separately . the residual rms phase variations are now 13@xmath63 . the scale factor changes by about 10@xmath50 over the 36 minutes . this ` empirically calibrated ' radiometric phase correction technique has been implemented successfully at the owens valley radio observatory and at the iram interferometer ( woody and marvel 1998 , bremer et al . , 1997 ) . a number of questions remain to be answered concerning this technique , including : ( i ) over what time scale and distance will this technique allow for radiometric phase corrections when switching between the source and the calibrator ? and ( ii ) how often will calibration of the t@xmath2 - @xmath3 relationship be required , ie . how stable are the radiometers and the mean parameters of the atmosphere ? water droplets present the problem that the drops contribute significantly to the measured t@xmath93 but not to w , thereby invalidating the model relating t@xmath93 and w. this problem can be avoided by using multichannel measurements around the water lines ( 183 ghz or 22 ghz ) , since t@xmath93 for the lines is not affected by water drops . alternatively , a dual - band system could be used to separate the effect of water drops from water vapor ( eg . 90 ghz and 230 ghz ) , since the frequency dependence of t@xmath93 is different for the two water phases . this later method requires a multi - band radiometer , which may be difficult within the context of the mma antenna design . the question of whether clouds will be a significant problem on high quality sites such as the mma site in chile remains to be answered . a final error introduced when using radiometric phase calibration is due to the different path through the troposphere seen by the radiometer with that seen by the astronomical receiver . if the radiometer is not the astronomical receiver itself , then the angular separation of the radiometer beam and the telescope beam , @xmath115 , can be of order a few degrees , depending on the lay - out of the receivers and the telescope optics . this angle corresponds to 100 m or so at a height of 2 km . the magnitude of the error introduced by such an observing path difference can be derived from the root phase structure function assuming b@xmath112 = h@xmath26 @xmath28 @xmath115 . fast switching phase calibration has been used extensively , and successfully , at the vla for diffraction limited imaging of faint astronomical sources on 30 km baselines at 43 ghz ( lim et al . , 1998 , wilner et al . , 1996 ) . however , the required switching times at higher frequencies ( 10 s of seconds or less ) has thus far precluded the use of fs for existing mm observatories . paired array calibration has been demonstrated effective in reducing tropospheric phase noise for interferometers ( asaki et al . , 1996 , 1998 ) , but has not yet been implemented as a standard technique for astronomical observing . radiometric phase correction has been used to varying degrees of success at different observatories , but mostly on an experimental basis ( welch 1994 , bremer 1997 , woody and marvel 1998 , elgered et al . , 1991 ) . we summarize the relative advantages and disadvantages of the three techniques . the advantages of fs are that : ( i ) fs uses the full array to observe the target source , and ( ii ) fs removes the long and short term electronic phase noise along with the tropospheric phase noise . the disadvantages are that : ( i ) fs places stringent constraints on telescope design in terms of slew rate , mechanical settling time , and electronic set - up time , and ( ii ) on - source observing time is lost due to frequent moves and calibration . simulations for the mma ( holdaway 1998 ) indicate that overall fast switching efficiencies ( including both sensitivity losses due to decreased observing time and increased decorrelation ) of about 0.75 should be possible , assuming a distribution of observing frequencies from 30 ghz to 650 ghz and using the measured distribution of atmospheric phase conditions at the chajnantor site . typical switching times range from 10 to 20 seconds . the advantages of pa calibration are that : ( i ) the ` target array ' observes the source continuously , and ( ii ) the demands on the antenna mechanics and electronics are less stringent than for fs . the disadvantages are that : ( i ) the electronic phase noise is not removed , ( ii ) the geometry of the array must allow for neighboring antennas , even in large ( sparsely populated ) arrays , and ( iii ) the number of visibilities from the target source array decreases quadratically with the decreasing number of antennas . these latter two effects can reduce significantly the fourier spacing coverage and sensitivity of the array . one possible solution to these problems is to have a ` calibration array ' of smaller , cheaper antennas strategically placed with respect to the antennas of the main array which is dedicated to tropospheric phase calibration by observing celestial calibrators at a fixed frequency ( eg . the solutions would then be extrapolated to the observing frequency of the main array using the linear relationship between tropospheric phase and frequency ( equation 1 ) . this would require a separate correlator and if system , and complications may arise due to the fact that the wet troposphere becomes dispersive at frequencies higher than about 400 ghz ( sutton and hueckstaedt 1997 ) . for pa and fs calibration , if the bulk of the phase fluctuations occur in a thin turbulent layer , it may be possible to perform an intelligent method of data interpolation , such as forward projection using a physical model for the troposphere and kalman filtering of the spatial time series , to account for the motion of the atmosphere across the array . the advantages of radiometric phase correction are that it : ( i ) alleviates constraints on telescope mechanics and array design , and ( ii ) the full array can observe the target source continuously . the disadvantages are that : ( i ) it places constraints on receiver lay - out such that the radiometer is always looking at the sky , ( ii ) it does not remove the electronic phase noise , and ( iii ) there remains significant questions about the viability of absolute radiometric phase correction , or the limitations imposed when using an ` empirically calibrated ' radiometric phase correction technique ( see sections 7.1 and 7.2 ) . all three calibration methods are being investigated for the mma , and we envision that all the methods will be employed to some degree at the mma , depending on the configuration , the observing conditions , and the scientific requirements of a given observation . we thank f. owen , o. lay , e. sutton , and j. carlstrom for useful comments on sections of this paper , and k. desai for allowing us to use a figure from a previous publication . this research made use of the nasa / ipac extragalactic data base ( ned ) which is operated by the jet propulsion lab , caltech , under contract with nasa . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . llllllll [ tbl-1 ] 1 & a@xmath84 mm@xmath29 & 0.0115 & 0.0073 & 2.79 & 0.670 & 0.053 & 0.0085 2 & @xmath82 & 0.0167 & 0.028 & 2.79 & 0.677 & 0.057 & 0.043 3 & t@xmath2 mk & 30 & 19 & 460 & 920 & 135 & 120 4 & t@xmath98 k & 40 & 100 & 100 & 100 & 100 & 40 5 & t@xmath99 k & 7 & 10 & 246 & 94 & 20 & 14 6 & t@xmath100 k & 47 & 110 & 346 & 194 & 120 & 54 7 & t@xmath116 mk & 1.4 & 3.5 & 11 & 6 & 3.8 & 1.7 8 & @xmath101gain & 1600 & 5800 & 750 & 210 & 890 & 450 9 & t@xmath100 k & & & 2000 & 2000 & 2000 & 10 & t@xmath117 mk & & & 63 & 63 & 63 & 11 & @xmath101gain & & & 4400 & 2200 & 15000 & 12 & @xmath101t@xmath24 k & 1.8 ( 0.7@xmath50 ) & 0.7 ( 0.3@xmath50 ) & 0.5 ( 0.2@xmath50 ) & 1.9 ( 0.7@xmath50 ) & 2.4 ( 0.9@xmath50 ) & 2.9 ( 1.0@xmath50 ) 13 & @xmath101@xmath118 k / km & 0.9 ( 14@xmath50 ) & 0.35 ( 5@xmath50 ) & 0.25 ( 4@xmath50 ) & 1.0 ( 15@xmath50 ) & 1.2 ( 18@xmath50 ) & 1.5 ( 22@xmath50 ) 14 & @xmath101h@xmath26 km & 0.28 ( 14@xmath50 ) & 0.11 ( 5@xmath50 ) & 0.08 ( 4@xmath50 ) & 0.29 ( 15@xmath50 ) & 0.37 & 0.45 ( 22@xmath50 ) 15 & @xmath101p@xmath24 mb & -5 ( 0.9@xmath50 ) & 2 ( 0.3@xmath50 ) & -7.5 ( 1.3@xmath50 ) & @xmath119 & 7 ( 1.2@xmath50 ) & -7.4 ( 1.0@xmath50 ) 16 & @xmath101h@xmath26 km & 0.07 ( 3@xmath50 ) & 0.03 ( 1.5@xmath50 ) & 0.1 ( 5@xmath50 ) & @xmath119 & 0.1 ( 5@xmath50 ) & 0.1 ( 5@xmath50 ) 17 & @xmath120 @xmath121 mb@xmath29 & -0.00133 & 0.00133 & -0.00133 & 0 & 0.00133 & -0.00133 18 & @xmath101(model w ) mm & 0.2 & 3 & 0.0025 & 0.01 & 3 & 4.8 * t@xmath100 : * total system temperature on sky . two different assumptions are made at a few frequencies , for cooled and uncooled systems . row 6 gives the case of a cooled receiver ( eg . using the astronomical receivers for radiometry ) . row 9 gives the case of an uncooled receiver . * @xmath101t@xmath24 : * required accuracy of the measurement of the atmospheric temperature ( in the turbulent layer ) in order to obtain t@xmath2 . values in parentheses indicate the percentage of the total . * @xmath101(model w ) : * differences between various model atmospheres involving different line shapes and different ` water vapor continuum fudge - factors ' relating the measured t@xmath93 to w ( sutton and hueckstaedt 1997 ) . |
we study the spiral arm influence on the solar neighbourhood stellar kinematics . as the nature of the milky way ( mw ) spiral arms is not completely determined , we study two models : the tight - winding approximation ( twa ) model , which represents a local approximation , and a model with self - consistent material arms named perlas . this is a mass distribution with more abrupt gravitational forces . we perform test particle simulations after tuning the two models to the observational range for the mw spiral arm properties . we explore the effects of the arm properties and find that a significant region of the allowed parameter space favours the appearance of kinematic groups . the velocity distribution is mostly sensitive to the relative spiral arm phase and pattern speed . in all cases the arms induce strong kinematic imprints for pattern speeds around ( close to the 4:1 inner resonance ) but no substructure is induced close to corotation . the groups change significantly if one moves only in galactocentric radius , but in azimuth . the appearance time of each group is different , ranging from 0 to more than 1 . recent spiral arms can produce strong kinematic structures . the stellar response to the two potential models is significantly different near the sun , both in density and kinematics . the perlas model triggers more substructure for a larger range of pattern speed values . the kinematic groups can be used to reduce the current uncertainty about the mw spiral structure and to test whether this follows the twa . however , groups such as the observed ones in the solar vicinity can be reproduced by different parameter combinations . data from velocity distributions at larger distances are needed for a definitive constraint . [ firstpage ] galaxy : kinematics and dynamics solar neighbourhood galaxy : evolution galaxy : structure galaxy : disc galaxies : spiral | 1106.1170 | the spiral arms of our galaxy have been typically characterised by radio observations of the 21-cm line of neutral hydrogen , giant hii regions , co emission , and optical data of young o and b stars ( e.g. @xcite ) . @xcite used fir and nir emission from the cobe / dirbe to fit a model for the spiral arms in the stellar component of the galaxy . recently , @xcite and @xcite , based on infrared data from the spitzer / glimpse survey , and @xcite , using trigonometric parallaxes and proper motions of masers in high - mass star - forming regions , reported new results about the spiral structure of the milky way ( mw ) . despite the effort , there are still several caveats in the spiral arm properties in our galaxy , such as pattern speed , strength , orientation and even the number of arms or their stellar or gaseous structure . furthermore , the nature of the spiral arms themselves , i.e. their origin or their lifetime , are nowadays a matter of debate ( @xcite ) . apart from direct methods to detect spiral arm over - densities , an alternative method is based on the analysis of kinematic groups that are induced by the spiral arms in the local velocity distribution . some of the moving groups in the solar neighbourhood were originally thought to be remnants of disrupted disc stellar clusters . however , the age or metallicity distribution of their stars are in contradiction with this hypothesis @xcite . @xcite suggested that a moving group could be a group of stars that crosses the solar neighbourhood following a certain type of orbit induced by the galactic bar gravitational potential . the demonstration that the bar resonances can produce a kinematic group , similar to the observed hercules group @xcite , has lead to the exploration of similar effects on the velocity distribution due to the spiral arms ( see @xcite for a review ) . this type of kinematic structures could depend strongly on some characteristics of the bar and spiral arms and , therefore , they result useful for our understanding of the mw large - scale structure and dynamics in its present and past form . few studies , however , have focused on the spiral arm effects on the disc velocity distribution . by integration of test particle orbits , @xcite showed that few , but intense stochastic spiral density waves produce structures which are arranged in branches , resembling those from observations @xcite . @xcite developed a method to quantify , through orbital integration , the velocity distribution that would result from the existence of spiral - induced families of periodic orbits in the solar neighbourhood . they found , for models with the sun in the outer limits of the 4:1 resonance and certain spiral arm orientation , two periodic orbits in the kinematic positions of coma berenices and hyades pleiades . the test particle simulations by @xcite showed that the combined effect of a bar and 4 weak spiral arms , is necessary to reproduce the main local moving groups ( hercules , hyades , pleiades , coma berenices and sirius ) . in the spiral - only models , more structures but less bold than those with the bar simulations were obtained . finally , @xcite showed that the angle - action variables of the stars in the hyades stream could be consistent with the effects of a recent inner lindblad resonance of a multi - arm ( number of arms @xmath4 ) and transient pattern . all this work has shown that the spiral arms are able to induce kinematic groups in the local velocity distribution . however , there is no study that explores the influence of each spiral arm property using potential models designed according to the latest observational evidences for the mw spiral structure . on top of that , previous studies have modelled the spiral arm potential with a cosine function , following the tight - winding approximation ( twa , e.g. @xcite ) . however , it is not clear to us whether the mw spiral arms satisfy the conditions for self - consistency of this model , in particular , regarding their tightness and weakness . moreover , the twa implies that the spiral arms are a steady global mode of the disc but , as mentioned , it is still a matter of discussion whether the spiral arms are long - lived or transient structures . motivated by the recent picture of the mw spiral arms and the possibility that the twa is not suitable for our galaxy , we study the spiral arm effects on the solar vicinity kinematics with the twa but also a different model , namely the perlas model @xcite . this is a 3d mass distribution from which more abrupt gravitational potential and the forces are derived . as proved in @xcite , this difference has far - reaching consequences on the gaseous dynamical behaviour . differences in the stellar response are expected too . in @xcite we presented a preliminary study of the kinematic effects of this model . here we aim to determine the conditions that favour the appearance of kinematic groups such as the ones that we observe near the sun . we explore the influence of the spiral properties and discuss to which extent we can use the kinematic imprints to constrain the spiral arm properties and nature . we also compare the local kinematic imprints of the perlas model with the ones of the twa to see whether these imprints are useful to test the spiral arm gravitational potential modelling . to do this , we first tune these two models to the latest observational determinations for the properties of the mw spiral arms . for some properties only a range of evidence is available in the literature . then we perform test particle simulations with both models considering these ranges , different initial conditions and integration times . section [ modelmw ] deals with the observed properties of the mw spiral structure . in section [ modeling ] we elaborate on the gravitational potential modelling of the spiral arms . section [ models ] presents the two spiral arm models and we fit them to the mw spiral arms . in section [ force ] we contrast the force fields of the two models . section [ sim ] describes the test particle simulations and initial conditions . afterwards , in section [ comparison0 ] we compare the response in density and kinematics to the two models . we explore the influence of spiral arm characteristics , such as pattern speed , orientation and integration time , on the local velocity distribution in sections [ patternspeed ] , [ maximum ] and [ time ] . in section [ constrain ] we discuss whether it is currently possible to constrain the spiral arm properties by using the local observed kinematic groups . finally , we summarise our main results and their implications . most of the properties of the mw spiral structure remain rather undetermined . here we establish a range of plausibility for each property in order to tune our spiral arm models to the mw spiral arms . table [ tab : mwarms ] shows these adopted ranges . * * the geometry of the mw spiral arms is still a matter of intense debate . different estimates of the pitch angle and even the number of arms can be found from different tracers ( see @xcite ) . maps of ob - associations and hii - regions , co emission , masers in high - mass star - forming regions ( @xcite , @xcite , @xcite and references therein , @xcite ) show a 4-armed pattern , usually referred as sagittarius carina , scutum centaurus ( or scutum crux ) , perseus and the outer ( or cygnus ) arms . however , according to cobe k - band observations @xcite and to the infrared spitzer / glimpse survey @xcite , only two of these four arms are major mw arms . these are traced by the stellar ( young and mainly old ) population as enhancements in the spiral tangencies . external galaxies often show different morphologies in blue and near - infrared colours ( grosbl , patsis & pompei 2004 ) . although it might not be a general property of galaxies @xcite , many galaxies classified as flocculent or multi - armed systems in blue , display a 2-armed grand design spiral in the k band ( @xcite , kendall , kennicutt & clarke 2011 ) . also hydrodynamic models have shown that it is possible to form arms of compressed gas , without increasing the stellar surface density , as a response to a 2-armed pattern @xcite . here we consider only these two stellar major arms ( @xmath5 ) as they trace the underlying mass distribution of our galaxy . and @xmath6 indicates the assumed solar position . open circles indicate several regions near the solar neighbourhood.,scaledwidth=40.0% ] the exact location of the stellar spiral arms is still unclear and part of the current models in the outer regions of the mw are an extrapolation of observations in the inner regions . @xcite , @xcite and @xcite consider that the stellar arm counterpart of scutum centaurus is the perseus arm . but according to the model by @xcite , the counterpart goes far beyond the perseus arm in the anti - centre direction . because of these discrepancies , we will adopt two different locus : the one reported by @xcite ( hereafter locus 1 , solid black curve in fig . [ f : mw ] ) , and the fitting by @xcite for the scutum and perseus arms ( hereafter locus 2 , dashed black curve in fig . [ f : mw ] ) . .assumed values or ranges for the properties of the mw spiral arms . [ cols="<,<,^",options="header " , ] to fit this model to the mw spiral arms we use the locus and parameters as discussed in section [ modelmw ] . additional parameters for this model are shown in table [ tab : arms11 ] . in this case , the spiral amplitude is quantified through the value @xmath7 which is the ratio of the spiral mass to the mass of the disc in the a&s model . in the perlas model the spiral arms are added as an mass enhancement to the axisymmetric background on the imposed locus . because of this , for the perlas model the mass of the spiral arms ( @xmath8 ) is globally subtracted from the original disc of the a&s model ( @xmath9 ) , which guarantees that the total mass of the model do not change and neither does the mean circular velocity . for this reason , the parameter @xmath10 is more suitable to be related with the @xmath7 , as it is the ratio of the surface density on the spiral arm ( axisymmetric disc and spheroids ) to the minimum density ( axisymmetric disc ) . as for the twa model , the density contrast @xmath10 decreases with radius . for locus 1 , the range of spiral mass ratio that fits the observational range of density contrast @xmath10 at @xmath11 is @xmath120.030.05 ( fig . [ f : contrastperlas ] ) . for locus 2 , the respective range is and for locus 2 with a more tightly wound arm , the arm longitude is larger than for locus 1 . this implies that a higher total spiral mass is necessary to reproduce the desired contrast . ] @xmath130.0350.06 additionally , the beginning of the spheroid superposition , i.e. the effective arm beginning , is at @xmath14 and @xmath15 , respectively for locus 1 and locus 2 . the superposition ends at @xmath16 for both cases . the arm half - width and height above the plane are taken to be @xmath17 and @xmath18 , respectively , according to the analysis of a sample of external galaxies @xcite and the discussion in @xcite . the main difference between the perlas spiral arms and the twa falls on the construction itself . in the twa , the spiral arms are a small perturbative term of the potential . this produces a cancellation of the contribution from the distant parts of the pattern to the local force . the perlas spiral arms correspond to an independent mass distribution . the contribution from the entire spiral pattern causes the spiral potential and force to adopt shapes that are not correctly fit by the simple twa perturbing term that has been traditionally employed . in this section we show these fundamental differences in detail by studying the force field exerted by these two models using the fourier decomposition . if not stated the contrary , the plots and values in this section are referred to the lower limit of the density contrast defined in section [ modelmw ] . for @xmath19 of equation ( [ eq : qr ] ) . long dashed line is the ratio of the radial axisymmetric part of the force of perlas model to the total axisymmetric radial force @xmath20.,scaledwidth=40.0% ] first we study the parameter @xmath21 , which is the @xmath22 term of fourier decomposition of the non - axisymmetric radial force scaled to the axisymmetric radial force . this is : @xmath23 where primes denote derivatives with respect to @xmath24 , and @xmath25 is the @xmath22-component of the fourier decomposition of the global potential , that is of the potential of the spiral arms plus the a&s potential . in particular for @xmath26 , @xmath27 corresponds to the axisymmetric part of the global potential . the twa model consists of a pure and simple @xmath5 component . therefore , the spiral arms only contribute in the numerator of equation ( [ eq : qr ] ) , and the axisymmetric model a&s is the denominator . by contrast , the perlas arms consist of a more complex potential structure with more than a @xmath5 term . in fig . [ f : decomp ] we show the fourier decomposition of the perlas potential . in particular , the long dashed line shows the ratio of the axisymmetric component of the radial force of the perlas model @xmath28 to the axisymmetric global radial force @xmath29 . we see that the axisymmetric radial force of the perlas model can be as large as 4% of the total axisymmetric force ( and up to 7% for high limit of the density contrast ) . this important @xmath26 component contributes , toguether with the a&s model , to the axisymmetric part of the global potential ( denominator in equation ( [ eq : qr ] ) ) together with the a&s model . in the same plot we see that for this model the @xmath5 component is dominant but it also has non - vanished @xmath4 terms @xmath21 . as a function of radius for the twa and the perlas model.,scaledwidth=40.0% ] in order to quantify the whole non - axisymmetric force of our spiral arm models we define the parameter : @xmath30 which is the ratio of all the non - axisymmetric radial force terms to the axisymmetric part of the global potential . which , contrary to the present study , only includes the @xmath5 component . ] for the twa , @xmath31 . but for the perlas model it is the contribution of all terms @xmath21 , with @xmath32 , of fig . [ f : decomp ] . [ f : qrsp ] shows @xmath33 as a function of radius for the two models . the range of the radial force amplitude of the two models is approximately the same as they are fitted to reproduce the same density contrast . however , the detailed shape of @xmath34 is different due to the important contribution of all terms @xmath32 in the perlas model . from this figure we also observe that the maximum radial force of perlas spiral arms with respect to the total axisymmetric radial force is 0.09 ( absolute value ) . around @xmath35 the parameter @xmath36 is @xmath37 . for the twa , the maximum is 0.14 and the value at solar radius is 0.03 . for the higher limit of the density contrast range , the shape of @xmath34 does not change . but in that case , the perlas model produces a maximum radial force with respect to the axisymmetric background of 0.16 and a value near the solar radius of 0.005 . for the twa the maximum value is 0.21 and the value at the solar radius is 0.04 . as a function of radius for the twa and the perlas model.,scaledwidth=40.0% ] the parameter @xmath38 was used in @xcite to quantify the tangential force of bars . here we use it for spiral arms as in @xcite . it is the ratio of the maximum spiral arm tangential force at a given radius to the mean total radial axisymmetric force at that radius : @xmath39 the parameter @xmath40 includes the tangential forces due to all terms @xmath32 for the perlas model . this parameter is shown in fig . [ f : qtsp ] . the large and small bumps at @xmath41 and @xmath16 for the black curve are due to the more abrupt beginning and end of the arms in the perlas model but have no consequences for our study that is focused on near solar radius . the amplitude range of @xmath42 is rather similar in the radii of interest in this study ( 710@xmath43 ) . the maximum of @xmath38 gives a single and quantitative measure of the torque or strength of the spiral arms @xcite and is called @xmath44 . for the perlas model we find that @xmath45 and for the twa it is 0.039 . for the maximum density contrast , @xmath46 is 0.17 for the perlas model and 0.060 for the twa . in @xcite the spiral strengths @xmath46 in a sample of 15 external galaxies range from 0 to 0.46 . according to this , the models for the mw spiral arms are below the median of external galaxies . for both models @xmath47 is @xmath48 for the lower density contrast limit and @xmath49 for the high case . for the twa and perlas model . the forces are scaled to the a&s radial axisymmetric force.,scaledwidth=45.0% ] , but as a function of azimuth for two different radius @xmath24.,scaledwidth=45.0% ] the parameters @xmath33 and @xmath40 measure average and maximum values at a given radius , respectively . we now study the detailed force profile . figs . [ f : frspiral ] and [ f : faspiral ] show the radial and tangential forces of the two models as a function of azimuth @xmath50 and radius @xmath24 scaled to the axisymmetric radial force of the a&s model . we see significantly different force fields . in general , the perlas spiral arms ( solid line ) presents more abrupt features than the twa . it also has different locations ( @xmath24 , @xmath50 ) for the minima and maxima for both the radial and tangential forces . the positions at which the force changes its sign is different as well . besides , the radial force of the perlas model is in general below the force for the twa , that is non - symmetric with respect to 0 and shifted to negative values . this is due to the inner enclosed spiral mass at each radius , that is the @xmath26 component of the perlas spiral force mentioned before . for instance , in the top right panel in fig . [ f : faspiral ] ( solar radius @xmath51 ) the radial force is negative for all @xmath50 , whereas the twa consists of a symmetric oscillation around 0 force . to conclude , it is quite different to model the gravitational potential by using the twa ( i.e. a cosine function , that fits a given locus and certain density contrast ) or by building a mass distribution on the same locus and contrast . we have seen that the two approaches to model the spiral arms gravitational potential give significantly different force profile . in next sections we study how these differences propagate to the kinematic stellar response . in order to study the effect of the spiral arms on the kinematic distribution near the sun , we perform numerical integrations of test particle orbits , as most of the studies up to now . these are simple models and their self - consistency is not assured ( section [ modeling ] ) . by contrast , models such as n - body simulations of galaxy formation could model self - consistently the spiral arm kinematic effects , as well as include naturally the bar or spiral arm evolution and other processes that might also sculpt the velocity distribution ( e.g. past accretion events , star formation bursts ) . however , n - body simulations with larger than current number of disc particles , better spatial and temporal resolution , and models similar to the mw are required . at this moment , test particle simulations allow us to use models that fit the mw spiral arms in a controlled manner . they are simpler and faster models that offer easy exploration and understanding of the influence of the spiral arm properties . for our simulations we adopt the potential models of section [ models ] and the initial conditions described in section [ ic ] . in section [ procedure ] we give more details about the method . our study focuses on the kinematic effects produced on or near the plane . for this we assume that the vertical motion is decoupled with the in - plane movement . this is reasonably true for nearly circular orbits that do not take larger height above the plane @xcite . as our initial conditions consist of rather cold discs ( see section [ ic ] ) , we adopt these assumptions and simplify our analysis by considering 2d simulations ( @xmath52 ) . the @xmath53 velocity reference system is used which is centred on a given position on the disc plane and moves following the regional standard of rest ( rsr ) . this is defined as the point located at a radius @xmath24 that describes a circular orbit around the galactic centre with a constant circular speed @xmath54 . @xmath55 is the radial velocity component , which is positive towards the galactic centre , and @xmath56 is the tangential component , positive in the direction of galactic rotation . we explore two different types of initial conditions : ic1 & ic2 . both are axisymmetric discs , following an exponential profile with scale length of @xmath57 . this is similar to the thin disc scale length found by @xcite from cobe / dirbe data . ic1 the velocity distribution relative to the rsr is adopted as a gaussian with low dispersions @xmath58 and constant for all radii . these values are similar to the induced gas velocity dispersions in the plane due to galactic spiral shocks @xcite and to the dispersion of the youngest hipparcos stars @xcite . with this ic we aim to simulate a cold young disc . ic2 for these ic , the phase space distribution function ( df ) is constructed , as discussed in @xcite . the velocity field is approximated using the moments of the collisionless boltzmann equation , simplified by the epicyclic approximation . following @xcite , we will adopt @xmath59 proportional to the surface density , and therefore , the radial velocity dispersion profile is : @xmath60 the local normalisation is chosen @xmath61 . the tangential velocity dispersion profile is derived from the epicyclic approximation ( equation 4.317 of @xcite ) . also we take into account the asymmetric drift ( equation 4.228 of @xcite ) . assuming that the disc is stationary , axisymmetric and symmetric about its equator , and that the orientation of the velocity ellipsoid is aligned with the coordinate axes , it is : @xmath62.\ ] ] this df confers to the ic2 disc the properties of an intermediate population ( @xmath63 ) of the mw thin disc @xcite . in our simulations the integration of each particle is initialised at a time @xmath64 ( as a convention ) and ends at t@xmath65 , being @xmath66 the particle exposure time to the potential . the time @xmath66 is chosen at random between @xmath67 and @xmath68 for each particle . this maximum integration time corresponds to @xmath69 revolutions of the spiral arms for a typical pattern speed of @xmath70 . with this procedure , the final velocity distributions result of a superposition of particles integrated different times , resembling the observed distribution , with a superposition of stars of different ages . in some cases , we separate the particles into different bins of integration times , to study the induced kinematic effects as a function of time . the integration of the motion equations is done with the bulirsch stoer algorithm of @xcite , conserving jacobi s integral within a relative variation of @xmath71 . the reference frame used for the calculations is the rotation frame of the spiral arms . the number of particles in each simulation is about @xmath72 . after the integration , we study the induced kinematic distribution near the sun . for the analysis , we focus on the @xmath55@xmath56 plane of the 15 circular regions indicated in fig . [ f : mw ] , which have @xmath73 of radius ( similar to the radius of the available observed local velocity distribution , see fig . 1 in @xcite ) . the centres of the regions are located at five different radius ( 7.3 , 7.9 , 8.5 , 9.1 and @xmath74 ) separated @xmath75 from the nearest ones , and at azimuths @xmath76 , @xmath77 and @xmath78 , which is found to be the approximate azimuth interval that shows significant differences between adjacent @xmath55@xmath56 planes . the symmetries of the galactic potential ( @xmath79=@xmath80 ) allow us to double the number of particles in each studied region . in all cases , the number of particles in the final velocity distributions is larger than @xmath81 , being statistically robust . fixed the spiral arm pattern speed , these 15 regions have different ratio @xmath82 , that is , the pattern speed scaled to the local azimuthal frequency of that region . these are representative of different pattern speeds , allowing us to explore small variations in this parameter . also they have different relative spiral phase @xmath83 which aims to explore the range of uncertainty in this parameter , as explained in section [ modelmw ] . we apply the wavelet denoising method ( wd ) to the velocity distributions . this method gives a smooth df from a discrete point distribution via a smoothing / filtering treatment at different scales , that eliminates poisson fluctuations ( for details , see * ? ? ? the results will be shown with logarithmic colour scale and contours representing , from inside out , 0.01 , 0.05 , 0.10 , 0.20 , 0.30 , 0.40 , 0.50 , 0.6 , 0.70 , 0.8 , 0.9 and 0.999 of the maximum density . we have run simulations for pattern speeds @xmath84 of 15 , 18 , 20 , 22 , 25 and @xmath85 ( inside the observational range of evidence , section [ modelmw ] ) for both models . we have used the two limits of density contrast , the two loci and the different initial conditions detailed in previous sections . here we compare the results for the two models . first , in section [ response ] we analyse the density response to the spiral models in our simulations . with this we aim to study the consistency between density response and gravitational potential of the models , and to relate the density response to the kinematic analysis of subsequent sections . [ f : response ] shows the density distribution of a simulation with the twa ( top ) and the perlas model ( bottom ) taking the same conditions ( pattern speed , locus , etc . ) for both models . darker regions show the denser regions with the same scale in all panels of the figure . first and second column correspond to short and long integration time , respectively . blue stars show the positions of maximum density in rings of @xmath86 of radius . the white curves indicate the locus of spiral arm model . inner and outer red dashed circles show the approximate positions of the 4:1 inner resonances and corotation , respectively . first row shows that the density response to the twa with the minimum density contrast and a pitch angle of @xmath87 does not follow the spiral locus imposed in the potential approximately beyond the 4:1 inner resonance . the spiral arms in this simulation show a more tightly wound pattern at outer radius for short integration time and high dispersion for long integration times . right panel also shows complex structures like rings and low density regions . the different pattern speeds and amplitudes in the ranges established for the mw give equivalent results . @xcite already studied the density response for several spiral models based on the twa , mainly focusing on periodic orbits . they concluded that strong spirals can not exist beyond the 4:1 inner resonance as in this disc region the spiral response is out of phase due to non - linear effects . by contrast , weak or tightly wound spirals can exist up to corotation , as the linear theory predicts @xcite . our results are in agreement with the results of @xcite for strong spirals . an exhaustive analysis of the response density is out of the scope of this paper . but the difficulties in using the twa tuned to the mw as a self - consistent model beyond the 4:1 inner resonance , poses important constraints in the use of the twa for the mw . for certain pattern speeds , this would imply a too short ending radius of the spiral arms ( e.g. the spiral arms could end at @xmath88 for a pattern speed of @xmath85 ) .. ] to conclude , the twa might not be a proper description for the mw spiral arms for some parameter combinations ( pattern speed , pitch angle , amplitude ) . the density response to the perlas model ( second row of fig . [ response ] ) shows also disrupting effects near the inner 4:1 resonance but , at larger radius the density response does follow the imposed spiral arms for small and large integration times . indeed , with larger pattern speeds , the response follow the imposed locus even beyond corotation . nevertheless , the orbital self - consistence of the perlas model is only assured for a certain range of parameters , specially in mass and pattern speed ( see @xcite ) . .,title="fig:",scaledwidth=40.0% ] .,title="fig:",scaledwidth=40.0% ] for the perlas model we also see complex density structures such as two extra arms for short integration times or low density regions and rings in the long integration case . all these patterns that are observed for both models could be related to the different resonances of the spiral arms and might well represent rings and spiral arm spurs that are seen in our galaxy and in external galaxies @xcite . for instance , two emptier regions are located on the 4:1 inner resonance at @xmath89 in the right panels . notice , however , that these appear at different azimuths depending on the model ( near @xmath90 for the perlas model but at an azimuth of @xmath91 for the twa ) . in fact , note that the induced density in our region of study ( near solar positions ) is significantly different between models . [ f : densitylocal ] is the local density response for all particles of the simulation . the twa model ( top ) presents softer density gradients , whereas the perlas model ( bottom ) shows the mentioned emptier region at solar azimuth . the local density response of the two models is different in all our examined cases , emphasizing the distinct force fields studied in section [ force ] and forecasting the differences in the induced velocity distribution near solar positions ( section [ comparison ] ) . simulations with ic2 show qualitatively equivalent results to the ones with ic1 for both models , except for the fact that the density features appear less sharp but more diluted , as expected for a hotter and , therefore , less responsive disc . for both models we observe a transient nature of the density patterns . in all cases the density response becomes stationary approximately between 400 @xmath92 and 1200 @xmath92 of integration time depending e.g. on the pattern speed ( except for outer radius @xmath93 , out of the region of study ) . similar time dependence will be observed for the kinematic distributions of these models ( section [ time ] ) . @xmath56 velocity distributions for the 15 regions of fig . [ f : mw ] for the simulations with the twa model with locus 1 , maximum density contrast , ic1 and a pattern speed of @xmath94 . the 15 panels are positioned in the figure similarly to their spatial location of fig . [ the scale in the right indicates the spatial position of the spiral resonances ( see text ) . ] @xmath56 velocity distributions for the 15 regions of fig . [ f : mw ] for the simulations with the perlas model with locus 1 , maximum density contrast , ic1 and a pattern speed of @xmath94 . ] in this section we compare the stellar kinematic response to the perlas model and the twa . as an example , we now adopt locus 1 , maximum density contrast , ic1 and a pattern speed of @xmath94 for both cases . the @xmath55@xmath56 plane for the regions of fig . [ f : mw ] are shown in figs . [ f:18twa ] and [ f : speed18 ] for the twa and perlas models , respectively . the 15 panels of each figure are positioned similarly to their spatial location of fig . . each panel corresponds to a region with different relative spiral phase @xmath83 and ratio @xmath82 , which are indicated in each panel . first we see that both models induce several and rich kinematic substructures near solar position . nevertheless , we see that the details and induced groups at each of the 15 regions are different for each model , in agreement with the differences between induced density ( section [ response ] ) . for instance , an elongated group at low @xmath56 appears for the left panel of the last row for the twa model but it is present in all panels in this row ( several spiral phases ) for the perlas model . the central part of these panels is also fairly different . the rest of the panels also show groups at different positions depending on the model . we have detected that the two models occasionally give a similar @xmath55@xmath56 plane when comparing different disc positions with comparable density distribution . for instance , we see similar velocity distributions near the emptier region on the 4:1 inner resonance ( section [ response ] ) , which is is located at different azimuths depending on the model . we attribute this occasional shift to the differences in the force fields of both models ( section [ force ] ) . we have seen that maxima , minima and 0 points of the force profiles ( both tangential and radial ) are locally shifted in radius and azimuth when comparing the two models . for example , in fig . [ f : faspiral ] we see that the maximum tangential force of the perlas model is shifted to larger azimuths compared to the twa . on the contrary , the minima are shifted to smaller azimuths . the link between phase space , force profile and resonance locations involves a large - scale study of the models which is out of the purpose of the present paper , but is currently under study ( antoja et al . , in preparation ) . as common aspects of the kinematic response of the two models , we notice variety of shapes , sizes , inclinations and positions of the induced substructures , for this and other pattern speed ratios . for example , notice more or less rounded groups but also thin elongated structures . these and other simulations have shown that both models induce structures at @xmath56 as high as @xmath95 and as low as @xmath96 . the @xmath55 component of the induced kinematic planes can range from -80 to @xmath97 . the number of induced groups depends on the region but is in most cases 2 or 3 . interestingly , the kinematic groups are found to be in general more symmetric with respect to the @xmath98 axis for the twa than in the perlas model . more important , by inspecting these kinematic plots , we see that the perlas model induces kinematic substructure where the twa model does not ( e.g. first and second rows ) . the twa gives substructure for a smaller range of pattern speeds . we examine this in detail in section [ maximum ] . in the following sections , we study the influence of the spiral arm properties on the kinematic distribution near the solar position . we focus basically on the perlas model but we contrast the results with the ones obtained by the twa model when important differences arise . but for a spiral pattern speed of @xmath99 . ] but for a spiral pattern speed of @xmath70 . ] figs . [ f : speed15 ] and [ f : speed20 ] show the contour plots of the @xmath55@xmath56 plane for the 15 regions shown in fig . [ f : mw ] of the simulations corresponding the pattern speeds of @xmath99 and @xmath70 , respectively , for the perlas model . together with the already presented fig . [ f : speed18 ] , these panels allows us to study the effects of several pattern speed ratios @xmath82 and relative spiral phases @xmath83 . in all cases we use locus 1 , maximum density contrast and ic1 . the panels show rich kinematic substructure near the solar position , strongly depending on the pattern speed ratio and relative phase of the spiral arm . the kinematic substructures change more slowly with azimuth ( along rows ) than with radius ( along columns ) . the changes are significant if one moves only @xmath1 in radius . but @xmath100 are needed in azimuth to detect important differences in the @xmath55@xmath56 plane . the kinematic groups slightly change position on the @xmath55@xmath56 plane with azimuth or radius . for instance , for a pattern speed of @xmath101 ( fourth row in fig . [ f : speed18 ] ) the velocity distributions show basically two groups , elongated in the @xmath55 direction , which appear in different positions in the @xmath55 component depending on the azimuth . there are equivalent @xmath55@xmath56 plane plots in different simulations . for instance , top left panel of fig . [ f : speed15 ] and third panel of the fourth row of fig . [ f : speed18 ] correspond to similar relative pattern speed @xmath82 and spiral phase @xmath83 . however , moving in radius is not strictly equivalent to changing the pattern speed . first the spiral strength or density contrast depends on radius ( see section [ models ] ) . second , although the initial velocity dispersion of particles in ic1 is flat with radius , at the end of the simulation this increases for inner radius , as we have seen with a simulation only including the axisymmetric part of the model . one example is the group at @xmath102 in the right panel in the last row fig . [ f : speed18 ] which is not present in the equivalent panel of fig . [ f : speed15 ] ( left panel of the second row ) . the smaller strength of the spiral at outer radius and/or the smaller velocity dispersion do not allow to crowd this structure for the simulation in fig . [ f : speed15 ] . but for locus 2 . ] now we compare the results with locus 1 and 2 ( fig . [ f : mw ] ) . [ f:18locus2 ] shows the same case as in fig . [ f : speed18 ] but using locus 2 . the small difference between the pitch angles of these loci ( @xmath87 and @xmath103 , respectively ) , produces essentially the same kinematic structures when we compare regions with the same ratio @xmath82 and relative spiral phase @xmath83 . few differences are observed such as the group at low @xmath56 for left panel of the last row in fig . [ f:18locus2 ] that does not appear in the equivalent relative position for locus 1 ( approximately middle panel of last row of fig . [ f : speed18 ] ) . these subtle differences are due to the tightness of the spiral arms for locus 2 . most of the 15 regions considered here correspond to interarm regions . when we analyse the velocity distributions in different regions of the galactic disc , we see that the induced kinematic substructure increases near and on the spiral arms ( antoja et al . , in prep . ) . for locus 2 , we find more substructure than for locus 1 for the twa because for this locus some of the regions are closer to the spiral arms . by contrast , for the perlas model and the same pattern speeds , rich substructure is found for both locus . but for the initial conditions ic2 . ] as expected the hotter population defined by the ic2 initial conditions ( section [ ic ] ) does not respond so strongly to the spiral perturbation . the main groups are still observed as the only changes are in the way each structure is populated but not in the orbital structure . the signal of some fine details slightly fades or appears less defined for the hotter case . we show an example in fig . [ f:18ic2 ] . notice that the bi - modality in the first panel of the first row in fig . [ f : speed18 ] that does not show up for ic2 . other structures manifested as separated groups are now stickied with particles filling the gaps in between . changing the pattern speed @xmath84 of the spiral arms corresponds to changing the position of the spiral resonances with respect to the regions under study . these resonances can be , as shown in other studies @xcite , a major influence on the orbital structure . a detailed orbital analysis would be needed to determine the exact influence of each of the resonance in sculpting the velocity plane . here , we examine the effects of the resonance proximity to the 15 regions of study . in the right part of e.g. figs . [ f : speed18 ] , [ f : speed15 ] and [ f : speed20 ] we show a scale in galactocentric radius ( configuration space ) where we indicate the approximate radius of the main resonances . we also show in this scale the limits of the regions corresponding to the 3 adjacent velocity distributions . we determine that clear and rich kinematic substructure is seen near solar azimuth for the perlas model for pattern speeds from 0.5 to 0.75 times the angular rotation rate . with a local circular frequency of @xmath104 , this is for pattern speeds from 13 to 19.5 @xmath105 at solar radius . this range of pattern speeds corresponds to being near the inner 3:1 , 4:1 and 5:1 resonances . we also see that for pattern speed ratio of 0.760.85 ( @xmath84 approximately between @xmath106 and @xmath107 ) we detect minor groups , that is , few and close to each other , or only small deformations of a unique clump centred in the @xmath55@xmath56 plane ( e.g. second and third rows in fig . [ f : speed20 ] ) . in these cases , the regions are located especially near the 6:1 inner resonance or high order @xmath22:1 resonances ( with @xmath108 ) . therefore , we attribute weaker effects to these resonances . for pattern speed ratio approximately between 0.86 and 1.2 ( @xmath84 between @xmath109 and @xmath110 ) we observe no substructure at the considered azimuths ( e.g. upper row in fig . [ f : speed20 ] ) . this is near corotation and high order resonances ( inner and outer @xmath22:1 resonances with @xmath111 ) , concluding that these resonances have no effect on the velocity plane at solar azimuths and for the maximum spiral arm density contrast . , which is also out of the mw range . in this latter case , the effects are due to the approaching main outer resonances 4:1 , 5:1 and 6:1 . ] but for the minimum spiral density contrast . ] if we compare the results for the maximum density contrast with the lower spiral mass limit , the velocity distributions are similar but we see in most cases less populated or / and smaller kinematic groups for the low density case . as an example we examine in fig . [ f:18lowmass ] the same case as in fig . [ f : speed18 ] but using the lower observational limit for the density contrast . notice how the structures are essentially the same but the groups in the extremes of the @xmath55@xmath56 plane ( far from the @xmath112 point ) disappear for the weaker spiral arms . for instance , the groups at low @xmath56 of last row of fig . [ f : speed18 ] are not populated in fig . [ f:18lowmass ] . we observe this trend also for other pattern speeds . the range of pattern speeds where we find significant kinematic substructure is reduced with respect to the maximum density contrast case ( 0.50.75 ) and it now approximately 0.580.75 . thus being near the 3:1 inner resonances has lower influence for weaker arms . by examining more simulations , we see that the twa gives less substructure than the perlas model , given the same density contrast . it also gives clear and rich kinematic groups near solar azimuth for a smaller range of pattern speeds . specifically , for the maximum density contrast , this range is 0.600.73 ( pattern speed approximately from 15.5 to 18.8 @xmath105 ) . this is for regions close to the 4:1 and 5:1 inner resonances . for smaller pattern speeds , only minor substructures or small deformations of the velocity distributions are observed , contrary to the perlas model . for larger pattern speeds , the induced kinematic groups disappear abruptly for the twa , but progressively for the perlas model . for a fixed pattern speed , all these lead to a smaller range of radius where substructure can be seen for the twa . for pattern speed ratios larger than 0.8 ( near corotation and high order resonances ) , we observe the same behaviour as in the perlas simulations . the comparison between both models for the minimum density contrast is more drastic . the twa model induces clear substructure only for ratios around 0.650.7 ( particularly close to the 4:1 resonance ) . for the rest of the pattern speeds , almost no substructure is induced at the considered azimuths . previous results referring to the pattern speeds that give important or minor substructure do not depend substantially on the initial velocity dispersion or on the use of locus 1 and 2 . a fundamental question is if spirals are long or short - lived features . the processes that induce spiral arms play a key role in this @xcite . with the aim to see whether the velocity distributions can show imprints of the spiral lifetime , in this section we study the dependence of the structures on integration time . in our simulations , we have not modelled the progressive appearance of the spiral arms but they are included all time . in fact , some experiments showed that , given the same initial conditions , the final orbital structure does not change significantly with the adiabatic introduction of the non - axisymmetric component ( pichardo , private communication ) or with different growth times @xcite . in fig . [ f : timeperlas ] , along columns , we show the time evolution of the @xmath55@xmath56 plane for 5 different representative cases ( the pattern speed and relative spiral phase are indicated in the first panel of each column ) . we show the results for time integration intervals of 400 @xmath92 from 0 to 2@xmath3 , which is approximately one revolution period of the spiral arms . in these panels we see more separated groups compared to previous plots and , in general , the structures become separated with time . structures in previous plots looked more continuous because they were composed by particles that had been integrated for a mix of times ( between 0 and 2@xmath3 ) . the @xmath55@xmath56 plane of these simulations changes with time . this is due to the ongoing phase - mixing in response to spiral arm potential . we see that the time of appearance is different for each group , ranging from 0 to @xmath113 . for instance , we observe two conspicuous groups in the first panel of second column ( integration time smaller than 400 @xmath92 ) . others appear later , as the branch at @xmath114 for the period 8001200 @xmath92 in the model of second column . we observe how the structures at low @xmath56 or the ones at the extreme parts of the @xmath55@xmath56 plane require more integration time , i.e. a larger spiral lifetime . this is because they correspond to more eccentric orbits and larger radial excursions that take more time to reach the current region . we have also found that the time of appearance does not depend strongly on the density contrast of the spiral arms . on the other hand , a hotter disc ( ic2 ) populates some structures at earlier times , specially at low @xmath56 . we see also that most structures change shape , size and position in the @xmath55@xmath56 plane with time . some are formed in the central parts of the @xmath55@xmath56 plane and become progressively detached with time . notice for example how the left structure in the third column shifts to negative more @xmath55 . see also the branch at low @xmath56 in the second column which becomes more populated and more extended in the @xmath55 direction for large integration times . in general after 1200 @xmath92 of integration time ( 34 revolutions of the spiral arms ) , the @xmath55@xmath56 plane becomes stationary ; for this and larger times all the groups have already appeared and preserve the same size , shape and position in the kinematic space . similarly , we saw that the global density structure became stationary after integration times between 400 and 1200 @xmath92 ( section [ response ] ) . in previous sections we have explored the effects of the spiral properties on the local velocity distribution . one would like to know which model parameter combination fits better the observed kinematic groups in the local @xmath55@xmath56 plane . here we discuss whether this constrain is currently possible . @xmath56 velocity distribution of the solar neighbourhood ( data from @xcite ) . right : wavelet transform for a scale of @xmath115 of the same distribution.,title="fig:",scaledwidth=23.0% ] @xmath56 velocity distribution of the solar neighbourhood ( data from @xcite ) . right : wavelet transform for a scale of @xmath115 of the same distribution.,title="fig:",scaledwidth=23.0% ] the observed heliocentric velocity distribution of the solar neighbourhood is shown in fig . [ f : observed ] ( left panel ) . this is obtained from the analysis in @xcite but limited to stars with good velocity determinations ( errors in all components @xmath116 @xmath117 ) . we have applied the wavelet transform ( see @xcite ) to this distribution with a scale of @xmath115 ( right panel of fig . [ f : observed ] ) , which highlights the kinematic structures of this size . the main groups of this distribution are sirius ( 1 ) , coma berenices ( 2 ) , pleiades ( 3 ) and hyades ( 4 ) , groups 5 and 6 that account for the elongation of hercules , and finally groups 7 and 8 . to obtain velocities with respect to the lsr , we consider a solar motion of @xmath118 ( schoenrich , binney & dehnen 2010 ) . error bars of @xmath119 in this plot account for the uncertainty in the definition of the exact group velocity . @xmath56 velocity distributions for several models for the perlas model with locus 1 , maximum density contrast and ic1 . panels in the bottom show examples of the same model with different integration times . orange points with error bars indicate the approximate positions of the local observed groups.,title="fig:",scaledwidth=15.0% ] @xmath56 velocity distributions for several models for the perlas model with locus 1 , maximum density contrast and ic1 . panels in the bottom show examples of the same model with different integration times . orange points with error bars indicate the approximate positions of the local observed groups.,title="fig:",scaledwidth=15.0% ] @xmath56 velocity distributions for several models for the perlas model with locus 1 , maximum density contrast and ic1 . panels in the bottom show examples of the same model with different integration times . orange points with error bars indicate the approximate positions of the local observed groups.,title="fig:",scaledwidth=15.0% ] @xmath56 velocity distributions for several models for the perlas model with locus 1 , maximum density contrast and ic1 . panels in the bottom show examples of the same model with different integration times . orange points with error bars indicate the approximate positions of the local observed groups.,title="fig:",scaledwidth=15.0% ] @xmath56 velocity distributions for several models for the perlas model with locus 1 , maximum density contrast and ic1 . panels in the bottom show examples of the same model with different integration times . orange points with error bars indicate the approximate positions of the local observed groups.,title="fig:",scaledwidth=15.0% ] @xmath56 velocity distributions for several models for the perlas model with locus 1 , maximum density contrast and ic1 . panels in the bottom show examples of the same model with different integration times . orange points with error bars indicate the approximate positions of the local observed groups.,title="fig:",scaledwidth=15.0% ] we find several models , all within the established range of evidence for the mw spiral arms , that fit some of the observed groups in the solar neighbourhood , without any preference between these different `` good '' models . we also find that one particular group can be induced by different parameter combinations . we show in fig . [ f : constrain ] several models with the perlas spiral arms that lead to kinematic groups similar to some of the observed ones . first panel in the first row fits coma berenices and presents elongations towards hyades and groups 7 and 8 . second panel corresponds to the same pattern speed but a slightly different spiral phase and reproduces better hyades and sirius , showing also a small elongation towards group 7 . third panel seems to fit coma berenices and deviations towards pleiades . other good fits have been found for the twa model . panels in the bottom of fig . [ f : constrain ] correspond to the same model but different integration times . we see that especially for integration times from 400 to @xmath120 the model is able to reproduce hyades and sirius . this shows that the spiral lifetime could be also constrained . but the fact that other combinations of parameters have reproduced successfully these kinematic groups prevents us to adopt this as the best mw fitting . notice that different authors have considered different observed groups and slightly different positions on the @xmath55@xmath56 plane due e.g. to different sample selections , or different solar motion determinations such as the one by @xcite ) . this ambiguity , however , does not change the conclusions of this section and , in fact , is an additional limitation to constrain the model parameters . this degeneracy is also noticed in the literature , where different models have led to the reproduction of the same group . for instance coma berenices have been related to the effects of the spiral arms @xcite and to the effects of the bar @xcite . our simulations have shown that the parameter space is still large to obtain a unique best - fitting to the observed velocity distribution . apart from the mentioned current degeneracy , we should not expect an exact equivalence between the positions of the observed groups and the velocity distributions from the simplified currently available models . we must take into account that other processes may have influenced the local velocity distribution : the galactic bar , external processes like past accretion events , or internal disc processes like star formation bursts or encounters with giant molecular clouds . but for a simulation including the potential of the galactic bar with a pattern speed of @xmath121 . the outer lindlbad resonance of the bar is indicated as -2:1 ( b ) in the right scale . ] the galactic bar is believed to influence considerably the local velocity distribution @xcite . one may wonder how the velocity distributions presented in this study change under the combined action of spiral arms and bar . several simulations have proved us that in such a combined case , the spiral arm imprints are still distinguishable . fig . [ f:18spiralbar ] are velocity distributions of the same simulation as in fig . [ f : speed18 ] , but now including also a galactic bar . the bar is modelled following mw observational constraints , as explained in @xcite , and has a pattern speed of @xmath121 and present inclination with respect to the line sun galactic centre of @xmath122 . the total central mass of the model ( bulge+bar ) is @xmath123 , approximately as estimated by @xcite . of this mass , 70% belongs to the bar whereas the rest corresponds to the bulge , with a similar proportion as the best model in @xcite . this mass gives maximums for the bar parameters @xmath33 and @xmath124 ( defined in section [ force ] ) of 0.25 and 0.37 , respectively . these values are larger than for the spiral arms but are achieved at inner radii . in general , we see that in the inner regions of the disk ( up to 3 or 4 @xmath43 ) the bar dominates in strength but at solar radius of @xmath125 the strength values are comparable ( @xmath33 ) or the spiral arms are stronger ( @xmath124 ) . as we see in fig . [ f:18spiralbar ] , the vast majority of structures created by the spiral arms are still seen in the combined case . in general , we see that the bar induces groups specially at the low part of the @xmath55@xmath56 plane . for instance , we have check that the new group at @xmath102 in the first row appears also in the simulations with only bar . whereas the central region of these 3 panels is rather similar to the spiral alone case . we find few structures that are not induced now ( e.g. structure at low @xmath56 of the third panel in the last row ) , structures that have changed their shape ( third panel in the fourth row ) or that have been shifted to a slightly different position of the kinematic plane ( structure at low @xmath56 of the first panel in the last row ) . the more severe differences are seen in the resonance overlap case ( fourth row ) . this is expected as , according to @xcite , this overlap can induce widespread chaos . also @xcite claimed that the potential parameter constraint is not possible in these regions . but notice that we find that in nearby regions ( @xmath126 in galactocentric radius ) the individual spiral effects can still be identified . to conclude , using simulations with spiral - only model is a valid first step to understand the isolated effects of this non - axisymmetric component on the velocity distribution . it is particularly interesting that the arms ( both models perlas and twa ) can populate branch - like groups at around @xmath127 ( section [ comparison ] ) . the @xmath56 velocity of these branches is consistent with the @xmath56 velocity of the observed hercules structure . this group has a @xmath56 heliocentric velocity between @xmath128 and @xmath96 ( @xcite , see also fig . [ f : observed ] ) , which with the solar motion by @xcite , corresponds to a @xmath56 velocity with respect to the lsr of @xmath129 to @xmath130 . up to now this @xmath55@xmath56 plane region has been believed to be exclusively populated by the effects of the bar resonances . we show here and in @xcite that spiral and unbarred models crowd structures at these negative @xmath56 . although its shape is not exactly equal to the hercules branch , especially regarding its observed inclination in the @xmath55@xmath56 plane and average radial motion @xmath131 , we have just shown that the galactic bar action can modify and shift the spiral - induced groups to more negative @xmath55 . as a bar - only model is also able to induce a group similar to hercules @xcite , it is difficult to favour the bar or the combination of spiral arms and bar as the cause of this group . the characteristics of the galactic bar are still quite uncertain . for instance , determinations of the bar s angle with respect to the line sun galactic centre range from @xmath132 @xcite to @xmath133 @xcite . moreover , the existence of one or two bars in the mw is currently being debated . it would be worth exploring in detail the kinematic effects of the combined spiral - bar allowed parameter space , although this is out of the scope of the present study . however , note that a different bar s orientation could produce some different kinematic groups , but will not change the conclusions of the section . this is not strictly true if we change the relative strength of the bar with respect to the arms or their pattern speeds . these may lead to kinematic distributions dominated by one of the patterns . two extremes cases would be when one pattern exceeds the other one in strength or when its resonances are much closer to the considered regions . however , the ranges for the parameters of the spiral arms and bar in our models based on literature seem to indicate that the mw is far from these regimes . simulations with other bar parameters , not presented in this study , have confirmed that the results of this section do not depend strongly on the parameters provided they are in the believed mw range . we could also consider including other kind of patterns in the disc as proposed in the literature . for instance , several models have shown that galaxies might have more than one spiral arm mode ( with different pattern speeds ) at different radius or overlapping in radius ( e.g. @xcite ) , or with pattern speed changing with radius ( e.g. @xcite ) . some of the modes could be corotating with the bar at the bar s end , or coupled with the bar through their respective resonances . these possibilities are being explored for other galaxies ( meidt , rand & merrifield 2009 ) but have hardly been addressed for the mw . these must be future points to take into account in our modelling . few studies have focused on the spiral arm effects on the local velocity distribution . moreover , all of them modelled the spiral arm potential following the twa . we have seen that , despite being uncertain , the observational evidences for the mw arms density contrast , and especially , the pitch angle , suggest that the assumptions for self - consistency of the twa in the mw case are , at least , doubtfully satisfied . here we have studied the spiral arm effects on the kinematics of the solar neighbourhood with the twa but also with a different spiral arm model , the perlas model . while in the twa , the spiral arms are a small perturbative term of the potential , modelled as a cosine function , the perlas spiral arms correspond to an independent 3d mass distribution , from which the gravitational forces are derived . we have seen that both radial and tangential forces of this model present more abrupt features and different disc locations for the minima , maxima and 0 points . here we aimed to determine the conditions that favour the appearance of kinematic groups , such as the ones that we observe in the solar neighbourhood . to do this , first we have tuned these two models to the latest observational determinations for the properties of the mw spiral arms . for many of these properties only a certain range of evidence could be established . next , we have performed test particle simulations with both models considering these ranges , different initial conditions and integration times . our analysis indicates that : 1 . a significant part of the allowed parameter space , especially for the perlas model , favours the triggering of kinematic groups with different shapes , sizes and inclinations , such as the observed ones in the solar neighbourhood this shows that it is feasible that some of the observed moving groups have a dynamical origin related with the spiral arms . the velocity distribution is certainly sensitive to the relative spiral arm phase and , especially , to the pattern speed . changes in the kinematic groups are significant if one moves only @xmath1 in radius . but the @xmath55@xmath56 plane changes more slowly with azimuth and @xmath100 are needed in this direction to detect important differences . due to this , the pattern speed could be better constrained using the observed kinematic groups ( within an error of @xmath134 ) than the relative spiral phase ( with a precision of @xmath135 ) . however , one would need to break the degeneracy mentioned below by other means such as reducing the free parameter space . 3 . for both models and for all density contrasts , within the observational mw range , the spiral arms induce strong imprints for pattern speeds around @xmath0 . this corresponds to regions close to the 4:1 inner resonance . no substructure at all is induced close to corotation or high order resonances ( @xmath111 ) , which corresponds to pattern speeds of 20.5 to 30 @xmath105 . changes in spiral strength produce no significant differences in most cases , which makes difficult to constrain this parameter . some groups in the extremes of the @xmath55@xmath56 plane are only populated for the maximum density contrast case , which could help to establish the spiral strength . the effects of a different pitch angle seem difficult to differentiate given the observational range for this parameter . the spiral arms induce groups in a large region of the @xmath55@xmath56 plane , including the low @xmath56 part . for instance , the arms , and not only the bar , can crowd the region of the hercules group . previous studies associated the spiral arm influence mainly to the central parts of the @xmath55@xmath56 plane , and the bar to low angular momentum structures , such as hercules @xcite . 6 . the kinematic groups that are dynamically induced by the spiral arms depend on integration time . the time of appearance is different for each group , ranging from 0 to 1200 @xmath92 . the structures at the extreme parts of the @xmath55@xmath56 plane require more integration time , i.e. a larger spiral lifetime . the early appearance of some groups demonstrates that even recent spiral arms ( @xmath136 ) may produce strong kinematic structures . structures also change shape , size and position in the @xmath55@xmath56 plane until a maximum time of @xmath113 , when the @xmath55@xmath56 plane becomes stationary . 7 each of the structures seen in the @xmath55@xmath56 plane is composed by particles with a wide range of integration times and a different minimum time . this is encouraging as the study of the stellar ages in the observed moving groups also reveal similar characteristics . for instance , see evolution in the sirius and hyades groups particularly for stars younger than @xmath68 ( fig . 13 in @xcite ) or the wide age distribution and different minimum age in a particular group ( fig . 14 in @xcite ) . 8 . models obtained from our simulations with the spiral arms that reproduce the local observed velocity distribution are degenerated . groups such as the ones that are observed in the solar neighbourhood can be induced by different model parameter combinations . for instance , several models create groups such as hyades , pleiades or sirius . 9 . in most of our simulations where both the spiral arms and the bar are included , individual imprints of the bar and the arms can still be identified in the final velocity distributions . this means that using simulations with spiral - only model is a valid first step to understand the isolated kinematic effects of this non - axisymmetric component . the stellar response near solar positions of the twa spiral arms and the perlas model is significantly different , both in configuration space and kinematics . both models are able to induce several kinematic groups but the velocity distributions and the groups are rather different . the perlas model gives more substructure than the twa given the same density contrast , and more substructure for a larger pattern speed range or , equivalently , a larger radius range . for the perlas model and both the maximum and minimum density contrast , we see clear and rich kinematic substructure near solar azimuth for pattern speeds from 15 to 19.5 @xmath105 ( regions near the inner 3:1 , 4:1 and 5:1 resonances ) . by contrast , this pattern speed range is smaller for the twa model even in the high density contrast case . on the other hand , for the minimum density contrast , the twa model induces some substructure only for a narrow range of pattern speeds ( around 16.718 @xmath105 ) . to conclude , the spiral arms induce strong imprints on the velocity distributions at solar regions and that these are sensitive to some arm properties . these indicate that kinematics can be used as one of the constraints on the current uncertainty about spiral structure of our galaxy , e.g. , in the pattern speed , strength , orientation and lifetime . the local distribution function is significantly sensitive to the used model , even if they are adjusted to reproduce the same observational constraints for the spiral structure of the mw . this stresses the importance of the specific spiral arm gravitational potential modelling for the mw studies , as this can induce significant deviations on their orbital dynamics . it is promising that the kinematic structures can help to improve this modelling or to test whether the mw spiral arms are weak and tightly wound , following the twa . the constraint of the mw spiral structure properties , however , is presently not straightforward . this is due , first , to the mentioned degeneracy . our simulations have shown that the parameters space is still too large to obtain a unique fitting to the observed velocity distribution . second , the uncertainty in the solar motion propagates to the velocities of the observed groups , making the fitting between observations and simulations imprecise . the problem is complex as , precisely , the presence of kinematic substructure complicates the solar motion determination . third , for this constraint we should take into account the other processes that have influenced and sculpted the real velocity distributions ( other non - axisymmetries , external accretion effects , dispersion by giant molecular clouds , etc ) . as an example , we have identified effects due to the combination of bar and spiral arms . for instance , structures that appear centred in the @xmath55 axis in the spiral - only models are shifted to negative @xmath55 due to the combined action , inducing a group similar to the observed hercules structure . nevertheless , none of the existing current galaxy models is complex enough to include all these processes at the same time . to break the degeneracy and use the kinematic groups to constrain the mw large scale structure , a smaller parameter space for the spiral properties and/or data from velocity distributions at different positions of the mw disc are needed . with the increase of knowledge on the mw structure and evolution , we will have to cope with a more complex scenario and a variety of processes that can play a role in the formation of moving groups . future models will be more realistic but more complex to interpret . besides , while now the spatial study of the observed moving groups is limited by the extension and precision of the current observational samples , this will soon change with the advent of data from new surveys ( usno , ucac3 , 2mass , rave , segue , panstars , lamost , gaia ) . explorations of the spiral arm effects like the present study will help us to interpret more sophisticated future models as well as velocity distributions in nearby regions coming from surveys such as gaia . we thank the referee , p. rautiainen , for useful suggestions that helped improve this manuscript . we acknowledge funding support from the european research council under erc - stg grant galactica-24027 , micinn ( spanish ministry of science and innovation ) - feder through grant aya2009 - 14648-c02 - 01 and consolider csd2007 - 00050 . some of the simulations were run at the hp cp 4000 cluster ( kanbalam ) in the dgsca / unam . |
the quantum hamiltonian generates in time a family of evolution operators . continuity of this family holds within any choice of representation and , in particular , for the weyl propagator , even though its simplest semiclassical approximation may develop caustic singularities . the phase jumps of the weyl propagator across caustics have not been previously determined . the semiclassical appproximation relies on individual classical trajectories together with their neighbouring tangent map . based on the latter , one defines a continuous family of unitary tangent propagators , with an exact weyl representation that is close to the full semiclassical approximation in an appropriate neighbourhood . the phase increment of the semiclassical weyl propagator , as a caustic is crossed , is derived from the facts that the corresponding family of tangent operators belong to the metaplectic group and that the products of the tangent propagators are obtained from gaussian integrals . the weyl representation of the metaplectic group is here presented , with the correct phases determined within an intrinsic ambiguity for the overall sign . the elements that fully determine the phase increment across a particular caustic are then analysed . | 1309.5068 | the construction of a path integral for the weyl propagator @xcite , that is , the evolution operator , @xmath0 , in the weyl representation , makes no restriction on the form of the hamiltonian that generates the continuous evolution of a quantum system . except for possible minor modifications due to ordering , the weyl symbol , @xmath1 , for the quantum hamiltonian , @xmath2 , is identified with its classical counterpart , @xmath3 , where @xmath4 is the @xmath5-dimensional classical phase space , @xmath6 . there is no need for the classical hamiltonian to be derived from a lagrangian , so that the path integral may just as easily be constructed for e.g. the kerr hamiltonian @xcite , @xmath7 , as for the usual form , @xmath8 . a classical trajectory coincides with a stationary phase of the weyl path integrand , just as with feynman path integrals in the position representation . thus , one obtains semiclassical ( sc ) approximations for the full path integral as a superposition of a few contributing classical trajectories . the main difference is that here the relevant trajectory is prescribed by a kind of boundary condition on the _ centre _ , @xmath9 , of its pair of endpoints , @xmath10 and @xmath11 , rather than by the pair of end - positions @xmath12 @xcite . in both the weyl and the position representations , contributing trajectories may coalesce , with the passage of time , or with continuous change in the boundary condition . at these _ caustics _ , the sc approximation has a singularity . in passing a caustic , sc approximations generally change their phase by @xmath13 , where @xmath14 is known as the maslov index @xcite . if @xmath15 , then , in the position representation , @xmath14 coincides with the morse index @xcite . a more general geometrical phase space treatment of the maslov index for periodic orbits is presented in references @xcite . the semiclassical approximation becomes exact in the limit where the hamiltonian is quadratic in the phase space variables . classically , such a hamiltonian generates a ( linear ) _ symplectic transformation _ between the endpoints of a classical trajectory : @xmath16 , where @xmath17 is a _ symplectic matrix_. the symplectic subgroup of classical canonical transformations , sp(@xmath5 ) , characterized by @xmath17 , corresponds to the metaplectic subgroup , mp(@xmath5 ) , of general quantum unitary transformations , @xmath18 , that is , the group of _ metaplectic operators _ @xcite . the amplitude of the weyl propagator for a metaplectic operator is constant throughout phase space , but it has a true caustic singularity at an instant when the sc form of the weyl propagator must be replaced by a dirac delta - function . the action of a weyl propagator on other operators involves phase space integrals . therefore , in the general sc scenario where the region of integration may encompass different maslov phases , it is crucial that these be correctly evaluated . indeed , such switches of phase can be more important than the possibly integrable sc singularities at the caustics themselves , as occurs with the _ initial value _ or _ final value _ algorithms recently proposed in @xcite . furthermore , there are applications where sc _ super propagators _ for wigner functions @xcite are specified by products of weyl propagators , each of which may traverse its own caustics @xcite , so that it is indispensible to determine the correct maslov phases in all cases . the purpose of the present paper is to establish the general connections of sc weyl propagators ( and of their fourier transforms ) through their caustics . in order to establish possible phase changes for caustic traversals , one needs to focus on the _ tangent map _ for the full canonical transformation in the neighbourhood of a given trajectory . in other words , one must study the symplectic approximation to the full canonical transformation between the neighbourhoods of the endpoints of a classical trajectory : @xmath19 , where the symplectic matrix for the transformations is labelled by the centre , @xmath20 , of the main trajectory . this is an essential ingredient of the sc approximation of the weyl propagator , @xmath21 . furthermore , one can also construct an exact metaplectic _ tangent operator _ , @xmath22 , represented by the weyl symbol @xmath23 , i.e. the _ tangent weyl propagator _ , based on the same symplectic matrix , @xmath24 . just as the tangent map is close to the full canonical transformation in the neighbourhood of its endpoints , we can define a small neighbourhood @xmath25 , to be measured from the centre , @xmath26 , and then we have @xmath27 . it is important to emphasize that this is a purely local approximation and that different arguments of the weyl propagator ( i.e. choices of the centre , @xmath26 ) are related to different trajectories with different linearized maps , so that the pure quantum tangent propagator is constructed entirely from a local classical map . hence , one can investigate the possible phase jumps of the sc approximation of the weyl propagator by analyzing , instead , the behaviour of the exact symbol for the tangent propagator . the tangent weyl propagators belong to the metaplectic group . the difficulty is that , even though sp(@xmath5 ) and mp(@xmath5 ) are isomorphic in the neighbourhood of the identity , the correspondence @xmath28 is not one to one : there is a double covering of the quantum group , i.e. a pair of metaplectic operators and , hence , a pair of metaplectic weyl propagators for each @xmath17 : @xmath29 . in particular , one should note that important particular symplectic matrices , such as @xmath30 , the identity matrix , and @xmath31 , the reflection at the origin , correspond to pairs of metaplectic operators , even though only one of these is commonly employed . since they only differ by an overall sign , the choice of different sheets of a metaplectic operator has no essential effect on how it acts on wave functions and this action completely cancels out for heisenberg operator evolution : @xmath32 . nonetheless , the choice of sheet is essential for sc applications and it lies in the focus of this paper , but we shall distinguish the sheet in the notation for @xmath33 only where it is essential . the notion of _ crossing a caustic _ needs to be carefully dissected in terms of corresponding families of classical maps and quantum operators : the quantum hamiltonian generates a continuous one - parameter family of evolution operators , @xmath34 , corresponding semiclassically to a family of canonical transformations , each of which may , in turn , be decomposed into its individual classical trajectories . in the neighbourhood of each of these trajectories , one can then define a family of tangent symplectic maps , @xmath35 , and hence , a continuous one - parameter family of tangent metaplectic operators : @xmath36 . rather than @xmath35 following the previous notation , because the centre , @xmath37 , of a given trajectory with fixed initial condition is itself time dependent . indeed , one also deals with caustics as the centre @xmath26 is varied while the time is fixed , but one needs to consider only one - parameter families of metaplectic operators , so this has been labelled as @xmath38 thoughout this paper . ] such a family does not form a subgroup of the metaplectic operators , because it is not closed with respect to multiplication , but the _ product rule _ for each pair of sequential time intervals is satisfied : for @xmath39 , one has the exact product @xmath40 , corresponding to symplectic matrices , which also satisfy @xmath41 , even though neither @xmath42 nor @xmath43 need belong to their respective family . continuity of the classical family of symplectic transformations entails the continuity of the family of tangent operators , but this crucial point will be masked by caustics in any given representation . a caustic of the tangent weyl propagator occurs for the same matrix @xmath35 as for the sc approximation of the full evolution operator @xmath34 , but it is a true singularity , at which the weyl symbol becomes a @xmath44-function @xcite . the fact that there is only a pair of metaplectic sheets , corresponding to the choice of a @xmath45-phase ( i.e. an overall sign ) for the weyl propagator , does not prevent a possible phase change of @xmath46 across a caustic . however , it will be shown that it is only the overall sign that needs to be determined by the history of each trajectory , because otherwise factors of @xmath47 for tangent propagators can be inferred from the dynamical properties of the symplectic matrix @xmath35 . fortunately , the caustics of different representations occur for different values of the trace of @xmath35 and thus at different parameter values @xmath38 . this allows for swapping representations , where the continuity in a @xmath38-interval for one representation supplies the sign change at the caustic of the other . an alternative method to evaluate the overall sign of the propagator beyond a caustic is based on the exact product for sequential time intervals . these can be chosen so that both propagators lie in the causticless neighbourhood , which guarantees that the generalized gaussian integral for their product , which lies beyond the caustic , supplies the correct sign . nontrivial phase changes may in principle also occur for the product of finite metaplectic operators , or for their product with translation ( heisenberg ) operators , or reflection ( parity ) operators . the specification of these phases is important for the sc treatment of operator and super operator evolutions , such as @xcite . the case of parity operators demands special care because they have been defined with different phases in different contexts , namely the basis operator for the weyl representation differs from the operators belonging to the metaplectic group that we consider here . after a review of basic classical formulae for the weyl propagators involving symmetric matrices and generating functions in , the weyl representation of metaplectic operators and its fourier transform , the chord representation , are presented in . the choices which are here made for the overall phases , that were ommited in @xcite , is justified by the general consistency of the propagators on switching representation ( ) on the one hand , and from their products ( sections [ sec : producttransrefl ] and [ sec : productsmetaplectic ] ) on the other hand . given the elegance with which the double sheeted topology of the metaplectic group is here seen to be rendered within the weyl representation , our aim is to provide a pragmatic recipe for determining the phase increment across each caustic , regardless of the metaplectic sheet . this is clearly synthetized in the concluding . a didactical appendix has been added to illustrate the theory in the quintessential example of products of harmonic oscilations . the _ centre generating function _ , @xmath48 , defines implicitly a canonical transformation , @xmath49 . this is achieved by a finite version of hamilton s equations , that is , defining the chord , @xmath50 , we have @xmath51 where @xmath52 in terms of @xmath53 blocks , that is , the standard symplectic matrix in hamilton s equations . the endpoints are given by @xmath54 for the canonical transformation generated by a hamiltonian during a short time , we thus have @xmath55 , where the third order correction is given in @xcite . the tangent map for the canonical transformation near a trajectory centred on @xmath56 is @xmath57 . it has the centre generating function corresponding to the symplectic matrix @xmath24 @xmath58 where @xmath59 is the ( symmetric ) hessian matrix for the full generating function evaluated at @xmath26 . ( henceforth , the origin will be shifted so that @xmath60 . ) indeed , a symmetric matrix @xmath61 defines the cayley parametrization of a symplectic matrix @xmath17 @xcite , and hence the symplectic transformation @xmath62 , by @xmath63 with the inverse @xmath64 one should recall the fundamental symplectic property @xcite , @xmath65 where @xmath66 is the transpose of @xmath17 . it is important to note that any @xmath5-dimensional symmetric matrix , @xmath61 , defines a symplectic transformation through , unless @xmath67 has an eigenvalue @xmath68 , whereas the condition for the symplectic matrix @xmath17 itself can only be verified a posteriori . the eigenvalues of @xmath67 must come in pairs , @xmath69 just as the eigenvalues of @xmath70 , which determine the symplectic flow for the hamiltonian , @xmath71 . they correspond to the same eigenvectors for which @xmath17 has a pair of eigenvalues @xmath72 . again , the fact that the matrix @xmath61 is real demands that any complex eigenvalues come in complex conjugate pairs ( see e.g. @xcite ) . notwithstanding that @xmath73 , one should keep in mind that the eigenvalues of @xmath61 are necessarily real , even if those of @xmath67 need not be . for @xmath74 , the transformation can be classified by @xmath75 , that is , for @xmath76 the transformation is elliptic , for @xmath77 it is simply hyperbolic and for @xmath78 hyperbolic with reflection . one should note that @xmath79/2 $ ] . the alternative classification of symplectic transformations in terms of @xmath80 or of @xmath75 , is illustrated in . the classification , for @xmath81 according to the invariants of @xmath67 is given in @xcite and , for larger dimensional phase spaces , a similarity transformation can always reduce the matrix @xmath67 , just as the symplectic matrix @xmath17 , into 2-dimensional or 4-dimensional blocks @xcite . or @xmath75.,scaledwidth=45.0% ] a continuous transition between elliptic and hyperbolic transformations may pass through the identity , but , in general the boundary between them lies along the phase space shears , for which the normal form will be cayley - parametrized by @xmath82 for some real parameter @xmath83 , such that @xmath84 for the identity . this is a parabolic transformation , such that its symplectic matrix has the form of a jordan block : @xmath85 the breakdown of the cayley parametrization , such that @xmath86 , for a one - parameter family of transformations , if @xmath74 , is specified by the limiting matrix , @xmath87 so that the family of symplectic transformations goes through a shear with a reflection through the origin . this is the generic form for a caustic , i.e. for a single degree of freedom we have @xmath88 , or equivalently , @xmath89 . but there is only one eigendirection , unless the parameter @xmath90 in , in which case the transformation reduces to a reflection at the origin . the neighbourhood of a caustic can be neatly treated with the aid of the complementary cayley parametrization @xcite , @xmath91 with its inverse : @xmath92 from and one obtains the explicit expression for the new cayley matrix as @xmath93 thus , @xmath94 and @xmath95 are related by a similarity transformation and can be diagonalized simultaneously , so that @xmath61 and @xmath95 have the same signature and their determinants have the same sign . it follows that @xmath96 at a caustic , which allows for the neighbourhood of the caustic to be described by @xmath97 such that @xmath98 for a pure reflection . thus , @xmath75 is small near the identity and diverges near the reflection , whereas the situation for @xmath99 is the reverse . just as the cayley matrix @xmath61 specifies a quadratic centre generating function , @xmath100 is the _ chord generating function _ for a symplectic transformation @xcite , such that @xmath101 and again @xmath102 . thus , whereas @xmath48 becomes singular at a reflection , because all the chords then have the same centre , @xmath103 becomes singular for a uniform translation , for which all centres have the same chord . one should note that , since the eigenvalues of both @xmath67 and @xmath104 come in pairs of opposite sign , it is generally true that @xmath105 . furthermore , one obtains from and that the eigenvalues for @xmath104 and @xmath67 are related by @xmath106 . hence , @xmath75 and @xmath99 change signs simultaneously at the boundary between hyperbolic and elliptic transformations , so that the classification in also holds for @xmath107 , providing that one interchanges @xmath108 . if @xmath74 , both signatures @xmath109 and @xmath110 for an elliptic transformation , whereas @xmath111 for both parametrizations in the case of a hyperbolic transformation . for the present purpose , the most appropriate way to consider the weyl representation of an operator @xmath112 , i.e. its _ centre symbol _ or _ weyl symbol _ , is @xmath113 where @xmath114 is the quantum operator corresponding to a reflection through the phase space point @xmath26 @xcite . within a translation , this is just the symplectic transformation with @xmath31 , @xmath115 , though its main cayley matrix @xmath61 has a singular limit . it is important to note that such an operator is only defined within an overall phase , but in all standard definitions , one imposes that @xmath116 . the metaplectic operator , corresponding to the symplectic transformation @xmath62 and cayley matrix @xmath61 , has as weyl symbol the _ weyl propagator _ , @xmath117^{1/2 } } \exp\!\left(\frac{\rmi}{\hbar}{\mathbf{x}}\cdot { \mathbf{b}}{\mathbf{x}}\right ) = \pm [ \det({\mathbf{i}}+ { \mathbf{j}}{\mathbf{b}})]^{1/2 } \exp\!\left(\frac{\rmi}{\hbar}{\mathbf{x}}\cdot { \mathbf{b}}{\mathbf{x}}\right ) , \label{uweyl}\ ] ] which was derived in @xcite only within a phase . thus , it should be noted that both amplitudes have here been specified by the square root of a determinant , rather than that of its modulus . there will then be a factor of @xmath47 , if the determinant is negative . in other words , the duplicity of the metaplectic sheet coincides with that of the riemann sheet for the square root , as shall be verified further on . the @xmath118 signs which distinguish the pair of sheets are kept as a reminder that either choice can result for a metaplectic operator labelled by the same @xmath17 , depending on its previous evolution . in all cases , but the special appeal of the present phase space representations is that they are specified by invariants of the full corresponding symplectic matrix , @xmath17 . ] one should keep in mind that for general motions the only difference between this tangent propagator and the approximate sc weyl propagator , corresponding to a nonlinear transformation , is that the action of the former is obtained as the homogeneous second order approximation to the full action @xmath48 . ( it follows that the amplitude of the tangent propagator is constant . ) in other words , may either be interpreted as the local phase of the sc propagator , or the exact phase of the tangent propagator . thus , in the neighbourhood of the origin , which corresponds to the midpoint of a trajectory for the full nonlinear motion , one can identify the overall sign of both propagators . the alternative forms for the determinantal amplitudes are immediately interpreted with the aid of the transformations between midpoints and endpoints : @xmath119 whereas @xmath120 thus , the jacobians for these transformations are @xmath121 which also allows for the expression of the amplitude in terms of @xmath122 , as in @xcite . in the present text , the choice of sign within the determinant in the lhs of is never altered in order to avoid any confusion with the issue of determining the overall maslov phases . any continuous family of symplectic transformations corresponds to a continuous family of metaplectic operators . if these are represented by their weyl symbol , then the latter must also be continuous , except at the caustics , where @xmath123 . recalling that the weyl representation of the operator @xmath124 is just the constant , @xmath125 , it follows that all the metaplectic operators that are continuously connected to @xmath124 are correctly described by the @xmath126 sign in . it will be verified further along that the pair of metaplectic sheets is correctly accounted for , including caustic traversals , by the square roots of these determinants , without taking their modulus . the weyl propagator for the harmonic oscillator @xmath127 , driven by the hamiltonian @xmath128 , provides an iluminating example . after a time @xmath38 , the cayley matrix for this typical elliptic propagator will be @xmath129 , so that the full weyl propagator is @xmath130 . \label{uho}\ ] ] one should note the simple form taken by the square root in the amplitude of . the maslov theory in the next section will confirm that the change of phase across the caustic is just that shown in , in which the sign ambiguity is fully accounted for by adding an even or odd factor of @xmath131 to the cosine phase . so there is a minus sign as the boundary @xmath132 is crossed , coinciding with the sign of the cosine . this signifies a change of metaplectic sheet . at @xmath133 , one obtains @xmath134 , that is , @xmath135 and it is only after a second rotation in phase space that the identity operator is regained . in the absence of the history of the evolution , the @xmath118 sign is undetermined , but our general concern are the particular phase jumps for particular caustic transitions . products of metaplectic operators for the harmonic oscillator are discussed in the appendix . in contrast , for the inverted harmonic oscillator , @xmath136 , the weyl propagator is @xmath137 . \label{uiho}\ ] ] in the case of such a quadratic hyperbolic hamiltonian , there is no caustic singularity . above , we have portrayed the propagators continuously connected to @xmath124 , whereas those in the neighbourhood of @xmath134 would all have a negative sign . the hyperbolic transformations with reflection ( see ) can not be reached continuously through the action of a hyperbolic hamiltonian . even so , to complete the types of symplectic transformation portrayed in , one may add the continuous family of operators ( for @xmath38 positive or negative ) : @xmath138 . \label{uriho}\ ] ] however , it should be noted that the corresponding hyperbolic transformations with reflection are continuously obtained from the reflection , for which @xmath31 and @xmath139 , while they can not be reached directly from @xmath140 , without traversing a caustic . for this reason the phase presented in anticipates the theory in the following sections . one should note that typical metaplectic operators for elliptic and hyperbolic transformations are simply obtained from these examples , owing to the invariance of the weyl representation with respect to symplectic transformations . the cayley matrix for the harmonic oscillator becomes singular for @xmath141 , when the transformation is just a reflection . the correct corresponding weyl propagator is a dirac delta - function , which is indeed singular . in general , a continuous family of metaplectic operators has a singular weyl representation as @xmath142 . if @xmath74 , this is equivalent to @xmath143 . the chord representation of an operator @xmath112 , i.e. its _ chord symbol _ , is @xmath144 where @xmath145 is a heisenberg operator , that is , the unitary operator corresponding to a phase space translation by the vector @xmath146 . ( it is referred to in various ways throughout the literature , but here we follow the notation in @xcite . ) this is the conjugate representation to the weyl representation , so that the _ chord propagator _ corresponding to the same metaplectic operator and , hence , the same classical symplectic matrix , @xmath17 , can be evaluated through the symplectic fourier transforms ( see e.g. @xcite ) , @xmath147 @xmath148 where we recall that the _ wedge product _ @xmath149 . to obtain the metaplectic chord propagator , one takes the fourier transform of the complex gaussian form of the weyl propagator , @xmath150^{1/2 } \int \frac{\rmd{\mathbf{x}}}{(2\pi\hbar)^{n } } \exp\!\left[\frac{\rmi}{\hbar } ( { \mathbf{x}}\cdot{\mathbf{b}}{\mathbf{x}}- { \mathbf{x}}\cdot{\mathbf{j}}\bxi)\right ] \label{tanchord}\ ] ] and uses so as to express the chord propagator in terms of the alternative cayley matrix , @xmath107 , defined by . then the amplitude of the chord propagator results from the equality @xmath151 if the number of negative eigenvalues of the @xmath5-dimensional matrix @xmath61 is @xmath152 , then its signature is @xmath153 , so that the gaussian integral gains the phase @xmath154 . on the other hand , one can also express @xmath155= \pi [ n_-(\mathrm{mod}\ ; 2)]$ ] , so that the new phase can be incorporated into @xmath156 , within a phase @xmath157 , for some integer @xmath158 . this leads to @xmath159^{1/2 } \exp\!\left(-\frac{\rmi}{4\hbar}\bxi\cdot \tilde{\mathbf{b}}\bxi\right ) \\ \nonumber & = \pm \frac{(2\rmi)^n}{[\det({\mathbf{i}}-{\mathbf{m}})]^{1/2 } } \exp\!\left(-\frac{\rmi}{4\hbar}\bxi\cdot \tilde{\mathbf{b}}\bxi\right ) , \end{aligned}\ ] ] where the overall @xmath118 sign need not coincide with the original weyl representation ( i.e. positive near the identity ) and all further phases are specified by the square roots of the determinant . one can also interpret both forms of the amplitude in terms of the relation between the endpoint and the chord : @xmath160\bi{x}^- \;\ ; \rm{or } \;\ ; \bi{x}^- = \frac{1}{2 } [ { \mathbf{j}}\tilde{\mathbf{b}}-{\mathbf{i}}]\bxi . \label{initialchord}\ ] ] in practice , the increment in phase resulting from the fourier transform , @xmath161 is just added onto the overall preexisting phase in the weyl representation ( i.e. zero , in a causticless neighbourhood of the identity ) , so as to determine the overall phase in . for example , one may perform directly the symplectic fourier integral on for the harmonic oscillator . the overall sign for the weyl propagator is ( + ) for @xmath162 , so the choice of sign for the chord propagator will correspond to the phase @xmath163 , which leads to @xmath164 . \label{uhochord}\ ] ] likewise , the fourier transform for the inverted oscillator leads to the chord propagator : @xmath165 . \label{uihochord}\ ] ] whereas the reflected hyperbolic chord propagator , obtained from , becomes @xmath166 . \label{urihochord}\ ] ] here , one should remark that both the oscillator , for @xmath167 , and the reflected hyperbolic family for @xmath168 correspond to a reflection about the origin , but this operator does not have the same phase as the operator @xmath114 employed in the definition of the weyl representation . indeed , the chord symbol of the standard reflection is @xmath169 , so that one should distinguish the present _ metaplectic reflection _ as @xmath170 ( i.e. @xmath171 ) , which satisfies @xmath172 . this subtle phase distinction has surely been anticipated in other contexts ( e.g. in @xcite ) , but it is for manipulations in the weyl representation that it is of crucial importance . returning to the full expression for the metaplectic chord propagator , one realizes that the limit @xmath173 must specify the metaplectic reflection , @xmath174 . indeed , all operators that are continuously connected to such a reflection will preserve the same overall sign and the boundary of this region is only reached as one of the eigenvalues of @xmath107 reaches the value @xmath175 . thus there is no caustic for the chord representation as an eigenvalue of @xmath107 goes through zero , even though there may be a switch of type ( elliptic@xmath176hyperbolic , if @xmath74 ) . in the previous examples , one thus verifies that the transition between the chord propagator for the harmonic oscillator ( as @xmath177 ) and the hyperbolic with reflection ( as @xmath178 ) is smooth furthermore , all chord propagators within a causticless neighbourhood of the reflection , @xmath179 , corresponding to the phase @xmath45 of the harmonic oscillator , share the overall phase @xmath180 . this scenario is just the complement of that previously observed for the weyl representation , in which all operators that are continuously connected to the identity must share the same overall sign . in other words , the continuous group of metaplectic operators is represented with a cut in the weyl representation that includes the reflection , while its cut in the chord representation includes the identity . each representation is smooth along the other s singularity . this is the basis of the maslov method of phase determination in the following section . the alternative approach to determine the maslov phases is to rely on products of metaplectic operators . at the passage of a one - parameter family of tangent operators through the singularity of its weyl representation , an operator , characterized by large @xmath181 or small @xmath182 , is multiplied by an operator close to the identity , i.e. with small @xmath181 . thus , the determination of the final overall sign should be included in the problem of establishing the overall phase for the product of any pair of metaplectic operators . this is obtained from the product rule in the weyl representation for arbitrary pairs of operators @xcite , @xmath183 : @xmath184({\mathbf{x } } ) = \int \frac{\rmd{\mathbf{x}}_2\rmd{\mathbf{x}}_1}{(\pi \hbar)^{2n } } a_2({\mathbf{x}}_2 ) a_1({\mathbf{x}}_1 ) \exp\!\left[\frac{2\rmi}{\hbar } ( { \mathbf{x}}_2 - { \mathbf{x } } ) \cdot { \mathbf{j}}({\mathbf{x}}_1 - { \mathbf{x } } ) \right ] . \label{prule}\ ] ] in the case of metaplectic operators , each weyl propagator is a complex gaussian , so that is a @xmath185-dimensional gaussian integral . this is reduced to a @xmath5-dimensional integral in the limiting case where either @xmath186 or @xmath139 , for one of the factors ( i.e. it is either a translation or a reflection ) . these special cases are considered in . continuity of the family of tangent operators ( corresponding to the neighbourhood of a trajectory with a given initial value ) implies the continuity of the family of tangent weyl propagators in a causticless neighbourhood of the identity operator . hence , the overall sign of the sc approximation to the full weyl propagator must be preserved in this region , regardless of the type of symplectic transformation ( hyperbolic or elliptic , if @xmath74 ) : a transition between these two types within this neighbourhood , at a tangent propagator characterized by , does not alter the overall sign . this scenario is in marked contrast to the chord propagator in this same region , which has its caustic exactly for those transformations characterized by . the change of phase across this boundary is then evaluated by just the procedure followed in the previous section , that is , by performing the fourier transform from the weyl propagator . this is a generalization of the maslov method @xcite that has already been tacitly employed to obtain the change of sign for the harmonic oscillator at @xmath187 . the same method can now be employed to explore the possible change of phase within the weyl representation across its caustic , characterized by . what then is the result of reversing the fourier transformation , so as to return to the tangent weyl propagator from the chord propagator ? there is certainly no change if no caustic has been crossed , since this procedure may be continuously connected to the simple task of performing and then reversing the same fourier transformation . in terms of gaussian integrals , this general conclusion follows from the important result in , that both cayley matrices for a given transformation @xmath17 , i.e. @xmath61 and @xmath107 , have the same signature . let us now consider a one - parameter family of tangent operators , which evolves from an original operator , characterized by @xmath188 , such that @xmath189 , to a new operator , characterized by @xmath190 , with @xmath191 . if a single weyl caustic is crossed , the continuity of this operator evolution is still maintained within the chord representation , so that there is no change of the overall phase of the chord propagator . the increment of the overall phase of the weyl propagator will then just depend on the pair of fourier integrals : @xmath192 = \frac{\pi}{2}(n_- -n'_- ) . \label{causticphase}\ ] ] there is no phase increment if @xmath193 . alternatively , @xmath194 need not equal @xmath157 ( with integer @xmath158 ) , if the evolution has changed the type of the transformation , being that the square roots in the amplitudes may be imaginary . this is just the case , if @xmath17 denoted an elliptic transformation ( harmonic oscillator ) , whereas @xmath66 becomes a reflected hyperbolic transformation , though both transformations are smoothly connected in the chord representation . in general , the sign of the determinant in the final weyl propagator already picks up the signatures for the pair of fourier transformations , so as to specify the phase of the weyl propagator within @xmath195 . that is , using and its reverse , as well as continuity of the chord propagator , one obtains @xmath196 & = \arg[\det({\mathbf{i}}+ { \mathbf{j}}\tilde{\mathbf{b}}')]-\arg[\det(\tilde{\mathbf{b } } ' ) ] \\ \nonumber & = \arg[\det({\mathbf{i}}+ { \mathbf{j}}{\mathbf{b}})]-\arg[\det({\mathbf{b}})]+\arg[\det({\mathbf{b}}')],\end{aligned}\ ] ] so that the phase increment due to the determinant is just @xmath197-\arg[\det({\mathbf{b}})]= \frac{\pi}{2 } [ ( { n'_- } -n_-)(\mathrm{mod}\ ; 2)].\ ] ] comparing with , one verifies that full knowledge of the signatures @xmath198 and @xmath199 only distinguishes the final @xmath118 sign . in this way , the postulation of the weyl and the chord representations of metaplectic propagators in terms of determinants without moduli in and is justified : the determinants automatically account for the correct phase changes across caustics , within an overall sign , which is then specified by . finally , one should note that the overall phase increment for crossing a caustic in the conjugate chord representation , for a continuous evolution that crosses no weyl caustic , is obtained in perfect symmetry to the above results . operators that correspond to translations or reflections through a point in phase space are the building blocks of , respectively , the chord and the weyl representation @xcite . just as the metaplectic operators , they can also be exactly rendered by semiclassical formulae . furthermore , the combination of translations and reflections form the quantum affine group , which , together with the metaplectic group , integrate the inhomogeneous metaplectic group @xcite , corresponding to the ( classical ) inhomogeneous symplectic group . the generating functions for these transformations add a linear term to the quadratic generating functions of the ( homogeneous ) symplectic transformations that we have so far considered @xcite . let us first consider the product of a metaplectic operator @xmath200 , represented by @xmath201 according to with the unitary ( heisenberg ) translation operator , @xmath202 which includes @xmath203 . assuming that @xmath200 is in the neighbourhood of @xmath124 for which the overall sign is positive , then , according to , the product transformation is given by @xmath204^{1/2 } \int \frac{\rmd{\mathbf{x}}_2\rmd{\mathbf{x}}_1}{(\pi\hbar)^{2n } } \exp\!\left(\frac{\rmi}{\hbar } \phi\right),\ ] ] where @xmath205 the integral over @xmath206 is just a dirac @xmath44-function , @xmath207^{1/2 } \int \frac{\rmd{\mathbf{x}}_1}{(\pi\hbar)^{2n}}\\\nonumber & \hspace{2em}\times(2\pi\hbar)^{2n } \delta(2{\mathbf{x}}_1 -2{\mathbf{x}}+\bxi ) \exp\left[\frac{\rmi}{\hbar } ( { \mathbf{x}}_1 \cdot{\mathbf{b}}_1{\mathbf{x}}_1- 2{\mathbf{x}}\cdot{\mathbf{j}}{\mathbf{x}}_1)\right],\end{aligned}\ ] ] so that the result is the inhomogeneous metaplectic transformation , @xmath208^{1/2 } \exp\left[\frac{\rmi}{\hbar}\big(({\mathbf{x}}-\bxi/2 ) \cdot{\mathbf{b}}_1({\mathbf{x}}-\bxi/2)+ { \mathbf{x}}\cdot{\mathbf{j}}\bxi\big)\right ] , \label{uxi}\ ] ] that corresponds to the inhomogeneous symplectic transformation generated by the centre generating function , @xmath209 through equations , while holding the translation chord , @xmath146 , as a fixed parameter . one notes that no new phase appears in the integral over the @xmath44-function , that is , the overall sign of @xmath210 is preserved . in the limit , @xmath211 , i.e. for the product with the identity operator , it is verified that @xmath212({\mathbf{x } } ) = u_1({\mathbf{x}})$ ] . the product of a metaplectic operator with the reflection , @xmath213 , through the phase space point @xmath214 also simplifies , because @xmath215 in the weyl representation @xcite . inserting this into then leads to @xmath216^{1/2 } \exp\!\left(-\frac{2\rmi}{\hbar } \bi{y } \cdot { \mathbf{j}}{\mathbf{x}}\right ) \int \frac{\rmd{\mathbf{x}}_1}{(\pi\hbar)^{n}}\\\nonumber & \hspace{5em}\times\exp\!\left[\frac{\rmi}{\hbar } ( { \mathbf{x}}_1 \cdot{\mathbf{b}}_1{\mathbf{x}}_1- 2{\mathbf{x}}_1 \cdot{\mathbf{j}}(\bi{y}-{\mathbf{x}})\right].\end{aligned}\ ] ] the integral has the same form as the fourier transform of a complex gaussian in the @xmath5-dimensional phase space which led to the chord propagator in . so this again leads to a metaplectic operator with its weyl symbol : @xmath217^{1/2 } \exp\!\left(-\frac{\rmi}{\hbar } s_\bi{y}({\mathbf{x } } ) \right ) , \label{urefl}\ ] ] where @xmath218 is the centre generating function for the corresponding inhomogeneous symplectic transformation specified by . in the case of a reflection through the origin ( i.e. a parity operator ) , @xmath219 and will be just the homogeneous metaplectic operator specified by @xmath220 , i.e. the product propagator adopts the cayley matrix of the original chord propagator . the factor @xmath221 may seem strange , since it does not fit in with the phases in . if one chooses @xmath201 for a harmonic osillator , the product may be matched to just a further rotation in phase space by @xmath222 , so that presumably should satisfy . thus , one should recall the conclusion in that the appropriate reflection included in the metaplectic group is not @xmath114 , but the operator @xmath170 . for the product of this metaplectic reflection with a general metaplectic operator , the factor of @xmath221 is indeed cancelled in the corresponding expression to . in applications where the product arises because evolution acts on an operator specified by its weyl symbol , this extra factor must be included , because the reflection operator implied will then indeed be @xmath223 , rather than @xmath224 . in practice , the phase increment due to the product by @xmath223 is just @xmath225 where @xmath226 is the number of negative eigenvalues of @xmath227 . this determines the @xmath118 sign and it is exactly the same phase increment as previously obtained in passing from the weyl propagator to the chord propagator in . for the sake of brevity , we note that in the chord representation products of translations or reflections with metaplectic operators are just fourier transforms of and . basically , the corresponding cayley matrices for factors and their products are given by . here again , one must carefully choose between @xmath223 and @xmath224 , depending on the context : their chord symbols are dirac @xmath44-functions with different phase factors . we now consider general products of pairs of arbitrary metaplectic transformations , @xmath228 , corresponding to symplectic transformations , such that @xmath229 , parametrized by cayley matrices , @xmath227 and @xmath230 , according to , so that both factors take on the form . for simplicity , we shall assume that both @xmath200 and @xmath231 have overall positive signs . inserting these ingredients into the product rule then leads to @xmath232^{1/2 } [ \det({\mathbf{i}}+ { \mathbf{j}}{\mathbf{b}}_2)]^{1/2 } \int \frac{\rmd{\mathbf{x}}_2\rmd{\mathbf{x}}_1}{(\pi\hbar)^{2n } } \exp\!\left(\frac{\rmi}{\hbar } \phi({\mathbf{x } } , { \mathbf{x}}_1 , { \mathbf{x}}_2)\right),\ ] ] where @xmath233 is a homogeneous quadratic form in its @xmath234 variables , or an inhomogeneous quadratic form in the @xmath185 variables @xmath235 if we fix the centre @xmath26 where the full propagator is evaluated . it is shown in @xcite that , upon obtaining @xmath236 and @xmath237 from the conditions , @xmath238 and @xmath239 , the resulting quadratic form in @xmath26 is just the centre generating function for the product transformation , that is , @xmath240 , corresponding to @xmath241 . adopting this stationary point of @xmath235 as the origin then leads to @xmath242^{1/2 } [ \det({\mathbf{i}}+ { \mathbf{j}}{\mathbf{b}}_2)]^{1/2 } \exp\!\left(\frac{\rmi}{\hbar}{\mathbf{x}}\cdot{\mathbf{b}}{\mathbf{x}}\right)\\\nonumber & \hspace{10em}\times\int \frac{\rmd{\mathbf{x}}_2\rmd{\mathbf{x}}_1}{(\pi\hbar)^{2n } } \exp\!\left(\frac{\rmi}{\hbar } \phi(0 , { \mathbf{x}}_1 , { \mathbf{x}}_2)\right),\end{aligned}\ ] ] so that our task is to determine the amplitude and phase for the @xmath185-dimensional complex gaussian integral in . one must evaluate the determinant of the @xmath185-dimensional matrix @xmath243 $ ] , where one defines the symmetric block matrices : @xmath244 the phase increment is then @xmath245 where @xmath246 is the number of negative eigenvalues . in the limit @xmath247 , one is left with @xmath248 , which has @xmath5 eigenvalues @xmath175 and @xmath5 eigenvalues @xmath68 , so that @xmath249 and @xmath250 . this confirms a basic assumption for previous results in this paper : there is no overall change of sign for the product transformation , as long as the matrix @xmath251 can be treated as a small perturbation of @xmath252 . indeed , for a continuous one - parameter family of matrices @xmath251 , it is only at @xmath253 , as an eigenvalue vanishes , that there may occur an overall change of phase , unless either of the matrices @xmath254 becomes singular . just as in the fourier transform between the weyl and the chord propagators in , the phase of the determinant of the relevant quadratic form already captures part of the information supplied by its signature , i.e. @xmath255=\pi[{\textsf{n}}_-(\mathrm{mod}\ ; 2)]$ ] . the block form of the @xmath185-dimensional determinant allows for the simplification : @xmath256 where we define , @xmath257 further understanding of the @xmath5-dimensional matrix with this determinant is gained by drawing back to the interpretation of the amplitude of the weyl propagators in terms of the matrix between the midpoint and the endpoint of the classical trajectory . for the product transformation with symplectic matrices @xmath241 , we then have @xmath258 the alternative form of arises similarly from the inverse product transformation to , i.e. @xmath259 is related to @xmath260 hence , the amplitude of the weyl propagator for the product operator follows from @xmath261 the conclusion is that the product of a pair of metaplectic operators in the weyl representation results in a weyl propagator of exactly the same form , but expressed in terms of the cayley matrix for the combined symplectic transformation . the indeterminacy of the overall sign in the square root of the determinantal amplitude can only be lifted by evaluating the phase @xmath194 in . in the case of a continuous one - parameter family of product transformations , i.e. a one - parameter family of double matrices @xmath251 ( or , in the simplest case , a one - parameter family of matrices @xmath254 , with the other matrix held fixed ) a possible change of the overall phase of the product only arises if there is a change of sign of @xmath262 . but , according to , this event entails a change of sign for the final @xmath263 , itself . so there is no change of phase unless the combined symplectic transformation crosses a caustic . it is often the case that @xmath231 can be identified with a member of a family @xmath34 for @xmath264 , while @xmath265 ; for instance , the symplectic hamiltonian flow : @xmath266 $ ] , where @xmath267 is the hamiltonian matrix . here , @xmath200 is an arbitrary passive metaplectic operator . then , if @xmath268 from @xmath187 until @xmath264 , the phase increment for a product can be obtained from the difference of signatures of @xmath5-dimensional matrices . one must be wary of the change of notation , since the transition of the cayley matrices in , which was @xmath269 , becomes here @xmath270 . the advantage of this alternative approach relying on continuity is that it only deals with @xmath5-dimensional matrices . in the case of the chord representation , even the products of metaplectic operators can be considered within a @xmath5-dimensional integral , because of its simpler product rule . the deduction of the double sheet structure of metaplectic operators has previously been achieved by littlejohn @xcite within the position representation and for the chord representation ( with different notation ) by de gosson @xcite . our object here is not merely to show how the group of metaplectic transformations is represented by weyl symbols , but above all to determine the phase increment in each specific case . it may happen that @xmath271 and hence @xmath272 has a double root for a family of operator products . an important instance of this occurs if the commutator @xmath273=0 $ ] . this includes the case where @xmath274 , or where @xmath275 and @xmath276 are generated by the same quadratic hamiltonian . then @xmath277 and @xmath278 can be simultaneously diagonalized , such that @xmath279 and @xmath280 are their eigenvalues for the common @xmath281-th eigenvector of both these matrices . hence , the caustic condition for the product transformation , @xmath282 , reduces to @xmath283 . this condition allows for elliptic transformations , but not direct hyperbolic transformations , for which @xmath284 . on the other hand , @xmath285 will also be a pair of simultaneous eigenvalues , so that @xmath271 has a double root . the change of phase at the caustic is then an integral multiple of @xmath45 , which is consistent with the need for a product of elliptic transformations to remain elliptic , i.e. there can be no transition to hyperbolic with reflection . an example of a single null eigenvalue occurs at the caustic of a family of products of elliptic with hyperbolic transformations : @xmath286 thus , defining the one - parameter family of transformations either by the parameter @xmath287 or by the parameter @xmath288 , there will be a single null eigenvalue , i.e. single zero of @xmath271 , at @xmath289 and another one at @xmath290 . the symplectic matrices corresponding to are @xmath291 for the caustic condition @xmath292 , one obtains @xmath293 so that @xmath294 , but @xmath17 is not a reflection . one can check that @xmath263 does change its sign as the product parameter @xmath295 passes the value 1 , which leads to a transition of the type of symplectic transformation for the product . in all cases , full account must be taken of continuity for the proper family of tangent propagators . indeed , one or both of the factor families may cross a caustic , without any effect on the overall phase of the product . an important instance is the _ fidelity _ , or _ loschmidt echo operator _ treated in @xcite , such that the product of finite evolutions remains close to the identity operator . continuity of the product operator constrains the continuity of its weyl representation , in spite of any possible discontinuities in the phase of the factor propagators . this may considerably simplify practical calculations . the way that the phase jumps cancel for the particular example of a product of harmonic oscillator evolutions is analyzed in the appendix . it should be remarked that the sc propagator for wigner functions may be considered as a special case of a compound propagator resulting from the product of evolution operators . this has the peculiarity that the dominant classical trajectory is precisely on a caustic , so that a higher uniform approximation has been developed , for which the phases were previously analyzed @xcite . in contrast , the mixed propagator presented in @xcite avoids the caustic , so that its tangent propagator for short times is given by with a positive sign . it is shown in @xcite that there are important applications where the caustic disappears within an integration , so that one needs only worry about the phase increment , which is the focus of the present study . notwithstanding the theoretical interest in the way that the weyl symbol renders the double sheeted metaplectic group , practical use of this phase space representation requires clear rules for phase increments . indeed , one should not need to be reminded about the abstract topology of this group so as to evolve quantum operators in the sc approximation . the difficulty lies in the crossing of caustics , either as time changes ( while the argument of the weyl propagator is fixed ) , or for different arguments of the same weyl propagator . in both cases , the phase increment coincides with that of the appropriate family of metaplectic operators , each of which corresponds to a definite classical trajectory . fortunately , one can indeed summarize the conclusions in a nutshell : 1 . continuity of the family of metaplectic operators , corresponding to the family of tangent maps neighbouring each classical trajectory , guarantees that there will be no sc phase jump , unless a caustic is crossed . this even holds for compound trajectories that are pieced together to form _ super propagators _ for wigner functions , or for sc evaluations of the quantum fidelity . no matter how many caustics may have been traversed by the factor operators , one only needs to worry about phase jumps if the product propagator has crossed a caustic . in a causticless neighbourhood of the origin , the sign for the tangent propagators in is positive . 2 . for each caustic traversal of the weyl propagator ( compound or not ) , one needs a sample of the classical cayley matrix for a pair of arbitrary instants before and after this sc singularity is met . this requires , either the quadratic approximation of the centre action ( i.e. the centre generating function ) or else the relation to the ( linearized ) symplectic transformation . then the phase jump is obtained by the difference in the signature between these cayley matrices . by expressing the amplitude of the weyl propagators in terms of a determinant without a modulus in , any factor of @xmath47 is already accounted for beyond the caustic . the remaining ambiguity concerning the final phase of @xmath45 is then resolved by adding the above phase to the phase preceding the caustic traversal . the required information is naturally available in any sc computation . examples of these rules for products of propagators corresponding to ( exactly metaplectic ) harmonic oscillators with different frequencies are discussed in the appendix . the symmetry between weyl propagators and their fourier transform leads to similar phase recipes for chord propagators . they are an equally valuable source of super operators for which one must determine the precise phase . the phase jump upon crossing a caustic is obtained by simply replacing @xmath61 and @xmath296 in by @xmath297 and @xmath298 . let us consider the product of evolution operators for a pair of harmonic oscillators @xmath127 , described by the weyl propagators . it should be recalled that these may be considered as tangent propagators for general nonquadratic hamiltonians , such that the corresponding classical tangent maps are both elliptic . the simplifying assumption here is that both maps can be diagonalized simultaneously . conveniently one keeps a common time , @xmath38 , as a single parameter for the product , while the frequency @xmath299 may differ from @xmath300 , even with respect to its sign , denoting a different sense of oscillation . the weyl propagator for the product is then @xmath301 } { \cos \left(\frac{\omega_1}{2 } t\right ) \cos \left(\frac{\omega_2}{2 } t\right ) } \int \frac{\rmd{\mathbf{x}}_2\rmd{\mathbf{x}}_1}{(\pi \hbar)^{2}}\\\nonumber & \hspace{5em}\times\exp\left[\frac{\rmi}{\hbar}(\omega_1 { { \mathbf{x}}_1}^2 + \omega_2 { { \mathbf{x}}_2}^2 + 2{\mathbf{x}}_2 \cdot { \mathbf{j}}{\mathbf{x}}_1 ) \right].\end{aligned}\ ] ] where @xmath302 the exponent in the above integral is just the quadratic form determined by the symmetric matrix , @xmath303 , for which @xmath304 while the characteristic equation , @xmath305 ^ 2 = 0,\ ] ] has double roots : @xmath306^{\frac{1}{2}}.\ ] ] hence , there are two alternatives for the overall change of phase increment , 1 . @xmath307 if @xmath308 . @xmath309 if @xmath310 . the simplest case is that of a product of forward oscillations ( @xmath300 and @xmath311 ) such that neither has individually reached the caustic at @xmath312 , but @xmath313 . this is just case ( ii ) , so that , if both @xmath314 , then @xmath315 . for the loschmidt echo , the returning motion is specified by @xmath316 , so that @xmath317 . for short times , both @xmath318 , so that @xmath319 , which is case ( i ) . there is no new phase and we obtain @xmath320 . if one now allows @xmath300 to be slightly larger than @xmath45 , though still @xmath321 , then @xmath322 and @xmath310 , which is case ( ii ) and hence @xmath323 . but this exactly cancels the phase coming from @xmath324 , so that the overall sign of the propagator remains positive , just as for small times , that is , @xmath320 . finally , if one allows both oscillators for the loschmidt echo to pass their caustic , we are back in case ( i ) since @xmath319 . but now both amplitudes , @xmath325 , as long as @xmath326 , so the overall sign is still positive and , once again , @xmath320 . the special simplicity of the harmonic oscillator allows for an appealing though untypical synthesis of these results , due to the identification of the previous conditions with the sign of @xmath327 in : 1 . @xmath328 ; then the sign coming from the integration is a minus if @xmath329 , but this converts @xmath330 into @xmath331 . 2 . @xmath332 ; then the integration sign is negative if @xmath333 , which converts @xmath334 into @xmath335 . in all cases the amplitude will be @xmath336 without any sign ambiguity , as portrayed in . one should be warned that this is a very special case : in general , one can not evaluate the final phase without keeping detailed track of the intial phases and the phase increment ! , while all points on the dash - dotted lines correspond to @xmath134.,scaledwidth=45.0% ] for a loschmidt echo , the product of oscillations as a function of time will define a vector in that grows from the origin , nearly parallel to the dashed line . thus , it will be a long time before it finally leaves the white strip and changes its sign . in contrast , the pair of oscillations in the same sense , that were first considered , runs in the direction of the other diagonal , so that the sign changes repeatedly as @xmath38 increases . we thank raul vallejos , eduardo zambrano and ji vanek for stimulating discussions . amoa thanks the hospitality of the university of augsburg and , reciprocally , gli is greatful to cbpf for its hospitality . partial financial support from the national institute for science and technology quantum information , faperj and cnpq ( brazilian agencies ) is gratefully acknowledged . 99 ozorio de almeida a m 1992 _ proc . r. soc . lond . a _ * 439 * 139 ozorio de almeida a m 1998 _ phys . rep . _ * 295 * 265 walls d f and milburn g j 1995 _ quantum optics _ ( berlin : springer ) toscano f , vallejos r o and wisniacki d 2009 _ phys . e _ * 80 * 046218 kirchmair g , vlastakis b , leghtas z , nigg s e , paik h , ginossar e , mirrahimi m , frunzio l , girvin s m and schoelkopf r j 2013 _ nature _ * 495 * 205 berry m v 1989 _ proc . lond . a _ * 423 * 219 maslov v p 1972 _ thorie des perturbations et mthodes asymptotiques _ ( paris : dunod ) maslov v p and fedoriuk m v 1981 _ semi - classical approximation in quantum mechanics _ ( boston : reidel ) gutzwiller m c 1990 _ chaos in classical and quantum mechanics _ ( new york : springer ) creagh s c , robbins j m and littlejohn r g 1990 _ phys . . a _ * 42 * 1907 robbins j m 1991 _ nonlinearity _ * 4 * 343 bargmann v 1961 _ comm . pure appl . _ * 14 * 187 kramer p , moshinsky m and seligman t h 1975 in _ group theory and its applications _ vol iii ed e m loebl ( new york : academic press ) guillemin v and sternberg s 1984 _ symplectic techniques in physics _ ( cambridge : cambridge university press ) voros a 1976 _ ann . h. poincar ( a ) _ * 24 * 31 voros a 1977 _ ann . h. poincar ( a ) _ * 26 * 343 littlejohn r g 1986 _ phys . rep . _ * 138 * 193 de gosson m 2006 _ symplectic geometry and quantum mechanics _ ( basel : birkhuser verlag ) ozorio de almeida a m , vallejos r o and zambrano e 2013 _ j. phys . theor . _ * 46 * 135304 dittrich t , viviescas c and sandoval l 2006 070403 ozorio de almeida a m and brodier o 2006 1790 dittrich t , gomez e a and pachon l a 2010 _ j. chem . * 132 * 214102 arnold v i 1978 _ mathematical methods of classical mechanics _ ( springer : new york ) rivas a m f , saraceno m and ozorio de almeida a m 2000 _ nonlinearity _ * 13 * 341 williamson j 1936 _ amer . _ * 58 * 141 grossmann a 1976 _ commun . phys . _ * 48 * 191 royer a 1977 _ phys . * 15 * 449 nicacio f , maia r n p , toscano f and vallejos r o 2010 _ phys . lett . a _ * 374 * 4385 _ paper in preparation _ |
star - forming regions that are visible at m and h but not in the _ u , g , r , i , z _ bands of the sloan digital sky survey ( sdss ) , are measured in five nearby spiral galaxies to find extinctions averaging mag and stellar masses averaging . these regions are apparently young star complexes embedded in dark filamentary shock fronts connected with spiral arms . the associated cloud masses are . the conditions required to make such complexes are explored , including gravitational instabilities in spiral shocked gas and compression of incident clouds . we find that instabilities are too slow for a complete collapse of the observed spiral filaments , but they could lead to star formation in the denser parts . compression of incident clouds can produce a faster collapse but has difficulty explaining the semi - regular spacing of some regions along the arms . if gravitational instabilities are involved , then the condensations have the local jeans mass . also in this case , the near - simultaneous appearance of equally spaced complexes suggests that the dust lanes , and perhaps the arms too , are relatively young . | 1310.7146 | direct evidence for spiral wave triggering of star formation has been difficult to find since density wave shocks were first identified by @xcite and @xcite . early proposals for triggering involved shock compression of incoming clouds @xcite , diffuse cloud formation in a phase transition driven by thermal instabilities @xcite , diffuse cloud formation in the magnetic instability of parker ( 1966 ; mouschovias et al . 1974 ) , cloud - cloud collisions @xcite , and giant cloud formation by gravitational instabilities in the shocked gas @xcite . color and age gradients downstream from the shock suggest star formation triggering @xcite , but differential compression of old stars and gas @xcite and between hi and co gas phases @xcite can produce structures that mimic age gradients . it is sometimes unclear whether there is a preferential formation of stars in spiral arm gas compared to interarm gas : some observations show a clear change in molecular cloud properties or star formation efficiencies inside spiral arms compared to the interarms @xcite while other observations do not @xcite . infrared observations of galaxies offer a better view of star formation than has been possible before . @xcite found bright young complexes and clusters in ten spiral galaxies using j , h , k photometry , and determined the ages , masses , and extinctions for hundreds of these regions . the extinctions ranged up to @xmath5 mag for the youngest objects , the masses were up to @xmath6 , and the ages were typically less than 10 myr . longer - wavelength observations with the _ spitzer _ space telescope penetrate dust even better than j , h , k photometry . @xcite studied embedded star formation on 500 pc scales in the _ spitzer_-based sings survey by comparing @xmath7 m emission with h@xmath1 . they found a median extinction of 1.4 mag in h@xmath1 , and that 4% of the @xmath7 m sources have h@xmath1 extinctions larger than 3 mag . here we select the most opaque regions in the optical bands of five nearby spiral galaxies and search the corresponding _ spitzer _ @xmath0 m images for sources without optical counterparts . the spatial resolution is @xmath8 pc or less . we find that many spiral arm dust filaments contain such sources and determine the positions , extinctions , and masses of the most prominent objects . the appearance of embedded star formation in spiral arm dust filaments suggests that gas collapse can be rapid in the cores of the filaments . this is in agreement with recent models of spiral shocks by @xcite . the observations reported here are part of the spitzer survey of stellar structure in galaxies ( s@xmath9 g ; * ? ? ? * ) , which is a survey at @xmath10 m and @xmath11 m using the infrared array camera ( irac ; * ? ? ? the survey comprises over 2300 nearby galaxies , including warm mission and archival data that were processed through the s@xmath9 g pipeline . the released images have @xmath12 pixels and a resolution of @xmath13 . five galaxies ( table [ galaxies ] ) representing different spiral arm classes but all with about the same luminosity , size , and hubble type were chosen for inspection of embedded star formation regions , using optical images from the sloan digital sky survey ( sdss ; * ? ? ? * ) for comparison . the assumed distances come from the nasa / ipac extragalactic database . the sdss database has images in @xmath14 filters , with @xmath15 pixels . the sdss images were repixelated and reoriented to match the _ spitzer _ images using the _ wregister _ task in the _ image reduction and analysis facility _ ( iraf ) . the most obvious embedded star - forming complexes were identified by eye on the @xmath16 @xmath10 m images by blinking these with the sdss @xmath17 images in iraf . photometric measurements were done for each identified source using the iraf routine _ imstat _ to draw rectangles around the complexes , which are often elongated . the average background intensity for each region and passband was measured from linear scans one pixel wide through the center of the region in a direction perpendicular to the local arm using the iraf routine _ pvec_. this background was then subtracted from the intensities of the regions to produce source magnitudes . ground - based observations with h@xmath1 narrow - band filters were also compared to the ir and optical images . ngc 4321 , ngc 5055 , ngc 5194 h@xmath1 continuum - subtracted images were obtained from the _ spitzer _ infrared nearby galaxies survey ( sings ) legacy project . h@xmath1 continuum - subtracted images of ngc 5248 and ngc 5457 were obtained from @xcite . the ngc 4321 , ngc 5055 , ngc 5194 h@xmath1 images have a resolution of @xmath18pixel , whereas ngc 5248 and ngc 5457 images have a resolution of 0.241@xmath19pixel . all of the information regarding the observations , data reduction and data products can be found in the sings webpage and in @xcite . the h@xmath20nii fluxes were measured for the same regions that were used to define the @xmath0 m sources , but the entire flux emission from these regions was included , not just the flux confined to the source rectangles ( the additional h@xmath1 flux from outside the rectangles , estimated to be 10 - 20% of the total , is assumed to be from the embedded sources because hii regions are often larger than the star complexes that excite them ) . we have estimated the contribution from the neighboring nii lines using integrated spectrophotometry of the galaxies from @xcite in the case of ngc 4321 and ngc 5194 , and an empirical scaling relation between the nii / h@xmath1 ratio and m@xmath21 described in appendix b of @xcite in the case of ngc 5055 , ngc 5248 and ngc 5457 . the resulting nii / h@xmath1 ratios are 0.430 ( ngc 4321 ) , 0.486 ( ngc 5055 ) , 0.590 ( ngc 5194 ) , 0.369 ( ngc 5248 ) and 0.389 ( ngc 5457 ) . after correcting for the nii contamination , the h@xmath1 fluxes , f(h@xmath1 ) , are converted to h@xmath1 luminosities , as @xmath22=4\pi d^{2 } f({\rm h\alpha})$ ] , with _ d _ being the distance to the galaxy in cm ( table 1 ) . 2mass images at j band ( @xmath23 m , skrutskie et al . 2006 ) of the same galaxies were measured as well , using the same box sizes as for the @xmath0 m sources and the same background regions for subtraction . the 2mass pixel sizes are @xmath24 . the 2mass data is used in our application of a technique proposed by @xcite to estimate the underlying old stellar population . as discussed more below , this technique assumes that the old stellar population in the galaxy disk has a flux density at @xmath23 m that is typically about twice the flux density at @xmath0 m . mentuch et al . then showed that the excess @xmath0 m emission above half the @xmath23 m emission correlates with h@xmath1 . this gives an independent measure of the intrinsic h@xmath1 luminosity of the source , presumably with less extinction than the direct h@xmath1 measurement . figures 15 show the galaxies studied here . the color composites in the top left panels use sdss @xmath17-band images for blue , sdss @xmath25-band images for green , and irac @xmath10 m images for red . they all have @xmath12 pixels and were combined into color images after first using iraf _ wregister _ to rotate , stretch , and align the optical images to match the infrared images . the top right panel shows an image remade from the _ u _ , _ r _ , and _ z _ sdss filters using data from the sdss webpage ( www.sdss.org ) . the middle left panel shows a sample region at @xmath10 m with circles highlighting @xmath10 m sources that do not appear in optical bands . these same circles are shown in the middle right panel on an sdss @xmath17-band image . for clarity , these circles are @xmath26 larger than the measurement rectangles . the bottom left image is the @xmath0 m full image with a rectangle indicating the sample region , and the bottom right image is the sample region again in h@xmath1 . comparison of the middle panels in figures 15 shows sources clearly visible at @xmath0 m but invisible at @xmath17-band . sometimes part of a @xmath0 m source is visible at @xmath17 inside the circles , but the measuring rectangle was made on the part of the @xmath0 m source that was not visible at @xmath17 or at any other sdss filter . measurement rectangles are typically several arcseconds in length and width . the identified sources appear to be embedded star - forming regions in the spiral arm dust lanes . they are visible at @xmath10 m because the extinction there is only @xmath27% of the @xmath17-band extinction @xcite . they are often visible in h@xmath1 too , which is strange because that is an optical band , but h@xmath1 emission can sometimes be more easily detected than optical starlight , and some of the h@xmath1 could be from regions that are between and around the densest parts of the obscuring clouds . higher resolution hst images show parts of these sources to be visible at optical bands , but the hst optical fluxes are still weak compared to the @xmath0 m fluxes because of a generally high obscuration . the origin of the @xmath0 m emission is understood only in general terms . some of it is from the rayleigh - jeans limit of the photospheric emission from young and old stars , some of it is from hot dust in young stellar disks and dense cores , and some is from pah emission associated with the hii regions . we assume that the pah part of the @xmath0 m emission is an independent measure of flux from the hii region . after subtracting the contribution from underlying stars @xcite , we can then determine the embedded cluster masses from the required ionizing fluxes . comparison with the sdss - band upper limits then gives a lower limit to the extinction . the sdss - band upper limit flux in a dust lane , @xmath28 , is taken to be the rms value of the sdss pixel counts , @xmath29 , multiplied by the counts - to - specific flux conversion factor , @xmath30 , divided by the square root of the number of pixels , @xmath31 , and multiplied by the student - t value @xmath32 giving 95% probability that the mean count is less than @xmath33 . that is , the probability that the mean flux from the source is less than the measured mean ( @xmath34 here ) plus the increment @xmath35 is 0.95 . here , @xmath29 comes from the iraf measurement of the source counts . the value of @xmath36 is for a typical number of pixels in a source , @xmath37 , so the number of degrees of freedom for the student - t statistic is @xmath38 , and the probability that the true mean @xmath39 is between @xmath40 and @xmath41 is 0.90 . this is the same as saying that @xmath42 with 95% confidence for @xmath32 . the limiting magnitude is obtained from the limiting source flux using the usual ab conversion , mag@xmath43 for @xmath28 in ergs @xmath44 s@xmath45 hz@xmath45 . table [ results ] lists the positions of the embedded sources , the number of pixels of @xmath12 size in the measured rectangles , their fluxes , @xmath46 , at @xmath10 m in @xmath39jy , ab absolute magnitudes in irac @xmath10 m , ab absolute magnitude upper limits in the sdss bands , and the h@xmath1 luminosities measured from ground - based observations . without optical detections of the embedded sources and only irac observations at @xmath13 resolution from _ spitzer _ and @xmath23 m observations at @xmath24 resolution from 2mass , we can not uniquely determine the source ages , masses and extinctions . however , the sources are all embedded in spiral arm dust lanes and likely to be young ( e.g. , * ? ? ? * ) , as also implied by the h@xmath1 , so we searched over a range of ages , and for each age determined the cluster mass and extinction required to give the observed @xmath0 m and @xmath23 m fluxes using the conversion to intrinsic h@xmath1 in @xcite : @xmath47 - 0.5 i[1.25\mu{\rm m}]\right)/0.043\ ] ] for @xmath48 in units of mjy / ster . that is , @xmath49 and @xmath50 in mjy / ster were measured from the background subtracted specific fluxes in each source ( in units of @xmath39jy from the image source counts ) , converted to mjy and divided by the area of the source in ster . the resulting @xmath51 was then multiplied by a typical h@xmath1 filter width in frequency units , @xmath52 , which comes from the filter width in wavelength units , @xmath53 , converted to frequency with the equation @xmath54 . this gives the source h@xmath1 flux in units of erg @xmath44 s@xmath45 . the intrinsic h@xmath1 flux derived in this way from the ir , and the directly measured h@xmath1 flux from ground - based observations , were converted to h@xmath1 luminosities @xmath55 by multiplying by @xmath56 for distance @xmath57 , and then to observed ionization rates @xmath58 using the equation @xmath59 s@xmath45 @xcite . these observed ionization rates were compared to stellar evolution models to determine intrinsic source properties . for the stellar models , we used tables in @xcite for ab magnitude versus time in _ u , g , r , i _ , and _ z _ bands , which are given for a unit of solar mass in an initial cluster with a fully sampled chabrier imf and solar abundances . we also used the tables for lyman continuum ionization rate versus time per unit initial solar mass , and residual stellar mass versus time per unit initial solar mass . the optical magnitudes were converted to luminosities per unit initial mass , and these along with the lyman continuum rate and the residual stellar mass , were integrated over time up to the assumed age using a constant star formation rate . this integral over the model lyman continuum rate per unit initial mass will be denoted by @xmath60 , and the integral over the model residual mass per unit initial mass will be denoted by @xmath61 . the ratio of the observed ionization rate @xmath58 to the model rate @xmath62 gives the correction factor to the model integrated mass , from which the required cluster mass can be derived : @xmath63 the integrated luminosities in each sdss band were found from the same ratio , e.g. , @xmath64)l_{\rm m , u}(t)$ ] for _ u_-band , and converted back to apparent magnitude . the resulting model magnitudes in all five sdss bands were then subtracted from the observed magnitude limits in each band to get the minimum extinction in each band . these were all converted to _ g_-band extinction using the extinction curve in @xcite , and the results were averaged together . this extinction curve should be more appropriate for embedded sources than the gray type of extinction that comes from a partially transparent screen @xcite . the different limiting g - converted extinctions varied among the passbands by an average of 0.5 mag , presumably because the sdss limits were not the same depth relative to the spectral energy distribution of the source . table 3 gives the derived properties of the embedded sources listed in table 2 for assumed ages of 3 myr . the masses and average _ g_-band - equivalent extinctions for the ground - based h@xmath1 measurements are in columns 3 and 4 , the masses and average _ g_-band - equivalent extinctions for the @xmath0 m measurements are in columns 5 and 6 , and the extinction needed in the hii region , measured at _ g_-band , to produce the observed h@xmath1 fluxes given the star forming regions measured at @xmath0 m , are in column 7 . the sources with lower limits have upper limits for the directly observed h@xmath1 fluxes ( more on this below and in the footnote to table 3 ) . all sources are observed at @xmath0 m . the derived cluster masses increase with assumed age in the range from 1 to 10 myr by about a factor of 2.5 for all sources ( not shown ) . this is small enough for the likely age range that we consider here only masses for an assumed age of 3 myr , as tabulated , with an uncertainty of a factor of 1.6 either way from unknown age effects . the mass differences obtained using the two methods , direct h@xmath1 observations and h@xmath1 inferred from @xmath0 m and @xmath23 m emissions , typically differed by a factor of 24 with the ir - based masses being larger . this difference amounts to an excess extinction of about 3.5 magnitudes for the direct h@xmath1 compared to the ir - derived h@xmath1 . table 4 gives the embedded source masses and extinctions for the two types of measurements and the equivalent @xmath17-band extinction in the hii region to give h@xmath1 consistent with @xmath0 m and @xmath23 m , averaged over all sources in each galaxy . the masses measured by direct h@xmath1 observations average @xmath65 , and the masses measured by @xmath0 m and @xmath23 m observations and converted to h@xmath1 average @xmath66 . the lower limits to the extinctions provided by the upper limits to the sdss fluxes and the ground - based h@xmath1 fluxes ( column 3 ) are about zero , meaning that the h@xmath1 measurements are consistent with the optical broad - band measurements in being highly extincted . in other words , given the stellar masses required to produce the ( highly extincted ) h@xmath1 fluxes that were directly observed , the expected @xmath67 fluxes are comparable to the upper limits within a few tenths of a magnitude ( the error bars in column 3 of table 4 ) . the lower limits to the extinctions derived from the @xmath0 m fluxes in comparison to the sdss bands ( column 5 ) are , on average , @xmath2 mag for all galaxies . this is comparable to the extinction in the hii region ( column 6 ) because the hii emission is extincted by about the same amount as the sdss bands . we conclude that these dusty regions are obscuring young stellar associations with an average mass of @xmath66 for an age of about 3 myr . the average mass would be @xmath68 for an age of 1 myr and @xmath69 for an age of 10 myr . the average extinction to these regions in @xmath17-band is @xmath2 magnitudes . we consider possible star formation mechanisms as summarized in the introduction . two of these mechanisms , the parker and thermal instabilities , involve diffuse gas without self - gravity . while they may contribute to interstellar structure , they are not involved with rapid star formation because that requires self - gravity and collapsing cloud cores . the other mechanisms , gravitational instabilities and cloud impacts or collisions , involve strong self - gravity from an early stage , the first because whole dust lanes can be unstable to fragmentation along their lengths , and the second because collisions increase a cloud s density and that can make the cloud gravitationally unstable . the observation of bright embedded infrared sources in galactic spiral shocks implies that massive star formation begins quickly in the shocked gas faster than the time delay usually inferred from the displacement between dust lanes or co emission and visible hii regions ( e.g. , * ? ? ? the observed mass of @xmath70 is comparable to the masses of more visible clusters and star - forming regions in normal galaxies . in what follows , we consider two possible formation mechanisms for the observed star complexes . for both , the dust lanes are assumed to be shocks in spiral density waves . this is consistent with the streaming motions that are often observed in co observations of these regions ( e.g. , * ? ? ? most of the molecular clouds in the milky way spiral arms @xcite seem related to spiral shocks too , considering their velocities and positions in the global spiral pattern @xcite . their concentration in spiral arms is apparent in the face - on view of the milky way shown by @xcite . evidently , these local co clouds would resemble dust lanes if they were viewed from outside the milky way . the two mechanisms considered are ( 1 ) gravitational collapse of gas that accumulates behind the shock front in a spiral arm , and ( 2 ) compression and collapse of pre - existing clouds that hit these shock fronts . in the following discussion , we suggest that gravitational collapse of spiral arm shock fronts is too slow on kpc scales to account for the rapid onset of star formation that appears to be required for these embedded sources . the densest parts of the shocked gas could have collapsed into stars , however . compression of interarm clouds that hit previously shocked gas could trigger the obscured sources faster . additional considerations are needed to determine which mechanisms apply . table [ averages ] suggests that a typical value of optical extinction to an embedded dust lane source is @xmath2 mag . an embedded source with @xmath71 mag of extinction may be inside a dust cloud that has @xmath72 mag of extinction all the way through it if the source is centered . this makes the total column density through the cloud @xmath73 pc@xmath74 for @xmath75 @xcite , ratio of total - to - selective extinction @xmath76 @xcite , and mean atomic weight 1.36 . a typical dust lane width is @xmath77 pixels in figures 15 ( 1 px@xmath78 ) , which is @xmath79 pc at a distance of 10 mpc . the dust lane linear density is the product of these , @xmath80 pc@xmath45 , or @xmath81 g cm@xmath45 . the critical line density for collapse of a filament is @xmath83 for 1-dimensional velocity dispersion @xmath84 @xcite . if @xmath85 km s@xmath45 in the shock , then @xmath86 is comparable to the observed line density . this suggests that spiral arm shock fronts are unstable to collapse along their lengths @xcite . the result of this collapse would be one or more giant cloud complexes with masses equal to @xmath87 for spacing @xmath88 between the complexes along the arm @xcite . taking @xmath89 from the figures here and from @xcite , the resulting cloud mass is @xmath90 . if we consider a typical efficiency of 2% for the star - forming complex , then a stellar mass of @xmath91 results , large enough to account for what we observe . the key to star formation in spiral arm dust lanes is that it has to happen quickly , before the spiral arm passes and the shocked gas decompresses in the interarm . this is a strong constraint if the spiral wave is steady . not all compressed regions will be part of a steady flow , however . if a dust lane is sufficiently self - gravitating , then it can remain intact as a giant cloud complex even when it enters the interarm region , making its lifetime longer than the arm crossing time . such shocks would not be steady but would cycle between gas build up , collapse , and dispersal . if the lifetime of the compressed state is long enough , then interarm protrusions and spurs result @xcite with star formation lingering in the interarm regions . @xcite observed m51 with 30 pc resolution and could see cloud build - up from small interarm co clouds to giant molecular associations as the gas flowed into and through the spiral arms . with 780 pc resolution , the excitation state of m51 co clouds changes from low in the interarm regions to high in the arms , corresponding to an increase in density and/or temperature with increasing star formation activity @xcite . there is also an indication in m51 of dense co gas just before prominent spiral arm star formation ; the map in figure 6 of @xcite shows regions with highly excited co and low @xmath7 m emission . these regions appear to be centered on dust lanes , although their exact spatial distribution is uncertain because the 780 pc resolution scale of the map is much larger than a dust lane . these regions avoid the parts of the dust lanes where we identify the brightest embedded sources , but that is expected from the lack of @xmath7 m emission in the @xcite sources . similarly , @xcite observed molecular clouds in ic 342 turn from weakly self - gravitating in the interarm regions to strongly self - gravitating and star - forming in the arms . we now consider when the collapse of density wave - shocked gas is fast enough to make stars in the steady state case . this requires a comparison between the collapse time of this gas and its dwelling time in the arm . the dwelling time for gas in a dust lane is the arm - to - arm time multiplied by the ratio of the gas mass inside the dust lane within some radial interval to the total gas mass from one arm to the next in that radial interval . this relationship is from the continuity equation . because of streaming motion along the arm in the dust lane , this dwelling time is longer than the purely geometric time obtained from the ratio of the angular thickness of the dust lane measured from the galaxy center divided by the angular rate of the spiral pattern relative to the gas . the ratio of the dust lane mass to the total arm - to - arm gas mass in a radial interval equals the ratio of the dust lane column density to the average column density , multiplied by the relative angular extent of the dust lane . in the star - forming part of the main disk of most spiral galaxies , v_-band extinction averaged in azimuth is typically about 1 magnitude . recall that the column density for general extinction @xmath92 equals @xmath93 pc@xmath74 in units of surface density , and an average value of @xmath94 pc@xmath74 for spiral galaxy disks is typical inside the molecular zone @xcite . for a general case like this , the ratio of column densities is then approximately @xmath95 for @xmath96 in the dust lane . the arm - to - arm distance is @xmath97 for a two - arm spiral at radius @xmath98 . taking @xmath99 kpc as representative , the relative dust lane thickness is @xmath100% with @xmath101 pc from above . thus the fraction of the gas mass in the dust lane , which is also the fraction of the arm - to - arm time in the dust lane , is approximately @xmath102 . the arm - to - arm time is @xmath103 myr for a position at @xmath104 kpc that is halfway to corotation where the galaxy rotation speed is @xmath105 km s@xmath45 . then the dwelling time for gas in the dust lane is @xmath106 myr . in general , the dust lane flow - through time is @xmath107 for dust lane width @xmath108 , azimuthal speed @xmath109 of the spiral arm relative to the gas , and average extinction through the disk at the radius of the dust lane , @xmath110 . this result does not depend on @xmath98 explicitly . if @xmath108 is in pc and @xmath111 is in km s@xmath45 , then the unit of this flow - through time is approximately in myr . the fraction of the time that gas spends in the vicinity of a spiral arm can be fairly high for a strong arm . for example , a co map of the galaxy shown in figure 4 , ngc 5248 @xcite , has two spiral arms midway out with @xmath112 k km s@xmath45 integrated intensities that occupy about 20%-25% of the azimuthal distance to the next arm , and it shows interarm regions with less than 3 k km s@xmath45 . thus the arms contain about half of the gas at that radius ( @xmath113 ) , and the arm dwelling time is half of the arm - to - arm time if most of the gas is molecular . this fraction would be lower if the @xmath114-factor that converts co to h@xmath115 is lower in the arms than in the interarms because of greater co temperatures in the arms . using @xmath116 km s@xmath45 , which is half of the rotation speed of ngc 5248 @xcite , assuming this position is half the distance to co - rotation , and assuming a galactocentric distance of 1 arcmin @xmath77 kpc , the arm - to - arm time is 126 myr and the arm dwelling time could be about 60 - 70 myr . in the spiral arms of m51 @xcite , the dust lanes appear as thin ribbons of co emission nearly resolved by carma . they occupy about 10% of the azimuthal angle and have 5 to 10 times the average interarm intensity in co. thus even the thin co arms in m51 contain 30% to 50% of the molecular mass at that radius , and because most of the gas in this galaxy is molecular @xcite , this is also approximately the fraction of the arm - to - arm time spent in the dust lanes ( unless the @xmath114-factor varies ) . the dwelling time in a spiral arm has to be compared with the timescale for self - gravitational collapse . the latter is about @xmath117 for length - to - width ratio @xmath118 along the dust lane and for density @xmath119 . the factor @xmath118 comes from the average equivalent density @xmath120 in a sphere containing the fragment , considering that the collapse time is @xmath121 for that average . in agreement with this , @xcite show one case of a filament with a collapse to a singularity in @xmath122 , which is @xmath123 , giving @xmath124 in their case ( see also * ? ? ? the average dust lane density is the extinction through the disk , @xmath96 , converted into grams @xmath44 , divided by the line - of - sight thickness , which is @xmath125 pc for a typical molecular disk . for @xmath126 in parsecs , this density is @xmath127 g @xmath128 , and the collapse time is @xmath129 myr , or some @xmath130 myr if @xmath125 pc and @xmath131 mag ( i.e. , using @xmath132 or @xmath133 from table 4 ) . for gravitational collapse in a dust lane , the collapse time has to be less than the flow - through time , which means @xmath134 or @xmath135 mag for @xmath136 , @xmath137 pc , @xmath138 mag , @xmath139 km s@xmath45 , and @xmath101 pc . this result implies that sufficiently dense spiral shocks , i.e. , those with extinctions through the disk greater than @xmath140 mag for these numbers , should have time to collapse along their lengths into new cloud complexes . because self - gravity is important from the start in such an instability , the cloud complexes that form will also be strongly self - gravitating , and should form stars at an accelerating pace while they collapse further . recall that our sources imply @xmath141 mag on average , considering we see the foreground extinction equal to half of this value . this extinction is slightly less than the critical value of 18 mag derived above , suggesting that the dust lanes we observe are only marginally unstable to collapse in the short time they have . if the line - of - sight cloud thicknesses were much smaller than the assumed @xmath125 pc , such as the size of a giant molecular cloud , @xmath142 pc , then @xmath143 pc , and this would be closer to our observations . in either case , the observed dust lanes do not appear to be strongly self - gravitating and in the process of rapid collapse . they could be forming stars only in the initially dense regions that were most unstable before they entered the spiral arm . such partial collapse is consistent with the spiral shock models by @xcite and with the growing evidence for star formation in marginally bound or unbound molecular clouds ( i.e. , but which are still unstable to collapse in their cores ; * ? ? ? @xcite ran simulations showing the steady accumulation of shocked gas in spiral arms , with dust lane density perturbations building up over several rotation periods from initially small amplitudes . eventually the irregularities caused by gaseous self - gravity led to the formation of giant , self - gravitating clouds . this is essentially the process we have in mind here for the formation of the observed embedded sources . @xcite showed examples of beads - on - a - string of star formation along spiral arms of all types , ranging from two - arm grand design spirals to multiple , long - arm spirals , to flocculent spirals ( see also elmegreen & elmegreen 1983 ) . beads - on - a - string patterns were studied in detail by @xcite using irac images like those considered here . in many images from the _ galaxy evolution explorer _ satellite ( _ galex _ ; martin et al . 2005 ) , one gets the impression that most star formation is in complexes that are strung out along spiral arms in a semi - regular fashion . interstellar magnetic fields may be involved in this regularity @xcite . evidently , there is a close resemblance between spiral arm star formation on kpc scales and filamentary star formation @xcite inside molecular clouds on parsec scales . compression and collapse of an incident interarm cloud can be faster than the collapse time calculated above for an initially uniform dust lane . the compression time @xmath144 is the cloud size @xmath145 divided by the relative perpendicular speed between the cloud and the spiral arm , @xmath146 . for a relative azimuthal speed between the cloud and the arm , @xmath139 km s@xmath45 as above , the perpendicular speed is @xmath147 km s@xmath45 when the spiral arm pitch angle is @xmath148 . then @xmath149 myr for @xmath150 pc . collapse would follow compression quickly because the compressed cloud density is high , approximately equal to the product of the interarm cloud density , @xmath151 ( in @xmath152pc@xmath153 for an interarm cloud extinction @xmath154 and line - of - sight thickness @xmath126 in pc ) multiplied by the square of the ratio of the perpendicular compression speed to the cloud thermal speed , @xmath155 . for @xmath156 mag , @xmath137 pc , @xmath157 km s@xmath45 , and @xmath158 km s@xmath45 , the compressed cloud density is @xmath159 atoms @xmath128 and the cloud collapse time , measured as @xmath160 , is @xmath161 myr . if such a density extends for @xmath137 pc along the line of sight , then the extinction through the disk would be @xmath162 mag and the extinction to an embedded source @xmath163 mag . this is significantly larger than what we observe here for the clouds on @xmath79 pc scales , but could apply in their cores . the original picture of cloud compression in spiral shocks was based on simulations by @xcite of a small diffuse cloud engulfed by a rapidly moving , infinitely - extended shock front . the cloud is compressed by the post - shock gas and forms a cometary structure , with the head collapsing into stars . many simulations have confirmed this basic process . spiral arms do not typically show cometary clouds and the dust lanes surrounding the embedded sources discussed here are not cometary . still , the basic model of cloud compression upon impact with an existing spiral arm dust lane ( from a previous shock ) seems generally valid . the interarm clouds in figures 15 are spiral - like with large cloud complexes distributed along their lengths . this interarm structure is also evident from well - resolved maps of ir @xcite and co @xcite , as in the case of m51 . such shapes are remnants of dust lane structures , secondary shocks , and spurs from the previous arms @xcite . interarm cloud complexes can be moderately massive : @xmath164 if @xmath150 pc and @xmath165 mag . the most opaque interarm clouds may contain @xmath4 and have a substantial molecular fraction ( e.g. , * ? ? ? the observed masses of the largest embedded complexes found here are consistent with interarm cloud compression . the pre - collision clouds would have to be @xmath163 times more massive than the observed stellar masses , considering a star formation efficiency on @xmath166 pc scales of 2% . to form a @xmath167 star complex inside a dust lane ( table 4 ) , the previous interarm cloud would have to contain @xmath168 . such a cloud , if spherical , would be @xmath169 pc in diameter if its average extinction is @xmath165 mag . with such a size in two dimensions , the cloud would form a @xmath169 pc long dust filament when it hits the spiral arm , and then the collapse to star formation would be like the dust lane collapse discussed above , with motions along its length and a factor @xmath89 geometric dilution of the collapse time . regular spacing of large star complexes along spiral arms , as occasionally seen in figures 15 and in other galaxies , implies that an instability might be involved and we are seeing the fastest growing wavelength of that instability ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the presence of several young sources in this string implies that the instability operates all along the length of the pattern with a nearly uniform starting time , to within some variation equal to the instability time or the arm dwelling time . such simultaneity requires that the dust lanes with regular star formation evolved within this time interval , that is , they built up , fragmented , and turned into stars along the length of the regular pattern within several tens of myrs . after this , another dust lane would be expected to build up to replace it using fresh interarm gas . this would be the case for a continuously moving spiral arm that accumulates gas , dumps it into star complexes and then accumulates more gas . alternatively , the whole stellar and gaseous arm could have formed recently as a result of a spiral instability in the disk @xcite . then the arm structure itself would be fairly young ( although the stellar populations in it would be mostly old ) and it could just now be forming its first string of star formation , i.e. , within the last @xmath170 myr . in both cases , the instability would be slow compared to the dynamics of star formation , so the ob associations that form would be in various stages of obscuration and break - out , ranging from embedded for the youngest to fully exposed with adjacent cloud debris for the oldest in the same arm segment . in cases without a regular spacing along spiral arms , a similar time sequence may occur with spiral - shocked gas followed by partial cloud collapse , star formation , and cloud dispersal , but with more of a distributed starting time . then there will be different phases in this process at different positions along the arm . there could also be other processes acting , such as direct compression of interarm clouds , which would be randomly placed with respect to the regular structures . a search for emission at @xmath0 m that has no associated optical counterpart in sdss images and only weak emission at h@xmath1 revealed 73 sources that are likely to be young star formation regions embedded in molecular clouds . the clouds appear as dark dust lanes in the spiral arms . the intrinsic h@xmath1 emission of each source was estimated from the @xmath0 m and @xmath23 m emissions using the method of @xcite , and the luminosity of the associated stars was determined after assuming an age in the range of 1 to 10 myr . this stellar luminosity , combined with upper limits from the sdss images , gave lower limits to the extinction and stellar mass . the observed h@xmath1 , compared with the h@xmath1 estimated from @xmath0 m and @xmath23 m , gave a consistent value for the extinction , which is apparently near the estimated lower limit ( or else the h@xmath1 would not be observed either ) . the average mass of an embedded source was found to be @xmath171 and the average extinction at @xmath17-band is @xmath172 mag . these sources are often parts of regular strings of star formation with additional embedded sources and other sources in various stages of lower obscuration . two mechanisms for triggering this star formation were considered : gravitational collapse of spiral density wave - shocked gas and gravitational collapse of compressed incident clouds . the spiral shock model suggests that the clouds we are observing are not strongly self - gravitating because they have not had enough time after their formation in the spiral arm . still , collapse should occur in dense cores of these clouds even if the overall cloud complexes are weakly bound . this is consistent with recent ideas that star formation occurs in the dense cores of generally unbound molecular clouds . the incident cloud model is also consistent with our observations if , again , the shock compression and collapse to star formation occurs primarily in unseen dense cores . the regular distribution of star formation along spiral arms , including the embedded sources as well as the exposed sources , suggest that a mild instability is operating to form weakly bound molecular clouds simultaneously along kpc lengths . simultaneity on these scales requires a uniform initial starting time for the instability , to within the unstable growth time or the flow - through time . in that case the arm could be relatively young , or the shock front in the arm could be relatively new . b.g.e . is grateful to e. mentuch for clarifications about the analysis method that her team proposed . e.a . and a.b thank the cnes for financial support . we acknowledge financial support to the dagal network from the people programme ( marie curie actions ) of the european union s seventh framework programme fp7/2007 - 2013/ under rea grant agreement number pitn - ga-2011 - 289313 . this research made use of the nasa / ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . this work is based in part on observations made with the spitzer space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . this research made use of the nasa / ipac extragalactic database which is operated by jpl , caltech , under contract with nasa . this research made use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation . funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s . department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium for the participating institutions . the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , university of cambridge , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington . ngc 4321 & sxs4 & 222 & 32 & 20.9 & 76 & 9.98 & -21.62 & -22.57 + ngc 5055 & sat4 & 377 & 55 & 7.54 & 27 & 9.03 & -20.36 & -21.37 + ngc 5194 & sas4p & 336 & 52 & 7.54 & 27 & 8.67 & -20.72 & -21.83 + ngc 5248 & sxt4 & 184 & 44 & 15.4 & 56 & 10.63 & -20.31 & -21.23 + ngc 5457 & sxt6 & 865 & 21 & 4.94 & 18 & 8.21 & -20.26 & -20.65 [ galaxies ] lccccccccccc ngc 4321 & & & & & & & & & & & + 1 & 12:22:50.498 & 15:50:24.03 & 30 & 282 & -13.83 & -6.09 & -6.44 & -6.60 & -6.80 & -7.05 & 38.47 + 2 & 12:22:50.238 & 15:50:19.91 & 36 & 240 & -13.65 & -6.37 & -6.74 & -7.13 & -7.02 & -7.36 & 38.66 + 3 & 12:22:48.705 & 15:50:14.28 & 16 & 220 & -13.56 & -5.88 & -6.04 & -6.54 & -6.74 & -7.20 & 38.19 + 4 & 12:22:48.471 & 15:50:11.66 & 24 & 254 & -13.71 & -6.23 & -6.77 & -7.21 & -7.30 & -7.80 & @xmath173 + 5 & 12:22:48.263 & 15:50:09.03 & 25 & 297 & -13.88 & -6.75 & -7.20 & -7.35 & -7.31 & -7.80 & 38.09 + 6 & 12:22:48.238 & 15:49:47.28 & 20 & 219 & -13.55 & -5.93 & -5.84 & -6.80 & -6.49 & -6.97 & 38.14 + 7 & 12:22:47.614 & 15:49:48.78 & 16 & 166 & -13.25 & -6.35 & -6.24 & -7.13 & -6.79 & -7.44 & 38.34 + 8 & 12:22:46.575 & 15:48:55.15 & 9 & 32 & -11.48 & -5.58 & -5.91 & -5.85 & -5.93 & -7.25 & @xmath174 + 9 & 12:22:46.445 & 15:48:50.65 & 6 & 32 & -11.45 & -6.98 & -5.23 & -5.62 & -6.63 & -7.74 & @xmath175 + 10 & 12:22:48.654 & 15:48:04.53 & 16 & 102 & -12.72 & -5.63 & -6.26 & -7.10 & -7.43 & -7.39 & @xmath176 + 11 & 12:22:56.241 & 15:48:10.16 & 48 & 1445 & -15.60 & -6.20 & -6.30 & -7.20 & -6.82 & -7.10 & 38.74 + 12 & 12:22:51.642 & 15:49:33.41 & 20 & 596 & -14.64 & -6.98 & -7.13 & -7.54 & -7.27 & -7.56 & 38.57 + 13 & 12:22:52.603 & 15:49:39.04 & 30 & 364 & -14.10 & -5.72 & -6.14 & -6.68 & -6.79 & -7.14 & 38.38 + 14 & 12:22:51.797 & 15:50:05.66 & 9 & 60 & -12.14 & -6.50 & -6.41 & -7.01 & -7.08 & -7.93 & 37.73 + 15 & 12:22:54.656 & 15:49:33.04 & 9 & 383 & -14.16 & -6.37 & -7.41 & -8.07 & -8.54 & -9.13 & 38.40 + ngc 5055 & & & & & & & & & & & + 1 & 13:15:50.565 & 42:01:16.91 & 6 & 63 & -9.98 & -7.05 & -6.38 & -7.38 & -7.94 & -9.15 & @xmath177 + 2 & 13:15:49.657 & 42:01:12.78 & 12 & 187 & -11.17 & -7.95 & -8.40 & -9.09 & -9.31 & -9.99 & 37.26 + 3 & 13:15:48.512 & 42:01:10.91 & 36 & 651 & -12.52 & -5.52 & -5.39 & -5.95 & -6.69 & -7.74 & 37.89 + 4 & 13:15:42.286 & 42:01:31.89 & 9 & 211 & -11.30 & -6.69 & -6.36 & -7.51 & -7.38 & -7.77 & 36.78 + 5 & 13:15:38.651 & 42:01:42.75 & 12 & 306 & -11.70 & -7.93 & -7.58 & -8.36 & -8.34 & -9.21 & 37.82 + 6 & 13:15:39.155 & 42:01:58.50 & 30 & 1239 & -13.22 & -6.58 & -7.25 & -7.59 & -7.57 & -8.06 & 38.13 + 7 & 13:15:38.886 & 42:02:05.63 & 12 & 507 & -12.25 & -5.42 & -5.42 & -6.68 & -6.09 & -7.34 & 38.40 + 8 & 13:15:39.861 & 42:02:19.51 & 6 & 129 & -10.77 & -6.52 & -5.98 & -6.56 & -7.27 & -8.40 & 37.77 + 9 & 13:15:39.895 & 42:02:19.13 & 9 & 201 & -11.25 & -7.98 & -8.24 & -8.72 & -8.74 & -9.07 & 37.58 + 10 & 13:15:40.534 & 42:02:40.14 & 20 & 555 & -12.35 & -7.11 & -7.11 & -7.61 & -7.81 & -8.56 & 38.23 + 11 & 13:15:47.401 & 42:02:19.91 & 16 & 175 & -11.09 & -5.47 & -5.45 & -6.74 & -6.51 & -7.45 & 37.99 + 12 & 13:15:50.364 & 42:02:25.53 & 6 & 118 & -10.67 & -7.16 & -7.68 & -8.14 & -8.60 & -9.43 & 37.59 + 13 & 13:15:53.326 & 42:02:22.53 & 12 & 157 & -10.98 & -8.44 & -8.35 & -9.36 & -8.49 & -9.29 & @xmath178 + 14 & 13:15:43.330 & 42:00:57.40 & 25 & 852 & -12.82 & -6.65 & -7.08 & -7.22 & -7.34 & -8.49 & 38.73 + 15 & 13:15:56.958 & 42:00:52.14 & 16 & 350 & -11.85 & -6.54 & -6.98 & -8.27 & -8.20 & -8.93 & 37.42 + 16 & 13:15:52.281 & 42:01:19.53 & 12 & 346 & -11.84 & -6.97 & -7.55 & -7.86 & -7.87 & -7.87 & 38.10 + ngc 5194 & & & & & & & & & & & + 1 & 13:29:54.725 & 47:12:36.45 & 16 & 994 & -12.98 & -7.76 & -8.25 & -8.86 & -8.67 & -9.03 & 38.50 + 2 & 13:29:52.038 & 47:12:42.82 & 20 & 1502 & -13.43 & -9.31 & -9.33 & -9.95 & -9.46 & -9.59 & 38.51 + 3 & 13:29:46.997 & 47:12:23.32 & 12 & 278 & -11.60 & -6.39 & -7.25 & -7.63 & -7.74 & -8.84 & 38.03 + 4 & 13:29:45.231 & 47:11:58.94 & 9 & 118 & -10.67 & -6.41 & -6.89 & -7.69 & -7.76 & -8.36 & 37.66 + 5 & 13:29:44.864 & 47:11:55.18 & 16 & 221 & -11.35 & -6.49 & -6.98 & -7.53 & -7.80 & -8.45 & 37.94 + 6 & 13:29:45.489 & 47:11:55.56 & 12 & 57 & -9.88 & -5.88 & -7.11 & -7.64 & -7.68 & -8.27 & 37.51 + 7 & 13:29:42.953 & 47:11:05.30 & 9 & 84 & -10.30 & -6.48 & -7.20 & -7.21 & -7.04 & -7.84 & 37.82 + 8 & 13:29:43.359 & 47:10:33.80 & 9 & 132 & -10.79 & -6.48 & -6.66 & -7.02 & -7.21 & -8.32 & 37.80 + 9 & 13:29:54.098 & 47:10:36.07 & 20 & 662 & -12.54 & -7.14 & -7.89 & -8.50 & -8.49 & -8.88 & 37.99 + 10 & 13:29:56.526 & 47:10:45.82 & 42 & 2465 & -13.97 & -8.45 & -9.23 & -9.62 & -9.67 & -9.81 & 38.58 + 11 & 13:29:56.453 & 47:11:15.82 & 30 & 322 & -11.76 & -7.26 & -7.15 & -7.79 & -7.70 & -8.64 & 38.09 + 12 & 13:29:55.828 & 47:11:44.70 & 12 & 1202 & -13.19 & -8.02 & -8.47 & -9.62 & -9.60 & -9.42 & 38.56 + ngc 5248 & & & & & & & & & & & + 1 & 13:37:29.691 & 8:54:30.29 & 28 & 466 & -13.71 & -9.08 & -9.15 & -9.71 & -9.57 & -10.22 & 38.44 + 2 & 13:37:29.033 & 8:54:20.91 & 24 & 766 & -14.25 & -8.95 & -9.59 & -10.10 & -10.25 & -10.46 & 38.53 + 3 & 13:37:28.780 & 8:53:27.66 & 42 & 1120 & -14.66 & -10.00 & -10.68 & -10.86 & -10.79 & -10.98 & 38.93 + 4 & 13:37:29.286 & 8:53:14.91 & 15 & 278 & -13.15 & -8.29 & -8.21 & -8.60 & -8.94 & -9.86 & 38.60 + 5 & 13:37:30.374 & 8:52:50.16 & 20 & 140 & -12.40 & -8.65 & -8.78 & -9.33 & -8.91 & -9.65 & 38.37 + 6 & 13:37:32.601 & 8:52:43.41 & 20 & 157 & -12.53 & -8.85 & -8.81 & -9.35 & -9.51 & -10.39 & 38.23 + 7 & 13:37:32.930 & 8:52:32.54 & 24 & 291 & -13.20 & -9.41 & -9.65 & -10.42 & -10.10 & -10.46 & 39.02 + 8 & 13:37:34.929 & 8:53:00.29 & 15 & 263 & -13.09 & -8.94 & -9.19 & -9.55 & -9.51 & -9.94 & 38.65 + 9 & 13:37:34.575 & 8:53:06.29 & 16 & 208 & -12.83 & -8.54 & -8.90 & -8.64 & -9.02 & -9.70 & 38.46 + 10 & 13:37:35.106 & 8:52:20.54 & 12 & 157 & -12.53 & -8.55 & -7.88 & -8.69 & -8.71 & -9.34 & 38.50 + 11 & 13:37:33.967 & 8:53:12.29 & 6 & 40 & -11.04 & -7.56 & -7.71 & -7.95 & -8.58 & -9.27 & 37.61 + 12 & 13:37:31.386 & 8:53:37.04 & 42 & 315 & -13.29 & -9.24 & -9.26 & -9.55 & -9.45 & -11.57 & 38.75 + 13 & 13:37:31.285 & 8:53:31.41 & 30 & 155 & -12.52 & -9.15 & -9.28 & -9.72 & -9.77 & -10.32 & 38.55 + ngc 5457 & & & & & & & & & & & + 1 & 14:03:07.750 & 54:22:43.86 & 20 & 227 & -10.46 & -5.71 & -5.56 & -6.35 & -7.35 & -7.76 & 37.62 + 2 & 14:03:05.818 & 54:22:45.35 & 12 & 108 & -9.65 & -5.75 & -5.14 & -5.28 & -5.93 & -6.90 & @xmath179 + 3 & 14:03:04.745 & 54:22:45.73 & 16 & 150 & -10.01 & -5.73 & -5.22 & -5.93 & -6.09 & -7.50 & @xmath180 + 4 & 14:03:03.843 & 54:22:47.22 & 25 & 246 & -10.55 & -6.51 & -6.58 & -7.17 & -7.12 & -7.76 & @xmath181 + 5 & 14:03:03.285 & 54:22:49.09 & 6 & 52 & -8.85 & -5.29 & -4.76 & -6.12 & -6.33 & -6.43 & @xmath182 + 6 & 14:03:03.198 & 54:23:01.84 & 20 & 206 & -10.35 & -6.13 & -6.24 & -7.28 & -7.92 & -8.44 & 36.22 + 7 & 14:03:29.293 & 54:22:02.54 & 25 & 270 & -10.65 & -6.86 & -6.85 & -7.23 & -7.46 & -7.95 & 37.41 + 8 & 14:03:28.604 & 54:21:44.92 & 81 & 1251 & -12.31 & -7.04 & -6.94 & -7.69 & -7.38 & -8.16 & 38.12 + 9 & 14:03:27.230 & 54:21:27.31 & 25 & 217 & -10.41 & -6.86 & -6.97 & -7.51 & -7.47 & -7.89 & 38.18 + 10 & 14:03:26.243 & 54:21:25.07 & 35 & 386 & -11.04 & -7.27 & -7.50 & -7.84 & -7.99 & -8.37 & 37.83 + 11 & 14:03:20.620 & 54:20:47.22 & 24 & 204 & -10.35 & -6.23 & -6.61 & -7.02 & -7.00 & -7.54 & 37.81 + 12 & 14:03:19.032 & 54:20:18.73 & 12 & 98 & -9.55 & -6.37 & -6.55 & -7.00 & -7.19 & -7.47 & 37.35 + 13 & 14:03:12.514 & 54:19:07.87 & 36 & 672 & -11.64 & -6.43 & -6.19 & -7.12 & -6.76 & -7.79 & 37.94 + 14 & 14:03:14.186 & 54:19:08.99 & 15 & 150 & -10.01 & -6.34 & -6.16 & -6.95 & -6.72 & -6.89 & 38.49 + 15 & 14:02:57.854 & 54:19:20.93 & 12 & 53 & -8.88 & -5.93 & -5.53 & -5.78 & -6.05 & -6.85 & 37.52 + 16 & 14:02:58.199 & 54:19:07.06 & 9 & 65 & -9.10 & -5.90 & -5.18 & -6.56 & -6.57 & -6.93 & 37.64 + 17 & 14:03:27.177 & 54:19:46.81 & 56 & 1118 & -12.19 & -7.32 & -7.53 & -8.13 & -8.02 & -8.45 & 38.24 [ results ] lcccccc ngc 4321 & 1 & 3.8 & @xmath183 & 5.3 & @xmath184 & 4.2 + & 2 & 4.0 & @xmath185 & 5.0 & @xmath186 & 2.8 + & 3 & 3.5 & @xmath187 & 5.3 & @xmath188 & 4.8 + & 4 & @xmath189 & @xmath190 & 5.0 & @xmath191 & @xmath192 + & 5 & 3.4 & @xmath193 & 5.3 & @xmath194 & 5.1 + & 6 & 3.4 & @xmath187 & 5.3 & @xmath195 & 4.8 + & 7 & 3.6 & @xmath196 & 5.3 & @xmath197 & 4.4 + & 8 & @xmath198 & @xmath199 & 4.4 & @xmath200 & @xmath201 + & 9 & @xmath202 & @xmath203 & 4.4 & @xmath204 & @xmath205 + & 10 & @xmath206 & @xmath207 & 5.0 & @xmath208 & @xmath209 + & 11 & 4.0 & @xmath210 & 6.2 & @xmath211 & 5.7 + & 12 & 3.9 & @xmath212 & 5.8 & @xmath213 & 5.1 + & 13 & 3.7 & @xmath214 & 5.5 & @xmath215 & 4.7 + & 14 & 3.0 & @xmath216 & 4.5 & @xmath217 & 4.0 + & 15 & 3.7 & @xmath218 & 5.6 & @xmath219 & 5.0 + ngc 5055 & 1 & @xmath220 & @xmath221 & 3.7 & @xmath222 & @xmath223 + & 2 & 2.5 & @xmath224 & 4.5 & @xmath225 & 5.1 + & 3 & 3.2 & @xmath226 & 5.0 & @xmath227 & 4.8 + & 4 & 2.1 & @xmath228 & 4.5 & @xmath229 & 6.4 + & 5 & 3.1 & @xmath230 & 4.6 & @xmath231 & 4.0 + & 6 & 3.4 & @xmath232 & 5.2 & @xmath233 & 4.9 + & 7 & 3.7 & @xmath234 & 4.8 & @xmath235 & 3.1 + & 8 & 3.1 & @xmath236 & 4.3 & @xmath237 & 3.3 + & 9 & 2.9 & @xmath238 & 4.5 & @xmath239 & 4.3 + & 10 & 3.5 & @xmath240 & 4.8 & @xmath241 & 3.5 + & 11 & 3.3 & @xmath242 & 4.3 & @xmath243 & 2.8 + & 12 & 2.9 & @xmath244 & 4.1 & @xmath245 & 3.4 + & 13 & @xmath246 & @xmath247 & 4.3 & @xmath248 & @xmath249 + & 14 & 4.0 & @xmath239 & 5.1 & @xmath250 & 3.0 + & 15 & 2.7 & @xmath251 & 4.7 & @xmath252 & 5.3 + & 16 & 3.4 & @xmath253 & 4.7 & @xmath254 & 3.6 + ngc 5194 & 1 & 3.8 & @xmath255 & 5.2 & @xmath256 & 3.4 + & 2 & 3.8 & @xmath257 & 5.3 & @xmath258 & 4.1 + & 3 & 3.3 & @xmath236 & 4.6 & @xmath259 & 3.5 + & 4 & 2.9 & @xmath260 & 4.2 & @xmath234 & 3.4 + & 5 & 3.2 & @xmath261 & 4.4 & @xmath262 & 3.3 + & 6 & 2.8 & @xmath263 & 3.9 & @xmath264 & 3.0 + & 7 & 3.1 & @xmath265 & 4.1 & @xmath266 & 2.8 + & 8 & 3.1 & @xmath267 & 4.3 & @xmath268 & 3.2 + & 9 & 3.3 & @xmath269 & 5.0 & @xmath270 & 4.5 + & 10 & 3.9 & @xmath271 & 5.5 & @xmath256 & 4.4 + & 11 & 3.4 & @xmath272 & 4.7 & @xmath273 & 3.6 + & 12 & 3.8 & @xmath274 & 5.3 & @xmath275 & 3.8 + ngc 5248 & 1 & 3.7 & @xmath276 & 5.4 & @xmath277 & 4.2 + & 2 & 3.8 & @xmath278 & 5.6 & @xmath279 & 4.7 + & 3 & 4.2 & @xmath280 & 5.8 & @xmath258 & 4.2 + & 4 & 3.9 & @xmath255 & 5.2 & @xmath256 & 3.7 + & 5 & 3.7 & @xmath281 & 4.9 & @xmath266 & 3.4 + & 6 & 3.5 & @xmath230 & 4.8 & @xmath214 & 3.5 + & 7 & 4.3 & @xmath282 & 5.2 & @xmath283 & 2.5 + & 8 & 3.9 & @xmath284 & 5.2 & @xmath285 & 3.5 + & 9 & 3.7 & @xmath284 & 5.0 & @xmath286 & 3.5 + & 10 & 3.8 & @xmath287 & 5.0 & @xmath258 & 3.3 + & 11 & 2.9 & @xmath288 & 4.3 & @xmath289 & 3.6 + & 12 & 4.0 & @xmath290 & 5.2 & @xmath291 & 3.1 + & 13 & 3.8 & @xmath257 & 4.3 & @xmath292 & 1.6 + ngc 5457 & 1 & 2.9 & @xmath293 & 4.0 & @xmath294 & 2.9 + & 2 & @xmath295 & @xmath296 & 3.7 & @xmath297 & @xmath298 + & 3 & @xmath299 & @xmath300 & 4.0 & @xmath301 & @xmath302 + & 4 & @xmath189 & @xmath303 & 4.1 & @xmath304 & @xmath305 + & 5 & @xmath206 & @xmath306 & 3.5 & @xmath307 & @xmath308 + & 6 & 1.5 & @xmath309 & 3.8 & @xmath310 & 6.1 + & 7 & 2.7 & @xmath311 & 4.1 & @xmath312 & 3.9 + & 8 & 3.4 & @xmath313 & 4.9 & @xmath314 & 4.0 + & 9 & 3.5 & @xmath292 & 4.0 & @xmath183 & 1.7 + & 10 & 3.1 & @xmath257 & 4.2 & @xmath315 & 3.1 + & 11 & 3.1 & @xmath287 & 4.0 & @xmath316 & 2.6 + & 12 & 2.6 & @xmath317 & 3.7 & @xmath318 & 3.0 + & 13 & 3.2 & @xmath232 & 4.6 & @xmath319 & 3.8 + & 14 & 3.8 & @xmath183 & 4.0 & @xmath320 & 0.9 + & 15 & 2.8 & @xmath321 & 3.4 & @xmath322 & 1.7 + & 16 & 2.9 & @xmath323 & 3.4 & @xmath324 & 1.6 + & 17 & 3.5 & @xmath325 & 4.8 & @xmath326 & 3.5 lccccc ngc 4321 & @xmath327 & @xmath328 & @xmath329 & @xmath330 & @xmath331 + ngc 5055 & @xmath332 & @xmath333 & @xmath334 & @xmath335 & @xmath336 + ngc 5194 & @xmath337 & @xmath338 & @xmath339 & @xmath340 & @xmath341 + ngc 5248 & @xmath342 & @xmath343 & @xmath344 & @xmath345 & @xmath346 + ngc 5457 & @xmath347 & @xmath348 & @xmath349 & @xmath350 & @xmath351 [ averages ] -band image for blue , sdss @xmath25-band image for green , and irac @xmath10 m image for red , all with @xmath12 pixels and combined by rotating , stretching , and aligning the optical images to match the infrared images . ( top right ) sdss color image remade from the _ u _ , _ r _ , and _ z _ sdss filters using data from the sdss webpage . ( middle left ) sample region at @xmath10 m with circles highlighting @xmath10 m sources that do not appear in optical bands . ( middle right ) the same circles on an sdss @xmath17-band image ( the circles are @xmath26 larger than the measurement rectangles ) . ( bottom left ) @xmath0 m full image with a rectangle indicating the sample region . ( bottom right ) sample region again in h@xmath1 . , width=528 ] |
the bulk of isotopes is created in the gamma processes mainly by sequences of photodisintegrations and beta decays in explosive conditions in type ia supernovae ( snia ) or in core collapse supernovae ( ccsn ) . the contribution of different stellar sources to the observed distribution of-nuclei in the solar system is still under debate . we explore single degenerate type ia supernovae in the framework of two - dimensional snia delayed - detonation explosion models . travaglio et al . ( 2011 ) ( hereafter trv11 ) discussed the sensitivity of-nuclei production to different snia models , i.e. delayed detonations of different strength , deflagrations , and the dependence on selected-process seed distributions . here we present a detailed study of-process nucleosynthesis occuring in snia with-process seeds at different metallicities . based on the delayed - detonation model ddt - a of trv11 , we analyze the dependence of-nucleosynthesis on the-seed distribution obtained from different strengths of thec - pocket . we also demonstrate thatpb - seed alone changes the-nuclei production considerably . the heavy- seeds ( 140 208 ) contribute with about 30 - 40% to the total light- nuclei production up toba ( with the exception ofmo andba , to which the heavy- seeds contribute with about 15% only ) . using a galactic chemical evolution code ( see travaglio et al . 2004 ) we study the contribution of snia to the solar stable-nuclei . we find that explosions of chandrasekhar - mass single degenerate systems produce a large amount of-nuclei in our galaxy , both in the range of light ( 120 ) and heavy-nuclei , at almost flat average production factors ( within a factor of about 3 ) . we discussed in details-isotopes such asmo with a behavior diverging from the average , which we attribute to uncertainties in the nuclear data or in snia modelling . li et al . ( 2011 ) find that about 70% of all sneia are normal events . if these are explained in the framework of explosions of chandrasekhar - mass white dwarfs resulting from the single - degenerate progenitor channel , we find that they are responsible for at least 50% of the-nuclei abundances in the solar system . | 1411.2399 | the origin of heavy nuclei was discussed by cameron ( 1957 ) , who called 35 species _ excluded isotopes_. indeed they are outside of both the @xmath1 and @xmath9 neutron capture paths , and they are typically 10@xmath101000 times less abundant than the corresponding @xmath1- and/or @xmath9-isotopes in the solar system . the origin of @xmath0-nuclei was investigated starting with the pioneering work of cameron ( 1957 ) and burbidge et al . ( 1957 ) , and later by audouze & truran ( 1975 ) and arnould ( 1976 ) . the first work analyzing the possibility of having efficient photodisintegrations in chandrasekhar - mass snia explosions was published by howard , meyer & woosley ( 1991 ) . the initial @xmath1-seed distribution they used was derived from helium flashes as calculated by howard et al . ( 1986 ) . in figure 2 of that work , the authors claim that they can reproduce the abundance pattern of all @xmath0-nuclei , including light-@xmath0 nuclei , within a factor of about three . however , they obtained an overproduction of @xmath11se , @xmath12kr , and @xmath13sr . the @xmath0-process abundances of these three isotopes are very sensitive to the proton density , which the authors considered rather uncertain . they also obtained a rather low production of @xmath6mo and @xmath14ru with respect to the other light @xmath0-nuclei . a detailed discussion on these light @xmath0-nuclei will be given in section 4 of the present paper . later goriely et al . 2002 , goriely et al . 2005 , and arnould & goriely 2006 , analyzed the @xmath0-process production in he - detonation models for sub - chandrasekhar mass wds . these authors considered as seeds @xmath1-process solar abundances . they found ca to fe to be overabundant with respect to @xmath0-nuclei ( with the exception of @xmath12kr ) by a factor of @xmath15100 . they concluded that a he detonation is not an efficient scenario for the production of the bulk solar - system @xmath0-isotopes . kusakabe et al . ( 2011 ) presented @xmath0-process nucleosynthesis calculations in a co - deflagration model of snia , i.e. , the w7 model of nomoto et al . similar to howard et al . ( 1986 ) , they assumed enhanced @xmath1-seed distributions using the classical @xmath1-process analysis , testing two different mean neutron exposures @xmath16 , a flat distribution for @xmath16 = 0.30 mb@xmath17 , and a decreasing @xmath1-process distribution corresponding to @xmath16 = 0.15 mb@xmath17 . they noticed that for a flat @xmath1-seed enhanced distribution the production factors of light @xmath0-nuclei show a strong deficiency in the range @xmath12se to @xmath18ru . from this , they concluded that snia may have contributed to the enrichment of @xmath0-nuclei more effectively than ccsne . in our previous trv11 paper , we obtained a consistent production of @xmath0-nuclei in the single - degenerate chandrasekhar mass wd explosion scenario . we presented two - dimensional hydrodynamic models of snia . the corresponding nucleosynthesis was calculated in a post - processing step following the thermal history of lagrangian tracer particles . we found a significant production of @xmath0-nuclei from these stellar explosions , at the same level compared to @xmath19fe for light as well as heavy @xmath0-nuclei . we demonstrated that our model is able to produce light and heavy @xmath0-nuclei in one single process . in our analysis , we assumed enhanced @xmath1-seed distributions directly obtained from a sequence of thermal pulses in the material accreted onto the exploding white dwarf from a normal companion star ( see gallino et al . 1998 ) . in the context of light-@xmath0 nuclei production , the major problem discussed in trv11 is the resulting abundance of @xmath6mo , found to be far too low relative to the abundances of the other light-@xmath0 nuclei . we also found an important contribution from @xmath0-process nucleosynthesis to @xmath20kr and @xmath21sr ( originally considered @xmath1-only nuclei ) , to the neutron magic @xmath22zr , and to the neutron - rich @xmath14zr ( due to neutron captures from the residual abundance of @xmath23ne during the explosive phase ) . concerning the heavy @xmath0-isotopes , the @xmath1-process nature of @xmath24gd has been later confirmed by different works on @xmath1-process nucleosynthesis ( gallino et al . 1998 ; bisterzo et al . 2010 ) , where a predominant @xmath1-process origin was demonstrated ( see also discussion in trv11 ) . in the present paper we investigate for the first time in the literature the effect of metallicity on @xmath0-process nucleosynthesis in snia , starting with a range of @xmath1-seed distributions obtained for different metallicities . using a simple chemical evolution code ( travaglio et al . 2004 ) , we estimate the contribution of snia to the solar @xmath0-process composition . the same study was recently done for radiogenic @xmath0-isotopes by travaglio et al . ( 2014 ) . it is currently impossible to discuss the interplay between the role of sneia and ccsne in the production of @xmath0-nuclei in the galaxy . infact a complete study of @xmath0-nucleosynthesis with metallicity is missed for ccsne . but also from the observational point of view , unfortunatey there is no way to measure observed chemical evolution of @xmath0-nuclei since they are too rare with respect to the @xmath1- and @xmath9-fractions . for example , there are recent attempts to observe elements like mo and ru ( very debated for @xmath0-process nucleosynthesis studies ) in field stars of the galaxy at different metallicities ( see e.g. hansen et al . 2014 ; peterson 2013 ) . only a small fraction of the elements mo and ru are @xmath0 : 14.84% of the total mo is @xmath25mo and only 9.25% is @xmath6mo ; for ru , only 5.5% of the element is @xmath14ru and 1.88% is @xmath18ru . the @xmath0-fractions indicated above are therefore too small to be of any interest when the elements are observed in field stars . only when will be possible to observe isotopes of mo and ru in the spectra then we will be able to give interesting indications on the details of their nucleosynthesis processes . for this work , the adopted snia model ( described in detail in trv11 ) is summarized in section 2 , together with a brief description of the tracer particles method used for nucleosynthesis calculations . in section 3 the @xmath1-process seed distributions considered in our study are described in detail . the resulting @xmath0-process production and the effect of metallicity on @xmath0-nuclei is discussed in section 4 , where also an analysis of the various production mechanisms of the @xmath0-nuclei will be performed , taking into account nuclear uncertainties . in section 5 we analyzed in detail how @xmath1-seeds with different atomic mass number contribute to the @xmath0-nuclei production . galactic chemical evolution calculations are presented in section 6 , and the contribution of snia to the solar system @xmath0-nuclei abundances is studied . in section 7 we discuss the contribution of snia to the production of radiogenic @xmath25nd , @xmath26sm and @xmath27tc . finally , conclusions and work in progress are drawn in section 8 . we used the snia - explosion model ( ddt - a ) described in detail by trv11 . it is a representative example of the single - degenerate scenario in which the wd has accreted material from a main - sequence or evolved companion star until it finally approaches the chandrasekhar mass and explodes as a delayed detonation . the model is based on the two - dimensional simulations presented by kasen et al 2009 . the explosion itself is simulated in 2d by means of the combustion code leafs ( reinecke et al . 1999 ; reinecke et al . 2002 ; rpke 2005 ; rpke & hillebrandt 2005 ) which follows the evolution with an eulerian grid . in order to compute for each zone of the star the history of temperature and density over time for each zone of the star we introduce a lagrangian component in the form of tracer particles . with the tracer - particle method it is possible to reconstruct the ensuing nucleosynthesis . the nuclear post - processing calculations are performed separately for each tracer . summing the chemical composition over all tracer particles gives the total yields . the tracer particles method was first introduced by nagataki et al . ( 1997 ) for ccssne , and by travaglio et al . ( 2004b , 2005 ) for snia . for 2d simulations it has been verified that 51,200 particles as used here , uniformly distributed in mass coordinates , give sufficient resolution ( seitenzahl et al . 2010 ) for each tracer particle we follow the explosive nucleosynthesis with a detailed nuclear reaction network for all isotopes up to @xmath28bi . we select tracers within the typical temperature range for @xmath0-process production , i.e. @xmath29 k , and analyze their behavior in detail , exploring the influence of different @xmath1-process seeds on the @xmath0-process nucleosynthesis . in order to determine the @xmath1-process enrichment prior to the explosion , we assume recurrent flashes occurring in the he - shell during the accretion phase with neutrons mainly released by the @xmath2c(@xmath30,n)@xmath31o reaction ( iben 1981 ; trv11 ) . this applies under the assumption that a small amount of protons are ingested in the top layers of the he intershell . protons are captured by the abundant @xmath32c and convert it into @xmath2c via @xmath32c(@xmath0,@xmath33)@xmath2n(@xmath34)@xmath2c at @xmath35 k. in our model , @xmath0-process nucleosynthesis occurs in snia starting from a pre - explosion @xmath1-process enriched seed composition . therefore , it is essential to determine the @xmath1-process enrichment prior to the explosion . here , we assume enhanced @xmath1-distributions produced directly by a sequence of thermal - pulse instabilities in the accreted material . this idea has been described in detail by trv11 and was previously discussed by iben ( 1981 ) , iben & tutukov ( 1991 ) , and howard & meyer ( 1993 ) . to be more specific , we assume recurrent flashes to occur in the he - shell during the accretion phase . the matter accumulated onto the carbon - oxygen white dwarf ( hereafter co - wd ) therefore becomes enriched in @xmath1-nuclei . however , the mass involved and the physical properties of the @xmath2c - pocket , providing the free neutrons for the @xmath1-process , still have to be considered as free parameters . since no physical models are available we explore different @xmath1-process distributions in order to better understand the dependence of our results on these initial seeds ( see also the discussion in trv11 and travaglio et al . 2012 ) . for the present work we calculate @xmath1-process distributions for 8 metallicities , i.e. @xmath360.02 , 0.019 , 0.015 , 0.012 , 0.011 , 0.010 , 0.006 , 0.003 ( a refined @xmath1-seed metallicity grid is necessary for chemical evolution calculations since the @xmath1-seeds are strongly dependent on metallicity ) , and we interpolate for all the other metallicities in between in order to calculate galactic chemical evolution . we also investigate the effect of the @xmath1-seeds with different @xmath2c - pocket properties ( st@xmath372 , [email protected] , st , st/1.5 , where @xmath384 @xmath37 10@xmath39 of @xmath2c in the pocket corresponds to the st case , gallino et al . 1998 ; see gallino et al . 1998 and bisterzo et al . 2010 for a detailed discussion of @xmath2c - pocket profiles ) . the @xmath1-process seeds adopted are shown in figure 1 for the st@xmath372 @xmath2c - pocket case , and metallicites of @xmath360.01 , 0.006 , 0.004 , 0.003 . in figure 2 we show the average of four @xmath2c - pockets at different metallicities . in all figures the s - seed abundances are normalized to the solar abundances of lodders ( 2009 ) . as discussed by gallino et al . ( 1998 ) and travaglio et al . ( 1999 ) , the synthesis of heavy nuclei requires neutron captures starting from fe seeds , so that the @xmath1-process is expected to decrease with decreasing metallicity , i.e. to be of _ secondary _ nature . however , the abundances produced depend not only on the initial fe concentration but also on the neutron exposure . the concentration of @xmath2c in the pocket is of _ primary - like _ nature ( it is built from h and freshly made c , and hence is independent of metallicity ) , while the abundance of the neutron absorber @xmath19fe varies linearly with @xmath40 . therefore , for a given amount of @xmath2c in the pocket , the neutron exposure ( proportional to the ratio @xmath2c/@xmath19fe ) is expected to increase linearly with decreasing metallicity . this dependence would compensate for the secondary nature of the @xmath1-elements , if the yields of @xmath1-nuclei were linearly dependent on both neutron exposure and @xmath40 . in addition , the behavior of @xmath3pb has to be considered separately ( travaglio et al . the gradual increase of the neutron exposure towards low metallicities masks the expected secondary behavior ( see also clayton 1988 ) , resulting in a rather complex dependence of @xmath1-process yields on metallicity . for the lower metal content the neutron flux feeds pb ( in particular @xmath3pb ) . a clear understanding of the @xmath1-seeds behavior versus metallicity , including the production of neutron - magic @xmath3pb at the termination of the @xmath1-process path , is very important for the nature of @xmath0-process ( see section 4 ) . figure 1 shows a variation by a factor of @xmath1510 for the abundances of the light @xmath1-only isotopes ( up to @xmath41140 ) when @xmath36 varies from 0.003 to 0.01 , and a spread of @xmath155 for heavy @xmath1-only nuclei ( with @xmath42140 ) . for @xmath3pb , the variation is by a factor of @xmath158 , but its trend is inverted with respect the behavior of other @xmath1-only isotopes ( i.e. lower @xmath40 gives higher @xmath3pb abundance ) . in figure 2 we plot the average of four different @xmath2c - pockets ( st@xmath372 , [email protected] , st , st/1.5 ) for the metallicities discussed above . the behavior of the @xmath1-only isotopes is different from that seen in figure 1 : still a rough dependence on @xmath40 for light @xmath1-nuclei in the range 90@xmath43 120 is observed , but the variation is now reduced by a factor of 3 . the spread is progressively decreasing for higher metallicities . further on , for heavy @xmath1-isotopes with 140 @xmath44 204 , an almost unique and flat distribution is obtained , independent of metallicity with an enhancement around 2000 times solar . eventually , for @xmath3pb a spread of about a factor of 4 stands out again , with the lowest metallicity showing the highest @xmath3pb abundance , at the level of 5000@xmath37 solar . the general trend sketched above can be understood in the light of the typical @xmath1-process enhancement occurring in agb stars . for a given @xmath2c - pocket strength , decreasing the metallicity the @xmath1-flow feeds more and more @xmath3pb at the termination of the @xmath1-path . at the same time a progressive depletion of the lighter @xmath1-process isotopes is found up to the magic neutron number @xmath45 82 ( @xmath46ba to @xmath47nd ) . in the region between the magic neutron number nuclei @xmath4882 and @xmath48126 , an almost flat @xmath1-process production factor ensues . a similar trend occurs at a fixed metallicity by increasing the @xmath2c - pocket strength . we recall here that case stx2 is around the maximum @xmath2c - pocket strength we can reach , beyond which further proton ingestion during a third dredge up episode from the envelope would result in a decrease of @xmath2c and production of the neutron poison @xmath49n ( see the review by busso , gallino , & wasserburg 1999 ) . in other words , with decreasing metallicity the otherwise flat @xmath1-process distribution near the neutron magic numbers n = 50 amd n = 82 , corresponding to atomic mass numbers around @xmath5090 and @xmath50140 , and also at the termination of the @xmath1-path at around @xmath50208 are progressively distorted . in figure 3 we show the @xmath1-seeds for the range of @xmath2c - pockets ( st@xmath372 , [email protected] , st , st/1.5 , st/2 ) and metallicities ( @xmath51 0.02 , 0.019 , 0.015 , 0.012 , 0.011 , 0.010 , 0.006 , 0.003 ) we cover in the galactic chemical evolution calculations ( see section 6 for discussion ) . the @xmath0-process nucleosynthesis is calculated using a nuclear network with 1024 species from neutrons and protons up to @xmath28bi combined with neutron , proton , and @xmath30-induced reactions and their inverse . the code used for this work was originally developed and presented by thielemann et al . we employ the nuclear reaction rates based on the experimental values and the hauser & feshbach statistical model non - smoker ( rauscher & thielemann 2000 ) , including the recent experimental results of maxwellian - averaged neutron capture cross section of various @xmath0-only isotopes ( dillmann et al . 2010 ; marganiec et al . theoretical and experimental electron capture and @xmath52-decay rates are from langanke & martnez - pinedo ( 2000 ) . we discuss in this section the sensitivity of @xmath0-process production to @xmath1-seeds at different metallicites . we analyze the _ primary_/_secondary _ nature of the resulting @xmath0-nuclei . according to rauscher et al . ( 2013 ) ( and references therein ) the @xmath0-process is of _ secondary _ nature and scales with the amount of seed nuclei in the star . in this work we present our results obtained with snia models which only partly confirm this statement . in figure 4 we plot the resulting @xmath0-process abundances , starting from @xmath11se , obtained by using different @xmath1-seeds at different metallicities . on the axis of ordinates the production factor of each isotope is plotted with respect to solar , normalized to @xmath19fe . note that the abundances of @xmath0-nuclei heavier than @xmath53 100 are much higher than a factor of @xmath383 times their solar value . however , in this figure we plot the nucleosynthesis resulting from one single star and not the integrated abundances over all the galaxy ( see section 6 for discussion ) . the choice for these @xmath2c - pockets and metallicities will be used for our best fit of galactic chemical evolution calculation . in order to better understand the dependence of @xmath0-nuclei production on metallicity and the @xmath2c - pocket , we show in two separate figures ( figure 5 and figure 6 ) the behavior of @xmath0-process nucleosynthesis as a function of metallicity and the @xmath2c - pocket , respectively . from figure 5 we can see that the first three @xmath0-only isotopes , @xmath11se , @xmath12kr , and @xmath13sr , show a secondary behavior : their abundances scale almost linearly with @xmath40 . they depend mainly ( for about 60 - 70% ) on the light ( up to @xmath41140 ) @xmath1-seeds that , as shown in figure 1 and figure 2 , are strongly dependent on metallicity . starting from @xmath22zr ( even if this isotope is not @xmath0-only it has to be included in the discussion , see also trv11 ) and @xmath25mo and up to @xmath46ba , the @xmath0-nuclei show a very weak dependence on @xmath40 . the isotopes in this atomic mass - number range are mainly produced by photodisintegration from the heavy @xmath1-seeds isotopes , thus showing a primary - like behavior ( see figure 6 ) . isotopes in the region from @xmath54ce up to @xmath55hg scale with metallicity but show the opposite trend , i.e. highest abundances for the lowest metallicities . trv11 ( see also dillmann et al . 2008a ) discussed the fact that pb - seeds are converted to nuclei of lower mass by photodisintegration sequences starting with ( @xmath33,n ) reactions . therefore , an important contribution to the heavy @xmath0-only isotopes is obtained . we also found , as discussed in trv11 , that the isotopes @xmath56 in , @xmath57sn , @xmath46la , @xmath24gd , and @xmath58ta , diverge from the average @xmath0-process production . among them , @xmath24gd and @xmath58ta have an important contribution from the @xmath1-process in agb stars ( arlandini et al . 1999 ) , or the neutrino process in ccsn ( woosley et al . 1990 ; wanajo et al . 2011 ) . both @xmath56 in and @xmath57sn are not fed by the @xmath0-process nor by the @xmath1-process . for these , we refer to the discussion of dillmann et al . ( 2008b ) and trv11 . the still puzzling @xmath6mo deserves special attention . can nuclear uncertainties account for the low @xmath6mo yield , compared to the other @xmath0 nuclei , in our models ? howard et al . ( 1991 ) demonstrated that most of the @xmath6mo is produced from @xmath18mo through a ( @xmath33,n ) chain . we confirmed that a similar reaction chain also acts in our models . to illustrate this , figure [ fig : flowtracers ] shows three exemplary trajectories for which time - integrated reaction flows are plotted in figures [ fig : flow94momax]@xmath10[fig : flow94momin ] . the flow shown in figure [ fig : flow94momax ] gives a maximum production of @xmath6mo ( the corresponding trajectory is labeled `` @xmath6mo@xmath59 '' in figure [ fig : flowtracers ] ) . the one shown in figure [ fig : flow94modrop ] also favors @xmath6mo production but at a lower level ( the corresponding trajectory is labeled `` @xmath6mo@xmath60 '' in figure [ fig : flowtracers ] ) . in both cases , @xmath6mo is fed through ( @xmath33,n ) reaction sequences , with the strongest flow originating at @xmath18mo , which , in turn , is replenished to some extent also by a small flow originating from @xmath61mo . it should be noted that the reactivities for this ( @xmath33,n ) sequence are experimentally well constrained , as they involve stable nuclei and the corresponding ( n,@xmath33 ) reaction cross sections have been measured ( dillmann et al . thermal population of excited states does not play a role in these nuclei and thus the measured ( n,@xmath33 ) cross sections allow to compute the ( n,@xmath33 ) and ( @xmath33,n ) rates without further theory uncertainties ( rauscher 2012 , 2014 ) . therefore this part of the flow does not bear large nuclear uncertainties , it is rather determined by the seed abundances received by the stable mo isotopes . some production of @xmath6mo is also found via @xmath62tc(@xmath33,p)@xmath6mo . its contribution to @xmath6mo , however , is an order of magnitude lower than that of the ( @xmath33,n ) sequence and thus it does not contribute appreciably to the uncertainty although the @xmath6mo(p,@xmath33 ) reactivities are unmeasured . consequently , the only significant nuclear uncertainty is found in the destruction of @xmath6mo by @xmath6mo(@xmath33,n)@xmath63mo . as already discussed by travaglio et al . ( 2014 ) , the sequence @xmath6mo(@xmath33,n)@xmath63mo(@xmath33,n)@xmath25mo , leads to production of @xmath25mo , with @xmath6mo(@xmath33,n)@xmath63mo being the slower , and thus determining , reaction here . a conservative estimate for the uncertainty of this rate ( taken from rauscher & thielemann 2000 ) is a factor of 2 . the comparison between the `` @xmath6mo@xmath59 '' and `` @xmath6mo@xmath60 '' trajectories in figure [ fig : flowtracers ] shows that `` @xmath6mo@xmath60 '' reaches a slightly higher peak temperature . this higher temperature not only increases the ( @xmath33,n ) flow , that destroys @xmath6mo without increasing the production notably but also enables further destruction of @xmath6mo by @xmath6mo(@xmath33,@xmath30)@xmath22zr . the latter reaction is offset by a slight enhancement in @xmath62tc(@xmath33,p)@xmath6mo but the flow through this reaction is still lower by a factor of 0.1 than the increased destruction flow . this illustrates that @xmath6mo can only be produced within a narrow temperature window : at low temperature , the ( @xmath33,n ) flow is small or non - existent , at too high temperature @xmath6mo is destroyed by additional reactions . prerequisite for the efficient production of @xmath6mo by photodisintegration in any site is that the mainly contributing trajectories spend as much time as possible in this temperature window . finally , figure [ fig : flow94momin ] shows the flows for a trajectory leading to a minimal production of @xmath6mo although the achieved peak temperature is similar to that in the `` @xmath6mo@xmath59 '' case ( the corresponding trajectory is labeled `` @xmath6mo@xmath64 '' in figure [ fig : flowtracers ] ; incidentally this is the same trajectory giving a maximum @xmath25nb production as shown in figure 5 of travaglio et al . inspection of the flows shows that all ( @xmath33,n ) flows in this region are significantly suppressed compared to the `` @xmath6mo@xmath59 '' case in figure [ fig : flow94momax ] . the @xmath6mo@xmath65mo flow is too small to show in the plotted range . the @xmath66zr@xmath67zr flow is even replaced by its reverse . the key to understanding the difference is the fact that the `` @xmath6mo@xmath64 '' trajectory reaches photodisintegration temperatures only at much higher density ( note the logarithmic scale of the horizontal axis in figure [ fig : flowtracers ] ) . at all times , @xmath68(n,@xmath33)@xmath69 and @xmath69(@xmath33,n)@xmath68 rates are competing . their relative strengths are on one hand determined by the reaction @xmath70-value ( which is given through the well - known nuclear masses ) but on the other hand ( n,@xmath33 ) rates also scale with the available neutron density whereas ( @xmath33,n ) do not ( rauscher 2011 ) . at the temperatures at which photodisintegration of nuclei in the mo region becomes possible , heavier nuclei are already significantly destroyed because they are less tightly bound ( their @xmath70-value for ( n,@xmath33 ) is lower than that of lighter nuclei ; see also rauscher et al . this allows ( n,@xmath33 ) reactions to occur in the lighter region and at high density they will be faster than their ( @xmath33,n ) counterparts . this illustrates the important point that significant production of @xmath0-nuclei in a @xmath33-process is possible only when the densities remain limited , i.e. it depends sensitively on the thermodynamic histories of the explosive layers . for example , a model in which more trajectories experience lower densities during @xmath33-processing would lead to increased @xmath6mo production . however , since this scenario will affect all ( n,@xmath33)/(@xmath33,n ) ratios and thus it is unclear whether it would lead to an enhancement of the final @xmath6mo with respect to the other light @xmath0-nuclides . this has to be investigated in detail in future calculations because the impact of enhanced ( n,@xmath33 ) rates at higher density is not trivial as it depends on the specific @xmath70-values in the ( @xmath33,n ) chains producing specific isotopes . furthermore , it will also impact other reaction types ( such as ( p , n ) , ( p,@xmath33 ) , and their reverses ) in different ways and thus the final outcome strongly depends on the actual reaction sequences producing and destroying a specific nucleus . this may change the ratio of the @xmath6mo abundance also relative to abundances of other @xmath0-nuclei but it remains an open question whether the required relative increase by a factor of 10 can be achieved . in order to understand in detail the @xmath1-seed origin of each @xmath0-nucleus , we performed the following study : fixing the metallicity ( @xmath36 0.006 , i.e. , to a value where we find the highest production of @xmath0-nuclei in our galactic chemical evolution calculations , see section 6 ) and fixing a @xmath2c - pocket ( i.e. , stx2 which is the highest value we used ) , we first tested the role of the @xmath1-seed @xmath3pb alone , by assigning solar abundances to all other @xmath1-seed nuclei . details of the resulting effect on @xmath0-nucleosynthesis are given in table [ tab : z0p006 ] ( third column for the @xmath3pb only case , and second column for the standard case ) . a @xmath3pb seed alone contributes about 30% to @xmath11se , and 10 - 20% ( or less ) to all the other light-@xmath0 nuclei up to @xmath71dy ( with the exception of @xmath72sm , where @xmath3pb alone accounts for about 45% ) . the contribution to the isotopes between @xmath73er and @xmath74pt from @xmath3pb is about 40% . finally , for the heaviest of the @xmath0-only nuclei , @xmath55hg , we find the highest share ( 60% ) in production originating from @xmath3pb . in table [ tab : snuc ] we list all @xmath1-isotopes which we find to have @xmath0-contribution . for these isotopes we report in the last column the effect of @xmath3pb . we find that most of these isotopes ( from @xmath20kr up to @xmath75hf ) , the @xmath3pb seed contributes about 10 - 20% ( or less ) . a similar analysis has been carried out accounting for either as @xmath1-nuclei either the only heavy @xmath1-nuclei ( 140 @xmath4 208 ) or the light @xmath1-nuclei ( 70 @xmath4 140 ) only . the results are shown in table [ tab : z0p006 ] : in column 6 the results for the heavy @xmath1-seed only case and in column 7 the light @xmath1-seed only case are given . not surprisingly , the heavy @xmath1-nuclei are most important for the heavy @xmath0-nuclei , in particular for @xmath76os ( 100% of contribution ) , @xmath71dy ( 89% ) , @xmath77ce ( about 84% and 75% , respectively ) , @xmath74pt ( about 73% ) , and @xmath78dy ( about 70% ) , but there are also significant contributions of about 40 - 60% are to the other heavy @xmath0-isotopes in the region from @xmath72sm up to @xmath55hg . we also find @xmath3830 - 40% of the light @xmath0-nuclei up to @xmath5ba to originate from heavy @xmath1-seeds , with the exception of @xmath6mo and @xmath7ba , which have a lower share of @xmath3815% . in contrast , light @xmath1-seeds are the main producers of the light @xmath0-isotopes , and they account for about 73% for @xmath11se and @xmath6mo ( where important contribution from other stellar sources or errors in reaction rates can be estimated ) , and finally @xmath7ba ( about 80% ) . all other light-@xmath0 isotopes are produced from light-@xmath1 seeds with typical sharesin the rage from 40 to 60% . the heaviest @xmath0-nuclei are almost unaffected by the light-@xmath1 seeds . in the fourth columns of table [ tab : z0p006 ] and table [ tab : snuc ] we show results of a second test for comparison , where we retain a metallicity of @xmath36 0.006 , but reduce the strenght of the @xmath2c - pocket by using the stx1.3 model . for this case we test the role of @xmath3pb seed for @xmath0-process nucleosynthesis , as we did in the above standard case ( see fifth column in tables [ tab : z0p006 ] , [ tab : snuc ] ) . comparing the two cases presented in table 1 ( column 3 and 5 ) , we notice that typically the case with stx2 @xmath2c - pocket and a @xmath3pb seed contributes by a factor of @xmath382 to the light-@xmath0 nuclei with respect to the case stx1.3 @xmath2c - pocket and @xmath3pb seed . for the abundance of @xmath72sm we obtain identical results for stx2 and stx1.3 cases . in contrast , however , we observe an increased production of the heavy-@xmath0 nuclei for the case stx2 and @xmath3pb seed by @xmath3810% . the only exception is for the heaviest @xmath0-nucleus @xmath55hg , where we find a highest abundance in the case stx1.3 @xmath3pb alone . the reason for this behavior is that at the same metallicity ( @xmath36 0.006 ) the @xmath2c - pocket stx1.3 is less efficient than the @xmath2c - pocket stx2 for the production of heavy-@xmath1 as well as for @xmath3pb . the main goal of this work is to provide predictions for galactic chemical evolution of @xmath0-nuclei . for this , we employ the galactic chemical evolution code presented by travaglio et al . 1999 , 2001 , 2004 . the model considers the galaxy as the evolution of three interconnected zones , halo , thick disk and thin disk . the matrix of the isotopes within the chemical evolution code was set to cover all the light nuclei up to the fe - group , and all the heavy nuclei along the @xmath1-process path up to @xmath28bi . for the present work we extended the matrix of the isotopes to account for the @xmath0-nuclei and we followed their evolution over time / metallicity until solar metallicity was reached . we included in the code the @xmath0-nuclei abundances obtained from our snia model at various metallicities as discussed in section 4 , and interpolate between them smoothly . in figure 7 we show the resulting @xmath0-process production factors taken at the epoch of solar system formation for nuclei in the atomic mass number range 70 @xmath79 210 . to be more clear we note that for these results we only include snia for the contribution to @xmath0-nuclei . our choice for the @xmath1-seeds was introduced in section 3 ( see figure 3 ) . as discussed in the previous section , it is clear that a few nuclei originally ascribed to the @xmath0-only group ( @xmath56 in , @xmath57sn , @xmath46la , @xmath24gd , and @xmath58ta ) are far below the average of the other @xmath0-nuclei . thus , if this model is correct , they should be ascribed to different astrophysical sources . as shown in figure 7 and detailed in table 3 , we find in our galactic chemical evolution calculations that all the other @xmath0-only isotopes are within a factor of about three produced at the solar system composition . the lightest @xmath0-nucleus @xmath12kr and the heaviest @xmath71dy and @xmath80w can be ascribed to the same production site when an additional uncertainty factor of two is included ( which seems reasonable given the large uncertainties in photodisintegration rates ) . therefore , for the first time , we are able to explain the synthesis of almost all @xmath0-isotopes production in one single scenario . the most striking problem we face is the very low relative abundance of the true @xmath0-only @xmath6mo with respect the the average of all other @xmath0-only nuclei , which is possibly related to the theoretical estimate of the neutron capture on unstable @xmath63mo . the effect of the nuclear uncertainties will be explored in a forthcoming paper . figure 7 and table 4 also show that neutron magic @xmath22zr and especially @xmath14zr receive an important contribution by the @xmath0-process ( about 20% and 40% , respectively ) . moreover , it is not excluded that the most proton - rich @xmath1-only isotopes for a given element may receive some contribution from the @xmath0-process . as illustrated in figure 7 , this is the case for @xmath20kr and @xmath21sr , with @xmath0-contribution of the order of 10% . however , before examining the problem in more detail , an analysis of the uncertainties in the involved reaction rates is necessary . out of the light @xmath0-nuclei the isotopes @xmath12kr , @xmath6mo , @xmath81cd , and @xmath82sn are by a factor of three or more ( for @xmath6mo ) less abundant compared to the solar value . among the heavy-@xmath0 nuclei @xmath71dy is also by more than a factor of three below the solar abundance . in contrast , we found @xmath80w higher by a factor of more than three with respect to solar . under the hypothesis that snia are responsible for 2/3 of solar @xmath19fe , and assuming that our ddt - a model represents the typical snia with a frequency of 70% ( li et al . 2011 ) , we conclude that they are responsible for at least 50% of all @xmath0-nuclei . with the same approach , travaglio et al . ( 2014 ) discussed the origin of the short - lived radionuclides @xmath25nb and @xmath26sm . they compared the value of the ratio between the abundances of these radionuclides and those of the corresponding stable reference isotopes @xmath25mo and @xmath72sm to what has been recently measured in meteorites ( rauscher et al . the conclusion is that snia can also play a key role in the production of @xmath25nb and @xmath26sm , but nuclear uncertainties have to be taken into account . in the framework of galactic chemical evolution of @xmath0-nuclei , the role of ccsn has also to be taken into account . rauscher et al . ( 2002 ) followed the @xmath33-process through the presupernova stages and the supernova explosion . as expected for the weak @xmath1-process component in massive stars , only @xmath1-nuclei in the mass range @xmath83 are produced and occurred in situ prior to the explosion phase . rauscher et al . ( 2002 ) presented results for 15 , 19 , 21 , 25 @xmath84 ccsne modeled in spherical symmetry and with initial solar metallicity . for the 15 , 21 , and 25 @xmath84 models , proton - rich heavy isotopes in the mass ranges @xmath85 and @xmath86 were produced in solar abundance ratios within about a factor of two relative to @xmath31o , the most abundant nucleus in the ejecta of ccsne . for mass numbers @xmath87 and @xmath88 the production of the @xmath0-isotopes is down by about a factor of @xmath89 . while the main @xmath33-process synthesizes @xmath0-nuclei through photodisintegration reactions during the sn shockfront passage , some of the models showed pre - explosive @xmath0-production due to a high entropy in the o / ne shell of the evolved star . most of this is wiped out again when the supernova shock sweeps through the layer . nevertheless , depending on the adopted convection model ( see also bazan & arnett 1994 ) , some light , strongly bound @xmath0-nuclei may survive from pre - explosive production . this behavior complicates predictions for the contribution of ccsne to the solar composition of @xmath0-nuclei . in addition , it has to be taken into account that the snia scenario discussed here may not explain all normal events . thus , alternative scenarios for @xmath0-process nucleosynthesis in binary systems should be explored , such as sneia from wd - wd mergers ( e.g. pakmor et al . 2010 , 2012 ) or double detonations in sub - chandrasekhar mass wds ( e.g. fink et al . we have presented results of detailed @xmath0-process nucleosynthesis calculations for two - dimensional models of delayed detonations in chandrasekhar - mass wds resulting from the single degenerate progenitor scenario . in these snia models , the nucleosynthesis was followed by the tracer - particles method . the initial @xmath1-seeds were assumed to be created during the mass accretion phase . since up to now no nucleosynthesis calculations of the accretion phase are available our hypothesis is based on the assumption that a small amount of protons are ingested at the top layers of the he intershell . following the work of trv11 on the nucleosynthesis for a solar metallicity snia model in the same scenario , we tested and discussed the consequences of different amounts of @xmath2c and different metallicites on the synthesis of @xmath0-nuclei . we demonstrated that the @xmath3pb @xmath1-seed alone plays an important role for @xmath0-nuclei production , due to photodisintegration chains starting from the heaviest nuclei and going down in mass number . we analyzed the dependence of all the @xmath0-isotopes on metallicity , and we identified the isotopes with a weak ( like @xmath25mo and @xmath46ba ) and a strong ( in particular the lightest @xmath0-isotopes , @xmath11se , @xmath90kr , and @xmath13sr ) dependence on @xmath40 . we discussed the still puzzling origin of @xmath6mo . clearly , nuclear uncertainties can not account for the factor of 10 deficiency in @xmath6mo abundance relative to other @xmath0-abundances . the @xmath6mo production was found to depend on the seeds in the mo isotopes as well as on the density at which the photodisintegration process occurs . this leaves room for possible variations in the hydrodynamic history of the mainly contributing explosive trajectories which could change the relative @xmath0-abundances . by means of a simple galactic chemical evolution code , including @xmath0-process contributions at different metallicities , we explored the snia contribution to the @xmath0-nuclei abundances in the solar system . we concluded that @xmath0-nuclei are mainly of primary - like origin , and that snia can contribute at least 50% to the solar abundance of all @xmath0-nuclei provided that they result from standard chandrasekhar - mass delayed - detonations forming in the single - degenerate progenitor channel . thus we identified a stellar source which , in principle , is able to produce light and heavy @xmath0-nuclei almost at the same level relative to @xmath19fe , including the much debated neutron magic @xmath25mo and @xmath91ru . the important contribution from @xmath0-process nucleosynthesis to the @xmath1-only nuclei @xmath20kr , @xmath21sr , and to the neutron magic @xmath22zr has also been elaborated . another relevant contribution is to the neutron rich @xmath14zr , due to neutron captures from the residual abundance of @xmath23ne during the explosive phase . with our galactic chemical evolution calculations , it was possible to predict a significant contribution from snia in the considered scenario to the extinct @xmath0-radionuclides @xmath25nb , @xmath26sm , and @xmath91tc in the early solar system as shown in travaglio et al . 2014 . alternative scenarios for @xmath0-process nucleosynthesis in binary systems have to be taken into account , such as sneia from wd - wd mergers , where we also expect some @xmath0-process production . a detailed analysis of different snia progenitors will be explored in a future work . this work has been supported by b2fh association . the numerical calculations have been also supported by regione lombardia and cilea consortium through a lisa initiative ( laboratory for interdisciplinary advanced simulation ) 2010 grant . c.t . thanks c. arlandini , p. dagna , and r. casalegno for technical support in the simulations . c.t thanks the university of washington ( institute for nuclear theory , seattle , wa , int program int-14 - 2b ) for the kind support in the last phase of this work . the work of fkr was supported by the deutsche forschungsgemeinschaft via the emmy noether program ( ro 3676/1 - 1 ) and the graduate school `` theoretical astrophysics and particle physics '' at the university of wrzburg ( grk 1147 ) additional funding was provided by the arches prize of the german ministry of education and research ( bmbf ) and the daad / go8 german - australian exchange program . tr is partially supported by the swiss nsf , the european research council ( grant ga 321263-fish ) . wh s work was supported by the transregional collaborative research centre trr33 `` the dark universe '' und the cluster of excellence `` origin and structure of the universe '' at munich technical university . arlandini , c. , kppeler , f. , wisshak , k. , gallino , r. , lugaro , m. , busso , m. , & straniero , o. 1999 , , 525 , 886 arnould , m. 1976 , , 46 , 117 arnould , m. , & goriely , s. 2006 , nucl . a , 777 , 157 audouze , j. , & truran , j.w . 1975 , , 202 , 204 bazan , g. & arnett , d. 1994 , , 433 , l41 bisterzo , s. , gallino , r. , straniero , o. , cristallo , s. , & kppeler , f. 2010 , , 404 , 1529 burbidge , e.m . , burbidge , g.r . , fowler , w.a . , & hoyle , f. 1957 , rev . phys . , 29 , 547 busso , m. , gallino , r. , & wasserburg , g. j. 1999 , annu . rev astrophys , 37 , 239 cameron , a.g.w . 1957 , , 62 , 9 clayton , d.d . 1988 , mnras , 234 , 1 dillmann , i. , heil , m. , kppeler , f. , plag , r. , rauscher , t. , & thielemann , f .- k . 2006 , aip conf . 819 , 123 ; kadonis ( karlsruhe database of nucleosynthesis in stars ) v0.2 , available at http://www.kadonis.org dillmann , i. , plag , r. , domingo - pardo , c. , heil , m. , kppeler , f. , rauscher , t. , & thielemann , f .- k . 2008a , proceedings `` international conference on nuclear data for science and technology 2007 '' , nice/ france dillmann , i. , rauscher , t. , heil , m. , kppeler , f. , rapp , w. , & thielemann , f .- k . 2008b , journal of physics g : nuclear and particle physics , 35 , 014029 dillmann , i. , domingo - pardo , c. , heil , m. , kppeler , f. , walter , s. , dababneh , s. , rauscher , t. , & thielemann , f .- k . 2010 , phys . c 81 , 015801 fink , m. , rpke , f.k . , hillebrandt , w. , seitenzahl , i.r . , sim , s.a . , & kromer , m. 2010 , , 514 , 53 gallino , r. , arlandini , c. , busso , m. , lugaro , m. , travaglio , c. , straniero , o. , chieffi , a. , & limongi , m. 1998 , , 497 , 388 goriely , s. , jos , j. , hernanz , m. , rayet , m. , & arnould , m. 2002 , , 383 , l27 goriely , s. , garca - senz , d. , bravo , e. , & jos , j. 2005 , , 444 , l1 hansen , c.j . , andersen , a.c . , & christlieb , n. 2014 , , 568 , 47 howard , w.m . , mathews , g.j . , takahashi , k. , & ward , r.a . 1986 , , 309 , 633 howard , w.m . , meyer , b.s . , & woosley , s.e . 1991 , , 373 , l5 howard , w.m . , & meyer , b.s . 1993 , in proceedings of the 2nd international symposium on nuclear astrophysics , held at karlsruhe , germany . ed . by f. kaeppeler and k. wisshak , ( bristol : iop publishing ) p.575 iben , i.jr . 1981 , , 243 , 987 iben , i.jr . , & tutukov , a.v . 1991 , , 370 , 615 kasen , d. , rpke , f.k . , & woosley , s.e . 2009 , nature , 460 , 869 kusakabe , m. , iwamoto , n. , & nomoto , k. 2011 , , 726 , 25 langanke , k. , & martnez - pinedo , g. 2000 , nucl . a , 673 , 481 li , w. , et al . 2011 , mnras , 412 , 1441 lodders , k. 2009 , , 591 , 1220 marganiec , j. , dillmann , i. , domingo pardo , c. , kppeler , f. , & walter , s. 2010 , phys . c 82 , 035806 nagataki , s. , hashimoto , n.a . , sato , k. , & yamada , s. 1997 , , 486 , 1026 nomoto , k. , thielemann , f .- k . , & yokoi , k. 1984 , , 286 , 644 pakmor , r. , kromer , m. , rpke , f.k . , sim , s.a ruiter , a.j . , & hillebrandt , w. 2010 , nature , 463 , 61 pakmor , r. , edelmann , p. , rpke , f.k . , hillebrandt , w. 2012 , , 424 , 2222 peterson , r.c . 2013 , , 768 , l13 rauscher , t. 2011 , int . e 20 , 1071 rauscher , t. 2012 , , 755 , l10 rauscher , t. 2013 , physical review letters , 111 , 061104 rauscher , t. 2014 , aip advances 4 , 041012 rauscher , t. , & thielemann , f .- k . 2000 , atomic data nucl . data tables , 75 , 1 rauscher , t. , heger , a. , hoffmann , r.d . , & woosley , s.e . 2002 , , 576 , 323 rauscher , t. , dauphas , n. , dillmann , i. , frhlich , c. , flp , z. , & gyrky , g. 2013 , rep . phys . , 76 , 66201 reinecke , m. , hillebrandt , w. , niemeyer , j.c . , klein , r. , & grbl , a. 1999 , , 347 , 724 reinecke , m. , hillebrandt , w. , & niemeyer , j. c. 2005 , , 386 , 936 rpke , f.k . 2005 , , 432 , 969 rpke , f.k . , & hillebrandt , w. 2005 , , 431 , 635 seitenzahl , i.r . , rpke , f.k . , fink , m. , & pakmor , r. 2010 , , 407 , 229 thielemann , f .- k . , nomoto , k. , & hashimoto , m. 1996 , , 460 , 408 travaglio , c. , galli , d. , gallino , r. , busso , m. , ferrini , f. , & straniero , o. 1999 , , 521 , 691 travaglio , c. , gallino , r. , busso , m. , & gratton , r. 2001 , , 549 , l346 travaglio , c. , gallino , r. , arnone , e. , cowan , j. , jordan , f. , & sneden , c. 2004 , , 601 , 864 travaglio , c. , hillebrandt , w. , reinecke , m. , & thielemann , f .- k . 2004b , , 425 , 1029 travaglio , c. , hillebrandt , w. , & reinecke , m. 2005 , , 443 , 1007 travaglio , c. , rpke , f.k . , gallino , r. , & hillebrandt , w. 2011 , , 739 , 93 travaglio , c. , gallino , r. , hillebrandt , w. , & rpke , f.k . 2012 , proceedings of science , 45 travaglio , c. , gallino , r. , rauscher , t. , dauphas , n. , rpke , f.k . , & hillebrandt , w. 2014 , , in press wanajo , s. , janka , h .- t . , & mller , b. 2011 , , 726 , l15 woosley , s.e . , & howard , w.m . 1990 , , 354 , l21 -seed abundances relative to solar for all @xmath2c - pockets cases and metallicities covered , used for our galactic chemical evolution calculation . colours and symbols are the same of figure 1 . see text for a detailed discussion . ] reaction flow for maximum @xmath6mo production ( trajectory @xmath6mo@xmath59 in fig . [ fig : flowtracers ] ) ; size and color of the arrows relate to the magnitude of the time - integrated flux on a logarithmic scale . only flows down to a factor 0.001 of the maximum flow are shown . ] reaction flow for maximum @xmath6mo production beyond the dropping edge ( trajectory @xmath6mo@xmath60 in fig . [ fig : flowtracers ] ) ; size and color of the arrows relate to the magnitude of the time - integrated flux on a logarithmic scale . only flows down to a factor 0.001 of the maximum flow are shown . ] reaction flow for a @xmath6mo production minimum ( trajectory @xmath6mo@xmath64 in fig . [ fig : flowtracers ] ) ; size and color of the arrows relate to the magnitude of the time - integrated flux on a logarithmic scale . only flows down to a factor 0.001 of the maximum flow are shown . ] @xmath11se & 5.9766d-08 & 32 & 7.7692d-08 & 22 & 37 & 73 + @xmath12kr & 2.2814d-08 & 14 & 2.6779d-08 & 5 & 22 & 60 + @xmath13sr & 1.6684d-07 & 15 & 1.7652d-07 & 5 & 24 & 63 + @xmath25mo & 3.3110d-07 & 20 & 2.5454d-07 & 10 & 33 & 46 + @xmath6mo & 1.5926d-08 & 10 & 1.5816d-08 & 4 & 16 & 74 + @xmath14ru & 1.5192d-07 & 22 & 1.1276d-07 & 11 & 35 & 44 + @xmath18ru & 5.6763d-08 & 20 & 4.4197d-08 & 10 & 33 & 49 + @xmath92pd & 1.9007d-07 & 21 & 1.4257d-07 & 11 & 35 & 45 + @xmath93cd & 3.4325d-07 & 22 & 2.5156d-07 & 11 & 37 & 43 + @xmath81cd & 7.3755d-08 & 19 & 5.7087d-08 & 10 & 33 & 49 + @xmath94 in & 2.3632d-09 & 14 & 1.9788d-09 & 6 & 24 & 59 + @xmath95sn & 4.4177d-07 & 21 & 3.2332d-07 & 11 & 36 & 43 + @xmath82sn & 1.2393d-07 & 17 & 9.9142d-08 & 8 & 30 & 54 + @xmath96sn & 1.3516d-10 & 8 & 1.2630d-10 & 3 & 14 & 75 + @xmath97te & 3.0791d-08 & 15 & 2.5063d-08 & 7 & 28 & 56 + @xmath98xe & 1.1681d-07 & 18 & 8.7859d-08 & 9 & 36 & 45 + @xmath99xe & 1.6423d-07 & 20 & 1.1991d-07 & 10 & 40 & 40 + @xmath7ba & 9.3644d-08 & 6 & 8.6408d-08 & 2 & 14 & 79 + @xmath5ba & 8.2042d-08 & 12 & 6.7356d-08 & 6 & 32 & 56 + @xmath100la & 1.5568d-10 & - & 1.5268d-10 & - & - & 99 + @xmath54ce & 2.9490d-08 & 16 & 2.0296d-08 & 9 & 84 & - + @xmath46ce & 6.1838d-08 & 25 & 3.7265d-08 & 16 & 75 & - + @xmath72sm & 4.7213d-07 & 44 & 1.9701d-07 & 40 & 54 & - + @xmath101gd & 4.2205d-10 & 6 & 2.8463d-10 & 3 & 94 & - + @xmath78dy & 4.5886d-09 & 29 & 2.3545d-09 & 22 & 71 & - + @xmath71dy & 2.4326d-09 & 11 & 1.5200d-09 & 7 & 89 & - + @xmath73er & 1.0113d-08 & 39 & 4.5391d-09 & 33 & 61 & - + @xmath102er & 3.2957d-08 & 40 & 1.4267d-08 & 35 & 59 & - + @xmath103yb & 3.2922d-08 & 40 & 1.4330d-08 & 34 & 60 & - + @xmath104hf & 2.3801d-08 & 37 & 1.0562d-08 & 31 & 63 & - + @xmath105ta & 1.0371d-11 & - & 6.4666d-12 & - & 100 & - + @xmath80w & 3.9411d-08 & 40 & 1.6578d-08 & 36 & 60 & - + @xmath76os & 3.8775d-09 & - & 2.2520d-09 & - & 100 & - + @xmath74pt & 5.2488d-09 & 27 & 2.3916d-09 & 23 & 73 & - + @xmath55hg & 9.9030d-08 & 58 & 3.5696d-08 & 61 & 41 & - + @xmath20kr & 1.2508d-07 & 12 & 1.5196d-07 & 4 & 19 & 58 + @xmath21sr & 4.2708d-07 & 13 & 4.8676d-07 & 4 & 20 & 67 + @xmath22zr & 3.2537d-06 & 17 & 2.8889d-06 & 7 & 28 & 55 + @xmath14zr & 1.6979d-07 & - & 2.2666d-07 & - & - & 100 + @xmath11se & 4.186d-10 & 0.41 + @xmath12kr & 9.968d-11 & 0.23 + @xmath13sr & 5.760d-10 & 1.93 + @xmath25mo & 3.643d-10 & 0.37 + @xmath6mo & 3.663d-11 & 0.06 + @xmath14ru & 1.431d-10 & 0.55 + @xmath18ru & 6.294d-11 & 0.71 + @xmath92pd & 7.251d-11 & 1.84 + @xmath93cd & 1.173d-10 & 2.02 + @xmath81cd & 3.067d-11 & 0.74 + @xmath94 in & 1.232d-12 & 0.05 + @xmath95sn & 1.451d-10 & 1.35 + @xmath82sn & 5.177d-11 & 0.69 + @xmath96sn & 8.794d-14 & 0.002 + @xmath97te & 1.254d-11 & 0.76 + @xmath98xe & 3.350d-11 & 1.41 + @xmath99xe & 4.281d-11 & 2.07 + @xmath7ba & 3.186d-11 & 1.83 + @xmath5ba & 2.364d-11 & 1.40 + @xmath100la & 5.825d-14 & 0.04 + @xmath54ce & 5.816d-12 & 0.72 + @xmath46ce & 1.045d-11 & 0.95 + @xmath72sm & 5.238d-11 & 1.66 + @xmath101gd & 8.258d-14 & 0.03 + @xmath78dy & 6.448d-13 & 0.65 + @xmath71dy & 4.439d-13 & 0.26 + @xmath73er & 1.231d-12 & 0.78 + @xmath102er & 3.792d-12 & 0.20 + @xmath103yb & 3.792d-12 & 2.75 + @xmath104hf & 2.802d-12 & 1.96 + @xmath105ta & 2.261d-15 & 0.18 + @xmath80w & 4.416d-12 & 4.48 + @xmath76os & 7.157d-13 & 1.42 + @xmath74pt & 6.640d-13 & 0.64 + @xmath55hg & 8.884d-12 & 1.65 + |
primes in the two complete associative normed division algebras and have affinities with structures seen in the standard model of particle physics . on the integers in the two algebras , there are two equivalence relations : a strong one , related to a and symmetry allowing to permute and switch signs of the coordinates of the integers , as well as a weak relation with origins from units in the algebra . weak equivalence classes within the strong equivalence classes of odd primes in case relate to leptons , the inert ones being neutrino like , and the split ones resembling electron - positron pairs . in the case , for odd primes , the equivalence classes come in groups of two or three , leading to a caricature of hadrons featuring either mesons built by a quark pair or then baryons obtained by quark triplets . we can now list for every rational prime all these particles and attach fractional charges to its constituents . | 1608.07175 | [ [ section ] ] when experimenting with primes in division algebras , an affinity of primes in @xmath0 and @xmath1 with the structure in the standard model of particle physics emerged : primes in @xmath0 resemble leptons , while equivalence classes of primes in @xmath1 have hadron like features @xcite . in @xmath0 , where besides the ramified gaussian primes of norm @xmath6 like @xmath7 , two other type of primes exist , there is a lepton structure : in @xmath0 we see cartoon versions of neutrini as well as electron - positron pairs . in @xmath1 , where besides the ramified primes above @xmath8 , only one type of primes exists , modulo @xmath9 gauge transformations , the structure of @xmath10-equivalence classes appear to be hadrons " : there are either mesons or baryons , where the individual elements of the equivalence classes resemble quarks . [ [ section-1 ] ] the algebras @xmath11 are naturally distinguished as these two are the only complete associative normed division algebras . not only the unit spheres @xmath4 of in @xmath11 but also the lie group @xmath3 acts on spheres in @xmath1 by diffeomorphisms rotating three complex planes hinged together at the real axes . this non - linear action allows to implement more symmetries or extend any quantum dynamics from @xmath0 to @xmath1 valued fields . these symmetries lead to finite norm preserving group actions on quaternion integers and can be implemented without mentioning lie groups : the weak equivalence is obtained by using multiplication by units in @xmath1 ; the strong symmetries on integers @xmath12 in @xmath1 is generated by the 24 permutations or sign changes . obviously this can be realized using both rotations or reflections in @xmath2 or then by permutations and sign changes in the space components @xmath13 of the quaternion . the upshot is that every integer has a strong equivalence class @xmath12 with non - negative entries where @xmath14 . if we look at weak equivalence classes in such integer classes , the result is that each class has either one , two or three elements , where the case with one element only happens if @xmath15 . [ [ section-2 ] ] the fact that the major gauge groups of the standard model appear as symmetries in @xmath1 suggests to look at * quaternion quantum mechanics * , where waves are @xmath1-valued . such a physics could be realized by looking at quaternion valued wave equations developed by fueter @xcite who also noticed that the dirac equation in quaternion analysis plays the role of the cauchy - riemann differential equations . alternatively , one can extend linear or nonlinear wave evolutions from @xmath0-valued fields in sub planes to quaternion - valued fields : just evolve simultaneously the classical waves on planes spanned by @xmath16 and one of the spacial units like @xmath17 in @xmath1 . if @xmath18 is any self - adjoint hamiltonian operator , then instead of quantum dynamics @xmath19 for a wave @xmath20 , one can consider two other waves @xmath21 and evolve all simultaneously using @xmath22 . this assures that @xmath23 stays invariant . the three evolutions can be merged together to a quaternion with this norm @xmath24 by putting @xmath25 and @xmath26 with sign @xmath27 chosen so that @xmath28 has positive jacobean . this assures that @xmath29 is time invariant . the second equivalence relation , the @xmath10 symmetry comes from the cayley - dickson picture expressing @xmath12 as a pair of two complex numbers @xmath30 for which @xmath31 with the complex matrix @xmath32 built by pauli matrices . these weak symmetries on @xmath33 can be implemented as determinant preserving unitary @xmath34 matrices and so by elements in @xmath10 . obviously , the weak and strong equivalence classes are different . looking at both together in @xmath1 gives the hadron structure on primes . this amusing fact is is pure finite combinatorics . [ [ section-3 ] ] one can ask now why number theory should matter in quantum dynamics , as a random quaternion is almost surely never an integer . one can look however at the dynamics on a geometry " , a space @xmath35 with exterior derivative and measure @xmath36 , where one classically evolves waves in @xmath37 , moving according to some dynamics in the hilbert space @xmath38 with norm @xmath39 . the quaternion quantum evolution preserves this norm but at individual points @xmath40 , the wave amplitude changes . gauge symmetry now can render integers relevant : if a wave amplitude reaches integer arithmetic @xmath1 norm @xmath41 at some point , the wave value @xmath42 can be gauged within the gauge groups to become @xmath1-integer valued at this point . its prime factorization structure features now particles at this point . the switch of factorizations could now be seen as particle processes in which fermions = odd primes exchange vector gauge bosons = units or ramified primes = neutral bosons . unit migration is an exchange of bosons and recombination is particle pair creation or annihilation , mending in some way also the particle - wave duality conundrum . [ [ section-4 ] ] this picture most certainly is just an allegory or caricature even in the realm of particle phenomenology alone when disregarding dynamics : the reason is that even the simplest processes like * beta decay of a neutron * @xmath43 to a proton , electron , electron neutrino , or a * pion decay * @xmath44 , where a meson decays into two leptons can not be explained by unit migration , recombination or meta commutation in @xmath1 alone . also , the way , how charge is defined below , no baryons do exist for which all three quarks have charge @xmath45 . this violates the existence of the @xmath46 baryon . we believe however that the definition of charge given here can be modified to incorporate this . we have just chosen an algorithm ( using lipschitz or hurwitz primes ) which gives us a deterministic charge . there is no other reason why @xmath47 charge triplets are excluded . + [ [ section-5 ] ] the picture drawn here is probably of little value for physics as the later by definition requires to be able to do * quantitative predictions * or * quantitative verifications * of experimentally observable processes . and the picture drawn here does neither . we feel however that the story is of mathematical interest and that it motivates to look more closely at higher arithmetic in @xmath1 as well at the mathematical structure of standard model which appears to suffer from a * lack of inevitability*. while the standard model is one of the most successful physical theories with excellent match between experiment and model , the structure of the involved lagrangians is complicated even if one looks at it from a non - commutative geometry point of view . [ [ section-6 ] ] affinities between mathematical structures and physical phenomena are sometimes useful , sometimes just amusing . here is an example of the more amusing type @xcite : when looking at the rubik group , a finite group , there are permutations available when disassembling and recombining the actual cube which resemble quarks : turn a corner cube by 120 degrees for example . rotating one corner and other corner in the opposite direction gives a quark - anti quark meson and this permutation is realizable . turning three corners by 120 degree can physically be realized and resembles a baryon . also this picture is useless for physics but it is of some value as it helps to learn more about the structure of that particular finite group . [ [ section-7 ] ] the standard model does not answer why the number of generations of hadrons or leptons is limited to three or why the gauge groups are not unified to a larger group like @xmath48 predicted by some grand unified theories . we have seen that @xmath1 naturally features the gauge groups of the standard model , even so the @xmath3 symmetry acts only by diffeomorphisms . the three generations can emerge naturally in a quaternion - valued quantum mechanics as we have to evolve waves in three different planes . since the time scales in the different planes are in general different ( as there is no reason why they should agree ) , a particle involving in a faster plane appears to be lighter . better than * imposing * a wave evolution is to let the system just evolve freely in its isospectral set . wave dynamics then emerges naturally @xcite in the form of isospectral lax deformations of exterior derivatives leading to three generations of geometries . as both leptons and hadrons move in at least one plane and more likely in 2 or 3 planes at the same time , the probability of having a light particles is larger . the heavier particles like the top quark are more rare . the neutrini in the intersection of these planes participate to any of three dynamics producing thought associations with the observed phenomenon of neutrino oscillations . [ [ section-8 ] ] number theory in @xmath1 has started with hurwitz @xcite who built the setup and established some factorization features . the fundamental theorem of arithmetic in @xmath1 was completed @xcite . the fact that quaternions were an outcast in part of the 20th century can historically be traced to the success of vector calculus , especially as formulated by gibbs and wilson @xcite , a book so successful that its content not only structured practically all modern calculus textbooks but also removed quaternions from the curricula . quaternion calculus was still cultivated , like by the swiss number theorist rudolf fueter who found a cauchy integral theorem for quaternions ( see @xcite ) . there is certainly much still to be explored . only recently , the permutation structure of the meta - commutation in the prime factorization has been studied for the first time @xcite . [ [ section-9 ] ] by the * hurwitz theorem * @xcite , the algebras @xmath49 form a complete list of all * normed division algebras*. related to the hurwitz theorem is the * frobenius theorem * @xcite which tells that the only associative real division algebras are @xmath50 and the * mazur theorem * which assures that the only * commutative banach division algebras * are @xmath51 and @xmath0 . the associative , algebraically complete division algebras are @xmath0 and @xmath1 . * algebraic completeness * in @xmath1 means that every polynomial @xmath52 has a solution @xmath40 @xcite . completeness in general fails : @xmath53 has no solution ( an example from @xcite ) or have unexpected solution sets : @xmath54 has an entire @xmath6-sphere as solutions and there are polynomials like @xmath55 for which @xmath1 is the solution set ( see @xcite ) . by choosing a basis , @xmath1 contains three linearly independent complex sub - rings generated by @xmath16 and a choice of spacial vectors @xmath56 satisfying @xmath57 . [ [ section-10 ] ] it is natural to ask for a relation between the frobenius and hurwitz theorem . note that the frobenius theorem does not assume normed `` but makes an associativity assumption . but there seems no obvious link between the frobenius and hurwitz statements , as the class of ' ' normed division algebras `` and ' ' division algebras `` is different and associativity is quite a strong assumption . the question asks whether there could be dependencies between the proofs of the theorems . the answer is probably ' ' no `` : going from frobenius to hurwitz requires to get rid of the attributes ' ' associative , finite dimensional `` and adding ' ' normed " instead . historically , hurwitz paper was only published posthumously and does not cite frobenius even so hurwitz ( 1859 - 1919 ) and frobenius ( 1849 - 1917 ) were contemporaries and both worked in zurich . frobenius was at eth zuerich between 1875 and 1892 . when frobenius took over kronecker s chair in berlin , hurwitz in turn took over his chair at eth from 1892 until his death in 1919 . [ [ section-11 ] ] the lattice @xmath58 of integers = * integer quaternions * is a maximal order in @xmath1 , where the notion of * order * in non - commutative algebras has been defined by emmy noether : @xcite : an order is just a sub - algebra which is a lattice ; it is * maximal * if it is not contained in a larger order . the lipschitz integers in @xmath1 form an order but only when adjoining the hurwitz quaternions , one gets a maximal order . the lattice is also known as the densest lattice packing of @xmath59 by unit spheres , the integer quaternions are partitioned into two different type of integers , the * lipschitz quaternions * and the * hurwitz quaternions*. the lipschitz quaternions consist of vectors @xmath60 with rational integers @xmath12 . the hurwitz quaternions are of the form @xmath61 with rational integers @xmath62 . the set @xmath63 of * quaternions primes * are integer quaternions for which the * arithmetic norm * @xmath64 is prime . for @xmath65 , the * conjugate prime * is defined as @xmath66 . the symmetric group @xmath67 of permutations on @xmath68 acts on integer quaternions and leaves primes invariant . also the group @xmath69 generated by elements in @xmath67 and the conjugation involution acts on @xmath63 . the group @xmath69 can not be realized within @xmath3 but would need the larger group @xmath9 and wipe out the charge information . having determinant @xmath70 , the conjugation involution @xmath71 is not realized in @xmath3 . [ [ section-12 ] ] any positive definite quadratic form @xmath72 in @xmath73 defines a lattice . in such a lattice , the * form - primes * are the integers for which @xmath74 is a rational prime . as hurwitz showed in @xcite , the primes in @xmath1 are in one - to - one correspondence with the form - primes in @xmath75 . the quaternions are also unique among division algebras in that rational and arithmetic primes are the same . for gaussian primes @xmath0 for example , real primes @xmath76 , primes of norm @xmath6 and primes of norm @xmath5 which is a prime of the form @xmath77 are distinguished . also the quadratic form @xmath78 on eisenstein integers generates only most of the eisenstein primes , the split ones . in that case , @xmath79 are the ramified primes . the inert primes , the primes on the integer axes @xmath80 or @xmath81 are not quadratic form primes among the eisenstein integers , because there , the square root of @xmath72 is prime and not the norm @xmath72 . [ [ section-13 ] ] a multiplication on @xmath1 is defined by writing @xmath82 with * space units * @xmath17 satisfying the hamilton relations @xmath83 and extending this linearly . the spacial part is generated by the units @xmath17 . together with @xmath16 , they form a basis . while the quadratic equation @xmath84 has the only solution @xmath85 in @xmath1 , the quadratic equation @xmath54 in @xmath1 already has already infinitely many solutions @xmath56 with @xmath57 . the rational unit @xmath16 paired with one of the space units @xmath56 forms a two - dimensional plane , which is also a complete sub - algebra isomorphic to @xmath0 . except for @xmath1 itself , these complex planes are the only complete normed division sub - algebras of @xmath1 . if we want the sub - algebra to have their integers as part of the integers in @xmath1 , there are only three sub planes left and these are the ones generated by @xmath16 and a choice of @xmath86 . [ [ section-14 ] ] if geometry " is a geometric space with an exterior derivative @xmath87 leading to laplacian @xmath88 , we have wave dynamics or isospectral lax deformations of the derivative @xmath87 . examples of geometries are compact riemannian manifold or a finite simple graphs . if we look at the wave equation and want to be able to reach from a point @xmath40 in the geometry to a point @xmath89 , we need to use a complex variable . the wave equation @xmath90 with laplacian @xmath91 has the explicit solution @xmath92 , where @xmath93 is the inverse of @xmath94 on the orthogonal complement of the kernel and @xmath87 is the exterior derivative . this can be written as a * complex schrdinger wave * @xmath95 with complex wave @xmath96 . since complex sub - algebras exist in @xmath1 , we also look at gaussian primes . real sub - algebras like @xmath51 generated by @xmath16 are not complete and some of their primes decay like @xmath97 which decays into @xmath98 . but primes which are initially in one of the three arithmetic complex wave subspaces , are also primes in @xmath1 . [ [ section-15 ] ] the linear map @xmath99 , mapping @xmath100 to @xmath101 satisfies @xmath102 . conversely , given @xmath103 , define @xmath104 by @xmath105 , leading to @xmath106 which satisfies @xmath107 . it satisfies @xmath108 . given @xmath109 , define @xmath110 . by construction , @xmath111 preserves the unit sphere in @xmath1 but it is neither linear nor invertible . any quaternion with prime amplitude @xmath5 can now be gauged to be an integer . the unitary group @xmath9 allows to realize any permutation of @xmath112 of the four coordinates of a quaternion as well as to perform sign changes like @xmath113 . in other words , modulo gauge transformations given by the @xmath9 action , every quaternion is equivalent to a quaternion @xmath12 with @xmath114 . the observation which led us to write this down was that the @xmath10 equivalence classes of these classes come now in groups of @xmath6 or @xmath115 for odd primes which looks like the meson and baryon structure for hadrons . [ [ section-16 ] ] we have seen that quaternions play a distinguished role among all algebraic structures : @xmath0 and @xmath1 are the only algebraically complete associative division algebras , and as @xmath1 is the maximal one containing @xmath0 as proper complete sub algebras : if @xmath35 is a sub - algebra of @xmath1 containing @xmath116 , then it must contain the real line @xmath51 and the square root of @xmath70 . in @xmath1 , any solution @xmath117 of @xmath118 x is of the form @xmath56 with @xmath57 . if it contains besides @xmath16 and a solution to @xmath118 a third linearly independent element , then by frobenius , @xmath119 , as there is no other algebra between @xmath0 and @xmath1 . [ [ section-17 ] ] the algebra @xmath1 has a natural representation in the algebra of complex @xmath34 matrices : every @xmath120 defines a matrix @xmath121 , where @xmath122 and @xmath123 are the * pauli matrices*. now @xmath124 and @xmath125 . the * cauchy - binet determinant formula * @xmath126 for matrices @xmath127 $ ] gives the identity @xmath128 . the unit sphere in @xmath1 is @xmath10 which happens to be a lie group . the normed division algebras @xmath0 and @xmath1 can be distinguished by the fact that their unit spheres are continuous lie groups as this property eliminates both @xmath51 and @xmath129 . the only lie groups which are spheres are @xmath130 for @xmath131 and the only continuous ones are @xmath130 with @xmath132 . ( a short proof of that statement was sketched by @xcite : if the lie group @xmath133 is abelian then the lie algebra must be @xmath73 so that it must be a universal cover of @xmath133 . this only works if @xmath133 is not simply connected , implying @xmath134 as this is the only not simply connected but connected sphere . if @xmath133 is non - abelian , define the 3-form @xmath135,z)$ ] which by the non - abelian assumption is not the zero form . one can show that it is left and right invariant and so closed but not exact , leading to a nontrivial cohomology @xmath136 . this forces @xmath137 as this is the only euclidean sphere with non - vanishing @xmath138 . ) [ [ section-18 ] ] there are 24 units in @xmath1 ; 16 of them are * hurwitz units * @xmath139 with @xmath140 and 8 of them are * lipschitz units * @xmath141 . the lipschitz units are permutations of @xmath142 , the hurwitz integers are permutations of @xmath143 . the finite set @xmath144 of units forms the * binary tetrahedral group * which can also be written as the special linear group @xmath145 over @xmath146 . as quaternions can be represented as @xmath34 matrices , @xmath144 is also a discrete finite subgroup of the unitary group @xmath10 . [ [ section-19 ] ] the group @xmath10 is also known as the compact symplectic group @xmath147 or the spin group @xmath148 . the finite group @xmath144 of units in @xmath1 can be identified as the semi - direct product of the * quaternion sub group * @xmath149 built by the @xmath150 lipschitz units and the cyclic group @xmath151 , generated by conjugation @xmath152 . it is finitely presented satisfying @xmath153 for the generators @xmath154 and @xmath155 . all cyclic subgroups have order @xmath156 or @xmath157 . the element @xmath158 for example generates a cyclic group of order @xmath157 . the only units for which all entries are non - negative are @xmath159 . the fact that each of the 16 regions like @xmath160 contains exactly one hurwitz unit will play a role combinatorially when looking at weak equivalence equations . [ [ section-20 ] ] in @xmath0 , the integers have only one type , but primes appear in three flavors : * inert * , * split * or * ramified * depending on how they factor in the field extension : the real ones with prime @xmath161 of the form @xmath162 , the ones with prime @xmath163 of the form @xmath164 and then the primes of norm @xmath165 . there are 4 units @xmath166 . the group @xmath144 of units is the cyclic group @xmath167 . there is an other group @xmath69 , the group of all permutations of the coordinates . in @xmath0 , it is the dihedral group @xmath168 . every equivalence class of @xmath69 can be represented as @xmath169 with @xmath170 . all @xmath69 equivalence classes have only one element . the @xmath144 equivalence classes have @xmath16 or @xmath6 elements . the units and the primes with norm @xmath6 have only one element , also the real primes have only one equivalence class . the other primes have @xmath6 elements in each equivalence class . an example is @xmath171 and @xmath172 . [ [ section-21 ] ] hurwitz showed that in @xmath1 and @xmath173 , there are exactly @xmath174 primes @xmath12 with @xmath175 . there are therefore @xmath176 classes of primes above each rational prime @xmath5 . for example , for @xmath79 we have the @xmath177 primes @xmath178 . by multiplying with units , we see that they are all not equivalent . [ [ section-22 ] ] while in @xmath0 , all factorizations are equivalent , the structure is more interesting in the non - commutative quaternion case , where conway and smith finalized in @xcite the factorization structure of @xmath58 : it consists of unit migration , recombination and meta commutation . recombination means to replace a pair @xmath179 in the prime factorization with an equivalent pair @xmath180 and move it to an other place . lets look next a unit migration : [ [ section-23 ] ] given an integer quaternion @xmath181 , the set of left conjugacy classes @xmath182 is different from the set of right equivalence classes @xmath183 . this means that if @xmath181 is an integer and @xmath184 is a unit , then @xmath185 can not be written as @xmath186 for a unit @xmath117 in general . what happens is that that the entries get scrambled around . this can be realized using permutations , respectively using elements of the group action of @xmath9 on @xmath1 . for any unit @xmath184 in @xmath1 , there exists a permutation @xmath187 of the four coordinates @xmath60 and an other unit @xmath117 such that for all @xmath188 the identity @xmath189 holds . since there are finitely many units , this can be checked case by case @xmath190 . here is the explicit list : + [ cols="<,<,<",options="header " , ] [ [ section-24 ] ] there are prime pairs which can be disappear and reappear anywhere as they temporarily become rational integers : a pair @xmath179 in the prime factorization can be moved anywhere and replaced with an other pair @xmath180 of the same norm . a rational number commutes with any number and @xmath179 is a rational prime if @xmath181 is a prime . [ [ section-25 ] ] as hurwitz already noticed , primes of norm @xmath6 play a special role . as in the complex @xmath0 case , all primes over the prime @xmath8 are equivalent . they especially do not have any charge . commuting with them does not change much of the quaternion . it only produces a rotation by 90 degrees in a two dimensional spacial sub plane ( 3 cases ) or a rotation by 180 degree reflection . the situation is otherwise similar to the unit migration : for any prime @xmath5 of norm @xmath6 , there is an other prime @xmath191 of norm @xmath6 and a rotation @xmath187 in a two - dimensional sub plane such that @xmath192 for all @xmath188 . here are the 6 cases + @xmath193 + @xmath194 + @xmath195 + @xmath196 + @xmath197 + @xmath198 + [ [ section-26 ] ] the fundamental theorem of arithmetic in quaternions requires to understand what happens if two odd primes are switched . this process is called meta commutation : for any two primes @xmath199 above @xmath200 there is a permutation @xmath187 of the @xmath176 prime classes above @xmath5 such that @xmath201 modulo @xmath144 . the integers of norm @xmath16 and @xmath6 have already been dealt with . the permutation structure has been studied in @xcite . if we use two @xmath144-equivalent primes @xmath202 in the factorization , then the moebius function @xmath203 is zero . + [ [ section-27 ] ] here is * fundamental theorem of arithmetic * for the complete associative maximal normed division algebra : up to unit migration , recombination and meta - commutation , the factorization of a quaternion integer into primes is unique . if the order of the primes is given , we have uniqueness up to unit migration . meta commutation takes care of the rest . see @xcite . [ [ section-28 ] ] how do we represent the @xmath9 equivalence classes of a prime modulo the group @xmath144 ? every prime @xmath149 in @xmath1 is @xmath144-equivalent to one or two or three @xmath9 representatives in the positive quadrant @xmath204 . let @xmath149 be the region in @xmath1 , where all coordinates @xmath62 are nonnegative . given a prime @xmath205 , we can write down all the 24 left conjugates @xmath185 and 24 right conjugates @xmath186 , where @xmath206 run over the 24 units . multiplying with lipschitz units different from @xmath85 gives @xmath207 and @xmath208 which are not in @xmath149 unless two coordinates are zero . but then the conjugate is in the same @xmath9 representative . for the @xmath209 hurwitz units @xmath184 , the list of @xmath210 is + @xmath211 + @xmath212 + @xmath213 + @xmath214 + @xmath215 + @xmath216 + @xmath217 + @xmath218 + @xmath219 + @xmath220 + @xmath221 + @xmath222 + @xmath223 + @xmath224 + @xmath225 + @xmath226 . + in order that we are in @xmath149 , we need to satisfy in each case 4 inequalities simultaneously . its possible for example in the last entry if @xmath227 or in the entry @xmath220 if @xmath228 . the list of @xmath229 is + @xmath230 + @xmath231 + @xmath232 + @xmath233 + @xmath234 + @xmath235 + @xmath236 + @xmath237 + @xmath238 + @xmath239 + @xmath240 + @xmath241 + @xmath242 + @xmath243 + @xmath244 + @xmath245 . there are again conditions for which this works like in the last case , if @xmath246 dominates . depending on whether we are in a class already covered in @xmath247 , we have now equivalence classes with 2 entries @xmath248 or equivalence classes with three entries @xmath249 . there are no other cases as from the list of 256 entries @xmath250 ( a list we omit ) , we are either equivalent to @xmath181 or not in @xmath149 . in the meson case , one of the entries is always zero . if two entries are zero like @xmath251 , the partner must be conjugated by a hurwitz unit and is then of the form @xmath252 . [ [ section-29 ] ] on @xmath80 where the set of rational primes @xmath253 are the fundamental building blocks , there is a natural multiplicative charge , the mbius function @xmath254 . for complex primes , we can define a charge by looking at the sign of the imaginary part when the argument is gauged to be in @xmath255 . the prime @xmath256 for example has charge @xmath16 while the prime @xmath257 has charge @xmath70 . the prime @xmath7 has zero charge as @xmath7 is conjugated to @xmath258 . for integers in @xmath0 it an be enhanced by defining @xmath259 as @xmath260 , where @xmath261 is the arithmetic norm and @xmath262 is the number of negatively charged primes in the factorization . also this is a multiplicative function . [ [ section-30 ] ] we can also define a charge function on the set @xmath263 of non - zero quaternion integers which are square free . we only take square free integers because quarks are fermions . here is a preliminary definition : lets assume that * charge * is a function @xmath264 satisfying : a ) @xmath265 for non - units , b ) @xmath266 , c ) @xmath267 if @xmath268 with @xmath269 , d ) primes which are @xmath69-equivalent but not @xmath67-equivalent have opposite charge , e ) if a lipschitz prime is @xmath144-equivalent to a hurwitz prime , the absolute charge of the lipschitz prime is not smaller than the hurwitz prime , f ) the sum of the charges of elements in a @xmath144-equivalence class is an integer , g ) @xmath270 is positive on the positive quadrant , and h ) @xmath270 is the smallest function of this kind . + [ [ section-31 ] ] a charge for units could be defined by @xmath271 if @xmath272 , but this needs independence of @xmath273 . a direct definition of charge for units does not work as @xmath274 and @xmath275 would imply @xmath276 . for example , the unit @xmath277 has the property @xmath278 and @xmath70 has charge @xmath279 . if the multiplicative nature would be extended to this , the unit would have charge @xmath279 . [ [ section-32 ] ] there is only one charge function : existence is shown by construction . because it takes the value @xmath16 for some equivalence classes , there can not be any smaller one . to get uniqueness , we can restrict to primes by property a ) . for the equivalence classes with one element , we have either units or prime @xmath6 . since conjugation leaves a prime @xmath6 invariant , this must have charge @xmath279 . lets look at the equivalence classes with one lipschitz and one hurwitz prime . if the two have opposite charge , then we get a @xmath280 or @xmath281 pair . if not , we get a @xmath282 pair . in an equivalence class with one lipschitz and two hurwitz primes , we have either the lipschitz one equivalent to one of the hurwitz , which gives @xmath283 or then the two hurwitz equivalent , which gives @xmath284 . [ [ section-33 ] ] now we an extend the charge function @xmath285 from primes to equivalence classes of primes . this can be done by adding up the charges of the elements of the equivalence class @xmath286 . this is motivated as follows . we look at a factorization into equivalence classes . then we sum over all possible paths of factorizations , we get for each hadron an integer charge , even so the charge of an individual quark is a fraction , a probability of hitting that particular quark . by definition , this is now an integer - valued function on equivalence classes of primes . + [ [ section-34 ] ] an other natural functional , additive with respect to the multiplication is the logarithm of the norm . we could abstractly define * mass * as a function @xmath287 satisfying : a ) @xmath288 , b ) @xmath289 , c ) @xmath290 if and only if @xmath181 is a unit , d ) @xmath291 takes integer values on integers , e ) @xmath292 is the smallest function of this kind . and show that there is only one mass function on @xmath263 , the function @xmath293 . whether mass " is a good name for such a functional will become only clear when looking at solutions of wave equations . [ [ section-35 ] ] we think of a square - free gaussian integer as a collection of * leptons * , where the individual gaussian prime equivalence classes are indecomposable * fermions*. a prime @xmath181 of type @xmath164 together with an opposite charged particle @xmath294 forms an * electron - positron * pair . we gauge the integers with units @xmath295 to be in the sector @xmath296 . the * charge * of a lepton @xmath181 is then defined as the sign of the argument of @xmath181 in the branch @xmath297 . we think of the logarithm of the norm @xmath74 or @xmath298 as * mass * , whatever is prime . a rational positive prime @xmath162 plays the role of a * neutrino*. it is lighter because its momentum @xmath298 is prime while for @xmath164 primes which are electrons or positrons , the energy @xmath74 is prime . a neutrino is * neutral * as it is located on the real axes . the largest known prime for example is a mersenne prime and so a neutrino . [ [ section-36 ] ] an integer @xmath299 is a * lepton configuration*. the fact that the gaussian primes form a * unique factorization domain * translates into the statement that any lepton configuration can be decomposed uniquely into such leptons as well as a bunch of neutral mystery particle @xmath6 which is its own anti - particle . the uniqueness holds only modulo * gauge transformations * which act here as multiplications by units . we will see that in the quaternion case , this fact is no more the case , because that is , where the hadrons will come in explaining why quarks form baryons and mesons . the electron - positron pair is * not bound * together : there is no unit which maps one into the other . factoring out the symmetry of units renders the factorization unique . the product @xmath300 for example is gauge equivalent to the product @xmath301 . let us now move from primes to rational integers and call a * rational integer * @xmath302 a * boson configuration * if it contains an even number of fermionic prime factors counted with multiplicity , otherwise it is a fermion . [ [ section-37 ] ] mathematically one could declare a natural number @xmath303 to be a fermion if its * jacobi symbol * @xmath304 is @xmath70 . otherwise , if @xmath305 , it is a boson . the gauss law of * quadratic reciprocity * tells now that two odd primes @xmath200 satisfy the * commutation relations * @xmath306 if at least one of them is a boson and that the * anti - commutation relation * @xmath307 hold exactly if both primes @xmath200 are fermions . in other words , if we look at the jacobi symbol as an operator @xmath308 , then bosons commute with everything else , but the sign changes , if we switch two fermions . [ [ section-38 ] ] the * two square theorem of fermat * telling that an integer @xmath309 can be represented as @xmath310 if and only if @xmath309 is a bosonic integer can be interpreted as the fact that a bosonic rational prime is actually composed of two leptons @xmath311 , where @xmath312 . the positron and electron are anti - particles of each other , but they are not equivalent , since there is no gauge from one to the other . if we factor out the * gauge symmetries * given by the * units * , then the factorization aka particle decomposition of the lepton set is unique . this is the * fundamental theorem of arithmetic * for gaussian integers . it can be proven from the rational case using the @xmath313 identification of @xmath314 with the set of rational primes . [ [ section-39 ] ] the * pauli exclusion principle * is encoded in the form of the * moebius function * @xmath315 which is equal to @xmath16 if a gaussian integer @xmath309 is the product of an even number of different gaussian primes , and @xmath70 if it is the product of an odd number of different gaussian primes and @xmath279 , if it contains two identical particles . again this * particle allegory * is already useful as a * mnemonic * to remember theorems like the two square theorem , or the quadratic reciprocity theorem : `` quadratic reciprocity means that only fermion primes anti - commute @xmath307 . fermat s two square theorem assures that bosonic rational primes @xmath316 are composed of two gaussian primes @xmath317 . the others are all real , light and neutral . '' [ [ section-40 ] ] hadrons are quaternion prime equivalence classes @xmath181 with norm @xmath74 different from @xmath6 . the prime @xmath6 is special also in quaternions : as hurwitz already showed one can despite non - commutativity place the factors @xmath6 outside : every integer quaternion @xmath181 is of the form @xmath318 where @xmath319 is an odd integer quaternion . there are two symmetry groups acting on hadrons . one is the group @xmath144 of units , the other is the group @xmath67 generated by coordinate permutations and conjugation . the group @xmath144 has 24 , the group @xmath67 has @xmath320 elements . the groups are no more contained in each other like in the complex case , where @xmath321 was a subgroup of @xmath322 . [ [ section-41 ] ] if we look at the @xmath144-equivalence classes first and then look at the orbits of @xmath67 , we see that some particles are fixed under @xmath67 or then that they come in pairs . we will interpret this as * particle anti - particle pairs*. we can however also look at the @xmath67 equivalence classes first and then look at the orbits of @xmath144 , then we see 1 or 2 or 3 particles combined . this is remarkable . lets repeat the statement : the @xmath3 equivalence classes of primes of positive charge are given by elements for which all coordinates are non - negative . if we call two such particles @xmath323 equivalent if there exists a unit in @xmath1 such that @xmath324 , then there are three type of equivalence classes : if @xmath165 , then there is one equivalence class . if @xmath181 is a unramified prime , then there are two cases : either there are two elements in each @xmath144 equivalence class or three elements in each @xmath144-equivalence class . so , it is important first to look at the strong equivalence classes " , then classify them according to weak equivalence " . if we would look at weak equivalence classes " first , then classify according to strong equivalence " , then we would get in the unramified case only two types . they tell the sign of the charge . [ [ section-42 ] ] a * hadron * is a quaternion prime equivalence class which contains more than one prime . as a matter of fact , each hadron to an odd prime @xmath5 consists either of two or three * quarks*. quarks can be either lipschitz or hurwitz primes . the * lagrange four square theorem * assures that there are no neutrini type particles among hadrons : particles for which three coordinates are zero . so , we can think of odd quaternion primes as * quarks * which form equivalence classes in the form of baryons or mesons . the conjugate quaternion is the * anti particle*. this is a probabilistic picture : take an integer @xmath325 . look at all possible prime factorizations of @xmath309 . we can look at such a factorization also modulo units meaning to look at equivalence classes . if we fix the order of the equivalence classes , then we see an ordered sequence of mesons ( equivalence classes with two primes ) , baryons ( equivalence classes with three primes ) and bosons related to @xmath6 ( equivalence classes with one prime ) . the latest primes can be moved anywhere without producing other changes as switching it with an other prime only produces a slight gauge transformation . switching the order of two other classes however can change the nature of the particles . [ [ section-43 ] ] as can be seen from the hurwitz factorization theorem which shows that in such a factorization , one can avoid have factors @xmath179 near each other . the allegory is that they would annihilate " into energy and re - emerge as particles . going from one factorization to an other is a rather complex interaction process changing the nature of some baryons involving gauge bosons . it does certainly not match with all particle processes as the number of primes in a factorization of @xmath181 is constant among all factorizations . this could just be the limit of particle phenomenology alone . the true nature of physics only comes with dynamics . [ [ section-44 ] ] baryons are fermions and mesons are bosons . like in the complex case , we have a mystery @xmath8 case , which has only one equivalence class . lets call it the * @xmath6 particle * even so we would like to associate it with something real . closest to mind comes higgs " which is neutral , light , its own anti - particle and can gives more mass to other particles . here it does so by multiplying with an other integer . [ [ section-45 ] ] the * group of units * contains @xmath150 particles of the form @xmath142 , @xmath326 , @xmath327,@xmath328 which modulo @xmath144 are all equivalent to the neutral @xmath329 , the @xmath330-boson . there are @xmath209 remaining units . modulo @xmath67 they are all equivalent to @xmath331 . this has a positive charge and is the * @xmath332 boson*. its conjugate is the * @xmath333 boson*. [ [ section-46 ] ] now lets look at a meson @xmath334 containing a lipschitz prime @xmath5 and hurwitz prime @xmath191 . since @xmath5 and @xmath191 are equivalent in @xmath144 , there exists a permutation , possibly with a conjugation , such that @xmath335 is gauge equivalent to @xmath191 . if a conjugation is involved , then @xmath5 and @xmath191 have the same sign of charge , otherwise opposite . lets postulate that the * charge * of a lipschitz quark is @xmath45 . since we have identified modulo @xmath67 we can assume that it is positive . in the meson case , the charge of the other particle is @xmath336 if it has the same charge and @xmath337 if it has opposite charge . [ [ section-47 ] ] in the baryon case , if the two other quarks have the same charge sign , they have charge @xmath338 . if one has the same charge than the lipschitz one , then both have charge @xmath45 and the other @xmath338 . the charge of an equivalence class is the sum of the charges . we have now assigned a charge in a gauge invariant way : a lipschitz quark @xmath12 has charge @xmath339 if @xmath14 and @xmath337 if it is obtained from that by switching two coordinates . the structure of the equivalence classes assures a compatible choice so that the total charge is an integer . [ [ section-48 ] ] in the meson case , we observe that one of the lipschitz primes is located on the three or two dimensional coordinate plane . if a coordinate is zero , then we can not perform all flipping operations and we have to see whether we have to flip it in the * gaussian sub - plane*. in this picture , all hadrons have charge @xmath340 . there are no hadrons of charge @xmath6 . particle physicists mention * beta @xmath341-hadrons * detected in experiments . also exotic 4 and 5 quark matter consisting of more than 3 quarks has been detected . it is likely just an analogue of 6 quark state which is known under the name deuterium " . but the quarks are the not in one bag " . its more like that a neutron and proton glued together by strong forces . [ [ section-49 ] ] the non - uniqueness of prime factorization allows to see the transition from one to an other factorization as a * particle process*. it involves the gauge bosons . it only becomes only unique modulo * unit migration * and recombination . ( see @xcite ) . this means that if @xmath40 is a quaternion integer and @xmath342 then @xmath343 , where the @xmath344 are units and @xmath345 are hurwitz primes . in other words , the factorization becomes unique if we look at it on the meson / baryon scale but it depends on the order . it becomes unique when including meta - commutation : the prime factorization of a nonzero hurwitz integer is unique up to meta - commutation , unit migration and recombination , the process of replacing @xmath346 with @xmath347 if @xmath348 have the same norm . + [ [ section-50 ] ] the @xmath8 is special @xmath349 and not included in the above remarks . it is neutral and equivalent to its anti - particle . it is `` higgs like '' , as in the lepton case and we can not place it yet . also , we like to think about a situation , where space is actually a dyadic group of integers ( a much more natural space as it is compact , features a smallest translation ) , here scaling by a factor @xmath6 is a symmetry and multiplication by @xmath6 makes the grid finer ( making the @xmath6-adic norm small ) . the mechanism of mass can anyway only be understood when looking at dynamical setups , where particles travel . when looking at wave equations in dyadic groups , the multiplication by 2 plays a special role and it can slow down particles , similar as mass does . + [ [ section-51 ] ] here are some examples . for @xmath79 , @xmath350 , @xmath351 form a meson of charge @xmath352 . for @xmath97 , we have a meson @xmath353 of charge @xmath16 . for @xmath354 , we have a baryon @xmath355 , @xmath356 , @xmath357 of charge @xmath279 and a meson @xmath358 , @xmath359 of charge @xmath16 . for @xmath360 we have a meson @xmath361 , @xmath362 of charge @xmath85 . and a meson @xmath363 , @xmath364 for which all coordinates are different . [ [ section-52 ] ] the integer units @xmath17 are the @xmath330 vector bosons while the @xmath365 etc are @xmath366 vector bosons . in the complex the @xmath367 generate photons . we can not have lipschitz primes @xmath368 as this is @xmath369 which is @xmath370 modulo @xmath371 . you see in the figures some pictures . we are able to attach to any prime a collection of baryons , for which the charges is determined . [ [ section-53 ] ] a square free integer is a collection of fermi particles . a given factorization is a particular particle configuration . going from one factorization to an other is an interaction process which invokes exchange of vector bosons and where particles change nature . say , if @xmath181 is a baryon and @xmath319 a meson , then @xmath372 and @xmath373 can represent a meson - baryon pair . [ [ section-54 ] ] gluons come from more general symmetry transformations . they do not have mass similarly like photons do not have mass . lets postulate that gluons are obtained from symmetry transformations when @xmath3 acts as a symmetry in @xmath1 . this is motivated from the complex place where @xmath374 , the multiplication with i , the square root of -1 , is realized as an element in the dihedral group , actually the cyclic group u. we can associate a gluon with a transformation @xmath375 which is a subgroup of @xmath3 the vector bosons @xmath376 are implemented by multiplications by a unit in @xmath10 . [ [ related - ideas ] ] related ideas + + + + + + + + + + + + + in 1952 , paul kustaanheimo @xcite explored the idea to do physics over a finite field @xmath377 rather than over the real or complex numbers . he mused about the size of the size of the prime . the situation is different however in that the approach is to take the integers in @xmath1 are the central point . we also should mention that particle phenomenology in division algebras has been worked on before , like in @xcite . the octonions could play an important role too but their combinatorics is more complicated as there are three type of primes : * gravesian integers * @xmath378 which play the role of the lipschitz primes in the hurwitz case , the * kleinian integers * @xmath379 which play the role of the hurwitz primes in @xmath1 . then there are the * kirmse integers * which includes elements for which @xmath371 of the entries are half integers . eight * maximal orders * have been identified by kirmse @xcite , a miscount which was later corrected by coxeter to seven @xcite ) . + [ [ quaternion - valued - waves ] ] quaternion valued waves + + + + + + + + + + + + + + + + + + + + + + + we have seen that any @xmath0-valued quantum evolution extends to an evolution of quaternion valued fields when seeing the evolution on @xmath380 three complex planes but then merge things in a common real line . it would be useful to explore this more systematically . the physics in division algebras is any new @xcite . one certainly also has to consider the possibility of octonion valued fields . we currently feel that the non - associativity in @xmath129 makes that algebra more difficult to use in a quantum mechanics frame work similarly as number theory is more difficult . but again , the frobenius - hurwitz theorem makes division algebras so special that it is only natural to consider them . particle physics today is in an interesting stage , as some urgent questions are open . are there more than three different flavors , are there larger symmetries like @xmath48 or will completely different approaches prevail and replace the standard model ? the structure outlined here hints that the answer could be no " in all three cases . [ [ stability - of - neutrini ] ] stability of neutrini + + + + + + + + + + + + + + + + + + + + + if @xmath381 or @xmath382 or @xmath383 is a gaussian prime for which @xmath384 is prime then it is also a lipschitz prime in @xmath1 . while @xmath0 contains also light primes " like @xmath385 , primes for which @xmath386 is prime , within @xmath1 : the initial inert prime @xmath79 would decay @xmath387 . the particle @xmath388 is equivalent to @xmath389 so that it is a meson . why does @xmath390 not decay but the prime @xmath391 does decay into an electron - positron pair ? one possibility to explain this is that @xmath390 is a different particle , evolving in a complex plane rather than the meson pair @xmath392 which evolves in the full algebra @xmath1 . however , here the stretch of the analogy becomes apparent already . [ [ tetra - and - penta - quarks ] ] tetra and penta quarks + + + + + + + + + + + + + + + + + + + + + + * tetra quarks * and * penta quark matter * seem have been confirmed by now . examples of such events were x(3872 ) , x(4140 ) . in one interpretation , these states of matter so special and are just pairs of mesons similarly as the * deuterium nucleus * is a familiar 6 quark state and the tritium nucleus a @xmath393 quark state . four and five quark states exotic because they are unusual mesons configurations which seem just be more rare . it is not necessary that a 4 quark state contains all the quarks together in the same way as in say the proton or neutron case . a deuterium is a 6 quark state , but the quarks are not glued as fundamentally as in a baryon . a baryon can only be separated by combining it with other particles , while a deuterium boson can be split into a neutron and proton , which are both fermions . [ [ number - of - electrons - and - neutrinos ] ] number of electrons and neutrinos + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we know that there are about an equal amount of @xmath394- than @xmath395-primes , with a slight * chebyshev bias * for smaller numbers . but we estimate about 1 billion times more neutrini than electron - positron pairs in the universe . let us assume the number of particles in the universe is finite . as there are infinitely many primes we have to assume we impose some * kustaanheimo threshold * @xcite number and look only at primes below that number and take this as a measure for the number of particles . if we count also the * virtual electron - positron pairs * which quantum field theory predicts and leads to manifestation like the * casimir effect*. using virtual particles one could imagine the two particle classes to have similar amount of elements . [ [ light - neutrini ] ] light neutrini + + + + + + + + + + + + + + we know that neutrini are much lighter than electrons . the neutrino mass is about @xmath396 ev which is about @xmath397 . the mass of an electron is about @xmath398 . current physics does not even tell how neutrinos acquire mass . the current understanding is that is that it does not happen through the higgs mechanism . that would be compatible with the fact that it is simply related to the amplitude of the wave . primes with larger norm have more mass . this is not unnatural since this is a dynamical mass interpretation . in number theory , there is a difference in the amplitudes : the real inert gaussian primes @xmath181 have prime @xmath386 while the split gaussian primes @xmath181 have prime @xmath74 . for example @xmath385 is a inert prime wile @xmath256 is a split prime . if we look at @xmath74 as a measure for mass , then @xmath74 is not prime for a real gaussian prime . it is the square root of @xmath74 which is prime . if there should be any relation with @xmath74 and mass and the mass relation measured is considered , this suggests that @xmath399 for primes modeling particles . by hurwitz , on a sphere @xmath400 this @xmath401 is also the number of different prime equivalence classes on such a sphere . there is no relation with @xcite since the later takes the more radical approach replacing the field @xmath0 with a finite field . [ [ the - unramified - primes ] ] the unramified primes + + + + + + + + + + + + + + + + + + + + + the prime @xmath8 is special . when looking for analogies , it comes closest to the higgs particle . the reason is that this particle is conjugate to its own anti - particle and therefore has no charge . the higgs is a boson unlike the neutral neutrini . what is the difference ? one could try the picture that @xmath6 is actually equivalent to a pair of fermions @xmath179 of norm @xmath6 , where @xmath181 and @xmath294 are conjugated . ramified as such , the two fermions @xmath402 are identified so that it becomes elementary and a boson . for the other primes like @xmath115 , there is no such factorization . for @xmath403 the factorization @xmath404 splits into two charged particles which are not equivalent . the commutation statistics for @xmath8 is described by the * second supplement of quadratic reciprocity * : @xmath405 for primes of the form @xmath406 and @xmath407 for the rest of odd primes . we can remember this as follows : while neutrini do not interact with electrons nor positrons , the prime @xmath6 interacts with both types . [ [ what - are - the - photons - in - this - allegory ] ] what are the photons in this allegory ? + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + photons are the units in @xmath0 . they play the role of gauge transformations . the photons @xmath408 satisfy @xmath409 and interact with electrons or positrons only and not with neutrini . indeed , by the first supplement of quadratic reciprocity , we have @xmath305 for electron for positrons primes but @xmath410 for neutrini primes : there is no solution to @xmath411 modulo @xmath5 if @xmath5 is a neutrino prime but there is a solution @xmath40 to @xmath411 modulo @xmath5 if @xmath5 is a charged lepton . all photons are gauge equivalent so that there is only one kind of a photon . a photon has zero charge as it is equivalent to the real element @xmath16 . [ [ seduced - by - analogies ] ] seduced by analogies + + + + + + + + + + + + + + + + + + + + to the end , one has to caution : tales like eddington explaining the * fine structure constant * by numerology or kepler obsessed with his * harmonices mundi * matching planetary motion with platonic solids or the thomson - tait approach to explain the * chemical elements * using knots , show that analogies do not always go well , even if done by eminent scientists . as an * amusing illustration * , in some online forum , a user discovered that the natural numbers @xmath412 , the complex numbers @xmath0 , the quaternions @xmath1 and the octonions @xmath129 are curiously linked to the basic constituents of * organic matter * like nitrogen @xmath72 , carbon @xmath285 , hydrogen @xmath413 and oxygen @xmath414 . this silly coincidence illustrates how scientific brains like to build connections and if necessary , bend things , if a match is not complete . while the allegory spun here could be of similar type , there is interesting combinatorics to be explored , as the figures below illustrate . how many different baryons are there for a given prime ? how many are there of each type ? is there a more natural way to associate charge to the quarks by looking at all possible factorizations . what exactly happens to the equivalence classes during meta commutation ? 10 j. kirmse . ber die darstellbarkeit natrlicher ganzer zahlen als summen yon acht quadraten und ber ein mit diesem problem zusammenhngendes nichtkommutatives und nichtassoziatives zahlensystem . , 76:6382 , 1925 . |
configurations of rigid collections of saddle connections are connected component invariants for strata of the moduli space of quadratic differentials . they have been classified for strata of abelian differentials by eskin , masur and zorich . similar work for strata of quadratic differentials has been done in masur and zorich , although in that case the connected components were not distinguished . we classify the configurations for quadratic differentials on and on hyperelliptic connected components of the moduli space of quadratic differentials . we show that , in genera greater than five , any configuration that appears in the hyperelliptic connected component of a stratum also appears in the non - hyperelliptic one . | 0705.3142 | we study flat surfaces having conical singularities of angle integer multiple of @xmath1 and @xmath2 linear holonomy . the moduli space of such surfaces is isomorphic to the moduli space of quadratic differentials on riemann surfaces and is naturally stratified . flat surfaces corresponding to squares of abelian differentials are often called _ translation surfaces_. flat surfaces appear in the study of billiards in rational polygons since these can be `` unfolded '' to give a translation surface ( see @xcite ) . a sequence of quadratic differentials or abelian differentials leaves any compact set of a stratum when the length of a saddle connection tends to zero . this might force some other saddle connections to shrink . in the case of an abelian differential this correspond to homologous saddle connections . in the general case of quadratic differentials , the corresponding collections of saddle connections on a flat surface are said to be _ _ omologous _ _ , usually denoted as @xmath3 , see section [ homologous ] . ] ( pronounced `` hat - homologous '' ) . according to masur and smillie @xcite ( see also @xcite ) , a `` typical degeneration '' corresponds to the case when all the `` short '' saddle connections are pairwise omologous ) . therefore the study of configurations of omologous saddle connections ( or homologous saddle connection in the case of abelian differential ) is related to the study of the compactification of a given stratum . a configuration of omologous saddle connections on a generic surface is also a natural invariant of a connected component of the ambient stratum . in a recent article , eskin , masur and zorich @xcite study collections of homologous saddle connections for abelian differentials . they describe configurations for each connected component of the strata of abelian differentials . collections of omologous saddle connections are studied for quadratic differentials by masur and zorich @xcite : they describe all the configurations that can arise in any given stratum of quadratic differentials , but they do not distinguish connected components of such strata . according to lanneau @xcite , the non - connected strata of quadratic differentials admit exactly two connected components . they are of one of the following two types : * `` hyperelliptic '' stratum : the stratum admits a connected component that consists of hyperelliptic quadratic differentials ( note that some of these strata are connected ) . * exceptional stratum : there exist four non - connected strata that do not belong to the previous case . in this article , we give the classification of the configurations that appear in the hyperelliptic connected components ( theorem [ confighyp ] ) . this gives therefore a necessary condition for a surface to be in a hyperelliptic connected component . unfortunately , this is not a sufficient condition since , as we show , any configuration that appears in a hyperelliptic connected component also appears in the other component of the stratum when the genus is greater than @xmath4 ( theorem [ config : non : hyp ] ) . we address the description of configurations for exceptional strata to a next article . we deduce configurations for hyperelliptic components from configurations for strata of quadratic differentials on @xmath0 ( theorem [ prop : config ] ) . configurations for @xmath0 are deduced from general results on configurations that appear in @xcite . note that these configurations are needed in the study of asymptotics in billiards in polygons with `` right '' angles @xcite . for such a polygon , there is a simple unfolding procedure that consists in gluing along their boundaries two copies of the polygon . this gives a flat surface of genus zero with conical singularities , whose angles are multiples of @xmath1 ( _ i.e. _ a quadratic differential on @xmath0 ) . then a generalized diagonal or a periodic trajectory in the polygon gives a saddle connection on the corresponding flat surface . we also give in appendix an explicit formula that gives a relation between the genus of a surface and the ribbon graph of connected components associated to a collection of omologous saddle connections . some particular splittings are sometimes used to compute the closure of @xmath5-orbits of surfaces ( see @xcite ) . these splittings of surfaces can be reformulated as configurations of homologous or omologous saddle connections on these surfaces . it would be interesting to find some configurations that appear in _ any _ surface of a connected component of a stratum , as was done in @xcite . i would like to thank anton zorich for encouraging me to write this paper , and for many discussions . i also thank erwan lanneau , for his many comments . here we first review standart facts about moduli spaces of quadratic differentials . we refer to @xcite for proofs and details , and to @xcite for general surveys . let @xmath6 be a compact riemann surface of genus @xmath7 . a quadratic differential @xmath8 on @xmath6 is locally given by @xmath9 , for @xmath10 a local chart with @xmath11 a meromorphic function with at most simple poles . we define the poles and zeroes of @xmath8 in a local chart to be the poles and zeroes of the corresponding meromorphic function @xmath11 . it is easy to check that they do not depend on the choice of the local chart . slightly abusing vocabulary , a pole will be referred to as a zero of order @xmath12 , and a marked point will be referred to as a zero of order @xmath13 . an abelian differential on @xmath6 is a holomorphic 1-form . outside its poles and zeroes , @xmath8 is locally the square of an abelian differential . integrating this 1-form gives a natural atlas such that the transition functions are of the kind @xmath14 . thus @xmath6 inherits a flat metric with singularities , where a zero of order @xmath15 becomes a conical singularity of angle @xmath16 . the flat metric has trivial holonomy if and only if @xmath8 is globally the square of an abelian differential . if not , then the holonomy is @xmath2 and @xmath17 is sometimes called a _ half - translation _ surface since the transitions functions are either translations , either half - turns . in order to simplify the notation , we will usually denote by @xmath6 a surface with a flat structure . we associate to a quadratic differential the set @xmath18 of orders of its poles and zeros . the gauss - bonnet formula asserts that @xmath19 . conversely , if we fix a collection @xmath20 of integers greater than or equal to @xmath12 satisfying the previous equality , we denote by @xmath21 the ( possibly empty ) moduli space of quadratic differentials which are not globally the square of any abelian differential , and having @xmath18 as orders of poles and zeroes . it is well known that @xmath21 is a complex analytic orbifold , which is usually called a _ stratum _ of the moduli space of quadratic differentials . we mostly restrict ourselves to the subspace @xmath22 of area one surfaces , where the area is given by the flat metric . in a similar way , we denote by @xmath23 the moduli space of abelian differentials of area @xmath24 having zeroes of degree @xmath25 , where @xmath26 and @xmath27 . a saddle connection is a geodesic segment ( or geodesic loop ) joining two singularities ( or a singularity to itself ) with no singularities in its interior . even if @xmath8 is not globally a square of an abelian differential we can find a square root of it along the saddle connection . integrating it along the saddle connection we get a complex number ( defined up to multiplication by @xmath12 ) . considered as a planar vector , this complex number represents the affine holonomy vector along the saddle connection . in particular , its euclidean length is the modulus of its holonomy vector . note that a saddle connection persists under small deformation of the surface . local coordinates on a stratum of abelian differentials are obtained by integrating the holomorphic 1-form along a basis of the relative homology @xmath28 , where _ sing _ is the set of conical singularities . equivalently , this means that local coordinates are defined by the relative cohomology @xmath29 . local coordinates in a stratum of quadratic differentials are obtained by the following way : one can naturally associate to a quadratic differential @xmath30 a double cover @xmath31 such that @xmath32 is the square of an abelian differential @xmath33 . the surface @xmath3 admits a natural involution @xmath34 , that induces on the relative cohomology @xmath29 an involution @xmath35 . it decomposes @xmath29 into an invariant subspace @xmath36 and an anti - invariant subspace @xmath37 . one can show that the anti - invariant subspace @xmath37 gives local coordinates for the stratum @xmath21 . it is well known that lebesgue measure on these coordinates defines a finite measure @xmath38 on the stratum @xmath22 . a hyperelliptic quadratic differential is a quadratic differential such that there exists an orientation preserving involution @xmath34 with @xmath39 and such that @xmath40 is a sphere . we can construct families of hyperelliptic quadratic differentials by the following way : to all quadratic differentials on @xmath0 , we associate a double covering ramified over some singularities satisfying some fixed combinatorial conditions . the resulting riemann surfaces naturally carry hyperelliptic quadratic differentials . some strata admit an entire connected component that is made of hyperelliptic quadratic differentials . these components arise from the previous construction and have been classified by m. kontsevich and a. zorich in case of abelian differentials @xcite and by e. lanneau in case of quadratic differentials @xcite . the strata of abelian differentials having a hyperelliptic connected component are the following ones . 1 . @xmath41 , where @xmath42 . it arises from @xmath43 . the ramifications points are located over all the singularities . @xmath44 , where @xmath42 . it arises from @xmath45 . the ramifications points are located over all the poles . in the above presented list , the strata @xmath46 , @xmath47 , @xmath48 and @xmath49 are the ones that are connected . the strata of quadratic differentials that have a hyperelliptic connected component are the following ones . 1 . @xmath50 where @xmath15 , @xmath42 and @xmath51 . it arises from @xmath52 . the ramifications points are located over @xmath53 poles . @xmath54 where @xmath55 , @xmath42 and @xmath56 . it arises from @xmath57 . the ramifications points are located over @xmath58 poles and over the zero of order @xmath59 . @xmath60 where @xmath55 , @xmath61 and @xmath51 . it arises from @xmath62 . the ramifications points are located over all the singularities in the above presented list , the strata @xmath63 , @xmath64 , @xmath65 , @xmath66 and @xmath67 are the ones that are connected . let @xmath68 be a flat surface and let us denote by @xmath31 its canonical double cover and by @xmath34 the corresponding involution . let @xmath69 denote the set of singularities of @xmath6 and let @xmath70 . to an oriented saddle connection @xmath71 on @xmath6 , one can associate @xmath72 and @xmath73 its preimages by @xmath74 . if the relative cycle @xmath75 $ ] satisfies @xmath75=-[\gamma_2 ] \in h_1(\widehat{s},\widehat{\sigma},\mathbb{z})$ ] , then we define @xmath76=[\gamma_1]$ ] . otherwise , we define @xmath76=[\gamma_1]-[\gamma_2]$ ] . note that in all cases , the cycle @xmath76 $ ] is anti - invariant with respect to the involution @xmath34 . [ def : homo ] two saddle connections @xmath71 and @xmath77 are omologous if @xmath76=\pm[\tilde{\gamma^{\prime}}]$ ] . consider the flat surface @xmath78 given in figure [ exemple_trivial ] ( a `` pillowcase '' ) , it is easy to check from the definition that @xmath72 and @xmath73 are omologous since the corresponding cycles for the double cover @xmath3 are homologous . consider the flat surface given in figure [ exempleplat ] ( at the end of section [ homologous ] ) , the reader can check that the saddle connections @xmath72 , @xmath73 and @xmath79 are pairwise omologous . consider two distinct saddle connections @xmath80 on a half - translation surface . the following assertions are equivalent : * the two saddle connections @xmath71 and @xmath77 are omologous . * the ratio of their lengths is constant under any small deformation of the surface inside the ambient stratum . * they have no interior intersection and one of the connected components of @xmath81 has trivial linear holonomy . furthermore , if @xmath71 and @xmath77 are omologous , then the ratio of their lengths belongs to @xmath82 . consider a set of omologous saddle connections @xmath83 on a flat surface @xmath6 . slightly abusing notation , we will denote by @xmath84 the subset @xmath85 . this subset is a finite union of connected half - translation surfaces with boundaries . let @xmath6 be a flat surface and @xmath86 a collection of omologous saddle connections . the _ graph of connected components _ , denoted by @xmath87 , is the graph defined by the following way : * the vertices are the connected components of @xmath84 , labelled by `` @xmath88 '' if the corresponding surface is a cylinder , by `` @xmath89 '' if it has trivial holonomy ( but is not a cylinder ) , and otherwise by `` @xmath90 '' if it has non - trivial holonomy . * the edges are given by the saddle connections in @xmath71 . each @xmath91 is located on the boundary of one or two connected components of @xmath92 . in the first case it becomes an edge joining the corresponding vertex to itself . in the second case , it becomes an edge joining the two corresponding vertices . in @xcite , masur and zorich describe the set of all possible graphs of connected components for a quadratic differential . this set is roughly given by figure [ fig : classification : of : graphs ] , where dot lines are chains of `` @xmath89 '' and `` @xmath88 '' vertices of valence two . the next theorem gives a more precise statement of this description . it can be skipped in a first reading . let @xmath17 be quadratic differential ; let @xmath71 be a collection of omologous saddle connections @xmath93 , and let @xmath87 be the graph of connected components encoding the decomposition @xmath94 . the graph @xmath87 either has one of the basic types listed below or can be obtained from one of these graphs by placing additional `` @xmath88''-vertices of valence two at any subcollection of edges subject to the following restrictions . at most one -vertex may be placed at the same edge ; a -vertex can not be placed at an edge adjacent to a -vertex of valence @xmath95 if this is the edge separating the graph . the graphs of basic types , presented in figure [ fig : classification : of : graphs ] , are given by the following list : * an arbitrary ( possibly empty ) chain of `` @xmath89''-vertices of valence two bounded by a pair of `` @xmath90''-vertices of valence one ; * a single loop of vertices of valence two having exactly one `` @xmath90''-vertex and arbitrary number of `` @xmath89''-vertices ( possibly no `` @xmath89''-vertices at all ) ; * a single chain and a single loop joined at a vertex of valence three . the graph has exactly one `` @xmath90''-vertex of valence one ; it is located at the end of the chain . the vertex of valence three is either a `` @xmath89''-vertex , or a -vertex ( vertex of the cylinder type ) . both the chain , and the cycle may have in addition an arbitrary number of `` @xmath89''-vertices of valence two ( possibly no `` @xmath89''-vertices at all ) ; * two nonintersecting cycles joined by a chain . the graph has no `` @xmath90''-vertices . each of the two cycles has a single vertex of valence three ( the one where the chain is attached to the cycle ) ; this vertex is either a `` @xmath89''-vertex or a -vertex . if both vertices of valence three are -vertices , the chain joining two cycles is nonempty : it has at least one `` @xmath89''-vertex . otherwise , each of the cycles and the chain may have arbitrary number of `` @xmath89''-vertices of valence two ( possibly no `` @xmath89''-vertices of valence two at all ) ; * `` figure - eight '' graph : two cycles joined at a vertex of valence four , which is either a `` @xmath89''-vertex or a -vertex . all the other vertices ( if any ) are the `` @xmath89''-vertices of valence two . each of the two cycles may have arbitrary number of such `` @xmath89''-vertices of valence two ( possibly no `` @xmath89''-vertices of valence two at all ) . each graph listed above corresponds to some flat surface @xmath6 and to some collection of saddle connections @xmath71 . ( 0,0)(0,0 ) ( -140,470.5 ) ( 0,0)(0,0 ) ( 38,-29)a ) ( 285,-53)b ) ( 132,-130)c ) ( 132,-362)d ) ( 132,-450)e ) two omologous saddle connections are not necessary of the same length . the additional parameters @xmath24 or @xmath96 written along the vertices in figure [ fig : classification : of : graphs ] represent the lengths of the saddle connections in the collection @xmath86 after suitably rescaling the surface . each connected component of @xmath97 is a non - compact surface which can be naturally compactified ( for example considering the distance induced by the flat metric on a connected component of @xmath97 , and the corresponding completion ) . we denote this compactification by @xmath98 . we warn the reader that @xmath98 might differ from the closure of the component in the surface @xmath6 : for example , if @xmath91 is on the boundary of just one connected component @xmath98 of @xmath99 , then the compactification of @xmath98 carries two copies of @xmath91 in its boundary , while in the closure of the corresponding connected component of @xmath97 , these two copies are identified . the boundary of each @xmath100 is a union of saddle connections ; it has one or several connected components . each of them is homeomorphic to @xmath101 and therefore defines a cyclic order in the set of boundary saddle connections . each consecutive pair of saddle connections for that cyclic order defines a _ boundary singularity _ with an associated angle which is a integer multiple of @xmath1 ( since the boundary saddle connections are parallel ) . the surface with boundary @xmath100 might have singularities in its interior . we call them _ interior singularities_. let @xmath102 be a maximal collection of omologous saddle connections . then a _ configuration _ is the following combinatorial data : * the graph @xmath87 . * for each vertex of this graph , a permutation of the edges adjacent to the vertex ( encoding the cyclic order of the saddle connections on each connected component of the boundary of @xmath100 ) . * for each pair of consecutive elements in that cyclic order , an integer @xmath55 such that the angle between the two corresponding saddle connections is @xmath103 . this integer will be referred as the _ order of the boundary singularity_. * for each @xmath100 , a collection of integers corresponding to the orders of the interior singularities of @xmath100 . following @xcite , we will encode the permutation of the edges adjacent to each vertex by a _ ribbon graph_. figure [ exemple ] represents a configuration on a flat surface . the corresponding collection @xmath104 of omologous saddle connections decomposes the surface into three connected components . the first connected component has four interior singularities of order @xmath12 , and its boundary consists of a single saddle connection with the corresponding boundary singularity of angle @xmath105 . the second connected component has no interior singularities . it has two boundary components , one consisting of a single saddle connection with corresponding singularity of angle @xmath106 , and the other consists of a union of two saddle connections with corresponding boundary singularities of angle @xmath107 and @xmath106 . the last connected component has no interior singularities , and admits two boundary components that consists each of a single saddle connection with corresponding boundary singularities of angles @xmath106 . figure [ exempleplat ] represents a flat surface with a collection of three omologous saddle connections realizing this configuration . [ generique ] when describing the configuration of a collection of omologous saddle connections @xmath108 , we will always assume that the quadratic differential is _ generic _ , and therefore , each saddle connection parallel to the @xmath91 is actually omologous to the @xmath91 ( see @xcite ) . a maximal collection of omologous saddle connections and the associated configuration persist under any small deformation of the flat surface inside the ambient stratum . they also persist under the well know @xmath109 action on the stratum which is ergodic with respect to the lebesgue measure @xmath38 ( see @xcite ) . hence , all admissible configurations that exists in a connected component are realized in a generic surface of that component . furthermore , the number of collections realizing a given configuration in a generic surface has quadratic asymptotics ( see @xcite ) . in this section we describe all admissible configurations of omologous saddle connections that arise on @xmath0 . to avoid confusion of notation , we specify the following convention : we denote by @xmath110 the set with multiplicity @xmath111 , where @xmath112 is the multiplicity of @xmath113 . we assume that @xmath114 for @xmath115 . for example the notation @xmath116 means @xmath117 . let @xmath118 be a stratum of quadratic differentials on @xmath0 different from @xmath119 . we give in the next example four families of admissible configurations for this stratum . in the next example , @xmath71 is always assumed to be a maximal collection of omologous saddle connections . we give in table [ graphcp1 ] the corresponding graphs and `` topological pictures '' . the existence of each of these configurations is a direct consequence of the main theorem of @xcite . [ excp1 ] * let @xmath120 be an unordered pair of integers with @xmath121 . the set @xmath71 consists of a single saddle connection joining a singularity of order @xmath122 to a distinct singularity of order @xmath123 . * let @xmath124 be an unordered pair of strictly positive integers such that @xmath125 ( with @xmath126 ) , and let @xmath127 be a partition of @xmath128 . the set @xmath71 consists of a simple saddle connection that decomposes the sphere into two one - holed spheres @xmath129 and @xmath130 , such that each @xmath100 has interior singularities of positive order given by @xmath131 and @xmath132 poles , and has a single boundary singularity of order @xmath133 . * let @xmath134 be an unordered pair of integers . let @xmath127 be a partition of @xmath135 . the set @xmath71 consists of two closed saddle connections that decompose the sphere into two one - holed spheres @xmath129 and @xmath130 and a cylinder , and such that each @xmath100 has interior singularities of positive orders given by @xmath131 and @xmath132 poles and has a boundary singularity of order @xmath133 . * let @xmath136 . the set @xmath71 is a pair of saddle connections of different lengths , and such that the largest one starts and ends from a singularity of order @xmath122 and decompose the surface into a one - holed sphere and a `` half - pillowcase '' , while the shortest one joins a pair of poles and lies on the other end of the half - pillowcase . now we will prove that the configuration described previously are the only ones for the stratum @xmath118 . we first start with several preliminary lemmas which are applicable to flat surfaces of arbitrary genus . let @xmath6 be a generic flat surface of genus @xmath137 in some stratum of quadratic differentials , and let @xmath71 be a maximal collection of omologous saddle connections on it . taking the natural compactification of each connected component of @xmath97 , we get a collection @xmath138 of compact surfaces with boundaries . the boundary of each @xmath100 is topologically a union of disjoint circles . we can glue a disc to each connected component of the boundary of @xmath100 and get a closed surface @xmath139 ; we denote by @xmath140 the genus of @xmath139 . [ genre ] let @xmath7 be the genus of @xmath6 , then @xmath141 . for each @xmath100 , we consider a collection of paths @xmath142 of @xmath100 that represent a symplectic basis of @xmath143 and that avoid the boundary of @xmath100 . when we glue the @xmath144 together , the @xmath145 provides a collection of cycles of @xmath146 . it forms a symplectic family because two paths arising from two different surfaces do not intersect each other . therefore we get a free family of @xmath146 , thus : @xmath147 in the appendix , we will improve lemma [ genre ] and give an exact formula in terms of the graph @xmath87 and the ribbon graph . [ genre_plus ] if @xmath148 is not a cylinder and has trivial holonomy , then @xmath149 . we recall that the initial collection of omologous saddle connections is assumed to be maximal , therefore there is no interior saddle connections omologous to any boundary saddle connection . let @xmath150 be the orders of the interior conical singularities of @xmath148 and @xmath151 be the orders of the boundary singularities . let @xmath152 be the closed flat surface obtained by gluing @xmath148 and a copy of itself taken with opposite orientation along their boundaries . if @xmath153 denotes the number of connected components of the boundary of @xmath148 and @xmath154 denotes the genus of @xmath152 , one can see that @xmath155 . the singularities of @xmath152 are of orders @xmath156 . furthermore , @xmath157 are nonnegative integers since @xmath152 has trivial holonomy . applying the gauss - bonnet formula for quadratic differentials , one gets : @xmath158 which obviously gives @xmath159 to conclude , we need few elementary remarks ( which are already written in @xcite ) about the order of the conical singularities of the boundary : * if a connected component of the boundary is just a single saddle connection , then the corresponding angle can not be @xmath1 otherwise the saddle connection would then be a boundary component of a cylinder . then the other boundary component of that cylinder would be a saddle connection omologous to the previous one ( see remark [ generique ] ) . so @xmath100 would be that cylinder contradicting the hypothesis . furthermore , the holonomy of a path homotopic to the saddle connection is trivial if and only if the conical angle of the boundary singularity is an odd multiple of @xmath1 . + therefore that angle is greater or equal to @xmath160 , and hence , the corresponding order @xmath161 of the boundary singularity has order at least @xmath96 . * if a connected component of the boundary is given by two saddle connections , then as before , the two corresponding conical angles can not be both equal to @xmath1 ( otherwise @xmath100 would be a cylinder ) and are of the same parity ( otherwise @xmath100 would have nontrivial holonomy ) . now we complete the proof of the lemma . we recall that the vertex corresponding to @xmath148 in @xmath87 is of valence at most four , and hence @xmath162 . the case @xmath163 is trivial . if @xmath164 then there is a connected component of the boundary of @xmath148 with one or two saddle connections . in both cases , the remarks a ) and b ) imply that @xmath148 admits a boundary singularity of order @xmath165 , and therefore @xmath166 . if @xmath167 , then there are at least two boundary components that consist of a single saddle connection . from remark a ) , this implies that @xmath148 admits two boundary singularities @xmath168 and @xmath169 of order greater than or equal to two . applying remarks a ) and b ) on the other boundary components , we show that @xmath148 admits at least an other boundary singularity of order @xmath170 . therefore @xmath171 finally , @xmath172 and the lemma is proven . now , we describe all the possible configurations when the genus @xmath7 of the surface is zero . [ prop : config ] let @xmath118 be a stratum of quadratic differentials on @xmath0 different from @xmath119 , and let @xmath71 be a maximal collection of omologous saddle connections on a flat surface in this stratum . then all possible configurations for @xmath71 are the ones described in example [ excp1 ] . it follows from lemmas [ genre ] and [ genre_plus ] that @xmath87 has no `` @xmath89 '' component . furthermore , a loop of the graph @xmath87 can not have any cylinder since this would add a handle to the surface . now using the description from @xcite of admissible graphs ( see figure [ fig : classification : of : graphs ] ) , we can list all possible graphs . for each graphs , we now describe the corresponding admissible configurations . \a ) a single `` @xmath90 '' vertex of valence two and an edge joining it to itself . + this can represent two possible cases : either the boundary of the closure of @xmath92 has two connected components , or it has only one . in the first case each connected component of the boundary is a single saddle connection . gluing these two boundary components together adds a handle to the surface . so this case does not appear for genus zero . + in the other case , the single boundary component consists of two saddle connections . the surface @xmath6 is obtained after gluing these two saddle connections , so @xmath71 consists of a single saddle connection joining a singularity of order @xmath122 to a distinct singularity of order @xmath123 . note that @xmath122 and @xmath123 can not be both equal to @xmath12 otherwise there would be another saddle connection in the collection @xmath71 ( see remark [ generique ] ) . b ) two `` @xmath90 '' vertices of valence one joined by a single edge . that means that @xmath71 consists of a single closed saddle connection @xmath72 which separates the surface in two parts . we get a unordered pair @xmath173 of one - holed spheres with boundary singularities of angles @xmath174 and @xmath175 correspondingly . the saddle connection of the initial surface is adjacent to a singularity of order @xmath176 . none of the @xmath133 is null otherwise the saddle connection would bound a cylinder , and there would exist a saddle connection omologous to @xmath72 on the other boundary component of this cylinder . now considering the interior singularities of positive order of @xmath129 and @xmath130 respectively , this defines a partition @xmath127 of @xmath177 . each @xmath100 also have @xmath178 poles , with @xmath179 . if we decompose the boundary saddle connection of @xmath100 in two segments starting from the boundary singularity , and glue together these two segments , we then get a closed flat surface with @xmath180 for the order of the singularities . the gauss - bonnet theorem implies : @xmath181 \c ) two `` @xmath90 '' vertices of valence one and a `` @xmath88 '' vertex of valence @xmath96 . this case is analogous to the previous one . \d ) a `` @xmath90 '' vertex of valence one , joined by an edge to a valence three `` @xmath88 '' vertex and an edge joining the `` @xmath88 '' vertex to itself . + the `` @xmath90 '' vertex represents a one - holed sphere . it has a single boundary component which is a closed saddle connection . the cylinder has two boundary components of equal lengths . one has two saddle connections of length 1 ( after normalization ) the other component has a single saddle connection of length @xmath96 . so , the only possible configuration is obtained by gluing the two saddle connections of length 1 together ( creating a `` half - pillowcase '' ) and gluing the other one with the boundary of the `` @xmath90 '' component . the boundary singularity of the `` @xmath90 '' component has an angle of @xmath182 ( equivalently , has order @xmath122 ) for some @xmath183 . \e ) a valence four `` @xmath88 '' vertex with two edges joining the vertex to itself . the cylinder has two boundary components , each of them is composed of two saddle connections . all the saddle connections have the same length . if we glue a saddle connection with one of the other connected component of the boundary , we get a flat torus , which has trivial holonomy and genus greater than zero . so , we have to glue each saddle connection with the other saddle connection of its boundary component . that means that we get a ( twisted ) `` pillowcase '' and the surface belongs to @xmath184 . + in each of these first four cases , the surface necessary has a singularity of order at least one . so , they can not appear in @xmath184 , which means that the fifth case is the only possibility in that stratum . in this section , we describe the configurations of omologous saddle connections in a hyperelliptic connected component . we first reformulate lanneau s description of such components , see @xcite . the hyperelliptic connected components are given by the following list : 1 . the subset of surfaces in @xmath185 , that are a double covering of a surface in @xmath186 ramified over @xmath187 poles . here @xmath188 and @xmath189 are odd , @xmath190 and @xmath191 , and @xmath192 . the subset of surfaces in @xmath193 , that are a double covering of a surface in @xmath186 ramified over @xmath187 poles and over the singularity of order @xmath189 . here @xmath188 is odd and @xmath189 is even , @xmath190 and @xmath194 , and @xmath195 . 3 . the subset of surfaces in @xmath196 , that are a double covering of a surface in @xmath186 ramified over all the singularities . here @xmath188 and @xmath189 are even , @xmath197 and @xmath194 , and @xmath198 . taking a double covering of the configurations arising on @xmath0 , one can deduce configurations for hyperelliptic components . this leads to the following theorem : [ confighyp ] in the notations of the classification theorem above , the admissible configurations of omologous saddle connections for hyperelliptic connected components are given by tables [ graphhyp3 ] , [ graphhyp1 ] , [ graphhyp2 ] and [ graphhyp_mp ] . no other configuration can appear . integer parameters @xmath199 in tables [ graphhyp3 ] , [ graphhyp1 ] , [ graphhyp2 ] are allowed to take values @xmath12 and @xmath13 as soon as this does not contradict explicit restrictions . in table [ graphhyp_mp ] , we list several additional configurations which appear only when at least one of @xmath200 is equal to zero . in the description of configurations for the hyperelliptic connected component @xmath201 with @xmath202 , the notation @xmath203 ( _ resp . _ @xmath204 ) still represents the orders of a pair of singularities that are _ interchanged _ by the hyperelliptic involution . for example in a generic surface in the hyperelliptic component @xmath205 , for @xmath206 , the second line of table [ graphhyp1 ] means that , between any pair of singularities that are interchanged by the hyperelliptic involution on @xmath6 , there exists a saddle connection with no other saddle connections omologous to it . but if @xmath71 is a saddle connection between two singularities that are not interchanged by the involution @xmath34 , then @xmath207 is a saddle connection omologous to @xmath71 ( see below ) , and which is different from @xmath71 . let @xmath208 be a hyperelliptic connected component as in the list of the previous theorem and @xmath209 the corresponding stratum on @xmath0 . the projection @xmath210 , where @xmath211 and @xmath34 is the corresponding hyperelliptic involution , induces a covering from @xmath208 to @xmath212 . this is not necessarily a one - to - one map because there might be a choice of the ramification points on @xmath0 . but if we fix the ramification points , there is a locally one - to - one correspondence . we recall to the reader that theorem of masur and zorich cited after definition [ def : homo ] says that two saddle connections are omologous if and only if the ratio of their length is constant under any small perturbation of the surface inside the ambient stratum . therefore , two saddle connections in @xmath211 are omologous if and only if the corresponding saddle connections in @xmath6 are omologous . hence the image under @xmath74 of a maximal collection @xmath213 of omologous saddle connections on @xmath214 is a collection @xmath71 of omologous saddle connections on @xmath6 . note that @xmath71 is not necessary maximal since the preimage of a pole by @xmath74 is a marked point on @xmath214 and we do not consider saddle connections starting from a marked point . however , we can deduce _ all _ configurations for @xmath208 from the list of configurations for @xmath212 . we give details for a few configurations , the other ones are similar and the proofs are left to the reader . -_first line of table [ graphhyp3 ] _ : the configuration for @xmath209 corresponds to a single saddle connection @xmath71 on a surface @xmath6 that joins a singularity @xmath215 of degree @xmath188 to the distinct singularity @xmath216 of degree @xmath189 . the double covering is ramified over @xmath216 but not over @xmath215 . therefore , the preimage of @xmath71 in @xmath214 is a pair @xmath217 of saddle connections of the same lengths that join each preimage of @xmath215 to the preimage of @xmath216 . the boundary of compactification of @xmath218 admits only one connected component that consists of four saddle connections . the angles of the boundary singularities corresponding to the preimages of @xmath215 are both @xmath219 , and the angles of the other boundary singularities are @xmath220 since @xmath217 are interchanged by the hyperelliptic involution . -_fourth line of table [ graphhyp3 ] _ : the configuration for @xmath209 corresponds to a single closed saddle connection @xmath71 on a flat surface @xmath6 that separates the surface into two parts @xmath129 and @xmath130 . each @xmath100 contains some ramification points , so the preimage of @xmath71 separates @xmath214 into two parts @xmath221 and @xmath222 that are double covers of @xmath129 and @xmath130 . one of the @xmath223 has an interior singularity of order @xmath224 , while the other one does not have interior singularities . the description from masur and zorich of possible graphs of connected components ( see figure [ fig : classification : of : graphs ] ) implies that @xmath221 and @xmath222 can not have the same holonomy . let @xmath222 be the component with trivial holonomy , and choose @xmath33 a square root of the quadratic differential that defines its flat structure . if @xmath222 has two boundary components , each consisting of a single saddle connection , then the corresponding boundary singularities must be of even order @xmath225 . if @xmath222 has a single boundary component , then integrating @xmath33 along that boundary must give zero ( @xmath33 is closed ) , which is only possible if the order @xmath225 of the boundary singularities are odd . applying lemma [ bord - revdouble ] below , we see that @xmath222 does not have interior singularity . hence , @xmath221 has an interior singularity of order @xmath224 . the order of the boundary singularities of @xmath221 are both @xmath226 , which is of parity opposite to the one of @xmath225 . applying again lemma [ bord - revdouble ] , we get the number of boundary components of @xmath221 . -_last line of table [ graphhyp3 ] _ : the configuration for @xmath209 corresponds to a pair of saddle connections on a surface @xmath227 that separate the surface into a cylinder @xmath228 and a one - holed sphere @xmath129 . the double cover @xmath221 of @xmath129 is connected , and applying lemma [ bord - revdouble ] we see that it has two boundary components . the double cover @xmath229 of the cylinder @xmath228 admits no ramification point . so _ a priori _ , there are two possibilities : @xmath229 is either a cylinder the same length of and a width twice bigger than the width of @xmath228 , or it is a pair of copies of @xmath228 . here , the first possibility is not realizable otherwise the double covering @xmath230 would be necessary ramified over @xmath188 . finally we get @xmath214 by gluing a boundary component of each cylinder to each boundary component of @xmath221 , and gluing together the remaining boundary components of the cylinders . + note that the preimage of the saddle connection joining a pair of poles on @xmath6 is a _ regular _ closed geodesic in @xmath214 , and hence in our convention , we do not consider such a saddle connection in the collection of omologous saddle connections on @xmath214 . when at least one of @xmath188 or @xmath189 equals zero , there is a marked point on @xmath0 that is a ramification point of the double covering . hence we have to start from a configuration of saddle connections on @xmath0 that might have marked points as end points : * if a maximal collection of omologous saddle connection on @xmath0 does not intersect a marked point , then the collection has already been described in theorem [ prop : config ] , and hence , the corresponding configuration in @xmath208 is already presented in tables [ graphhyp3 ] and [ graphhyp2 ] . * if a non - closed saddle connection in a collection admits a marked point as end point , then this saddle connection is simple since we can move freely that marked point . hence the corresponding configuration in @xmath208 is already written in tables [ graphhyp3 ] and [ graphhyp2 ] . * if a closed saddle connection admits a marked point as end point , then it is a closed geodesic . this corresponds to a new configuration on @xmath0 and the corresponding configuration in @xmath208 is described in table [ graphhyp_mp ] . the proof is analogous to the other cases . this concludes the proof of theorem [ confighyp ] . [ bord - revdouble ] let @xmath231 be a flat surface whose boundary consists of a single closed saddle connection and let @xmath232 be the order of the corresponding boundary singularity . let @xmath223 be a connected ramified double cover of the interior of @xmath231 and let @xmath233 be interior singularities . the sum @xmath234 is even and : * if @xmath235 is even , then the compactification of @xmath223 has two boundary components , each of them consists of a single saddle connection , with corresponding boundary singularity of order @xmath236 . * if @xmath235 is odd , then the compactification of @xmath223 has a single boundary component which consists of a pair of saddle connections of equal lengths , with corresponding boundary singularities of order @xmath236 . by construction , the boundary of the compactification of @xmath223 necessary consists of two saddle connections of equal lengths . it has one or two connected components . now we claim that @xmath237 where @xmath238 is the number of connected components of the boundary of @xmath223 . this equality ( that already appears in @xcite ) clearly implies the lemma . to prove the claim , we consider as in lemma [ genre_plus ] the surface @xmath239 of genus @xmath240 obtained by gluing @xmath223 and a copy of itself with opposite orientation along their boundaries . the orders of the singularities of @xmath239 are @xmath241 , so we get @xmath242 and therefore @xmath243 given a concrete flat surface , we do not necessary see at once whether it belongs or not to a hyperelliptic connected component . indeed , there exists hyperelliptic flat surfaces that are not in a hyperelliptic connected component . as a direct corollary of theorem [ confighyp ] , we have the following quick test . let @xmath6 be a flat surface with non - trivial holonomy and let @xmath71 be a collection of omologous saddle connections on @xmath6 . if one of the following property holds , then the surface @xmath6 does not belong to a hyperelliptic connected component . * @xmath97 admits three connected components and neither of them is a cylinder . * @xmath97 admits four connected components or more . following @xcite , given a fixed stratum , one can get a list of all realizable configurations of omologous saddle connections . nevertheless it is not clear which configuration realizes in which component . in the previous section we have described configurations for hyperelliptic components . in the section we show that _ any _ configuration realizable for a stratum is realizable in its non - hyperelliptic connected component , provided the genus @xmath7 is sufficiently large . we will use the following theorem which is a reformulation of the theorem of kontsevich - zorich and the theorem of lanneau cited in section [ bas_def ] . the following strata consists entirely of hyperelliptic surfaces and are connected . * @xmath46 , @xmath47 , @xmath48 and @xmath49 in the moduli spaces of abelian differentials . * @xmath63 , @xmath64 , @xmath65 , @xmath66 and @xmath67 in the moduli spaces of quadratic differentials . any other stratum that contains a hyperelliptic connected component admit at least one other connected component that contains a subset of full measure of flat surfaces that do not admit any isometric involution . [ simple_dans_nonhyp ] let @xmath212 be a non - connected stratum that contains a hyperelliptic connected component . if the set of order of singularities defining @xmath212 contains @xmath244 , for some @xmath206 , then there exists a non - hyperelliptic flat surface in @xmath212 having a simple saddle connection joining two different singularities of the same order @xmath122 . here we call a saddle connection `` simple '' when there are no other saddle connections omologous to it . according to masur and smillie @xcite , any stratum is nonempty except the following four exceptions : @xmath245 , @xmath246 , @xmath247 and @xmath248 . according to masur and zorich @xcite ( see also @xcite ) , if @xmath249 , then there is a continuous path @xmath250}$ ] in the moduli space of quadratic differentials , such that @xmath251 and @xmath252 is in @xmath253 for @xmath254 , and such that the smallest saddle connection on @xmath252 , for @xmath254 is simple and joins a singularity of order @xmath188 to a singularity of order @xmath189 . we say that we `` break up '' the singularity of order @xmath255 into two singularities of order @xmath188 and @xmath189 . we first consider the stratum @xmath256 . by assumption , @xmath212 is non - connected , so , either the genus is greater than @xmath95 , or @xmath257 and @xmath258 . hence the stratum @xmath259 is nonempty . now , we start from a surface @xmath260 in that stratum , and break up the singularity @xmath261 of order @xmath262 into two singularities @xmath215 and @xmath216 of orders @xmath263 and @xmath189 respectively ( see figure [ simple_non_hyp ] ) . we get a surface @xmath129 with a short vertical saddle connection @xmath72 between @xmath215 and @xmath216 . since the `` singularity breaking up '' procedure is continuous , there are no other short saddle connections on @xmath129 . then , we break up the singularity @xmath215 of order @xmath263 into a pair of singularities @xmath264 and @xmath265 of orders @xmath188 . we get by construction a surface @xmath130 in the stratum @xmath212 with a simple saddle connection @xmath73 between @xmath264 and @xmath265 , and of length very small compared to the length of @xmath72 . the fact that the `` singularity breaking up '' procedure is continuous implies that there persists a saddle connection @xmath266 between @xmath216 and one of the @xmath267 ( see figure [ simple_non_hyp ] ) . by construction , we can assume there is no other saddle connection of length @xmath268 , where @xmath269 denotes the length of @xmath266 and @xmath270 . hence , @xmath266 is simple by theorem of masur and zorich cited after definition [ def : homo ] . according to theorem [ confighyp ] , this can not exist in the hyperelliptic connected component since the corresponding configuration is not present in table [ graphhyp1 ] . thus @xmath130 belongs to the non - hyperelliptic connected component and we can assume , after a slight perturbation , than @xmath130 is not hyperelliptic . since by construction , the saddle connection @xmath73 is simple and joins two singularities of order @xmath271 , the lemma is proven for the stratum @xmath185 . the proofs for @xmath193 and for @xmath196 are analogous : note that these case do not occur for the genera @xmath24 or @xmath96 , because all corresponding strata are connected . therefore the genus is greater than or equal to @xmath95 and the stratum @xmath272 is nonempty . [ config : non : hyp ] let @xmath212 be a stratum of meromorphic quadratic differentials with at most simple poles on a riemann surface of genus @xmath273 . if @xmath212 admits a hyperelliptic connected component , then @xmath212 is non - connected and any configuration for @xmath212 is realized for a surface in the non - hyperelliptic connected component of @xmath212 . the fact that @xmath212 is non - connected follows directly from the theorem of lanneau . let @xmath6 be a flat surface in the hyperelliptic component of @xmath212 and let @xmath71 be a maximal collection of omologous saddle connections . the hyperelliptic involution @xmath34 maps @xmath71 to itself and hence , induces an involution @xmath35 on the set of connected components of @xmath97 . we claim that @xmath35 does not interchange two connected components @xmath274 of @xmath97 , for otherwise we can continuously deform @xmath100 outside a neighborhood of its boundary and reconstruct a new flat surface @xmath275 in the same connected component . by construction , such surface @xmath275 is not any more hyperelliptic . therefore , if @xmath6 is in a hyperelliptic component , then @xmath34 must induce an isometric and orientation preserving involution on each connected component of @xmath97 . using the formula for the genus of a compound surface proved in the appendix and the list of configurations for hyperelliptic connected components given in the previous section , we derive the following fact : if @xmath6 has genus @xmath273 and if @xmath71 is a maximal collection of omologous saddle connections , then at least one of the following three propositions is true . * @xmath97 admits a connected component @xmath276 of genus @xmath277 , that has a single boundary component and whose corresponding vertex in the graph @xmath87 is of valence @xmath96 . * @xmath97 admits a connected component @xmath276 of genus @xmath278 , that has exactly two boundary components and whose corresponding vertex in the graph @xmath87 is of valence @xmath96 . * @xmath97 is connected and the corresponding vertex in the graph @xmath87 is of valence @xmath279 . the proof follows from lemmas [ lem_b ] , [ lem_a ] , [ lem_c ] to situations a ) , b ) , c ) correspondingly . [ lem_b ] let @xmath6 be a flat surface in a hyperelliptic connected component and let @xmath71 be a maximal collection of omologous saddle connections . we assume that @xmath97 admits a connected component @xmath276 of genus @xmath277 , whose corresponding vertex in the graph @xmath87 is of valence @xmath96 , and such that @xmath260 has a single boundary component . then there exists @xmath280 that has the same configuration as @xmath281 , with @xmath275 in the complementary component of the same stratum . the boundary components of @xmath260 consists of two saddle connections of the same length and the corresponding boundary singularities have the same orders @xmath206 . identifying together these two boundary saddle connections , we get a hyperelliptic surface @xmath282 . if we continuously deform this surface , it keeps being hyperelliptic since we can perform the reverse surgery and get a continous deformation of @xmath6 . hence , @xmath282 belongs to a hyperelliptic component , and the hyperelliptic involution interchange two singularities of order @xmath283 . the genus of @xmath282 is greater than @xmath95 , so the corresponding stratum admits an other connected component . now we start from a closed flat surface @xmath152 in this other connected component . according to lemma [ simple_dans_nonhyp ] , we can choose @xmath152 such that it admits a simple saddle connection between the two singularities of order @xmath283 . now we cut @xmath152 along that saddle connection and we get a surface @xmath129 that have , after rescaling , the same boundary as @xmath260 . by construction , @xmath129 admits no interior saddle connection omologous to one of its boundary saddle connections . so , we can reconstruct a pair @xmath280 such that @xmath77 has the same configuration as @xmath71 in @xmath6 . the surface @xmath129 admits a nontrivial isometric involution if and only if @xmath152 shares this property . so , we can choose @xmath152 in such a way it admits no nontrivial isometric involution , and therefore the surface @xmath275 is non - hyperelliptic . this argument also works when @xmath282 is in the stratum @xmath284 ( here @xmath285 and @xmath286 ) . in any other case for @xmath287 , it is not possible to replace @xmath260 by a surface @xmath129 with no involution . [ lem_a ] let @xmath6 be a flat surface in a hyperelliptic connected component and let @xmath71 be a maximal collection of omologous saddle connections . we assume that @xmath97 admits a connected component @xmath276 of genus @xmath288 , that has two boundary components , and whose corresponding vertex in the graph @xmath87 is of valence @xmath96 . then there exists @xmath280 that has the same configuration as @xmath281 , with @xmath275 in the complementary component of the same stratum . each boundary component of @xmath260 consists of one saddle connection and the corresponding boundary singularities have the same orders @xmath206 . now we start from a closed flat surface @xmath152 with the same holonomy as @xmath276 and whose singularities consists of the interior singularities of @xmath276 and two singularities @xmath289 and @xmath290 of order @xmath291 . we can always choose @xmath152 such that it admits a saddle connection @xmath292 between @xmath289 and @xmath290 . now we construct a pair of holes by removing a parallelogram as in figure @xmath293 and gluing together the two long sides . note that the holes can be chosen arbitrary small , and therefore , the resulting surface with boundary does not have any interior saddle connection omologous to one of its boundary components . we denote by @xmath129 this surface , and up to rescaling , we can assume that @xmath260 and @xmath129 have isometric boundaries . hence replacing @xmath260 by @xmath129 in the decomposition of @xmath6 , we get a new pair @xmath294 that have the same configuration as @xmath281 . we now assume that @xmath129 admits a nontrivial isometric ( orientation preserving ) involution @xmath34 . then this involution interchanges the two boundary components of the surface . it is easy to check that we can perform the reverse surgery as the one described previously and we get a closed surface that admits a nontrivial involution . hence if @xmath152 belongs to a stratum that does not consist entirely of hyperelliptic flat surfaces , then we can choose @xmath152 such that @xmath275 is not in a hyperelliptic connected component . the hypothesis on the genus , the theorem of kontsevich - zorich and the theorem of lanneau imply that this arguments works except when @xmath152 belongs to @xmath48 , @xmath295 , @xmath65 , or @xmath67 . we remark that if @xmath296 , then @xmath260 must have nontrivial linear holonomy and no interior singularities . according to the list of configurations for hyperelliptic connected components given in section @xmath95 , this can not happen . we exhibit in figure [ b_case ] three explicit surfaces with boundary that corresponds to the three cases left . we represent these three surfaces as having a one - cylinder decomposition and by describing the identifications on the boundary of that cylinder . the length parameters can be chosen freely under the obvious condition that the sum of the lengths corresponding to the top of the cylinder must be equal to the sum of the lengths corresponding to the bottom of the cylinder . bold lines represents the boundary of the flat surface . now we remark that a nontrivial isometric involution must preserve the interior of the cylinder , and must exchange the boundary components . this induces some supplementary relations on the length parameters . therefore , we can choose them such that there is no nontrivial isometric involution . [ lem_c ] let @xmath6 be a flat surface of genus @xmath297 with nontrivial linear holonomy that belongs to a hyperelliptic connected component and let @xmath298 be a maximal collection of omologous saddle connections on @xmath6 . if @xmath97 is connected , then there exists @xmath280 that has the same configuration as @xmath281 , with @xmath275 in the complementary component of the same stratum . since @xmath99 is connected , the graph @xmath87 contains a single vertex , and it has valence four . according to theorem [ confighyp ] , two different cases appear : \a ) the surface @xmath299 has one boundary component . in this case we start from a surface in @xmath300 and perform a local surgery in a neighborhood of the singularity , as described in figure [ local4 ] ( see also @xcite , section @xmath4 ) . we get a surface and a pair of small saddle connections of length @xmath301 that have the same configuration as @xmath71 . the stratum @xmath300 admits non - hyperelliptic components and the same argument as in the previous lemmas works : if we start from a generic surface in a non - hyperelliptic component , then the resulting surface after surgery does not have any nontrivial involution . \b ) the surface @xmath299 has two boundary components , each of them consists of a pair of saddle connections with boundary singularities of order @xmath302 and @xmath303 . we construct explicit surfaces with the same configuration as @xmath71 , but that have no nontrivial involution . let @xmath304 and we start from a surface @xmath260 of genus @xmath305 in @xmath306 , that have a one - cylinder decomposition and such as identification on the boundary of that cylinder is given by the permutation @xmath307 when @xmath305 is even , and otherwise by the permutation @xmath308 we assume that @xmath188 and @xmath189 are odd and we perform a surgery on @xmath260 to get a surface @xmath129 with boundary as pictured on figure [ c_case ] . the surface @xmath129 admits two boundary components that consist of two saddle connections each and which are represented by the bold segments . each symbol , , , represents a different boundary singularity . it is easy to check that the boundary angles corresponding to and are both @xmath219 and that the angles corresponding to and are @xmath220 . hence after suitable identifications of the boundary of @xmath129 , we get a surface @xmath275 and a pair of omologous saddle connections @xmath77 that have the same configuration as @xmath281 . however , @xmath275 does not admit any nontrivial involution if the length parameters are chosen generically . note that this construction does not work when @xmath309 , but according to section @xmath95 , and since @xmath188 and @xmath189 are odd , we have @xmath310 , which is greater than or equal to @xmath95 by assumption . the case @xmath188 and @xmath189 even is analogous and left to the reader ( note that in this case , @xmath311 , and the construction works also for @xmath309 ) . here we improve lemma [ genre ] and give the relation between the genus of a surface and the genera of the connected components of @xmath97 , where @xmath71 is a collection of omologous saddle connections . we first remark that this relation depends not only on the graph of connected components , but also on the permutation on each of its vertices ( _ i.e. _ on the ribbon graph ) . indeed , let us consider a pair of omologous saddle connections that decompose the surface into two connected components @xmath129 and @xmath130 . then either both @xmath129 and @xmath130 have only one boundary component , or at least one of them has two boundary components . in the first case , @xmath6 is the connected sum of @xmath221 and @xmath222 , so @xmath312 , while in the second case , one has @xmath313 . let @xmath281 be a flat surface with a collection of omologous saddle connections . pure _ ribbon graph associated to @xmath281 is the 2-dimensional topological manifold obtained from the ribbon graph by forgetting the graph @xmath87 , as in figure [ purerg ] . here we do not assume that the collection @xmath71 is necessary maximal . when @xmath87 has a single vertex , then we prove the proposition using direct computation and the description of the boundary components corresponding to each possible ribbon graph . we refer to @xcite for this description . then our goal is to reduce ourselves to that case by removing successively from the collection @xmath322 some @xmath91 whose corresponding edges joins a vertex to a distinct one . we define a new graph @xmath323 , which is a deformation retract of the pure ribbon graph : the vertices of @xmath323 are the boundary components of each @xmath100 , while the edges correspond to the saddle connections in @xmath71 ( see figure [ exemple_genre_exact ] ) . for each vertex , there is a cyclic order on the set of edges adjacent to the vertex consistent with the orientation of the plane . if the initial pure ribbon graph does not embed into the plane , then it is also the case for @xmath323 . by construction , the euler characteristic of @xmath323 is the same as the pure ribbon graph associated to @xmath281 , and is easier to compute . let @xmath87 contains at least two vertices . choose a saddle connection representing an edge joining two distinct vertices of @xmath87 , and up to renumeration , we can assume that this saddle connection is @xmath72 . let us study the resulting configuration of @xmath324 . the saddle connection @xmath72 is on the boundary of two surfaces @xmath325 and @xmath326 . then the connected components of @xmath327 are the same as the connected component of @xmath99 except that the surfaces @xmath325 and @xmath326 are now glued along @xmath72 , and hence define a single surface @xmath328 . the genus of @xmath328 ( after gluing disks on its boundary ) is @xmath329 . furthermore , if @xmath72 was in a boundary component of @xmath129 ( resp . @xmath130 ) defined by the ordered collection @xmath331 ( resp . @xmath332 ) . then the cyclic order in the corresponding boundary component of @xmath328 is defined by @xmath333 . therefore @xmath330 is obtained from @xmath323 by shrinking the edge corresponding to @xmath72 and removing an isolated vertex that might appear ( see figure [ exemple_genre_exact ] ) . it is clear that the difference @xmath334 between the euler characteristic of @xmath323 and its number of connected component is constant under this procedure . one can also remark that if @xmath323 is connected and does not embed into the plane ( case @xmath24 of the proposition ) , then this is also true for @xmath330 . forgetting successively these @xmath91 will lead to the case when @xmath87 has a single vertex . at each steps of the removing procedure , the numbers @xmath314 and @xmath335 do not change , and the sum of the genera associated to the vertices does not change either . this concludes the proof . a. eskin , h. masur , a. zorich , _ moduli spaces of abelian differentials : the principal boundary , counting problems , and the siegel - veech constants _ , publ . hautes tudes sci . * 97 * ( 2003),61179 . a. zorich , _ flat surfaces_. in collection `` frontiers in number theory , physics and geometry . volume 1 : on random matrices , zeta functions and dynamical systems '' , p. cartier ; b. julia ; p. moussa ; p. vanhove ( editors ) , springer - verlag , berlin , ( 2006 ) , 439 - 586 . |
production rates of dileptons and photons from the quark - gluon ( qgp ) phase are calculated taking into account effects of confinement and spontaneous chiral symmetry breaking (sb ) not far from the transition temperature . we find that the production rates of dileptons with large momenta and of photons originating from the qgp around are suppressed by thesb effect . we also discuss to what extent information about details of the chiral transition , such as its characteristic temperature range and the steepness of the crossover , are reflected in these quantities . | 1505.03869 | in the limit of massless quarks with @xmath2 flavours , quantum chromodynamics ( qcd ) has an exact chiral @xmath3 symmetry . dynamical quark mass generation implies that this symmetry is spontaneously broken . in addition , non - zero current quark masses in the qcd lagrangian break chiral symmetry explicitly . spontaneous chiral symmetry breaking ( @xmath0sb ) is manifest in the temperature dependence of the chiral ( quark ) condensate , @xmath4 . recent determinations of @xmath5 from lattice qcd with @xmath6 flavours and physical current quark masses display a chiral crossover at a transition temperature around @xmath7 mev , slightly lowered from a previous determination @xcite , resulted from improvements in controlling the continuum limit . ] schematic models of the nambu & jona - lasino ( njl ) type @xcite link the dynamically generated quark masses , @xmath8 , to the chiral condensate : @xmath9 this concept is adopted in the present work , as will be explained in the next section . the other important aspect of qcd at temperatures around and above the transition temperature ( @xmath1 ) is the deconfinement transition . at zero temperature quarks and gluons are confined inside hadrons . above the deconfinement temperature the quarks and gluons are released and become active degrees of freedom . an order parameter for this transition to the quark - gluon plasma ( qgp ) is the polyakov loop . in the pure gauge case ( @xmath10 ) deconfinement emerges as a first - order phase transition , whereas it becomes a continuous crossover in the presence of dynamical quarks . thus , in full qcd , the polyakov loop is not an order parameter in the strict sense but it nonetheless serves as a monitor for deconfinement as a rapid crossover transition . motivated by this scenario , models have been designed ( referred to as pnjl models ) @xcite that unify the polyakov - loop - induced suppression of color non - singlet degrees of freedom below @xmath1 with the njl mechanism of spontaneous chiral symmetry breaking . high - energy heavy - ion collisions ( hic ) are the experimental tool for investigating properties of the qgp . electromagnetic probes such as dileptons and photons @xcite produced in these collisions are particularly important messengers for the physics of the hot and dense qcd matter since they are barely affected by successive interactions on their way through the surrounding expanding medium . at the relativistic heavy ion collider ( rhic ) , the phenix experiment has measured @xcite a large elliptic anisotropy of the produced photons characterized by the quantity @xmath11 . since the anisotropy is considered to be produced by the collective flow at later stages of the hic , it is natural that the produced hadrons , mainly generated at the later stages , have large @xmath11 . by contrast , photons are generated also at early stages of the hic ( in the initial state @xcite , or in the thermalization process @xcite ) . it is therefore not easy to explain @xcite the large photon @xmath11 , and currently no theory has completely succeeded in doing it . recently , a new mechanism to generate a large photon @xmath11 was suggested @xcite . it is based on the suppression of the photon production in the qgp phase due to the confinement effect . nevertheless , the @xmath0sb effect was not taken into account in this analysis , so if this effect further suppresses the photon production from the qgp phase , one can expect that the photon @xmath11 is more enhanced . motivated by these issues , we calculate the production rates of dileptons and photons in the qgp phase , using a model in which the confinement and the @xmath0sb effects are both taken into account . the present work focuses in this context primarily on the role of the latter effect , in terms of dynamically generated constituent quark masses around @xmath1 . it is an extension of the previous refs . @xcite in which corresponding studies have been performed with quark masses set equal to zero . this paper is organized as follows : the next section focuses on the implementation of the effects of confinement and @xmath0sb in our model . the dilepton production rate is calculated and the @xmath0sb effect on this rate is discussed in sec . [ sec : dilepton ] . section [ sec : photon ] is devoted to the calculation of the photon production rate , and to the discussion on how the @xmath0sb affects this rate . for this purpose , the photon production rate in the case of massless quark is also calculated beyond leading - log order if @xmath12 is very small . in the case of photon production , @xmath13 , where @xmath14 is the photon energy , @xmath15 the coupling constant , and @xmath16 the temperature , respectively . the production rate in the massless case at leading - log order was already calculated in refs . @xcite . ] . we summarize and present concluding remarks in sec . [ sec : summary ] . the detailed derivation of the photon production rate for vanishing quark masses , @xmath17 , is given in the appendix . this section describes the treatment of confinement and spontaneous chiral symmetry breaking within the framework of the present model . the confinement effect is taken into account as a modification of the quark and gluon thermal distribution functions , in a so - called semi - qgp model @xcite . the starting point is qcd with @xmath18 colors and with finite averages of the temporal component of the background gluon field ( @xmath19 ) , given by : @xmath20 where @xmath21 is the coupling . @xmath19 is assumed to be diagonal in color space and constant in space and time . the spatial components of the background gluon field , @xmath22 , are set equal to zero . the tracelessness of the gauge field implies @xmath23 . neglecting fluctuations of @xmath24 , this quantity is related to the expectation value of the polyakov loop ( @xmath25 ) by @xmath26 where @xmath27 is the wilson line in the temporal direction in imaginary time , with @xmath28 , and @xmath29 denotes path ordering . when performing calculations involving quarks and gluons , the background gluon field acts as an imaginary chemical potential coupled to color charges . the quark and gluon distribution functions , @xmath30 and @xmath31 , are modified as @xcite @xmath32 where @xmath33 and @xmath34 are color indices in the double - line notation @xcite , running from @xmath35 to @xmath18 , and @xmath36 . it might appear that the distribution functions have imaginary parts , but in fact , after summing over color indices , these distributions turn out to be real as they should . to see this , consider the color - averaged quark distribution function : @xmath37 where we have introduced @xmath38 . since @xmath39 is real , it follows that @xmath40 is real . next we demonstrate the suppression of the distribution functions for color - non - singlet degrees of freedom in the confined phase . from eq . ( [ eq : average - quark - distribution ] ) , the averaged quark distribution function becomes @xmath41 where we have used the following expression , valid in the confined phase for odd , but with an unphysical result . for example , when @xmath42 , the baryonic excitations appearing in the confined phase are diquarks , i.e. bosons , whereas the distribution function eq . ( [ eq : nf - confined ] ) is fermionic . for this reason , we restrict ourselves to the odd-@xmath18 case . ] @xmath43 : @xmath44 non - trivial contributions to @xmath39 in the expansion ( [ eq : average - quark - distribution ] ) arise whenever @xmath45 is an integer multiple of ( odd ) @xmath43 . we see that @xmath40 is suppressed compared to the distribution in the deconfined limit , @xmath46 , and it vanishes at large @xmath18 . for @xmath47 one notes that @xmath40 represents color - singlet 3-quark objects ( with energy @xmath48 and baryon number @xmath49 ) , although these structures are not localised or clustered in space . sometimes this property is called statistical confinement " . the gluon distribution function is likewise suppressed after taking the color sum . for the @xmath50 case @xmath51 can be written in the representation @xmath52 where @xmath53 with @xmath54 is the only independent quantity . the polyakov loop reads : @xmath55 in the asymptotic deconfined phase ( @xmath56 ) , we have @xmath57 , while in the confined phase , @xmath58 for the pure glue case ( @xmath59 ) . a correspondingly similar expression holds for @xmath39 : @xmath60 we add the following remark concerning the interpolating ansatz , eq.([eq : correction - polyakov - q ] ) : if the perturbative correction is not considered ( i.e. @xmath74 ) , then @xmath75 gives the upper bound for the polyakov loop . on the other hand , using eq.([eq : correction - polyakov - q ] ) for not too small @xmath76 presumably overestimates the perturbative correction and is expected to give a lower bound for @xmath70 . quark mass generation in qcd implies spontaneous @xmath0sb . the dynamical ( constituent ) quark mass @xmath77 is not directly accessible in lattice qcd computations , but it figures as a well - defined quantity in studies of the quark propagator using dyson - schwinger equations @xcite and schematic models of njl type @xcite . starting from massless current quarks , such approaches feature , in the form of a characteristic gap equation , a proportionality between the dynamical quark mass @xmath77 and the chiral condensate @xmath78 , realised also at finite temperatures @xmath16 as indicated in eq . ( [ eq : condensate ] ) . the transition from the @xmath0sb ( nambu - goldstone ) phase to chiral symmetry restoration in the wigner - weyl phase with @xmath79 is second order in the limit of vanishing current quark masses . this transition becomes a continuous but rapid crossover when the explicit chiral symmetry breaking by small non - zero current quark masses is included . in the present work we make use of the typical behaviour of @xmath8 derived from njl and pnjl models . the input in such models is designed to reproduce vacuum properties of the pion ( its decay constant @xmath80 and its physical mass @xmath81 starting from non - zero current quark masses ) . in the two - flavour case the emerging dynamical @xmath82 and @xmath83 quark masses can be well represented by the following expression : @xmath84~. \label{eq : njlmass}\end{aligned}\ ] ] a standard fit to two - flavour njl results gives @xmath85 mev , @xmath86 mev and a transition temperature @xmath87 mev . the pnjl model features a characteristic entanglement of @xmath8 reflecting spontaneous ( dynamical ) @xmath0sb , with the polyakov loop representing confinement . while values of @xmath88 found in pnjl calculations are typically larger than those found in njl models , the corresponding profiles of @xmath8 are qualitatively similar in all such approaches , apart from shifting the @xmath16-scale @xcite . the @xmath89 njl and pnjl model calculations yield values of @xmath88 that are usually higher than the transition temperature determined in recent @xmath90 lattice qcd computations @xcite . since one of the aims of the present paper is to explore the systematics of quark mass effects on dilepton and photon production rates , we allow variations of @xmath88 between 150 and 180 mev , and we also examine the influence of different slopes of the chiral crossover controlled by the parameter @xmath91 . four parameter sets will be used , as given in table [ tab : parameter ] . the constituent quark mass @xmath92 at @xmath93 will be kept fixed ( at @xmath94 mev ) for all sets a - d. set b with @xmath95 mev is motivated by the chiral transition temperature @xmath96 mev found in the lattice qcd computations of refs . sets c and d with different values for @xmath91 are used in order to analyze the sensitivity to the steepness of the crossover . with these parameter options the behavior of @xmath8 in both njl and pnjl models can be covered , their main difference being a shift in @xmath88 . the temperature dependent quark masses @xmath8 with these parameter sets are shown in fig . [ fig : mass - various ] . for later convenience , we also plot the thermal quark mass in the deconfined phase ( with @xmath50 ) : @xmath97 with @xmath15 determined by eq . ( [ eq : running - coupling ] ) . the dynamically generated quark mass tends to exceed the thermal mass at temperature @xmath98mev . .parameters ( in mev ) used in the expression ( [ eq : njlmass ] ) for the dynamical quark mass @xmath8 . [ cols="<,^,^,^,^",options="header " , ] ) as function of @xmath16 , evaluated using parameter sets a , b , c , and d. also plotted for comparison is the thermal quark mass , eq . ( [ eq : thermalmass ] ) . , scaledwidth=50.0% ] at leading order of the strong and electromagnetic coupling , the dominant process contributing to the dilepton production rate is the quark - antiquark pair annihilation process followed by pair production of the leptons via the photon . the amplitude of this process factorizes into a part describing the @xmath99 process ( shown in fig . [ fig : dilepton - process ] ) , and a part describing the conversion of the photon into the dilepton @xcite . the differential production rate is is defined so that it is proportional to the phase space element @xmath100 of the lepton pair , and therefore carries mass dimension four . ] @xmath101 where @xmath102 and @xmath103 are the four - momenta of the leptons , @xmath104 , and @xmath105 . here @xmath106 is the wightman photon self - energy , which includes the information about the dynamics of the @xmath99 process . the leptons are on - shell ( @xmath107 neglecting the small lepton masses ) , hence @xmath108 is time - like ( @xmath109 ) . at leading order we have the following expression for @xmath110 : @xmath111 where @xmath112 is a flavour index running from @xmath35 to @xmath2 , and @xmath113 are the quark and antiquark energies , @xmath8 is the temperature - dependent constituent quark mass and @xmath114^{-1}$ ] is the distribution function for anti - quarks @xcite with a sign change of the imaginary chemical potential as compared to + @xmath115^{-1}$ ] for quarks . the squared matrix element is @xmath116 where we have used @xmath117 , and @xmath118 are the electric quark charges in units of @xmath119 for each flavour @xmath112 . using the four parameter sets ( a , b , c , and d ) of table [ tab : parameter ] . the energy @xmath121 is chosen to be 700 mev . the curves connecting the computed points are drawn to guide the eye . , scaledwidth=50.0% ] consider first the case with zero total three - momentum of the lepton pair . when @xmath120 the integration in eq . ( [ eq : dilepton - expression ] ) can easily be performed . the leptons are produced back - to - back . the energies of the quark and anti - quark satisfy @xmath122 . as a result eqs . ( [ eq : dilepton - formal ] ) and ( [ eq : dilepton - expression ] ) yield , because we have assumed a common @xmath16-dependent dynamical quark mass @xmath77 for all ( light ) flavours throughout the present work . for @xmath123 we have @xmath124 . ] @xmath125 in the case @xmath126 the dilepton production rate vanishes for obvious kinematic reasons . the physical interpretation of this expression is evident . it can conveniently be factorized into three parts , coming from the phase space integral , the matrix element squared , and the distribution functions . for @xmath50 ( [ eq : dilepton - result - p=0 ] ) becomes @xmath127 where @xmath128e^{-\beta e}+1\over e^{-2\beta e}+\left[3l(q)-1\right]e^{-\beta e}+1 } ~,\nonumber \end{aligned}\ ] ] come from these three parts , respectively . note that @xmath129 and @xmath130 become unity when chiral symmetry is completely restored ( @xmath17 ) , while @xmath131 when there is no confinement effect ( @xmath56 ) . for @xmath17 our expression reduces to that in one of the author s previous works @xcite . only the parts @xmath129 and @xmath130 depend explicitly on the quark mass @xmath8 . the term @xmath132 depends on the quark and antiquark energy @xmath14 , but this is in turn fixed by the lepton energy , i.e. @xmath133 . we also note that the product @xmath134 is always less than one as long as @xmath135 . this implies that finite @xmath77 always reduces the dilepton production rate . by contrast , the background gluon field @xmath136 enters directly in part @xmath132 through the sum over products of quark and antiquark distributions , @xmath137 . this separation of terms in eq . ( [ eq : dilepton - result - p=0-n=3 ] ) helps making the analysis of @xmath0sb and confinement effects transparentsb as it is realized in the pnjl model , @xmath77 and @xmath53 are not independent quantities . it is nonetheless instructive to analyze the effects of confinement and @xmath0sb on the dilepton production separately , and so it is convenient in the present context to treat @xmath77 and @xmath53 as independent quantities . ] , especially in comparison with previous results of a semi - qgp model calculation @xcite in which the @xmath0sb effect was not considered . the difference is then primarily in the function @xmath134 . the quantities @xmath134 and @xmath132 are shown as functions of @xmath16 in the left panel of fig . [ fig : dilepton - abc ] . in the given example @xmath121 is fixed at 700 mev , where we have used parameter set a for the evaluation of the quark mass @xmath77 . the @xmath134 plot displays the @xmath77 dependence while the latter plot contains information about the dependence on @xmath25 . from the @xmath134 plot one observes that this quantity is significantly smaller than unity in the region @xmath138 mev : the constituent quark mass suppresses the dilepton production rate by modifying the available phase space and the matrix element . when @xmath139 mev @xmath134 quickly approaches unity because around this temperature , @xmath77 decreases rapidly as shown in fig . [ fig : mass - various ] and becomes negligible compared to the energy @xmath140 mev appearing in @xmath134 . from the plot of @xmath132 one observes that the effect of the polyakov loop does not suppress the dilepton production , but even slightly enhances it . this unexpected behaviour has already been discussed previously in refs . @xcite , referring to the fact that the initial state ( the quark - antiquark pair ) is color - singlet . the product @xmath141 indicates the modification of the dilepton production rate compared to the case with @xmath56 and @xmath17 . it shows the combined effect of @xmath77 and the polyakov loop . since @xmath142 , the suppression coming from finite @xmath77 in the region @xmath138 mev is balanced by the enhancement due to the polyakov loop . for @xmath143 180 mev the factor @xmath134 approaches unity and @xmath141 is nearly equal to @xmath132 . with increasing delepton energy @xmath121 the influence of @xmath77 becomes less significant . this is illustrated by comparing the left and right panels of fig . [ fig : dilepton - abc ] . at @xmath144 gev , @xmath134 differs only slightly from unity even when @xmath138 mev . we also examine the dependence of the dilepton production rate on the four different parameter sets summarized in table [ tab : parameter ] . the result for @xmath134 is shown in fig . [ fig : dilepton - fitdependence ] using @xmath145 mev . lowering @xmath1 increases the dilepton rate as the comparison between set b " and set a " results shows . this trend is expected since , as @xmath1 is lowered , @xmath77 becomes smaller in the relevant range and so the suppression effect on dilepton production is reduced . on the other hand , if the chiral crossover transition proceeds with a steeper slope around @xmath1 , the quark mass @xmath77 below the transition temperature tends to be larger and the suppression effect induced by @xmath77 is enhanced , as can be seen from the comparison of the plots for set a " , set c " , and set d " . above the transition temperature the quark mass quickly decreases and the effect is not visible in the figure . for non - vanishing three - momentum of the lepton pair , @xmath146 , using @xmath147 where @xmath148 and @xmath149 is the angle between @xmath150 and @xmath151 , eqs . ( [ eq : dilepton - formal ] ) and ( [ eq : dilepton - expression ] ) give : @xmath152 where @xmath153 kinematics implies that the dilepton production rate vanishes for @xmath154 . performing the integration one arrives at : @xmath155 . \end{split}\end{aligned}\ ] ] in contrast to the @xmath120 case , the effects of the polyakov loop and of the quark mass @xmath77 are not well separated any more . in the limit @xmath17 we reproduce the corresponding expression in ref . @xcite . in the @xmath50 case , we have @xmath156 \tilde{b}(m ) f_{l\overline{l}}(q , m ) , \end{split}\end{aligned}\ ] ] where @xmath157 and @xmath158 , with @xmath159 note that @xmath160 when @xmath17 , and @xmath161 becomes unity when @xmath56 and @xmath17 . the effect of the polyakov loop @xmath25 is analyzed in refs . @xcite . here we focus on the role of the dynamical quark mass , @xmath8 . the quantity @xmath162 , which is the ratio of the dilepton production rate at finite @xmath77 to that at @xmath17 ( @xmath163 ) , is shown in fig . [ fig : dilepton - finitep ] as a function of @xmath164 at three temperatures around @xmath88 ( 140 , 180 , 190 mev ) , and at one temperature ( @xmath165 mev ) that is significantly larger than @xmath88 . in the left panel , @xmath121 is fixed at 1gev and the parameter set a is used . suppression effects on dilepton production due to non - zero quark masses are marginal at low momentum @xmath164 but become significant at higher momenta . complete suppression occurs at the limit , @xmath166 , due to the kinematic constraint . at lower temperature the quark mass @xmath8 is larger than at higher temperature , hence the suppression is more significant . the parameter dependence of the dilepton production rate encoded in @xmath162 is plotted in the right panel of fig . [ fig : dilepton - finitep ] as a function of @xmath164 , at @xmath167 170 mev and @xmath1681 gev , using the parameter sets a , b , c , and d. the differences in @xmath8 translate directly to the momentum ( @xmath169 ) at which @xmath162 vanishes because of the kinematic constraint : @xmath170 , and @xmath171 is larger than @xmath172 . since @xmath173 is determined entirely by @xmath8 , this tendency is understood in the same way as explained in the discussion of the quantity @xmath134 in the previous subsection . at leading order the dominant contributions to the photon production rate come from compton scattering @xmath174 and pair annihilation @xmath175 @xcite . the landau - pomeranchuk - migdal ( lpm ) effect @xcite also contributes at the same order when @xmath17 . in this paper we neglect the contribution from the lpm mechanism . this is justified when @xmath77 is large compared to the thermal quark mass @xmath176 . it is also justified , even when @xmath17 , in the large @xmath18 limit @xcite . the two basic processes , quark - gluon compton scattering and pair annihilation of quarks , are illustrated in fig . [ fig : photon-2to2 ] . the amplitudes of these processes are infrared - singular in the limit of small energy and momentum of the exchanged quark . these singularities are naturally regularized when the constituent quark mass is finite . the resulting photon production rate has a characteristic logarithmic dependence on @xmath77 as will be shown later . even when @xmath17 the thermal quark mass acts to regularize the amplitudes such that they depend logarithmically on @xmath177 . which mass acts prominently as a regulator depends on which one is larger at a given temperature . with the purpose of this paper to investigate finite - mass effects systematically , we calculate the photon production rates with finite dynamical quark masses and compare versus results with @xmath17 but using a resummed quark propagator , which contains the information on the thermal quark mass . consider first the contribution from the compton scattering process . within the kinetic theory , the photon production rate from this mechanism is expressed as @xmath178~ , \end{split}\end{aligned}\ ] ] where the sum is over quark flavours @xmath112 , the spins of all the particles , and color indices ( @xmath179 ) . @xmath180 is the matrix element of the compton scattering process , the explicit expression of which will be given later . we consider a real photon with @xmath181 and focus on the @xmath182 case relevant for high - energy heavy - ion collisions . the boltzmann approximation will be used in the initial state : in eq . ( [ eq : compton - expression ] ) , we replace @xmath183 by @xmath184 $ ] . this is justified when @xmath185 , @xmath186 . it is known that this treatment gives the correct result at leading - log order @xcite . with this approximation eq . ( [ eq : compton - expression ] ) leads to the following expression : @xmath187 here we have introduced the mandelstam variables : @xmath188 and @xmath189 . the variables with tilde are defined as @xmath190 and @xmath191 , and we have introduced @xmath192 and @xmath193 . for details concerning this change of integration variables , see refs . @xcite . the squared matrix element after spin summations is given by : @xmath194 ^ 2 \biggl[\frac{1}{\tilde{s}^2}\left(-\frac{\tilde{s}\tilde{t}}{2}+m^2\tilde{s}+2m^4\right ) \\ & + \frac{1}{\tilde{t}^2}\left(-\frac{\tilde{s}\tilde{t}}{2}+m^2\tilde{t}+2m^4\right ) + \frac{m^2}{\tilde{s}\tilde{t}}\left(\tilde{s}+\tilde{t}+4m^2\right ) \biggr ] , \end{split}\end{aligned}\ ] ] where @xmath195 with @xmath196 encodes the color structure of the quark - gluon vertex in double - line notation . the first term is the contribution from the @xmath197-channel process , the second from @xmath198-channel exchange , and the third term comes from the interference of these two amplitudes , respectively . one finds : @xmath199 where @xmath200 and @xmath201 in the deconfined phase with @xmath202 , this expression becomes : @xmath203 in the limit @xmath204 the logarithmic term in @xmath205 dominates over the other pieces : @xmath206 this expression agrees with the one in leading - log order when regarding @xmath207 as the infrared cutoff @xmath208 used in ref . @xcite . it also suggests that @xmath77 works as an infrared regulator , as was mentioned . in the large @xmath18 limit , eq . ( [ eq : photon - pair - result ] ) becomes @xmath223 inserting @xmath50 , this expression becomes @xmath224 with the dimensionless function @xmath225 ^ 2 \,h(s , m^2 ) . \end{split}\end{aligned}\ ] ] and @xmath226 as functions of @xmath16 with @xmath227 gev . the parameter set a is used . shown for comparison are results obtained in the leading - log approximation ( ll ) and in the completely deconfined phase ( @xmath228 ) . , scaledwidth=50.0% ] and @xmath229 at various temperatures @xmath16 with @xmath227 gev . the parameter set a is used . the solid ( red ) line represents @xmath230 while the dotted ( blue ) line represents @xmath229 . , scaledwidth=50.0% ] we proceed by first analyzing separately the contributions from compton scattering and pair annihilation . in eqs . ( [ eq : compton - result ] ) and ( [ eq : photon - pair - result - n=3 ] ) , the prefactors @xmath231 are common , so we plot the @xmath16-dependent functions @xmath232 and @xmath226 in fig . [ fig : photon - xy ] . the energy is set to @xmath233 gev , and the parameter set a is used to generate @xmath8 . as polyakov loop input a fitting function for the lattice data of @xmath234 , plotted in fig . [ fig : polyakov - q ] , has been used . comparing these quantities with the corresponding ones at the leading - log order ( eqs . ( [ eq : photon - compton - ll ] ) and ( [ eq : photon - pair - ll ] ) at large @xmath18 limit ) , we see that the @xmath16 dependence can be well described by the leading - log approximation when @xmath16 is larger than about 250 mev . at such temperatures @xmath77 becomes negligible which makes the leading - log approximation more accurate . also shown in the same figure are @xmath232 and @xmath226 in the deconfined limit with @xmath235 ( eqs . ( [ eq : photon - compton - deconfine ] ) and ( [ eq : photon - pair - result - deconfine ] ) at large @xmath18 limit ) . one observes that the polyakov loop effect suppresses these quantities significantly at all @xmath16 . note that the contribution from pair annihilation is larger than that from the compton scattering process in the limit @xmath235 , whereas this is not the case when incorporating the polyakov loop . this behaviour derives from the fact that @xmath232 is linear in @xmath39 while @xmath226 involves @xmath236 , which makes the suppression of @xmath226 due to the polyakov loop more significant . the @xmath16-dependent functions @xmath230 and @xmath229 are shown in fig . [ fig : photon - xy - z ] . one observes that @xmath230 is significantly smaller ( larger ) than @xmath229 when the temperature is smaller ( larger ) than about 190 mev , close to the transition temperature @xmath88 of parameter set a. this behavior of @xmath230 can be understood as follows . in the low temperature region , @xmath77 is of order or larger than @xmath16 . from the exponential part of eq . ( [ eq : photon - compton ] ) it follows that the dominant contribution to @xmath232 comes from the region @xmath242 so that the suppression factor @xmath243 results in @xmath232 . as for the pair annihilation contribution , the lower bound of the integral in eq . ( [ eq : photon - pair - result ] ) gives the exponential suppression factor @xmath244 in @xmath226 when @xmath77 is large . on the other hand , when @xmath16 exceeds about 190 mev , @xmath77 becomes small and makes the leading - log approximation ( eqs . ( [ eq : photon - compton - ll ] ) and ( [ eq : photon - pair - ll ] ) ) more accurate . this leads to an enhancement of @xmath230 compared to @xmath229 because of the factor @xmath245 . calculated using the parameter sets @xmath246 as a function of @xmath16 at @xmath227 gev . , scaledwidth=50.0% ] while these considerations remain valid , the detailed high-@xmath16 behavior needs nonetheless to be examined with greater care . when @xmath77 is much smaller than the thermal quark mass , which is realized for @xmath247 mev according to fig . [ fig : mass - various ] , this thermal mass works as an infrared regulator instead of @xmath77 , and thus the role of the leading factor @xmath245 appearing in @xmath230 is now to be replaced by @xmath248 in @xmath229 . at temperatures well above 200 mev it is therefore appropriate to use @xmath229 rather than @xmath230 . in summary , for practical applications we suggest to use @xmath230 for @xmath249 and @xmath229 for @xmath250 , where @xmath251 is the temperature at which @xmath230 equals @xmath229 . next we examine how variations of the temperature @xmath88 and of the steepness @xmath91 of the chiral crossover transition influence the photon production rate . [ fig : photon - parameterdependence ] shows the quantity @xmath230 , using the parameter sets @xmath129 , @xmath130 , @xmath132 , and @xmath252 , as functions of @xmath16 . we focus on the low @xmath16 region ( @xmath249 ) where the results including the dynamical quark mass @xmath77 are expected to be reliable . varying @xmath91 changes the photon production rate only moderately , as can be seen by looking at @xmath230 using the parameter sets @xmath129 , @xmath132 , and @xmath252 with @xmath253 mev . by contrast , varying @xmath88 induces a more significant change in @xmath230 as can be seen by using the parameter set @xmath130 with @xmath254 mev . the shift of the quark mass profile @xmath8 to a lower transition temperature implies an increase of @xmath230 already around @xmath255 where the quark mass starts to drop rapidly . in this work dilepton and photon production rates in the quark - gluon phase close to the transition temperature @xmath1 have been calculated using a model that takes into account effects of confinement and spontaneous chiral symmetry breaking ( @xmath0sb ) , mainly focusing on the latter . implications of @xmath0sb are realised in the form of dynamically generated quark masses as the temperature approaches @xmath1 from above . through its induced kinematic constraint , this effect significantly suppresses the production of dileptons with large three - momenta . for dileptons emitted back - to - back with zero momentum the production rate is reduced more modestly . we have also examined the dependence of the dilepton production rate on the parameters that characterise the chiral crossover transition , suggesting that information about the transition can indeed be encoded in the production rate of dileptons with large three - momentum . in particular variations of the transition temperature imply different spatial momenta at which the dilepton production vanishes as a consequence of phase space suppression . a comparison of our calculated dilepton production rate with the results of other nonperturbative calculations is useful at this point . the dilepton rate ( eq . ( [ eq : dilepton - result - p=0-n=3 ] ) ) for @xmath123 at @xmath120 and @xmath256 mev is plotted as a function of @xmath121 in fig . [ fig : dileptonrate_comparison ] . the parameter set a was used to calculate @xmath8 . note that in this case @xmath257 , the invariant mass of the dilepton , and we use @xmath180 as the variable in this figure . for comparison the results from a pnjl model calculation @xcite , lattice qcd @xcite , and a perturbative calculation are also shown . the perturbative ( born approximation ) result is obtained using eq . ( [ eq : dilepton - result - p=0-n=3 ] ) with @xmath258 . we note that our expression for the dilepton production rate ( eq . ( [ eq : dilepton - result - finitep ] ) ) has the same functional form in terms of @xmath25 and @xmath77 as that ) in our paper by considering the case of zero chemical potential . to confirm this , one needs to rewrite @xmath259 in the two equations in ref . @xcite to @xmath260 by changing the integration variable @xmath108 . ] obtained by using the pnjl model @xcite . this is not surprising since the leading dilepton production amplitude ( @xmath261 ) is the same in both approaches . beyond leading order , loop corrections enter , which make the results depend on details of the model used . at @xmath256 mev the dynamical quark mass @xmath8 , in our approach as well as in the pnjl model , is negligible compared with @xmath16 . the difference in fig . [ fig : dileptonrate_comparison ] between our result and the pnjl model comes solely from the difference of the polyakov loop values used in these calculations . our smaller polyakov loop stems from subtracting the perturbative correction , so that our dilepton production rate is enhanced compared to that of the pnjl model , as explained in ref . the overall enhancement of the dilepton production rate induced by the polyakov loop is also explicit in the fact that the perturbative ( born ) rate is smaller than both the pnjl and semi - qgp results . at small invariant mass @xmath180 one observes that our calculated dilepton rate is significantly lower than the one deduced from lattice qcd . one should note that the lattice results display considerable sensitivity with respect to the ansatz chosen for the form of the vector current spectral function in the low - mass region , as discussed in ref . @xcite : a `` free - field '' ansatz @xcite implies a very small dilepton production rate at small @xmath180 , whereas a `` breit - wigner '' ansatz @xcite yields a large dilepton rate reflecting substantial contributions from hadronic sources . the photon production rate is found to be sensitive to @xmath0sb effects when @xmath16 is less than about 190 mev . we have studied the parameter dependence of this rate . it is found to vary significantly at low @xmath16 around @xmath1 depending on the choice of the transition temperature . comparing our calculated photon rate with corresponding results using other methods is once again of some interest . [ fig : photonrate_comparison ] shows our result , eq . ( [ eq : photon - result ] ) , as a function of the photon energy @xmath14 in comparison with the semi - qgp photon rate @xcite taken at leading - log order and using @xmath17 , and with the perturbative qgp result @xcite . here we have set @xmath262 mev , @xmath89 , and used the parameter set a. the large-@xmath18 limit is taken consistently in all three case studies . the semi - qgp approach at leading - log order and @xmath17 corresponds to eq . ( [ eq : photon - result - m=0 ] ) with @xmath263 while the perturbative qgp result is given with @xmath229 in eq . ( [ eq : z - result - largen ] ) at @xmath264 . as seen in fig . [ fig : photonrate_comparison ] the perturbative photon rate is larger than the semi - qgp rate over the whole energy range . the reason is that both polyakov loop and constituent quark mass effects suppress the photon production as discussed in sec . [ ssc : photon - result ] and ref . @xcite . at the lower energies @xmath265 gev ) our result exceeds the one calculated at leading - log order and with @xmath266 . this difference comes primarily from the term beyond leading - log . at high photon energies the leading - log form is approached asymptotically . indeed , at sufficiently large @xmath14 , the expressions for @xmath232 and @xmath226 can be approximated by their leading - log forms , eqs . ( [ eq : photon - compton - ll ] ) and ( [ eq : photon - pair - ll ] ) , so that @xmath230 equals @xmath229 as given in eq . ( [ eq : z - ll ] ) , except for the replacement @xmath267 . at large @xmath14 , this difference is negligible , and so the photon production rates in the two approaches become approximately equal . a few concluding remarks are in order concerning possible @xmath0sb effects in high - energy heavy - ion collisions ( hic ) . dilepton production in hic is commonly discussed using the invariant mass @xmath180 instead of @xmath164 . with inclusion of dynamical quark masses , the kinematical constraint for dilepton production reads @xmath268 . in the njl model the dynamical quark mass is still quite large ( @xmath269 mev ) at temperatures around @xmath270 mev , so dileptons with @xmath271 mev are forbidden in this temperature range . at slightly higher temperatures where the dynamical quark mass decreases the kinematic constraint is correspondingly less stringent . while such effects can in principle be observed in hic , one recalls , on the other hand , that the low - mass dilepton spectrum with @xmath272 gev results primarily from hadronic currents involving pseudoscalar and vector mesons rather than liberated quark - antiquark pairs @xcite . in this range of invariant masses , dilepton spectra are governed by hadronic sources . an interesting point of discussion is the photon elliptic flow @xmath11 observed at rhic @xcite . as demonstrated in ref . @xcite , the suppression of the photon production rate from the qgp phase due to the confinement effect can increase the total elliptic flow , a welcome feature in order to understand the unexpectedly large @xmath11 for photons @xcite . our results suggest that the effect of @xmath0sb can add to increasing the total photon @xmath11 further at @xmath16 between 150 and 190 mev by its mechanism of suppressing the photon production in that temperature range . of course , other sources of photon production in hic also need to be considered , such as photons coming from the initial state @xcite , from the thermalization process @xcite , and from the hadron phase @xcite . a complete evaluation of the photon @xmath11 requires summing the contributions from all the photon sources in extended hydrodynamic simulations @xcite . at @xmath120 and @xmath256 mev as a function of the dilepton invariant mass @xmath180 . the solid ( red ) line represents our result ( eq.([eq : dilepton - result - p=0-n=3 ] ) ) , the dashed ( green ) line is the result of a pnjl model calculation @xcite , the dotted ( blue ) line is the result from lattice qcd @xcite . the dash - dotted ( magenta ) curve shows the perturbative ( born ) dilepton rate ( using eq.([eq : dilepton - result - p=0-n=3 ] ) with @xmath258 ) . , scaledwidth=50.0% ] at @xmath262 mev as a function of the photon energy @xmath14 . solid ( red ) line : present result ( eq . ( [ eq : photon - result ] ) ) ; dashed ( black ) line : semi - qgp model with @xmath17 and using leading - log approximation @xcite ; dotted ( blue ) line ( labeled qgp with @xmath266 and @xmath273 ) : perturbative qcd calculation including resummations @xcite . , scaledwidth=50.0% ] the authors thank robert lang for providing numerical results of the temperature - dependent njl constituent quark mass . in this appendix we derive the photon production beyond leading - log order in the limit of vanishing quark mass , @xmath266 . the result at leading - log order has been reported previously @xcite , but the derivation beyond this order has not been performed yet . the infrared singularities in the contributions from compton scattering and pair annihilation require introducing a cutoff @xmath241 as a regulator . in principle @xmath241 is an arbitrary quantity , but it should satisfy @xmath274 . the phase space integral appearing in the compton scattering process and the pair annihilation contributions are restricted by the cutoff @xmath241 @xcite : @xmath275 and @xmath276 . in the phase space excluded by @xmath241 , the momentum of the exchanged quark is smaller than @xmath241 , and thus much smaller than @xmath16 . such soft quarks should be treated using the hard thermal loop ( htl ) resummed propagator @xcite , which contains the information of the thermal mass of the quark , so we calculate the resummed one - loop diagram drawn in fig . [ fig : photon - htlresum ] in order to obtain the photon production rate due to the soft momentum exchange . the hard parts from compton scattering and pair annihilation are given by eqs . ( [ eq : compton - expression ] ) and ( [ eq : pair - expression ] ) , setting @xmath17 , and the phase space integration is constrained by the cutoff @xmath241 as explained above . the resulting expressions are as follows @xcite : @xmath277 ^ 2 \\ & ~~~\times \ln\left(1+e^{-\beta({s}/(4e)-iq_b)}\right ) \left(\frac{{t}}{s } + \frac{{s}}{t } \right),\\ \nonumber \left . e\frac{d\vargamma}{d^3{{\mathbf{p}}}}\right|_{\text{pair } } & \simeq -\frac{\alpha\alpha_s}{16\pi^4}\frac{t}{e}e^{-\beta e}\sum_{f , a\sim d}q^2_f \int^\infty_{2\mu^2 } d{s } \int^{-\mu^2}_{-s+\mu^2 } d{t } \\ \nonumber & ~~~\times \frac{1}{s } e^{i\beta(q_a - q_b - q_c+q_d ) } [ ( t^{dc})_{ba}]^2\\ & ~~~\times \ln\left(1-e^{-\beta s/(4e)}e^{i\beta(q_c - q_d)}\right ) \left(\frac{u}{{t } } + \frac{t}{{u}}\right).\end{aligned}\ ] ] here we have used the boltzmann approximation for the initial state . by performing these integrations , we get @xmath278,\\ \nonumber \left . e\frac{d\vargamma}{d^3{{\mathbf{p}}}}\right|_{\text{pair } } & \simeq \frac{\alpha\alpha_s}{4\pi^4}t^2 e^{-\beta e}\sum_{f } q^2_f \sum^\infty_{n=1 } \frac{1}{n^2}\left[({n_c}l_n)^2 - 1\right ] \\ & ~~~\times \left[\ln\left(\frac{4et}{\mu^2}\right)-1-\gamma_e-\ln n\right].\end{aligned}\ ] ] in the large-@xmath18 limit these expressions become @xmath279,\\ \nonumber \left . e\frac{d\vargamma}{d^3{{\mathbf{p}}}}\right|_{\text{pair } } & \simeq \frac{\alpha\alpha_s}{4\pi^4}t^2 e^{-\beta e}\sum_{f } q^2_f { n_c}^2 \sum^\infty_{n=1 } \frac{1}{n^2}(l_n)^2\\ & ~~~\times \left[\ln\left(\frac{4et}{\mu^2}\right)-1-\gamma_e-\ln n\right],\end{aligned}\ ] ] with @xmath50 they reduce to : @xmath280,\\ \nonumber \left . e\frac{d\vargamma}{d^3{{\mathbf{p}}}}\right|_{\text{pair } } & \simeq \frac{\alpha\alpha_s}{4\pi^4}t^2 e^{-\beta e}\sum_{f } q^2_f \\ \nonumber & ~~~\times\bigl[\left(\ln\left(\frac{4et}{\mu^2}\right)-1-\gamma_e\right)\frac{3\pi^2}{2}\\ \nonumber & ~~~\times\left(8q^2-\frac{16}{3}q+1\right)\\ \nonumber & ~~~-2.8127 -2\sum^\infty_{n=1 } \frac{1}{n^2}\\ \label{eq : photon - pair - largen - n=3 } & ~~~\times\left(2\cos(2\pi nq)+\cos(4\pi nq)\right)\ln n\bigr],\end{aligned}\ ] ] by using eq . ( [ eq : ln - q ] ) , and @xmath281 where @xmath282 is the glaisher constant . on the other hand , the contribution from the soft part reads @xcite @xmath283\\ & ~~~\times{\mathrm{tr}}\left [ \rho^*_a(\omega,{{\mathbf{k } } } ) \rho(\omega - e,{{\mathbf{k}}}-{{\mathbf{p } } } ) \right ] , \end{split}\end{aligned}\ ] ] which corresponds to the one - loop diagram drawn in fig . [ fig : photon - htlresum ] . the summation includes the color indices @xmath33 . here the bare ( htl - resummed ) quark spectral functions are : @xmath284 where @xmath285 is the sign function . the htl spectral function after decomposition reads @xmath286\\ & ~~~+\theta(|{{\mathbf{k}}}|^2-\omega^2)\rho^{\text{spacelike}}_{a\pm}(\omega,{{\mathbf{k } } } ) \end{split}\end{aligned}\ ] ] here @xmath287 is the residue , @xmath288 the dispersion relation , and @xmath289 is the thermal quark mass . their respective expressions are : @xmath290,\end{aligned}\ ] ] where @xmath291 . the explicit form of @xmath292 is irrelevant in our analysis , so we do not write it here . the thermal quark mass is given by @xcite @xmath293 , \end{split}\end{aligned}\ ] ] where @xmath294 . at large @xmath18 , the last term is neglected . the production rate is reduced to the following form @xcite : @xmath295 . \end{split}\end{aligned}\ ] ] the integration in the equation above does not depend on @xmath33 , and its numerical value approximately @xmath296 @xcite . the final result in the large @xmath18 limit is given by the sum of the compton , pair annihilation , and soft parts , namely eqs . ( [ eq : photon - compton - largen - n=3 ] ) , ( [ eq : photon - pair - largen - n=3 ] ) , and ( [ eq : photon - soft - largen - n=3 ] ) , when @xmath50 . a. bazavov , t. bhattacharya , m. cheng , c. detar , h. t. ding , s. gottlieb , r. gupta and p. hegde _ et al . _ , phys . d * 85 * , 054503 ( 2012 ) [ arxiv:1111.1710 [ hep - lat ] ] . s. borsanyi , g. endrodi , z. fodor , a. jakovac , s. d. katz , s. krieg , c. ratti and k. k. szabo , jhep * 1011 * , 077 ( 2010 ) [ arxiv:1007.2580 [ hep - lat ] ] . a. bazavov , t. bhattacharya , m. cheng , n. h. christ , c. detar , s. ejiri , s. gottlieb and r. gupta _ et al . _ , phys . d * 80 * , 014504 ( 2009 ) [ arxiv:0903.4379 [ hep - lat ] ] . y. nambu and g. jona - lasinio , phys . 122 * , 345 ( 1961 ) ; phys . * 124 * , 246 ( 1961 ) ; u. vogl and w. weise , prog . part . * 27 * , 195 ( 1991 ) ; t. hatsuda and t. kunihiro , phys . rept . * 247 * , 221 ( 1994 ) [ hep - ph/9401310 ] . k. fukushima , phys . b * 591 * , 277 ( 2004 ) [ hep - ph/0310121 ] ; c. ratti , m. a. thaler and w. weise , phys . rev . d * 73 * , 014019 ( 2006 ) [ hep - ph/0506234 ] ; k. fukushima , phys . d * 77 * , 114028 ( 2008 ) [ erratum - ibid . d * 78 * , 039902 ( 2008 ) ] [ arxiv:0803.3318 [ hep - ph ] ] ; t. hell , s. roessner , m. cristoforetti and w. weise , phys . d * 79 * , 014022 ( 2009 ) [ arxiv:0810.1099 [ hep - ph ] ] . g. vujanovic , c. young , b. schenke , r. rapp , s. jeon and c. gale , phys . c * 89 * , no . 3 , 034904 ( 2014 ) [ arxiv:1312.0676 [ nucl - th ] ] . c. gale , y. hidaka , s. jeon , s. lin , j. f. paquet , r. d. pisarski , d. satow and v. v. skokov _ et al . _ , phys . lett . * 114 * , no . 7 , 072301 ( 2015 ) [ arxiv:1409.4778 [ hep - ph ] ] . a. adare _ et al . _ ( phenix collaboration ) , phys . lett . * 109 * , 122302 ( 2012 ) . s. turbide , r. rapp and c. gale , phys . c * 69 * , 014903 ( 2004 ) [ hep - ph/0308085 ] . b. schenke , s. jeon and c. gale , phys . c * 82 * , 014903 ( 2010 ) [ arxiv:1004.1408 [ hep - ph ] ] . b. schenke , s. jeon and c. gale , phys . lett . * 106 * , 042301 ( 2011 ) [ arxiv:1009.3244 [ hep - ph ] ] . p. aurenche , m. fontannaz , j. p. guillet , e. pilon and m. werlen , phys . d * 73 * , 094007 ( 2006 ) [ hep - ph/0602133 ] . h. van hees , c. gale and r. rapp , phys . c * 84 * , 054906 ( 2011 ) [ arxiv:1108.2131 [ hep - ph ] ] ; g. basar , d. kharzeev , d. kharzeev and v. skokov , phys . lett . * 109 * , 202303 ( 2012 ) [ arxiv:1206.1334 [ hep - ph ] ] ; a. bzdak and v. skokov , phys . lett . * 110 * , no . 19 , 192301 ( 2013 ) [ arxiv:1208.5502 [ hep - ph ] ] ; f. m. liu and s. x. liu , phys . c * 89 * , no . 3 , 034906 ( 2014 ) [ arxiv:1212.6587 [ nucl - th ] ] ; o. linnyk , v. p. konchakovski , w. cassing and e. l. bratkovskaya , phys . c * 88 * , 034904 ( 2013 ) [ arxiv:1304.7030 [ nucl - th ] ] ; b. muller , s. y. wu and d. l. yang , phys . d * 89 * , no . 2 , 026013 ( 2014 ) [ arxiv:1308.6568 [ hep - th ] ] ; o. linnyk , w. cassing and e. l. bratkovskaya , phys . c * 89 * , no . 3 , 034908 ( 2014 ) [ arxiv:1311.0279 [ nucl - th ] ] ; a. monnai , phys . rev . c * 90 * , no . 2 , 021901 ( 2014 ) [ arxiv:1403.4225 [ nucl - th ] ] ; h. van hees , m. he and r. rapp , nucl . a * 933 * , 256 ( 2015 ) [ arxiv:1404.2846 [ nucl - th ] ] . y. hidaka , s. lin , r. d. pisarski and d. satow , arxiv:1504.01770 [ hep - ph ] . y. hidaka and r. d. pisarski , phys . d * 78 * , 071501 ( 2008 ) [ arxiv:0803.0453 [ hep - ph ] ] ; phys . d * 81 * , 076002 ( 2010 ) [ arxiv:0912.0940 [ hep - ph ] ] . e. gava and r. jengo , phys . b * 105 * , 285 ( 1981 ) ; n. brambilla , j. ghiglieri , p. petreczky and a. vairo , phys . d * 82 * , 074019 ( 2010 ) [ arxiv:1007.5172 [ hep - ph ] ] . y. burnier , m. laine and m. vepsalainen , jhep * 1001 * , 054 ( 2010 ) [ erratum - ibid . * 1301 * , 180 ( 2013 ) ] [ arxiv:0911.3480 [ hep - ph ] ] . k. kajantie , m. laine , k. rummukainen and m. e. shaposhnikov , nucl . phys . b * 503 * , 357 ( 1997 ) [ hep - ph/9704416 ] . a. c. aguilar and j. papavassiliou , phys . d * 83 * , 014013 ( 2011 ) [ arxiv:1010.5815 [ hep - ph ] ] ; d. binosi , l. chang , j. papavassiliou and c. d. roberts , phys . b * 742 * , 183 ( 2015 ) [ arxiv:1412.4782 [ nucl - th ] ] . l. d. mclerran and t. toimela , phys . d * 31 * , 545 ( 1985 ) ; h. a. weldon , phys . rev . d * 42 * , 2384 ( 1990 ) ; c. gale and j. i. kapusta , nucl . b * 357 * , 65 ( 1991 ) . g. staadt , w. greiner and j. rafelski , phys . d * 33 * , 66 ( 1986 ) . r. baier , h. nakkagawa , a. niegawa and k. redlich , z. phys . c * 53 * , 433 ( 1992 ) . j. i. kapusta , p. lichard and d. seibert , phys . d * 44 * , 2774 ( 1991 ) [ erratum - ibid . d * 47 * , 4171 ( 1993 ) ] . p. aurenche , f. gelis and h. zaraket , phys . d * 62 * , 096012 ( 2000 ) [ hep - ph/0003326 ] . p. b. arnold , g. d. moore and l. g. yaffe , jhep * 0111 * , 057 ( 2001 ) [ hep - ph/0109064 ] ; jhep * 0112 * , 009 ( 2001 ) [ hep - ph/0111107 ] ; jhep * 0206 * , 030 ( 2002 ) [ hep - ph/0204343 ] . g. baym , j. -p . blaizot , f. gelis and t. matsui , phys . b * 644 * , 48 ( 2007 ) [ hep - ph/0604209 ] . c. a. islam , s. majumder , n. haque and m. g. mustafa , jhep * 1502 * , 011 ( 2015 ) [ arxiv:1411.6407 [ hep - ph ] ] . r. d. pisarski , phys . lett . * 63 * , 1129 ( 1989 ) ; e. braaten and r. d. pisarski , phys . lett . * 64 * , 1338 ( 1990 ) ; nucl . b * 337 * , 569 ( 1990 ) ; phys . d * 42 * , 2156 ( 1990 ) ; phys . d * 46 * , 1829 ( 1992 ) ; r. kobes , g. kunstatter and k. mak , phys . d * 45 * , 4632 ( 1992 ) . |
we review the next generation japanese infrared space mission , astro - f . astro - f will be the first survey of the entire sky at infrared wavelengths since the iras mission almost 20 years ago . astro - f will survey the entire sky in 4 far - infrared bands from 50 - 200microns and 2 mid - infrared bands at 9 and 20microns to sensitivities of 10 - 1000 times deeper than the iras satellite at angular resolutions of 25 - 45arcsec ( c.f . iras 2 - 5arcmins ) . astro - f can be considered a super - iras . using the galaxy evolution model of pearson ( 2001 ) we produce expected numbers of sources under 3 different cosmological world models . we predict that astro - f will detect of the order of 10 s millions of sources in the far - infrared wavelength bands , most of which will be dusty lig / uligs of which as many as half will lie at redshifts greater than unity . we produce number - redshift distributions , flux - redshift and colour - colour diagrams for the survey and discuss various segregation and photometric redshift techniques . furthermore , we investigate the large scale structure scales that will be accessed by astro - f , discovering that astro - f and sirtf - swire probe both different scales and redshift domains and concluding that the 2 missions will supplement rather than supplant one another . [ firstpage ] cosmology : source counts infrared : source counts , surveys galaxies : evolution . | astro-ph0210292 | in order to understand such fundamental cosmological questions such as type dependent galaxy evolution , star formation history , and the large scale structure of universe , large data sets of homogeneous data are required . most recently , these large data sets have been mainly accumulated at nir / optical wavelengths ( e.g. the sloan digital sky survey , sdss , @xmath0 galaxies over 1/4 of the sky , @xcite , the 2 degree field galaxy redshift survey , 2dfgrs , @xmath1 galaxies over @xmath22000sq.deg . of the sky @xcite ) . to investigate galaxy evolution and starformation in the high redshift z@xmath31 universe , due to finite observation times , a trade off must be made between the area and the depth of a given survey , that is to say , the greater the sensitivity or depth required , the narrower the survey has to be , e.g. the hubble and subaru deep fields , williams et al . , maihara et al . . hence , although the two hdf fields have produced excellent images and some @xmath41000 spectroscopic redshifts with @xmath330 galaxies at z@xmath32 ( cohen et al . , dennefeld , they are effectively limited to 2 lines of sight ( hdf - n , 5.3sq.arcmin . , williams et al . and hdf - s , 0.7sq.arcmin . , williams et al . , gardner et al . ) and are therefore somewhat susceptible to statistical uncertainty ( e.g. from large scale structure ) . for example , almost one quarter of the total [email protected] 0.65@xmath6 m luminosity comes from just 4 galaxies . in the past few years considerable progress has been made in the understanding of the star formation rate of high redshift galaxies using the lyman drop out technique to isolate and identify galaxies at redshifts of @xmath22@xmath44 ( steidel et al . , madau et al . ) . well over 500 lyman - break systems have now been observed at z@xmath33 . these are uv - bright , blue spectra , actively star - forming galaxies . estimations of star formation rates of a few to @xmath7 led to the popular madau plot @xcite and the somewhat premature implication that the star formation rate of the universe peaked at z@xmath21@xmath42 . however , the early analysis of the optical / uv star formation rate neglected the effects due to dust . subsequent analysis correcting for extinction did indeed imply corrections to the star formation rates of lyman - break galaxies of factors of 2 - 7 ( pettini et al . , dickinson , meurer et al . , calzetti & heckman , steidel et al . , goldader et al . ) . indeed , studies with the infrared space observatory , iso @xcite , of the hubble deep fields have revealed star formation rates at least comparable to or higher than those of the optical / uv studies ( rowan - robinson et al . , mann et al . ) . at sub - millimetre wavelengths , surveys with scuba on the jcmt @xcite revealed a large ( @xmath33000/sq.deg . at @xmath8 ) strongly evolving population of sources with bolometric luminosities @xmath9 and star formation rates of @xmath10 lying at redshifts @xmath31 ( smail et al . , hughes et al . , barger et al . , blain et al . , barger et al . , dunne et al . fox et al . ) . the straight forward interpretation of these results is that most massive star formation occurs in dense molecular clouds that obscure the view at optical wavelengths . what is seen at optical / uv wavelengths represents star formation at the periphery of the clouds and is but the tip of the iceberg . of course this revelation is not new . the first significant evidence came from the iras satellite , launched in 1983 , which surveyed almost the entire sky in four wavebands , 100@xmath6 m , 60@xmath6 m , 25@xmath6 m & 12@xmath6 m , detecting more than 25,000 galaxies and providing a huge legacy of data . iras revealed a new population of galaxies in the ir indicating that most of the star formation history had in fact been missed by the optical observations . the local ratio of @xmath11 was found to vary from @xmath12 for normal quiescent galaxies ( rowan - robinson et al . , corbelli , salpeter & dicky to as much as @xmath13 for luminous @xmath14 and ultraluminous @xmath15 galaxies ( e.g. soifer et al . . further evidence of a darker side to star formation has come from measurements , detections and upper limits of the integrated extragalactic background light at 2.2@xmath6 m , 3.5@xmath6 m , 140@xmath6 m and 240@xmath6 m to 2000@xmath6 m by the cobe , iso and irts missions ( puget et al . , fixsen et al . , hauser et al . , lagache et al . , lagache & puget , gorijian , wright & chary , matsumoto et al . , wright & reese , finkbeiner et al . ) . these data have indicated that the far - infrared background is comparable to the optical near - infrared ( nir ) background , i.e. that 50@xmath16 of the integrated rest frame optical / uv emission from hot , young stars is reprocessed by dust and re - radiated at mid - infrared ( mir ) and far - infrared ( fir ) wavelengths . therefore the rest frame optical / uv luminosities of galaxies only contribute a lower limit to the star formation history of the universe @xcite . in order to tightly constrain general mass evolution of galaxies and the star formation history of the universe , large area , wide field infrared surveys to redshifts 1@xmath42 are required . for example , although the optical / uv studies of lyman - break galaxies sample large volumes at _ high _ redshift ( the volume encompassed at 2@xmath5z@xmath510 is 20x the volume at z@xmath51 , @xmath17 ) most cosmological evolution takes place at lower redshifts . the iso and sirtf @xcite missions are observatories making surveys of the order of 10 s sq.deg .. the iras mission produced a pioneering survey of the infrared sky but could only reach out to relatively low redshifts ( e.g. [email protected] for the pscz , @xcite ) and is now more than 20 years old . to extend unbiased measurements of the star formation rate out to higher redshifts a new all sky infrared survey is required . astro - f , also known as the infra - red imaging surveyor ( iris , @xcite ) will be the second infrared astronomy mission of the japanese institute of space and astronautical science ( isas ) . astro - f is a 67 cm cooled telescope and will be dedicated for all sky and large area surveys in the infrared , one of its primary objectives being to map the entire sky in 4 fir bands ranging from 50 - 200@xmath6 m . with its timely launch almost on the 20th anniversary of the iras mission astro - f can indeed be regarded as a _ super - iras_. in this paper , we review the astro - f mission its objectives and predicted results ( see takeuchi et al . for an earlier discussion of the astro - f mission ) . note that although astro - f will also provide equally valuable information in the areas of galactic star - formation , planetary systems and innumerable other fields of galactic astronomy and astrophysics , here we will concentrate solely on and its extragalactic all sky survey , as this is one of the primary motivations behind the mission . in section [ sec : astro - f ] we introduce the astro - f mission and the infrared all sky survey . section [ sec : model ] explains the model parameters used to predict the survey predictions . in section [ sec : survey ] we present the predictions of the all sky survey and in section [ sec : colours ] we discuss the simple methods of colour discrimination of sources in the all sky survey . a summary of results and the conclusions are given in section [ sec : conclusions ] . the astro - f infrared space telescope @xcite incorporates a 67 cm diameter silicon carbide ritchey - chretien design telescope ( 11 kg primary mirror ) cooled to 5.8k ( the detectors to [email protected] ) by @xmath2170l of liquid helium contained within a light weight cryostat @xcite . the temperature of the outer wall of the cryostat is expected to be maintained below 200k via radiation cooling and a pair of 2-stage stirling - cycle coolers ensure minimum heat flow from the outer wall of the cryostat , thus almost doubling the lifetime of the helium providing a mission lifetime of more than 550 days . furthermore , the use of mechanical coolers means that the near - infrared detectors will still be usable even after helium exhaustion . astro - f is scheduled to be launched in mid 2004 into a sun - synchronous polar orbit with an altitude of 750 km with isas s m - v launch vehicle . @lllll filter & wavelength band & array size & pixel size ( diffraction beam ) & @xmath19 sensitivity + n60 & @xmath20 m & @xmath21x@xmath22 & @xmath23 & @xmath24mjy + n170 & @xmath25 m & @xmath26x@xmath22 & @xmath27 & @xmath28mjy + wide - s & @xmath29 m & @xmath21x@xmath30 & @xmath31 & @xmath21mjy + wide - l & @xmath32 m & @xmath26x@xmath30 & @xmath33 & @xmath34mjy + + [ fis ] @lllll filter & wavelength band & fov & spatial ( in pixel ) resolution & @xmath19 sensitivity + irc - mir - s9w & @xmath35 m & @xmath36x@xmath36 & @xmath37x@xmath38 & @xmath39mjy + irc - mir - l20w & @xmath40 m & @xmath36x@xmath36 & @xmath37x@xmath38 & @xmath41mjy + + [ irc ] astro - f incorporates 2 focal plane instruments covering the entire ir region from the k - band to 200@xmath6 m . the first focal plane instrument is the far - infrared surveyor ( see table [ fis ] @xcite @xcite ) which will survey the entire sky simultaneously in 4 fir bands from 50 to 200@xmath6 m with angular resolutions of 25 - 45 arcsec . the fis has 2 kinds of detector arrays ; a short wavelength array ( 50 - 110@xmath6 m ) comprised of unstressed ge : ga detectors and a long wavelength array ( 110 - 200@xmath6m)comprised of stressed ge : ga detectors . these 2 arrays are divided into 2 further narrow and wide bands known as n60 , wide - s , n170 , wide - l respectively ( see table [ fis ] ) . the two arrays are tilted against the scan direction by 26.5deg . such that the interval between the scan paths traced by the detector pixels is half of the physical pitch of the pixels . the long and short wavelength arrays observe almost the same area of sky , the only difference being due to the difference in the size of the arrays . as well as the all sky survey mode the fis may also be used for pointing observations and includes a fourier - transform spectrometer mode that can be operated via the wideband arrays with a resolution of @xmath42 allowing imaging spectroscopy of selected sources . in survey mode , astro - f will perform a continuous scan of the sky . astro - f spins around the sun pointed axis once every orbit of 100 minutes , keeping the telescope away from the earth . the result is to trace out a great circle with a solar elongation of 90deg scanning the sky at a rate of 3.6arcmin / sec . the second focal - plane instrument is the infrared camera ( irc , @xcite,@xcite , @xcite ) . the irc employs large - format detector arrays and will be used to take deep images of selected sky regions in the near - infrared and mid - infrared range after the all sky survey has been performed . the irc has 3 cameras each with 3 filters . the irc - nir has bands in k , l , m plus a grism and prism from 2.5 - 5@xmath6 m & 2.0 - 5@xmath6 m respectively . the irc - mir - s has narrow bands at 7@xmath6 m & 11@xmath6 m and a wide band at 9@xmath6 m , plus 2 grisms at 5 - 8@xmath6 m & 7 - 12@xmath6 m . the irc - mir - l has narrow bands at 15@xmath6 m & 24@xmath6 m , a wide band at 20@xmath6 m , plus 2 grisms covering the range 11 - 19@xmath6 m & 18 - 26@xmath6 m . pearson et al . made detailed predictions for small , deep and large , shallow area survey strategies using the irc instrument in pointing mode . the possibility of adding the 2 wide band channels of the irc - mir ( s9w , s20w)in scan mode is currently under investigation @xcite . in all sky survey mode the irc would utilize just one line ( perpendicular to the scan direction of the telescope beam ) of the mid - ir detector arrays of the mir - s and mir - l channels . data from these channels would be sampled and downloaded to the ground simultaneously with the fis all - sky survey data . typical angular resolutions of @xmath410are expected ( since integrated signal will be read out every 10 . a significant portion ( but not all ) of the sky could be surveyed . in this paper we offer a preliminary analysis of an additional simultaneous all sky survey using these 2 detector channels . the specifications of the irc all sky survey mode are shown in table [ irc ] . the astro - f all sky survey will be far superior to that of the iras survey . optimally , astro - f will have almost 2 orders of magnitude better sensitivity at 100@xmath6 m . the detection limits for pointing observations in the near - mid - infrared are expected to be @xmath43 and 10 - 100mjy in the far - infrared ( c.f . @xmath40.5 - 1.5jy for iras at 12 , 25 , 60@xmath6 m & 100@xmath6 m respectively @xcite ) . astro - f will also offer 4 times the resolution over 10 times the detector area compared to the isocam instrument on iso . spatial resolution with iras was limited to 2 - 5arcmin . astro - f will attain spatial resolutions of 50 - 70arcsec at wavelengths of 50 - 200@xmath6 m respectively . [ sensitivity ] plots the sensitivities expected for the fis all sky survey in table [ fis ] against the seds of two sources from the the canada - u.k . deep scuba survey ( cudss , lilly et al . , sed fits from takagi et al ): cudss10a and cudss03a , which have redshifts of 0.55 & @xmath41.6 and 850@xmath6 m fluxes of 4.8mjy & 3.4mjy respectively . both of these sources would be detected in the all sky survey ( cudss03a in only the 2 most sensitive fis bands ) and both sources would be easily detectable in all the 9 bands of the irc instrument in pointing mode ( see pearson et al . for irc sensitivities in pointing mode ) . nasa s sirtf mission @xcite , due for launch in early 2003 , offers similar sensitivities to astro - f . however these two missions should be seen as complementary rather than rivalling each other . astro - f will be able to reach such detection limits over a wider area ( 10 - 20sq . deg . due to the larger fov , c.f . , the sirtf - swire survey @xcite ) in the mid - infrared and the entire sky at far - infrared wavelengths , providing the first all sky survey at 170@xmath6 m . astro - f will be sensitive to larger scales at lower redshifts while sirtf while conversely be sensitive to smaller scales at higher redshifts . astro - f is a surveyor , while sirtf is an observatory ( a similar comparison may be made between the iras and iso missions ) . at 2.5 - 5 years , sirtf has a longer projected lifetime than astro - f , although it has no capability at near - infrared wavelengths ( @xmath44 ) . astro - f has this added capability and will in fact be able to continue to use this instrument even after liquid helium exhaustion . at present , the astro - f mission can be divided into 4 distinct phases . 1 . _ phase 0 - performance verification _ - @xmath460 days - initial check out and calibration of telescope and instruments . a mini - survey may also be attempted . phase 1 - all sky survey _ - @xmath4180 days - astro - f will scan the entire ecliptic longitude in survey mode . the all sky survey will have first priority although there may be as many as 1500 independent irc pointings and @xmath4300 fis pointings planned . phase 2 - pointing observations _ - @xmath4300 days - any supplemental survey observations will be performed . the rest of this phase will be dedicated to pointing observations with the irc and fis instruments . this phase will last until helium exhaustion during which @xmath46000 pointings are expected . _ phase 3 - helium boil off _ - @xmath3365 days ? - even after the liquid helium has been exhausted the mechanical coolers will keep the temperature low enough to enable the continuation of observations using the irc - nir detector . in this phase there is hope to be as many as 10500 pointings . the galaxy evolution models of pearson ( hereafter cpp ) are used to estimate the source counts , the number of sources detected and other observables such as number - redshift distributions and expected galaxy colours in the astro - f all sky survey . the cpp model is an improved extension of the original model of pearson and rowan - robinson model that provided an extremely good fit to iras data . the model incorporates a 4 component parameterization segregated by iras colours of galaxies @xcite consisting of normal , starburst , ultraluminous galaxies @xcite , defined at 60@xmath6 m and an agn population defined at 12@xmath6 m . the cool ( high s(100@xmath6m)/s(60@xmath6 m ) ) and warm component ( low s(100@xmath6m)/s(60@xmath6 m ) ) 60@xmath6 m luminosity functions of saunders et al . are used to represent the normal , starburst & ulig galaxies respectively , the uligs comprising the high luminosity tail of the warm luminosity function ( @xmath45 ) . the agn population utilizes the 12@xmath6 m luminosity function of lawrence et al . modelled on the sample of rush et al . ( this iras selected sample is consistent with the iso subsample of spinoglio et al . ) . spectral energy distributions are required both for the k - corrections of sources and to transform the luminosity functions from their rest wavelength to the observation wavelength ( i.e. by a factor @xmath46 where @xmath47 is the wavelength at which the luminosity function is defined and @xmath48 is the wavelength of observation . these model spectra templates are taken from the new radiative transfer models of efstathiou , rowan - robinson & siebenmorgen and efstathiou & rowan - robinson for the starburst & ultraluminous galaxy components and the normal galaxy component respectively . the agn spectral template is represented by the dust torus model of rowan - robinson . the cpp model assumes that the starburst , ulig and agn components evolve with cosmic time . the normal galaxy population does not evolve . the starburst and agn components undergo positive luminosity evolution of the form @xmath49 to a redshift of @xmath50 @xcite which provides a good fit to the source counts , at least at brighter fluxes from the radio to the x - ray regime ( e.g. oliver et al . , serjeant et al . , pearson et al . , pearson & rowan - robinson , benn et al . , boyle et al . ) . however , the subsequent strong evolution evident in the galaxy population at fir wavelengths as detected by isophot on iso ( dole et al . , efstathiou et al . ) , at mir wavelengths with isocam ( especially at 15@xmath6 m where the source counts span 2 orders of magnitude in flux ( oliver et al . , flores et al . , altieri et al . , aussel et al . , gruppioni et al . , biviano et al . , elbaz , serjeant et al . ) and in the sub - mm with scuba on the jcmt ( hughes et al . , smail et al . , scott et al . ) has introduced the need for some extreme evolutionary scenarios in galaxy evolution models . therefore , the ulig population is assumed to undergo extreme evolution in both density and luminosity ( see pearson for details of these evolutionary models ) . the density evolution is of the form @xmath51}$ ] with @xmath52 , @xmath53 , @xmath54 to @xmath55 and decaying quickly towards higher redshifts . the luminosity evolution is of the form @xmath56}$ ] with @xmath57 , @xmath58 , @xmath59 to @xmath55 and decaying quickly to higher redshifts . in this scenario the luminosity evolution corresponds to the accretion of matter onto a core initiating star - formation at high redshifts . the density evolution corresponds to major mergers at a later epoch . locally , uligs have a space density comparable to optically selected qsos ( @xmath60 ) , although analysis of the iras faint source catalogue leads to the conclusion that they should be much more numerous at higher redshift @xcite . moreover both the iso 15@xmath6 m differential source counts @xcite and the dust enshrouded star formation rate @xcite , @xcite exhibit a peak due sources at [email protected] . the evolutionary models of cpp provide an excellent fit to the source counts , number - redshift distributions and the integrated background radiation from sub - mm to nir wavelengths ( pearson and references there in ) . although the fir sed of galaxies is relatively featureless , the spectral templates are convolved with each individual fis filter bandwidth to achieve smoothed model spectral templates for all components in all astro - f fis wavebands . the seds are also convolved with the 2 mir wide bands . this convolution is more significant as unidentified infrared bands ( uibs probably due to pah features in the galaxy seds @xcite , @xcite , @xcite , @xcite ) may significantly affect any mir observations @xcite @xcite @xcite . indeed , the equivalent widths of these unidentified bands can be as much as 10@xmath6 m , comparable to the astro - f irc - mir wide filter band passes of s9w(6 - 11.5@xmath6 m ) and l20w(14 - 26@xmath6 m ) ( see table [ irc ] ) . following growing consensus between observation and theory ( e.g. balbi et al . , freedman et al . , kravtsov et al . ) , the model assumes @xmath61 unless otherwise explicitly stated . the iras all sky survey probed out to redshifts of @xmath20.03 - 0.2 , astro - f is hoped to extend this range out to redshift @xmath621 . in order to assess the visibility of sources in the astro - f all sky survey , we plot in fig . [ lz ] the luminosity - redshift distributions for each model galaxy component at each all sky survey wavelength . in essence fig . [ lz ] plots the rest 60@xmath6 m luminosity that a source must have if it is to be detected in the survey in a given wavelength band with limiting flux _ s_. we see that the normal galaxy population is quickly shifted beyond @xmath63 ( the arbitrary limit for the normal galaxy population in cpp ) at redshifts of @xmath40.3 & 0.8 in the n60 & n170 bands respectively , i.e. there will be no high redshift ( z@xmath31 ) normal galaxies in the sample . however , the starburst population ( starbursts , ligs , uligs ) enjoy a stronger _ negative _ k - correction in the longer wavelength bands as the band pass climbs the rayleigh - jeans slope towards the dust peak around 170 - 60@xmath6 m in the observed frame . thus a @xmath64ulig ( corresponding to a total fir luminosity of @xmath65x@xmath66 @xcite ) should be detectable out to redshifts of z@xmath41 in the long wavelength bands . similarly the agn have an almost flat luminosity curve out to redshifts of 2.5 - 3 as the wavelength band pass slowly move up the long wavelength sed towards the peak at around 30@xmath6 m . figure [ lz ] also emphasises the constraints placed on the mid - ir all sky survey by their relatively low sensitivity . at mid - ir wavelengths almost all the sources would be expected at z@xmath671 except for the agn that have high luminosities and are also enjoying a sampling wavelength closer to the peak in their sed . in summary , from fig . [ lz ] we may broadly expect 4 regimes of increasing redshift encompassing the normal , starburst , ulig and agn populations respectively . @lccccc band & flux & confusion source density & + & 5@xmath68 mjy & ( number of sources ) & @xmath69 & @xmath70 & @xmath71 + @xmath72 & @xmath24 & @xmath73 & @xmath74 & @xmath74 & @xmath74 + @xmath75 & @xmath28 & @xmath41 & @xmath76 & @xmath77 & @xmath78 + @xmath79 & @xmath21 & @xmath80 & @xmath21 & @xmath21 & @xmath81 + @xmath82 & @xmath34 & @xmath83 & @xmath84 & @xmath85 & @xmath86 + @xmath87 & @xmath39 & @xmath88 & @xmath89 & @xmath89 & @xmath90 + @xmath91 & @xmath41 & @xmath92 & @xmath93 & @xmath93 & @xmath94 + [ confuse ] the astro - f all sky survey will provide a huge database for the statistical study of galaxies . the zeroth order moment of this statistical data base ( for a complete sample of sources ) is simply the number counts of detected sources . integral galaxy number counts as a function of flux will indeed form the _ front line _ of the results from the astro - f mission . [ counts ] shows the predicted source counts calculated from the model explained in section [ sec : model ] . numbers for the all sky survey are calculated using 3 different cosmologies assuming a hubble constant of @xmath95 , a flat @xmath17 universe , an open @xmath96 universe and a flat @xmath97 universe ( note , changing the value of the hubble constant will have no effect on the source counts ) . the source counts in the n60 and n170 narrow bands can be compared directly with observations by iras and iso respectively whilst counts in the wide far - ir bands will probe down to and below the confusion limit due to point sources . in general , the ultimate sensitivity for any far - ir telescope at high galactic latitude will be determined by one of 3 contesting noise values . the instrument noise , the noise due to diffuse ir cirrus emission @xcite which is spatially dependent or the confusion limit due to point sources @xcite . in table [ confuse ] we calculate the limiting flux sensitivity corresponding to the confusion limit due to point sources for each of the astro - f all sky survey bands for 3 different cosmological world models . we assume the classical confusion criteria of a source density of 1 source per @xmath440 beams @xcite , @xcite of the observing instrument , where the beam diameter is given by @xmath98 , where @xmath99 is the telescope diameter ( 67 cm ) . here we calculate the source confusion assuming a beam diameter equivalent to the fwhm of the airy disc , @xmath98 such that the beam diameter at 170@xmath6 m is @xmath41arcmin corresponding to a source confusion density of @xmath2100 sources ( where the subsequently calculated limiting flux will of course be model dependent ) . from table [ confuse ] we see that for the shorter fis bands and the mir bands the confusion limit due to point sources is not really dependent on the world model assumed and is on or well below the expected sensitivity capabilities of the instruments . for the longer wavelength bands ( n170 , wide - l ) we would expect that the galaxy counts will be quite severely restricted by the source confusion due to the extremely steep slope of the source counts expected ( @xmath43.3 compared to 2.5 expected for a non - evolving population in a euclidean universe ( kawara et al . , dole et al . ) ) . indeed , observations at 175@xmath6 m from the iso firback survey @xcite and japanese lockman hole survey @xcite place even more severe limits on the source confusion level suggesting that the confusion due to point sources would become severe at flux levels of @xmath445mjy . although sirtf boasts a larger diameter mirror ( 85 cm ) than astro - f and also has much better sensitivity at 70@xmath6 m ( 2.75mjy ) and 160@xmath6 m ( 17.5mjy ) , it too will feel the pinch of the source confusion . estimated limits for sirtf are 11mjy at 70@xmath6 m and 60mjy at 160@xmath6 m . it should be noted that various innovative fluctuation analysis methods can be used to push below these limits ( e.g. matsuhara et al . ) . the source counts in fig . [ counts ] are plotted by galaxy component and tabulated in fig . [ numbers ] . the source counts are rapidly increasing down to fluxes of 10 - 20mjy where they begin to flatten off . approximately , 1.6 million sources will be detected in the n60 band compared to more than 10 million in the n170 band . the n170 band samples higher up the source count slope ( closer to the turn over ) and is more sensitive to higher redshift galaxies due to the negative k - corrections ( see fig . the dust hump at the peak of the emission spectrum at 100@xmath6 m and 60@xmath6 m for normal and starburst galaxies passes through the n170 band at redshifts of 0.7 and 1.8 respectively where as the emission in the n60 band will already be decreasing . in the wide bands , approximately 22 million and 47 million galaxies may be detected in the wide - s and wide - l bands respectively . the vast majority of sources detected in all bands will be luminous and ultraluminous galaxies . expected numbers range from 1 - 40 million from n60 - wide - l respectively although the exact numbers will be dependent on the assumed evolutionary scenario . the longer wavelength far - ir bands ( n170/wide - l ) will be particularly sensitive to these galaxies . agn will be preferentially detected in the short wavelength fir bands ( n60/wide - s ) as the agn emission peaks at @xmath425@xmath6 m meaning that the longer wavelengths are still climbing the the rayleigh - jeans slope of the sed out to redshifts of @xmath46 ( see fig . approximately 7000 agn will be detected in the n60 band , with as many as 40,000 in the superior wide - s band . the n170 band will detect between 700 - 1000 while the wide - l band should see as many as 4500 ( these numbers being dependent on the assumed cosmology ) . comparison with the iras integral counts at 60@xmath6 m ( n60 & wide - s bands ) shows that the astro - f survey will probe between 3 - 10 times deeper than the various iras surveys ( 0.5jy for iras - psc and 120mjy for the vfss gregorich et al . , bertin et al . ) . moreover , a comparison with the iras 60@xmath6 m differential source counts ( hacking & houck , rowan - robinson et al . , saunders , gregorich et al . ) emphasizes that astro - f will be reaching very interesting flux levels where some discrimination between competing evolutionary scenarios will be possible . at the limit of the iras - psc the counts are euclidean ( i.e. flat in the normalized differential counts - the middle panel of figure [ counts ] ) , the iras - vfss survey extended this flux limit down to @xmath2120mjy and suggested an upturn in the differential source counts at @xmath4200 - 300mjy similar to that observed in the differential counts at radio wavelengths due to the emergence of a radio sub - mjy population of starburst galaxies @xcite . kim & saunders re - analysed the iras - fss data concluding that there was evidence of much stronger evolution at higher redshifts to z@xmath41 . the nature of this evolution can not be constrained by the iras source counts at 60@xmath6 m with both luminosity evolution or density evolution being equally acceptable ( pearson , oliver et al . ) . astro - f will be able to detect and quantify the upturn in the differential counts at fluxes @xmath5200mjy . astro - f will be the first ever all sky survey at 170@xmath6 m . since the peak of the energy output of dust enshrouded star formation lies in the fir at @xmath460@xmath6 m the n170 and wide - l bands will be particularly sensitive to high redshift ( z@xmath31 ) galaxies as the peak of the sed emission is redshifted to 170@xmath6 m at [email protected] producing the phenonemon where galaxies are brighter at higher z than lower z as made apparent in fig . [ lz ] , @xcite,@xcite . to date the largest survey at 170@xmath6 m was carried out by the firback team covering 4sq.deg . down to 180mjy @xcite . the astro - f all sky survey will be aptly placed below the limit of the firback survey in both the n170 and wide - l bands . although the wide - l band will almost certainly be source confused , valuable work can still be carried out on and below the confusion limit ( c.f . the fluctuation analysis of iso 170@xmath6 m source counts @xcite ) . the sensitivity of the fir bands will also be good enough to provide discrimination between different evolutionary scenarios in the differential counts . the firback differential counts have already shown a steep non - euclidean slope of [email protected] at fluxes less than 500mjy ( c.f . a euclidean slope of 2.5 ) . astro - f should detect any turnover in the differential source counts . the lower panels of figure [ counts ] shows the predictions for the possible additional survey in the 2 wide mir wavebands at 9@xmath6 m and 20@xmath6 m . in this case , the survey sensitivities will be somewhat lower than compared to the fir bands . this is due to constraints on the data transmission rate of the astro - f satellite . nominal sensitivities are currently 50mjy & 100mjy at 9@xmath6 m & 20@xmath6 m respectively with some possibility of increasing this to 30mjy and 70mjy respectively by using a lower sampling rate . predicted numbers in the wide-9@xmath6 m band are approximately 17,000 sources in total , mostly at local redshifts with approximately 2000 starburst galaxies and 2000 agn . results from the longer wide-20@xmath6 m band are similar with approximately 10,000 sources in total of which around half would be star forming galaxies . note that in the case of iras only 893 galaxies were obtained from the iras - fss ( @xmath10125 , @xmath30.22jy at 12@xmath6 m @xcite , @xcite ) . furthermore , around 1000 sources detected by iso were used to construct the iso number counts at 15@xmath6 m ( elbaz et al . , gruppioni et al . , serjeant et al . ) . astro - f will obtain an order of magnitude larger sample than the iras sample and the iso 15@xmath6 m samples . thus the local luminosity function of ir galaxies can be accurately derived since all galaxies at 9@xmath6 m must be detected by the fis survey . astro - f will also be an ideal tool for the discovery of rare objects such as hyperluminous ( @xmath102 , @xcite ) and primeval galaxies , c.f . iras f10214 + 4724 discovered at the limit of the iras - fss with @xmath103=0.2jy , at z=2.286 and far - ir luminosity @xmath104 @xcite . totani & takeuchi have suggested that starbursting primordial ellipticals may have higher dust temperatures ( 40 - 80k ) than previously assumed . such star bursting primordial ellipticals may also be observed in the astro - f all sky survey . the best strategy for such detections are ( i ) wide area shallow as opposed to narrow , deep surveys , since these objects are rare but bright with source densities of between 0.001 - 0.01per sq.deg . depending on the evolutionary scenario and cosmological world model . ( ii ) far - ir wavelengths since the bulk of any emission ( 90 - 99@xmath16 ) should emerge at these wavelengths due to the copious amounts of dust expected due to starformation . ( iii ) multi - wavelength survey for ir colours for photometric redshifts , sed fitting and discrimination from other less luminous sources . an f10214 type hyperluminous galaxy with luminosity of order a few @xmath105 would be detectable out to redshifts @xmath44 in the fis bands and at z@xmath31 in the irc - mir bands . depending on the cosmology , we could expect of the order of 300 such sources over the entire sky detected in the wide - s and wide - l bands and between 40 - 100 in the wide-9 and wide-20 mir bands respectively . corresponding follow up and cross correlations with future mm wave instruments such as lmt ( www.lmtgtm.org , e.g. a large area survey of a few 1000sq.deg . ) would be particularly sensitive , will further constrain the detections of such objects since a detection at mm wavelengths with no corresponding astro - f counterpart will automatically imply a high redshift ( and therefore high luminosity ) for any given source . @lcccccccccccccccc band & flux & & & & & + & 5@xmath68 mjy & mean & mode & med & mean & mode & med & mean & mode & med & mean & mode & med & mean & mode & med + @xmath72 & @xmath24 & @xmath106 & @xmath107 & @xmath108 & @xmath109 & @xmath107 & @xmath109 & @xmath110 & @xmath111 & @xmath93 & @xmath112 & @xmath113 & @xmath114 & @xmath115 & @xmath107 & @xmath116 + @xmath75 & @xmath28 & @xmath117 & @xmath118 & @xmath119 & @xmath94 & @xmath107 & @xmath120 & @xmath121 & @xmath107 & @xmath122 & @xmath123 & @xmath118 & @xmath124 & @xmath125 & @xmath126 & @xmath127 + @xmath79 & @xmath21 & @xmath128 & @xmath129 & @xmath130 & @xmath131 & @xmath107 & @xmath132 & @xmath133 & @xmath134 & @xmath135 & @xmath136 & @xmath129 & @xmath137 & @xmath138 & @xmath107 & @xmath139 + @xmath82 & @xmath34 & @xmath140 & @xmath141 & @xmath142 & @xmath143 & @xmath107 & @xmath144 & @xmath145 & @xmath134 & @xmath146 & @xmath147 & @xmath141 & @xmath148 & @xmath149 & @xmath150 & @xmath151 + @xmath87 & @xmath39 & @xmath109 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath109 & @xmath107 & @xmath107 & @xmath120 & @xmath107 & @xmath109 + @xmath91 & @xmath41 & @xmath152 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath152 & @xmath107 & @xmath109 & @xmath111 & @xmath107 & @xmath109 + [ mean ] figure [ nz ] shows the expected number redshift distributions for the astro - f all sky survey in the 4 fis bands along with predictions for the mir all sky survey currently under consideration . beyond a redshift of @xmath40.3 the number redshift distributions in all the fis bands are dominated by luminous / ultraluminous infrared galaxies consistent with the results of deep surveys with iso ( e.g. elbaz et al . , dole et al . ) . the narrow fis bands ( n60,n170 ) are only sensitive to normal and starburst galaxies out to redshifts of @xmath40.5 , the shortest band only detecting a few percent of these galaxies above a redshift of unity . the longer wavelength bands , due to the large k - corrections encountered ( see fig . [ lz ] ) , do better at detecting sources at z@xmath31 with about half the total number of detections expected to be at redshifts higher than unity . the wide - l band ( and to a lesser extent the wide - s band ) , due to their better sensitivity are also sensitive to starburst galaxies at z@xmath621 with between 1/20 - 1/8 and 1/3 - 2/3 depending on the cosmology being found at z@xmath31 in the wide - s and wide - l bands respectively . due to the nature of the assumed agn sed almost all agn detected in the long wavelength bands will be at z@xmath32 ( 30 - 50@xmath16 for the short wavelength bands ) , thus providing a unique sample of far - infrared , high redshift agn especially at 170@xmath6 m where the low detection rate means that only a large area ( all sky ) survey such as the astro - f survey can provide us with a statistically viable sample of such objects . in summary , the astro - f survey will provide the database for a comprehensive redshift survey out to redshift @xmath41 for normal and starburst galaxies and to redshift 2 and higher for ulig and agn sources thus bridging the dark gap between the iras observations in the local universe and the optical and sub - mm studies in the high redshift regime . to investigate any correlation between the number - redshift distributions further , we plot in fig . [ residual ] , the n - z residual . that is , the difference between the n - z distributions at 2 wavelengths . residual numbers are plotted as @xmath153 . a positive value indicates that @xmath154 , a negative value that @xmath155 and a value of zero indicates that all galaxies are seen in both wavelength bands , i.e. that the n - z distributions are identical . we conclude that normal galaxies are preferentially detected in the long wavelength bands due to the fact that the peak of their sed is around 100@xmath6 m as are the lig / ulig sources due to the strong k - corrections incurred at redshifts up to @xmath41.8 . for starburst galaxies , at lower redshifts we are sampling the dust emission hump at @xmath460@xmath6 m , thus the sources are preferentially detected in the short wavelength bands , moving to higher redshift shifts the residuals in favour of the longer wavelength bands . for agn , the shorter wavelength bands are preferred due to the flat nature of the sed from 30@xmath43@xmath6 m . the 60@xmath6 m band climbs out of the rayleigh - jeans region of the sed at z@xmath21 whereas the 170@xmath6 m remains on the rayleigh - jeans slope until z@xmath34 . note , the residuals corresponding to the wide - narrow band passes simply show the increase in detections due to the enhanced sensitivity of the wide band filters over the narrow band filters . we find average redshifts , median(mean ) , of 0.29(0.34 ) , 0.99(0.96 ) , 0.63(0.67 ) , 0.97(0.98 ) for the n60 , n170 , wide - s , wide - l bands respectively . by comparison the iras qdot sample had a median redshift of@xmath4=0.03 @xcite and even the iras very faint source survey analysis was limited to [email protected] @xcite . in table [ mean ] a detailed analysis of the expected redshift values ( mean , mode , median ) is tabulated for all bands and all components for the the astro - f all sky survey . distinct populations sample distinct redshift regimes at factors of 2 - 10 deeper than the iras surveys . the number - redshift distribution of the proposed mir all sky survey at 9@xmath6 m and 20@xmath6 m shown in the bottom pair of panels in fig . [ nz ] is somewhat limited to redshifts @xmath50.5 ( z@xmath51 in the sparse sampling rate scenario - see the figures in italics in fig . [ numbers ] ) . however , even this relatively shallow sample will enable a more accurate determination of the local 9@xmath6 m and 20@xmath6 m luminosity functions . previous recent studies of the 12@xmath6 m and 25@xmath6 m luminosity functions with iras have been limited to [email protected] at @xmath156 ( fang et al . , shupe et al . ) . even with the limitations of the proposed all sky mir survey ( median redshift @xmath20.05 , the mir luminosity function of galaxies will be accurately determined out to at least a factor of 2 higher in redshift and a factor of 2 - 5 better sensitivity in flux . the astro - f all sky survey will be an invaluable tool in the study of large scale structure . the iras survey mapped the local large scale structure of the universe out to redshifts @xmath20.01 - 0.1 in a series of studies including 35000sq.deg.down to 600mjy at 60@xmath6 m and 20000sq.deg . down to 200 - 250mjy at 12,25,60@xmath6 m ( e.g. lonsdale et al . , rush et al . , oliver et al . , shupe et al . , saunders et al . ) using the iras psc @xcite and fss @xcite data sets . wide sky coverage is a necessity for large scale structure studies , as studies on smaller scales can be effected by localized fluctuations . astro - f will probe to 10 - 100 times deeper than iras over 500x the area of the sirtf - swire program @xcite . a detailed analysis of the usefulness of astro - f in the study of large scale structure is beyond the scope of this present work and will be addressed in a later paper ; in this work , we briefly outline the potential of such studies using the all sky survey data . we approach our assessment via the density function and derived visibility function of galaxies in the astro - f all sky survey ( see von hoerner & condon for a more detailed derivation and explanation of the visibility function ) . in general , the total number of sources observable down to some limit @xmath157 is given by , @xmath158 where @xmath159 is the local differential luminosity function per decade in luminosity @xmath160 . integrating over the volume element @xmath161 out to some limiting redshift @xmath162 , the total number of sources . discarding the second half of the integral gives @xmath163 gives the number density of galaxies in a given redshift bin at redshift @xmath162 . this is effectively the galaxy luminosity function in a bin of width @xmath164 centred on a redshift @xmath162 , we call this the density function @xmath165 . the density function for the astro - f all sky survey is shown in fig . [ vis]a for all the fis and mir bands . the density function is essentially a measure of the survey sensitivity and plotting the corresponding results for the sirtf - swire survey the superiority of the latter s sensitivity can be clearly seen . by taking logarithmic steps in the density , a threshold redshift @xmath166 can be read from the density function in fig . [ vis]a . using this redshift , @xmath167 , a corresponding cosmological volume is calculated which will be the sampled volume above a given galaxy density for the proposed survey , where @xmath168 . convolving this volume with the total coverage of the survey gives us the quantity we refer to as the visibility function of sources in the all sky survey and is a function of both the depth ( sensitivity ) and the coverage of any given survey . the visibility function is plotted in fig . [ vis]b and clearly shows the power and potential of the astro - f all sky survey for large scale structure studies with 2 - 3 orders of magnitude in the volume being sampled at densities of @xmath169 compared to the equivalent sirtf - swire bands . in truth , the astro - f and sirtf - swire surveys supplement rather than supplant one another and are extremely complementary with astro - f covering scales from @xmath410 - 1000@xmath170mpc from [email protected] - 1 and the sirtf - swire survey covering the range from scales of @xmath410 - 100@xmath170mpc from redshifts @xmath30.1 - 5 . galaxy surveys at a single wavelength provide insight into source density and variation , background contributions and serendipitous discoveries but are by definition rather monochromatic . astro - f will survey in 4 fir ( + 2 mir ) bands allowing multi - band correlations and fir ( + mir ) colours producing as many as 6 new points on the ir seds of galaxies and source counts as a function of galaxy type ( via colour discrimination similar to the way the iras galaxies were segregated into cirrus , starburst and agn populations on the basis of their ir colours @xcite ) . moreover , it may be also be possible to obtain rough photometric redshifts from the fir data alone . in fig . [ ccz ] the variation of galaxy colours with redshift calculated for the astro - f all sky survey fis bands ( _ top panels _ ) and mir wide-20@xmath6 m & wide-9@xmath6 m bands ( _ middle panels _ & _ bottom panels _ respectively ) , for a normal galaxy of @xmath171 , a starburst galaxy of @xmath172 , an ultraluminous galaxy of @xmath173 and an agn of @xmath174 . for the normal and starburst ( including lig & ulig sources ) galaxies , there is a general increase of the long wavelength ( n170 & wide - l ) flux as the sed climbs through the rayleigh - jeans region resulting in redder n170/n60 & wide - l / wide - s colours and bluer n60/wide - l colours . as the long wavelength band moves over the dust hump in the sed , the colours decrease in magnitude . for a normal galaxy the dust emission peaks at @xmath4100@xmath6 m while the emission peak for starburst galaxies is at the shorter 60@xmath6 m wavelength due to the higher dust temperatures in these galaxies . thus , this effect occurs at [email protected] and @xmath41.9 for normal and starburst galaxies respectively . since the dust torus sed we use for our agn template is rather flat from 3 - 30@xmath6 m and declines steeply to longer wavelengths , the short wavelength band flux must always be higher than the corresponding longer wavelength fluxes at least out to z@xmath21 , resulting in an inversion of the agn colour variation with redshift compared to the other galaxy populations . although this may seem at first sight a useful discriminator , one implication may be that the highest redshift agn will in fact have similar colours to the low redshift normal galaxy population ( see fig . previous work with iras / iso data has also shown that seyfert galaxies are indistinguishable from normal galaxies in 100 - 60@xmath6 m colours or 200 - 100@xmath6 m colours @xcite . however the astro - f n170 band may provide additional discriminators that will allow us to unravel the 2 populations somewhat . the addition of the 2 wide mir bands to the astro - f all sky survey may provide further constraints on the colours of galaxies ( bottom 2 panel rows of fig . [ ccz ] ) . however , their value as a statistical tool ( for example to collect a large sample of photometric redshifts ) will be severely limited by their low sensitivity , restricting such studies to [email protected] where there would be @xmath175 sources over the entire sky corresponding to 1/700 and 1/90 n170 & n60 sources respectively . in general , the low redshift starburst population will have decidedly redder 20@xmath6m/@xmath176 colours than the normal galaxies . however , the colours of the low redshift agn ( z@xmath177 ) will be significantly redder ( especially in 20@xmath6m / n170 , 20@xmath6m / wide - l , 9@xmath6m / n170 , 9@xmath6m / wide - l ) than all other sources . again the is due essentially to the flat nature of the agn sed from 3 - 30@xmath6 m and may be useful as a discriminator of low redshift agn . @lllll survey & reference & area & band & sensitivity + poss - ii & reid et al . 1991 & northern sky & optical & b@xmath178=22.5 , r@xmath179=20.8 , i@xmath179=19.5 + ukss & yentis et al . 1992 & southern sky & optical & b@xmath178=21 + 2mass & jarret et al . 2000 & all sky & 2@xmath6 m & j=15 ( 1.6mjy ) , k@xmath180=13.5 ( 2.9mjy ) + rosat & voges et al . 1999 & all sky & x - ray & 9x10@xmath181erg @xmath182 s@xmath183 + nvss & condon et al . 1998 & 82@xmath16 sky & radio & 2.5mjy + galex & bianchi et al . 1999 & all sky & uv & 20mjy ( m@xmath184=20 - 21 ) + sdss & stoughton et al . 2002 & 10000sq.deg . & optical & u=22,g=22.2,r=22.2,i=21.3,z=20.5 + vista & www.vista.ac.uk & 10000sq.deg . & optical & u=25.6,b=26.7,v=26.5,r=25.3,i=24.7 + vista & www.vista.ac.uk & 5000sq.deg . & nir & j=23.4,h=21.9,k=20.5 + ukidss & warren & hewitt 2002 & 4000sq.deg . & nir & y=20.5 , j=20 , h=18.8 , k=18.4 + + [ followup ] @llll survey & reference & area & band & sensitivity + pscz & saunders et al . 2000 & all sky & 60@xmath185jy + sdss & stoughton et al . 2002 & 10000sq.deg . & [email protected] + 2dfgrs & colless et al . 2001 & 2000sq.deg . & b@xmath186 + 6dfgrs & colless 2002 & 17000sq.deg . & k@xmath187 , h@xmath188 , j@xmath189 , i@xmath190 , b@xmath191 , s@xmath192 + aa@xmath193 2df & www.ast.cam.ac.uk/aao & 2deg.fov & b@xmath523 + + [ redsurvey ] @llllllllllll & & + component & n60 & k & i & b & radio & x - ray & k & i & b & radio & x - ray + & mjy & & & & mjy & erg/@xmath194/s & & & & & + normal & 73 & 12.2 & 14.0 & 15.8 & 0.5 & 1.1x10@xmath195 & 8330 & 8330 & 16600 & 180 & 360 + starburst & 67 & 16.3 & 16.2 & 19.5 & 0.7 & 0.6x10@xmath195 & 42 & 512 & 260 & 230 & 170 + ulig & 58 & 17.9 & 19.9 & 22.0 & 0.6 & 0.5x10@xmath195 & 6 & 13 & 24 & 230 & 160 + agn & 61 & 13.3 & 19.0 & 15.0 & 0.01 & 90x10@xmath195 & 1750 & 2220 & 11100 & 4 & 166660 + + [ identifcation ] initial redshift estimates for the astro - f all sky survey will be made via associations with photometric surveys at other wavelengths and extensive use of existing optical / near - ir redshift surveys and catalogues . in table [ followup ] & table [ redsurvey ] we list large scale photometric surveys at other wavelengths and 4 of the largest present or near future large area redshift surveys respectively . these are of particular interest since it is in the domain of large area / volume statistics that astro - f / sirtf missions will be most discriminated from each other . hirashita et al . have suggested that typical astro - f galaxies will have magnitudes [email protected] & [email protected] meaning that they will be detectable in the poss - ii and ukss catalogues ( [email protected] , [email protected] , i@xmath519 ) although they may be too faint for the sdss ( [email protected] ) and 2dfgrs ( [email protected] ) redshift surveys . the pscz & sdss comprise of 15,411 & 10@xmath197 galaxy redshifts respectively while the 2df has presently obtained more than 200,000 @xcite . however , these surveys are limited to [email protected] , 0.4 , 0.2 respectively and additional redshift surveys / follow up may be utilized ( e.g. 6df @xcite , aa@xmath1932df , vlt - virmos @xcite ) . the astro - f all sky survey will produce a catalogue of sources numbering in the millions or even tens of millions . identification of these sources may prove to be elusive in some cases . in this work we make a very brief initial analysis of the identification problem that astro - f may face . similar problems are encountered by many fir - submm facilities ( a prime example being the difficulty in identifying scuba sources at other wavelengths due to the size of the scuba beam @xmath198 @xcite , @xcite ) since it is very difficult to identify sources at these wavelengths when the source densities are greater than 1 per beam ( or conversly one beam / source ) . in table [ identifcation ] we tabulate the estimated median fluxes of astro - f galaxies in the n60 band . using a sample of template seds and the flux ratios between 60@xmath6 m and other wavebands from radio to x - rays ( helou et al . , condon et al . , white & becker , rowan - robinson et al . , corbelli et al . , pearson & rowan - robinson , soifer et al . , green et al . ) we calculate the median fluxes of these sources at 1.4ghz , johnson k , i , b bands and the rosat 0.5 - 2kev x - ray band . immediately by comparison with table [ followup ] we see that normal galaxies will be easily identified by both 2mass or poss - ii / ukss with around 1 misidentification in @xmath22000 sources . these normal galaxies are quite bright and are predicted to be relatively local with their median redshift being @xmath20.06 . the starburst galaxies are also readily identifiable with poss - ii / ukss , although we predict a slightly brighter b - band flux of 19.5 compared to the [email protected] of hirashita et al . . however , the lig / ulig sources will be more problematical being too faint for 2mass , nvss and possibly poss - ii / ukss as well . the majority of the agn detected by astro - f should be picked up by the hamburg - eso quasar survey in the area it covers @xcite . we then predict and tabulate in table [ identifcation ] the number of astro - f beams per source at these wavelengths and flux levels using published surface densities of sources from the literature ( compilations from condon , afonso et al . , fioc & rocca - volmerange , boyle et al . , mchardy et al . ) . from table [ identifcation ] we see that if they are bright enough to be detected , most sources should be readily identifiable , i.e. , more than a few beams per source . however the uligs will be elusive in the k , i , b bands as will the agn in the radio . we would stress that this is a rather preliminary analysis of the identification problems and procedures that will be encountered with the astro - f all sky survey sources . more sophisticated analysis , such as likelihood ratio ids @xcite are beyond the present scope of this work and will be considered at a later time . in addition , photometric redshifts from the 4 fis - fir bands ( for sources [email protected] ) plus an additional 2 mir bands for sources [email protected] will be used to provide rough constraints on the redshifts of sources . in figures [ cc ] & [ ccmir ] we plot the colour - colour diagrams for fis bands and fis+irc bands respectively . we find in general that different galaxy types occupy distinct regions of the colour - colour plane and that by combining one or more of these plots that we should be able to obtain some degree of segregation in the galaxy populations both in type and in redshift . to illuminate some examples , for starburst , lig & ulig sources there is a trend towards redder n170/wide - l - n170/n60 colours with redshift with galaxies at z@xmath32 having colours an order of magnitude redder than those at z@xmath51 . the n60/wide - l - n60/wide - s colour - colour distribution of normal galaxies are similarly elongated with increasing redshift . however , as stated previously , there is some degeneracy between the low redshift colours of normal galaxies and the high redshift colours of starburst / lig / ulig sources . none of the colour - colour plots in fig . [ cc ] are capable of confidently separating these source populations from each other . the agn , however , are very well segregated in both redshift and from the other galaxy populations by their colours . the low redshift agn occupy very distinct areas in both the n60/wide - l- n60/wide - s and n170/n60 - wide - l / wide - s colour - colour diagrams . it is a possibility that the colours of high redshift agn may be confused with with the colours of high redshift normal ( cool ) galaxies although it is dubious that any any such cool sources could be detected in the astro - f all sky survey since a normal galaxy at redshift @xmath41 would have to have an ir luminosity @xmath9 to be detected in the all sky survey ( see fig . [ lz ] ) . therefore it seems reasonable to assume that any sources with lg(s(n60)/s(wide - l))@xmath1990 in the n60/wide - l - n60/wide - s plane would indeed be high redshift agn . combining the fis - fir colours in fig . [ cc ] with the irc - mir colours we find that we can obtain a set of more powerful discriminating tools . in fig . [ ccmir ] we additionally plot the wide-20/wide-9 - wide-9/wide - l & wide-20/wide-9 - wide-9/wide - s colour - colour diagrams ( note that swapping the 2 mir bands in these colour - colour plots will simply reflect the shape of the plots in the x - axis plane ) . with the addition of the mir bands we find we can break the degeneracy between the low - z cool objects and high - z star bursting objects . in these diagrams , the normal galaxies lie blueward of the starburst populations in wide-20/wide-9 colours and in general have redder wide-9/fir colours than the star forming populations . although the wide-9/fir colours in both the low - z normal galaxies and high - z star bursting objects are both approximately constant and unchanging ( since both the fir flux the mir - nir flux is on a black body curve for both low - z normal and high - z starburst galaxy populations ) , the 20@xmath6 m flux is moving through the uib emission features while the 9@xmath6 m flux climbs the normal galaxy sed while falling down through the starburst sed at higher redshifts causing a degeneracy between the 2 galaxy populations . we have reviewed the astro - f all sky survey and its expected results . astro - f will carry out the first all sky survey at infrared wavelengths for 20 years and will probe down to sensitivities between 10 - 1000 times deeper than the iras mission to more than 5 times the angular resolution in 4 far - ir bands from 50 - 200@xmath6 m and 2 mid - ir bands at 9 & 20@xmath6 m . expected numbers of sources range from 10 s millions in the longest wavelength bands where they will be source confusion limited , to 10,000 at mid - ir wavelengths ( a number of an order of magnitude higher than that detected by iras or iso at mid - ir wavelengths ) . most of the sources detected in the far - infrared bands will be dusty ligs & uligs of which , in the longest wavelengths bands , @xmath350@xmath16 will be at redshifts greater than unity . the ultimate goal of the astro - f all sky survey is to achieve approximately 95@xmath16 sky coverage by the end of the survey to a reliability of @xmath399@xmath16 for bright sources . absolute flux uncertainties of 10 and 20@xmath16 are expected for point and diffuse sources respectively . pointing accuracy is hoped to be @xmath200 . note that the orbit of astro - f will be sun - synchronous polar with the fov always perpendicular to the sun . this will have significant effects on the visibility of sources . sources near the ecliptic plane will be visible only 2 days ( 29 orbits ) every half year whereas targets near the ecliptic poles will be will be observable on many orbits . the north ecliptic pole is indeed under investigation as a target area for deep pointed surveys with the irc @xcite and will complement well with the proposed planck deep nep survey . ultimately , it is expected that the astro - f all sky survey will produce several catalogue products which could then serve as input catalogues and cross correlation databases for later missions such as herschel @xcite & planck ( including the all sky survey with hfi @xcite ) due for launch in 2007 . these catalogues can be broken down as follows ; 1 . _ astro - f flux of known sources _ - measurement of sources used for the input source catalogue and fluxes of iras point sources . 2 . _ bright source catalogue _ - a catalogue of sources that can be extracted easily such as bright sources or those at high ecliptic latitude . 3 . _ faint source catalogue _ - the all sky point source catalogue to the optimum sensitivity . image maps _ - crude maps , square degree imagelets and 10 s sq.deg . image atlases . the astro - f surveyor mission should be seen as supplementing not supplanting the sirtf observatory mission . the two are complementary . sirtf will cover relatively small areas ( 1-70sq.deg . ) to high sensitivity while astro - f will cover huge areas to more moderate sensitivities . in the study of large scale structure sirtf - swire will be sensitive to small scales over large redshift while conversely astro - f will probe larger scales to shallower depths . however , one of many clear distinctions between the two missions is that the astro - f all sky survey is the only way that the infrared dipole work carried out with iras can be checked and extended ( yahil et al . , rowan - robinson et al . , rowan - robinson et al . ) . astro - f will also provide an ideal companion to planck which will survey the whole sky at sub - mm wavelengths and also down to 200@xmath6 m which will provide an important longer wavelength channel to join fir measurements by astro - f . further associations and cross correlations with large scale surveys at other wavelengths ( swire , pos - ii , ukss , sdss , nvss , galex , rosat , ukidss , 6df , 2mass , lmt , etc . , see tables [ followup ] , [ redsurvey ] ) will result in a fully comprehensive unified view of galaxy evolution and star formation from radio - mm - ir - optical - uv - x - ray wavelengths . the simulations for the astro - f all sky survey discussed in this paper can be accessed through the world wide web at @xmath201 . other information on the astro - f mission can be found at @xmath202 . during this work , cpp was supported by a fellowship of the royal society . cpp would like to thank steve serjeant for reading through this manuscript and providing many useful comments . |
the catalina real time survey ( crts ) has found over 500 cataclysmic variable ( cv ) candidates , most of which were previously unknown . we report here on followup spectroscopy of 36 of the brighter objects . nearly all the spectra are typical of cvs at minimum light . one object appears to be a flare star , while another has a spectrum consistent with a cv but lies , intriguingly , at the center of a small nebulosity . we measured orbital periods for eight of the cvs , and estimated distances for two based on the spectra of their secondary stars . in addition to the spectra , we obtained direct imaging for an overlapping sample of 37 objects , for which we give magnitudes and colors . most of our new orbital periods are shortward of the so - called period gap from roughly 2 to 3 hours . by considering the cross - identifications between the catalina objects and other catalogs such as the sloan digital sky survey , we argue that a large number of cataclysmic variables remain uncatalogued . by comparing the crts sample to lists of previously - known cvs that crts does not recover , we find that the crts is biased toward large outburst amplitudes ( and hence shorter orbital periods ) . we speculate that this is a consequence of the survey cadence . | 1207.3070 | in cataclysmic variable stars ( cvs ) , a white dwarf primary star accretes matter by way of roche lobe overflow from a binary companion , which resembles a main - sequence star . the variety of cv behaviors leads to a complicated taxonomy @xcite . many cvs undergo dwarf nova outbursts , thought to be caused by accretion disk instabilities which greatly increase the rate at which matter moves inward in the disk . other cvs , called novalike variables , remain in a bright state for years at a time . in still others , called am her stars or polars , the matter that is transferred becomes entrained in a strong white - dwarf magnetic field , and is funneled directly onto the white dwarf s magnetic pole . the main driver of cv evolution is thought to be a gradual loss of orbital angular momentum . this causes the roche critical lobe of the secondary star to shrink , leading to a shortening of the orbital period @xmath0 , and driving mass transfer on long ( gyr ) timescales . the mechanisms by which angular momentum is lost are not fully understood . it is often supposed that magnetic braking of the secondary star predominates at longer periods ( @xmath1 3 hr ) , and that magnetic braking becomes inefficient at short period , so that gravitational radiation predominates . around @xmath0 = 70 min , the secondary becomes degenerate and its radius begins to _ increase _ with mass , leading to a slow _ increase _ in the orbital radius . this turnaround is often called the _ period bounce _ , even though it is thought to take place very slowly . the histogram of cv orbital periods shows a significant dip at roughly 2 hr @xmath2 3 hr , known as the _ gap _ @xcite . this is often explained as follows . as the secondary loses mass , its thermal timescale increases to become comparable to the time for the orbit to evolve , with the result that the secondary exceeds its equilibrium radius . at about three hours , the secondary becomes fully convective , reducing the efficiency of magnetic braking . as the orbital evolution slows , the secondary detaches from its roche lobe , shutting down mass transfer . the detached system continues to evolve to shorter periods , crossing the gap and eventually re - establishing contact with the roche lobe near @xmath3 hr . while the mechanism by which this happens is somewhat speculative , @xcite shows that a discontinuity in the secondary stars radii occurs across the gap . in a steady state , the number of stars with a given @xmath0 should be inversely proportional ( roughly ) to @xmath4 , the rate at which the period changes . if @xmath4 really is very slow at short periods , then there should be a large population of short - period cvs , and the gradual turnaround at the period bounce around 70 minutes should lead to a ` spike ' in the distribution @xcite . efforts to confront theories such as this with observation have often been frustrated , because the sample of known cvs is incomplete in ways that are difficult to quantify . the discovery channels for cvs include the following : ( 1 ) dwarf nova outbursts are conspicuous they last for days or weeks and typically have amplitudes of several magnitudes . ( 2 ) nearly all cvs have unusual colors compared to normal stars , most conspicuously ultraviolet excesses arising from accretion processes or ( in some cases ) the underlying white dwarf . ( 3 ) with the exception of some novalike variables and dwarf novae near the peak of outburst , nearly all cvs show emission lines , especially in the balmer sequence ; these can be strong enough to be noticed in surveys such as iphas @xcite . ( 4 ) a great many other cvs have been discovered as optical counterparts of x - ray sources . ( 5 ) cvs at minimum light can be rather faint ( @xmath5 ) , and a small handful have turned up in proper motion surveys . new , large samples of cvs with consistent selection criteria are potentially useful for clarifying issues such as the space density and orbital period distribution of these objects . because the colors of cvs overlap those of quasars , the sloan digital sky survey ( sdss ) turned up a large number of spectroscopically confirmed cvs ( @xcite ; hereafter referred to as szkodyi - viii ) . @xcite compiled orbital periods for 137 of these ; with the sdss sample , they were finally able to discern the long - predicted period spike . like the sdss , the catalina real time survey ( crts ; * ? ? ? * ) has discovered many new cvs . while the sdss cvs were originally selected by color they were chosen for spectroscopy largely because their colors overlap those of quasars the crts selects entirely by variability . briefly , the crts surveys most of the accessible sky at galactic latitudes @xmath6 and declination @xmath7 every lunation , using a 0.7 m schmidt telescope in the catalina mountains near tucson , arizona . they search for variability using a master catalog that reaches to @xmath8 . objects that show abrupt outbursts of @xmath9 mag amplitude lasting less than few weeks are classified as likely cvs , making the crts a prolific source of dwarf novae in particular . the crts maintains a catalog of `` confirmed / likely '' cvs on the world wide web . we downloaded this catalog on 2012 march 7 , when it contained 584 objects , and used this data set for the present analysis . the crts cv sample is of great interest because of its depth and selection criteria , so it is a natural choice for follow - up studies similar to @xcite . @xcite describe high - speed photometry of 20 objects , mostly from the crts , and give orbital periods for 15 , including two eclipsing dwarf novae and two superhumpers . only two of their sample had periods longer than the 2 - 3 hour period gap . the period distribution of their sample also showed the spike just above the turnaround . in addition , they found that dwarf novae with more recorded outbursts tended to have longer orbital periods than those with fewer outbursts . here , we report on followup spectroscopy and photometry of the crts cvs listed in table [ tab : star_info ] . most of the crts cvs are too faint for us to follow up , so we selected this sample based largely on apparent brightness . operational considerations , such as the ease with which observations could be interleaved other programs , the nearness of the target to opposition at midnight , and ( for radial velocity targets ) the strength and tractability of emission and absorption spectra also entered into target selection . we find spectroscopic periods for 8 objects , one of which was independently measured by @xcite . in addition , we obtained exploratory spectra of 28 others , largely to assess their feasibility for radial - velocity studies . we also obtained standardized magnitudes for 37 objects ( 17 of which also have spectra ) . _ plan of this paper . _ we describe our equipment and techniques in section [ sec : techniques ] . table [ tab : spectro_summary ] summarizes all of our spectroscopy . in section [ sec : explore ] we describe the spectra of the 28 objects for which we have only a only a quick spectrum . the eight objects for which we have radial velocity time series are discussed in section [ sec : rvs ] ; table [ tab : velocities ] gives the velocities , and table [ tab : parameters ] lists parameters for sinusoidal fits to the velocity time series . table [ tab : photometry ] gives magnitudes and colors of objects for which we have standardized photometry . finally , in section [ sec : statistics ] we consider how the crts cv samples overlaps with other lists of cvs , discuss the apparent selection biases of crts and the implications for the cv population in general . all our data are from mdm observatory on kitt peak , ariznoa . nearly all are from the 2.4 m hiltner telescope , but a single photometric measurement was kindly taken by j. halpern at the 1.3 m mcgraw - hill telescope . the spectroscopic observing setups were as follows . _ modular spectrograph . _ for most of the spectra we used the modular spectrograph and a 600 line mm@xmath10 grating . the detector was either ` templeton ' , a 1024@xmath11 thinned site ccd that gave 2 pixel@xmath10 from 4600 to 6700 , or ` nellie ' , a thick 2048@xmath11 ccd that gave 1.7 pixel@xmath10 from 4460 to 7770 , with vignetting toward the ends of the range . the choice of detector was dictated by availability during a series of controller upgrades . with the modspec , targets are centered using the image reflected from the polished slit jaws . for this , we used new ( 2010 ) slit viewing optics and a self - contained andor ikon ccd camera unit ; this has greatly improved the acquisition of faint targets with this instrument . all of our radial - velocity time series were taken with the modular spectograph . for wavelength calibration we obtained comparison lamps in twilight and used the night - sky @xmath12 5577 line to track spectrograph flexure during the night . we also observed flux standards in twilight when the sky was clear , and bright o and b - type stars in order to correct approximately for telluric absorption . _ ohio state multi - object spectrograph ( osmos ) . _ this versatile instrument @xcite images the focal plane onto a 4 k @xmath13 4 k ccd , at a scale of 0.273 arcsec pixel@xmath10 . filters can be placed in the parallel beam of the reducing camera for direct imaging , or volume - phase - holographic grism disperser can be inserted for spectroscopy . in spectroscopic mode , one can insert slits in two different locations in the focal plane , which yield different wavelength coverage ; we used the ` inner ' slit , which gives coverage from 3960 to 6875 at 0.75 pixel@xmath10 . osmos has no slit - viewing camera , so targets are acquired by taking a direct image ( without slit and disperser ) and then moving the telescope to align the target with the known location of the slit . we inserted a @xmath14 filter for the acquisition exposures , and therefore obtained a rough @xmath14 magnitude ( without a color transformation ) for our targets . some humid weather in 2011 september intermittently caused condensation in the center of the detector window ; fortunately , the important h@xmath15 feature was outside the affected region . for most of the spectroscopic reduction we used iraf routines . to extract one - dimensional spectra from the modspec images , we used a local implementation of the optimal - extraction algorithm of @xcite . we computed synthetic @xmath14 magnitudes for our spectra using the passband tabulated by @xcite . our slit was usually 1.1 arcsec wide , which means that an unknown fraction of the light was lost ; also , many of our spectra were taken through thin cloud . experience suggests that our synthetic magnitudes are accurate to @xmath16 mag in good conditions ; in poor conditions , they can be considerably too faint . we selected eight of our targets for time - series radial velocity observations , with the aim of determining orbital periods . for these , we pushed observations to large hour angle in order to avoid daily cycle count aliases . to measure emission - line radial velocities we used the convolution algorithms described by @xcite , in which an antisymmetric function is convolved with the line profile and the zero of the convolution is taken as the line center . we used methods described by @xcite to tune the convolution function for maximum signal . we estimated uncertainties in the velocities by propagating the counting - statistics errors in the spectral channels ; these estimates do not include possible systematic effects . for absorption - line velocities , we used the @xcite convolution algorithms as implemented by @xcite . to search for periods we used a ` residual - gram ' method described by @xcite . once we had established a period , we fit the time series with sinusoids of the form @xmath17 $ ] . uncertainties in the fit parameters were estimated from the scatter around the best fit using the procedure described by @xcite . for direct imaging , we mostly used the ` nellie ' ccd , which gave 0.24 per pixel . this detector is insensitive in the ultraviolet , so we used @xmath18 filters ( although @xmath19 and @xmath20 were sometimes skipped ) . we derived photometric transformations from observations of @xcite standard stars . the scatter in the transformations was generally @xmath21 mag . as noted earlier , we also have some direct images from osmos . table [ tab : spectro_summary ] shows that we obtained only brief exposures of most of our spectroscopic targets , generally taking one or two exposures in a single visit . our main purpose was to verify that the objects showed spectra typical of cvs . fig : idspecs1 ] and [ fig : idspecs2 ] show these exploratory spectra . of the 28 objects surveyed , 27 appear to be bona fide cvs , the exception being css0350 + 35 . here are brief descriptions . _ css0051 + 20 . _ our one spectrum has modest signal - to - noise , but shows clear h@xmath15 emission and confirms that this is a cv . the continuum has a shape that hints at an m - dwarf contribution , but we can not be sure this is real . _ css0208 + 37 . _ the balmer emission lines are strong and appear double - peaked , suggesting a low orbital inclination . a prominent blue continuum may be from an accretion disk , or may be from a white dwarf photosphere ; on the other hand , the balmer decrement appears extreme , with h@xmath22 much stronger than h@xmath15 , suggesting that an instrumental effect might be enhancing the blue end of the spectrum . dwarf novae declining from outburst can sometimes show blue continua , but our synthesized and measured magnitudes ( @xmath23 and @xmath24 , respectively ) are both a bit fainter than the crts minimum magnitude of 17.9 , ruling out this interpretation ; the filter photometry was obtained only two days before the spectrum . _ css0328 + 28 . _ this is an unremarkable dwarf nova spectrum . _ css0350 + 35 . _ the spectrum is a good match to an m1 dwarf , and shows only a narrow h@xmath15 line similar to a dme star . while this might be a cv in which mass transfer has stopped , there is no sign of a white dwarf in the spectrum , so the spectrum is consistent with a flare star . the light curve at the crts does not rule out either a cv or flare star , so the flare star classification appears likely . this is , notably , the _ only _ object in our spectroscopic sample that appears not to be a cv . _ css0401 + 08 . _ our spectrum is consistent with a dwarf nova in outburst , with weak h@xmath15 emission and weak , broad h@xmath22 absorption ( with a hint of a central reversal ) on a blue continuum . the magnitude synthesized from the spectrum ( @xmath25 ) is also well above the 18.5 mag minimum listed in the crts . _ css0411 + 23 . _ the emission lines are relatively narrow ( h@xmath15 has a fwhm near 300 km s@xmath10 ) , suggesting a relatively low orbital inclination . weak helium emission is present at @xmath26 , and 5015 . _ css0440 + 02 . _ the emission lines are broad ( @xmath27 km s@xmath10 fwhm ) and just show double peaks at our signal to noise ratio . the system is probably close to edge - on . _ css0447 + 09 . _ this shows evidence for a k - type contribution , in the form of an absorption dip near 5170 , and absorption in the na d lines . the signal - to - noise ratio is not good enough to warrant a detailed decomposition of the spectrum , but the absorption features suggest that about half the light comes from a mid - k star , probably in the range k4 to k6 . this suggests that the orbital period is likely to be @xmath28 h or greater @xcite , but a much shorter period is possible if the secondary has lost much of its mass @xcite . _ css0505 + 08 . _ we detect h@xmath15 at rather low signal - to - noise ratio , confirming the cv nature of the object . _ css0514 + 08 . _ both h@xmath15 and h@xmath22 are detected in emission . the fwhm of h@xmath15 is around 1000 km s@xmath10 , suggesting an intermediate orbital inclination . no secondary star is seen at our signal - to - noise ratio . _ _ this object shows an unusual spectrum , with a blue continuum and a narrow ( @xmath30 km s@xmath10 ) , relatively weak h@xmath15 emission line . the nad lines are detected in absorption , but the continuum does not show any other convincing late - type features , so the nad may be interstellar . the spectrum s synthetic magnitude ( @xmath31 ) suggests that the object was somewhat brighter than minimum light ( @xmath32 ) , but not dramatically so . the crts light curve shows many outbursts of @xmath33 magnitudes . this may be a z cam - type dwarf nova , with a persistent plateau state near @xmath32 , or could be some other kind of novalike variable . _ css0545 + 02 . _ the red digital sky survey and the sdss finding chart ( linked from the crts site ) show a nebulosity around this object , strongest along a northeast - southwest axis , extending to a radius of about 20 arcsec from the object . in our spectrum , the apparent sharp absorption features at h@xmath15 ( 6563 ) , [ nii ] ( @xmath34 6548 and 6584 ) , and [ sii ] ( @xmath34 6717 and 6730 ) are strong nebular emission features in the sky that have been over - subtracted . the long - slit spectrum shows extended weak h@xmath15 , h@xmath22 , [ nii ] and [ sii ] emission along 2.6-arcmin length of the slit , and [ oiii ] @xmath34 4959 and 5007 emission extending to @xmath35 arcsec from the star . the stellar spectrum shows h@xmath15 emission with a width of @xmath36 km s@xmath10 , very much resembling a cv . the crts light curve hangs mostly at @xmath37 or so , but the sdss has @xmath38 , in good agreement with our @xmath39 and @xmath40 synthesized from our spectrum ; the nebula may be affecting the crts measurement . if this really is a cv , the nebulosity is especially intriguing ; it is also possible that this is a planetary nebula central star , in which case the broad h@xmath15 might arise from a stellar wind . the crts light curves shows some faint upper limits , so the system may eclipse . _ css0558 + 00 . _ our single , low signal - to - noise spectrum shows weak , broad h@xmath15 on a blue continuum . the synthesized magnitude ( @xmath41 ) is well above the minimum @xmath42 found by crts . it appears we caught this in a relatively rare outburst ; the crts light curve shows only three outbursts , even though the object is well - observed . our filter photometry , obtained 9 days earlier than the spectrum , shows the object much fainter , at @xmath43 . this is a dwarf nova . _ css0905 + 12 . _ this appears to be an ordinary dwarf nova observed near minimum light . h@xmath15 has a fwhm of 1200 km s@xmath10 , suggesting an intermediate orbital inclination . _ css1055 + 09 . _ the spectrum shows a contribution from an m dwarf which is presumably the secondary star . because of our limited spectral coverage and signal - to - noise , we can only roughly estimate the secondary s spectral type to be m3 @xmath44 . the secondary appears to contribute at least half the continuum light in the region of our spectrum . the easy visibility of the secondary suggests that the orbital period lies above the 2 - 3 hour gap . _ css1139 + 45 . _ our spectrum is very poor , but shows broad h@xmath15 emission , confirming that this is a cv . the synthesized magnitude and spectrum both indicate it was at minimum light . _ _ this relatively bright object shows strong balmer and hei lines . h@xmath15 has a fwhm of 800 km s@xmath10 , indicating a fairly low orbital inclination , and the spectrum and synthesized magnitude indicate that the spectrum was taken near minimum light . _ _ our spectrum has poor signal - to - noise , but does show broad h@xmath15 and h@xmath22 emission . the synthesized @xmath45 is fainter than the crts minimum , but conditions were partly cloudy at the time . _ css1616@xmath2918 . _ the spectrum is typical of dwarf novae in outburst , with narrow h@xmath15 emission , broad h@xmath22 absorption , and a blue continuum . the synthesized @xmath46 , while the crts lists variation between @xmath47 and @xmath48 . _ css1649 + 03 . _ broad h@xmath15 emission is clearly detected , but the poor signal - to - noise ratio precludes further analysis . our photometric measurements and the synthesized magnitudes both agree well with the magnitude at minimum listed in the crts . _ css1720 + 18 . _ strong , broad h@xmath15 and h@xmath22 confirm the cv status . our spectrum has poor signal - to - noise , which is unsurprising given the object s faintness . _ css1727 + 13 . _ h@xmath15 has a fwhm of 1200 km s@xmath10 , indicating an intermediate inclination . the spectrum is typical of dwarf novae at minimum light , and has a synthesized @xmath49 . it was taken only 3 days after photometric measurements showing @xmath50 , indicating that the star had faded quickly following an outburst . _ css1735 + 15 . _ our spectrum was taken in partly cloudy conditions , but does show h@xmath15 in emission . in addition , nad absorption is present , and the k - star features near @xmath51 are just detected . the orbital period is probably well longward of the 2 - 3 hr gap . we think that the hump in the continuum near 5300 is probably an artefact . _ css1752 + 29 . _ the h@xmath15 emission has a modest equivalent width , and the continuum is blue , suggesting that the system was in outburst . magnitudes from the osmos acquisition image and the synthesized spectrum agree nicely at @xmath52 , while the crts lists @xmath53 for minimum light , so the system was likely declining from outburst . again , the continuum hump near 5300 is probably an artefact . _ css2029@xmath2915 = sy cap . _ @xcite found superhumps with @xmath54 = 0.063759(22 ) d = 91.8 min . using the @xmath0-@xmath54 relation derived by @xcite , we predict @xmath55 min . our spectrum shows narrow h@xmath15 ( fwhm @xmath56 400 km s@xmath10 ) , indicating a low orbital inclination . _ css2059@xmath2909 . _ the h@xmath15 line has a fwhm of 1500 km s@xmath10 , suggesting a fairly low orbital inclination . the crts light curve shows a gradual brightening of the minimum magnitude over time , with outbursts superposed . our spectrum and acquisition image were taken in partly cloudy conditions . _ css2213 + 17 . _ the h@xmath15 portion of our spectrum was unaffected by a dewar condensation problem , and shows a broad emission line , confirming that this is a cv . _ cs2227 + 28 . _ this spectrum was also affected by the condensation , but h@xmath15 was again in the clear portion , and is nicely detected with a somewhat triangular profile , indicating a moderate orbital inclination . we obtained time series spectroscopy for eight targets . figs . [ fig : montage1 ] and [ fig : montage2 ] show the average spectra , periodograms , and folded velocity curves , and table [ tab : parameters ] gives the parameters of the sinusoidal fits . we discuss the objects in order of ra . the spectrum is typical of dwarf novae at minimum light . the lines are single - peaked , and h@xmath15 has a fhwm of 1100 km s@xmath10 . we adopt @xmath57 min . an alternate choice of daily cycle count gives 116.7 min , but the monte carlo test described by @xcite assigns this a probability below 1 per cent . at this orbital period , this is likely to be an su uma star showing superhumps and superoutbursts . the equivalent widths of the emission lines are rather smaller than in most dwarf novae at minimum light , suggesting that we caught the system somewhat above minimum . the lines are single - peaked and the fwhm of h@xmath15 is @xmath58 km s@xmath10 , indicating a fairly low inclination . our @xmath59 min , places this in the period range of the su uma stars . the spectrum shows a conspicuous contribution from a late - type star . using spectra of stars classified by @xcite , we find that the companion s spectral type to be k4.5 @xmath60 1 subclass , and that its contribution to the spectrum is equivalent to @xmath61 . in fig . [ fig : montage1 ] , the upper spectral trace shows the average spectrum , and the lower shows the result of subtracting a scaled k4 star from the average . most of our observations are from 2011 march , but we also have velocities from january and september . an unambiguous cycle count over the whole interval yields @xmath62 hr . the emission- and absorption - line velocities are both modulated ; the emission - line modulation is @xmath63 cycles out of phase from the absorption , consistent with the half - cycle offset expected if the emission lines trace the white dwarf motion . if we _ assume _ that this is the case , then the mass ratio ( secondary to white dwarf ) is @xmath64 . with this mass ratio and a typical white dwarf mass of @xmath65 m@xmath66 , the orbital inclination @xmath67 would be near 35@xmath68 . while this is only meant to be illustrative , the rather low secondary velocity amplitude ( @xmath69 km s@xmath10 ) implies an orbital inclination low enough that eclipses are very unlikely . we estimate a distance using the secondary s contribution as follows . if we fix @xmath0 at its measured value and assume that the secondary star fills its roche critical lobe , then the secondary s radius , @xmath70 depends almost entirely on its mass @xmath71 ; furthermore , the dependence of @xmath70 on @xmath71 is weak . evolutionary calculations by @xcite suggest that a k4 star in an 9-hour orbit should have @xmath72 m@xmath66 . as a check , taking our mass ratio at face value and assuming a broadly typical @xmath73 m@xmath66 for the white dwarf gives @xmath74 m@xmath66 , in reasonable agreement . with this mass range , the approximation given in eqn . 1 of @xcite constrains @xmath70 to be @xmath75 r@xmath66 . from data in @xcite , we estimate the surface brightness of a k4.5 dwarf to be such that it would have @xmath76 if it had @xmath77 r@xmath66 . scaling this to the secondary s radius yields @xmath78 . the secondary s synthetic @xmath14 magnitude is 18.0 , but this is probably a little too faint , because the synthetic magnitude of the average spectrum ( @xmath79 ) is fainter than the @xmath80 we find from the more accurate filter photometry . discrepancies this large are to be expected ( section [ sec : techniques ] ) . correcting for these losses gives @xmath81 for the secondary contribution . at this celestial location ( @xmath82 ) , @xcite estimate the total galactic extinction to be @xmath83 , which taking @xmath84 makes @xmath85 , assuming the star lies outside the dust layer . putting all this together yields @xmath86 , or a distance @xmath87 pc . notice that we did not assume that the secondary star follows a main - sequence mass - radius relation , but rather combined the roche size constraint with the surface brightness . this is an su uma - type dwarf nova , and was observed in superoutburst by @xcite , who detected superhumps . the superhump period @xmath54 was not determined cleanly , but appeared to be 0.0763 d. photometry obtained by @xcite gave @xmath88 h , or 0.0748 d , in rough accordance with expectations based on the superhump period . our spectroscopic period is essentially identical , at 0.07485(5 ) d. the emission lines are barely double - peaked , suggesting that the inclination is not small , but @xcite make no mention of an eclipse . as far as we know , this is the only system studied here which has a period determination in the literature . like css0647 + 49 , this object also has a strong contribution from the secondary star . by comparing the spectrum with stars classified by @xcite , we estimate the secondary s spectral type to be k7 @xmath60 1 subtype , and find that the secondary s contribution is nominally equivalent to @xmath89 . however , many of the exposures used in the average spectrum were taken in partly cloudy conditions and mediocre seeing . our average spectrum has synthesized @xmath90 , but our best exposures have @xmath91 . the crts light curve shows that the source is fairly steady when not in outburst , at @xmath32 , consistent with our best exposures , so it appears that the mean spectrum from which the secondary magnitude was derived is about 0.6 mag too faint . we therefore adopt @xmath92 for the secondary contribution . the emission - line radial velocities did not yield a period , but the absorption spectrum showed an unambiguous modulation at @xmath93 hr . the relatively small velocity amplitude of the secondary ( @xmath94 km s@xmath10 ) constrains the inclination to be fairly low for any realistic white dwarf mass , so eclipses are not expected . we can once again estimate a distance using the secondary s contribution , following the same procedure we used for css0647 + 49 . using @xcite as a guide , we estimate @xmath95 m@xmath66 , from which we infer @xmath96 r@xmath66 at this period . the surface brightness for k7 @xmath60 1 is equivalent to @xmath97 for a 1 r@xmath66 star @xcite . at this location ( @xmath98 ) the @xcite extinction map gives @xmath99 mag . combining this with the apparent magnitude of the secondary gives @xmath100 pc . the emission lines are notably broad the fwhm of h@xmath15 is nearly 1700 km s@xmath10 and the lines show incipient double peaks , so the inclination is likely to be fairly high . even so , the radial velocity amplitude is modest . we detect a modulation in the emission line velocity , but with the available the choice of daily cycle count is ambiguous , so we give two possible periods in table [ tab : parameters ] . both possible periods are well below the 2 - 3 hour gap , so it is likely that this will prove to be an su uma - type dwarf nova with superoutbursts and superhumps . the spectrum is typical of dwarf novae at minimum light . we obtained radial velocities on a single night in 2011 june , after which the star went into outburst , suppressing the h@xmath15 emission and ending the measurements . in 2012 may we obtained velocities on three consecutive nights . from the combined data we adopt a period near 0.0582 d , but periods near 0.111 d and 0.0552 d ( the latter being a daily cycle count alias of our adopted period ) are not completely ruled out . the @xmath56 345 d gap in the time series created fine - scale ringing in the periodogram . to derive the period uncertainty in table [ tab : parameters ] , we shifted the 2012 may data back in time by an integer number of periods , removing the gap . the exact period used in this artificial shift has essentially no effect on the resulting period uncertainty . the spectrum shows weak m - dwarf features , in particular the extra continuum around 5950 - 6170 , and the band head near @xmath101 . the features are too weak to derive good constraints on the spectral type , but an m1.5 dwarf contributing around 25 per cent of the flux at 6500 gives a reasonably good match . as one might expect from the spectrum , the period is relatively long ; the best - fitting @xmath0 is 7.12(3 ) hr , or 3.37 cycle d@xmath10 ; however , we can not entirely rule out an alias at 4.45 cycle d@xmath10 , or 5.40 hr . as noted earlier , the crts cv sample is of great interest in characterizing the cv population . in this section , we consider what the crts sample can tell us about the completeness of the available cv sample . the number of non - cvs included in the crts sample appears to be small . in the sample of 36 objects for which we obtained spectra , we found only a single apparent non - cv . the fainter objects in the crts sample were beyond our magnitude limit , and it is possible that the fainter end of the sample includes a greater fraction of interlopers . however , the selection criterion outbursts of more than 2 mag should be fairly robust even for faint objects . we assume , then , that essentially all crts cvs are real cvs . we compared the crts list to several other samples of cvs , which we describe here . _ the sdss sample . _ szkodyi - viii list 286 cvs . although the sdss data release 8 covers some areas at low galactic latitude that crts does not @xcite , all of the sdss cvs lie within the nominal sky coverage of the crts , so for practical purposes the sdss coverage is entirely contained within the crts coverage . szkodyi - viii do not tabulate the subtypes of their objects , though they do give limited information on this . to enable more detailed comparisons , we classified the objects in szkodyi - viii , primarily on the basis of their spectra , supplemented by the information in the text of the papers . our classification scheme was as follows : dn : : this class was used for objects known to be dwarf novae , and for objects whose spectra resemble those of dwarf novae . the spectra classified this way tended to have strong , broad balmer emission ( with h@xmath15 equivalent width usually greater than 30 ) , relatively flat disk continua ( in @xmath102 vs. @xmath12 ) , and weak or absent heii @xmath124686 . objects showing blue continua and white - dwarf absorption wings around h@xmath22 were classified as dn - w ; if a k or m - type secondary was present , we classified the object as dn-2 . in some cases the sdss spectrum shows the object in outburst . dwarf novae in outburst can be difficult to distinguish from novalike variables , but in these cases the flux level in the spectrum will usually be much greater than expected based on the imaging data . nl : : this class included spectra showing blue continua , without white - dwarf absorption , and relatively weak emission lines , or stronger emission lines and substantial heii @xmath124686 ( typically half the strength of h@xmath22 in those cases ) . the balmer absorption wings in a novalike variable can superficially resemble white - dwarf absorption , but with experience the distinctive white - dwarf line profile can usually be distinguished from the disk absorption lines seen in ux - uma type variables . am : : these showed heii @xmath124686 similar in strength to h@xmath22 , or other evidence of a strong magnetic field such as cyclotron humps . ncv : : we assigned 14 objects to this non - cv " class . this heterogeneous group include objects whose spectra resembled reflection - effect white dwarf systems , subdwarf b stars , and chromospherically active m dwarfs . one , sdss j1023 + 00 , has proven to be a binary containing a millisecond radio pulsar @xcite . removing 14 apparent non - cvs from the sdss sample leaves us with 272 objects . because of the limited information especially the lack of long - term light curves the classifications should be considered somewhat rough . they are given in table [ tab : sdssclasses ] . _ the ritter - kolb catalog . _ @xcite have maintained a catalog ( hereafter rkcat ) of cvs and related object with known or suspected orbital periods . some of these were discovered in the crts , but entries that do not match with crts objects clearly were not . in our comparisons , we used the cataclysmic binary list , ` cbcat ' , from version 7.16 of the ritter - kolb catalog ( hereafter rkcat ) , which contains 926 objects , of which 582 are in the crts survey area . rkcat provides subclassifications similar to the ones we invented for the sdss sample . _ the general catalog of variable stars . _ we downloaded the general catalog of variable stars @xcite ; the 2012 january version contains 45678 entries , of which 753 are classified as some kind of cv ( types n , na , nb , nc , nl , ug , ugss , ugsu , or am ) . of these , 287 are in the crts survey area . table [ tab : samplestats ] shows the number of cross - matches , and non - matches between the crts sample and the lists detailed in the previous section . there are only 44 cross - matches between the crts and the sdss cv samples . [ fig : sdsscrtshist ] shows histograms of the sdss sample and the crts objects that lie in the sdss footprint . note the following : 1 . at minimum light , most of the crts objects are too faint to have been identified as cvs by sdss . the fainter crts objects are for the most part detected in the sdss imaging data ( provided they were in the sdss footprint ) , but were too faint to be selected for spectra . 2 . among the brighter crts objects , somewhat more than half are not in the sdss sample . the top panel shows that only a rather small fraction of sdss objects are recovered by crts . why does crts miss so many cvs ? as noted earlier , one expects crts to preferentially select dwarf novae , and to be less sensitive to other cv subtypes . consistent with this , 40 of the 44 cross - matches between sdss and crts are objects that we classified as ` dn ' . the rkcat and gcvs comparisons ( table [ tab : samplestats ] ) also show this tendency for crts to select dwarf novae ; of the 134 crts objects that are listed in sdss , rkcat , or gcvs , 123 of them are classified as dwarf novae in at least one of these catalogs , and only 11 are other kinds of cv . we therefore confirm that , as one might expect , cvs that are not dwarf novae are largely passed over by crts . the objects recovered by crts are mostly dwarf novae , but are the objects _ not _ recovered by crts _ not _ dwarf novae ? in the latter part of the table , we give the numbers of cvs that lie in the nominal crts survey area , but which are _ not _ recovered by crts . as expected , dwarf novae constitute a smaller portion of the unrecovered objects ; the aggregate dwarf nova fraction is 348/602 , rather than 123/134 . however , these figures also imply that over half the _ unrecovered _ objects actually _ are _ dwarf novae . somehow , 348 dwarf novae that we know of in the crts survey area have slipped through its net . how can we account for this ? some of these non - recoveries ( or ` misses ' ) are to be expected , because dwarf novae from one subclass the wz sge stars erupt very infrequently , in some cases on timescales of decades or more . an even more extreme example , the star gd 552 @xcite , appears identical to a short - period dwarf nova at minimum light , but it has _ no _ observed outbursts . some fraction of the wz sge stars will have been missed simply because they have not outburst during the @xmath103 years that crts has operated . to explore the effect of outburst interval on crts selection , we used data from rkcat , which gives an average outburst interval ( which they denote @xmath104 ) for some of the listed dwarf novae . we arbitrarily chose 700 days , roughly half the time span of the crts survey , as the dividing line between short and long outburst intervals . in addition , we assigned to the long - interval group any dwarf nova subclassified as ` wz ' . many dwarf novae could not be assigned because because @xmath104 was not given ( and they were not classed as wz ) , but sufficient information existed to classify 137 dwarf novae from the crts footprint ; 95 of these were in the short - interval group , and 42 in the long - interval group . the objects that _ were _ recovered in crts included 18 from the short - interval group , and 9 from the long - interval group ; among rk dwarf novae that were _ not _ recovered by crts , there were 77 short - interval systems , and 33 long - interval systems . the fact that so many short - interval systems are _ not _ recovered by crts argues strongly that long outburst intervals are not the main reason crts is missing so many dwarf novae ( although it must account for some cases ) . indeed , the ratio of short to long interval dwarf novae for crts / rk matches is remarkably similar to the ratio in the group of rk dwarf novae that are not matched to crts . the main reason for the incompleteness must therefore lie elsewhere . dwarf nova outbursts will be missed if they occur between observations . perusal of crts light curves suggests that some parts of the sky are covered rather infrequently . a good number of dwarf nova outbursts must therefore ` slip through the cracks ' in the coverage . this is exacerbated by the 2-magnitude criterion ; not only must the object be caught in outburst , but during that part of the outburst where it is more than 2 magnitudes above minimum . to see whether outburst amplitude might be an important selection factor , we again turned to rkcat , this time finding the outburst amplitude @xmath105 . for objects that did not have superoutburst magnitudes listed , we let @xmath106 ( in their notation ) , but if superoutburst magnitudes were available , we used @xmath107 . [ fig : outbcdf ] shows the cumulative distribution functions of @xmath105 for the recovered and unrecovered dwarf novae . as expected , crts is nearly blind to objects with @xmath108 ; more interestingly , there is a significant bias against smaller outburst amplitudes extending all the way up to @xmath109 . this effect probably arises because , in any given snapshot , a large - amplitude object will have a greater likelihood of being caught at @xmath110 than a small - amplitude object , which would have to be fairly close to its peak brightness in order to exceed the survey threshold . it seems , then , that the crts survey is biased toward large outburst amplitudes . how might this affect other quantities ? [ fig : amplvsper ] we plot @xmath105 against @xmath0 , for those dwarf novae in the crts footprint that have both quantities tabulated in rkcat . while there is a great deal of scatter at any given @xmath0 , there is a clear trend for short - period dwarf novae to have greater outburst amplitude , and hence a greater likelihood of being discovered by crts . the crts sample should therefore be biased toward shorter orbital periods . this must contribute at some level to the preponderance of short periods found in this paper and by @xcite . the fact that so many of the crts cv sample are new discoveries , and its continuing effectiveness in finding new ones , both show that a great many cvs remain undiscovered . future synoptic surveys with faster cadence , and less - stringent variability criteria should discover many more cvs . we obtained spectra of 36 crts cv candidates , and confirmed that all save one appear to be _ bona fide _ cvs . for eight of the objects we obtained spectroscopic periods , and found that three of them had @xmath0 longward of the 2 - 3 hour gap . in addition , we examined the overlap between the crts cv sample and others previously - existing samples . most crts cvs are new discoveries , but crts has not recovered even a majority of the known dwarf novae in its footprint . this suggests that a great many cvs remain undiscovered . analysis of the recovered and unrecovered samples shows that the crts sample is biased toward large outburst amplitudes , which in turn biases it toward shorter orbital periods . we gratefully acknowledge support from nsf grants ast-0708810 and ast-1008217 , and thank dartmouth undergraduates jason spellmire and erin dauson for conscientious and cheerful assistance at the telescope . samus n.n . , durlevich o.v . , kazarovets e v. , kireeva n.n . , pastukhova e.n . , zharova a.v . , et al . general catalog of variable stars ( gcvs database , version 2012jan ) , moscow : sternberg astronomical institute . http://www.sai.msu.su/gcvs/gcvs/index.htm lrrrrrr css091026:005153 + 204017 & css0051 + 20 & 3 & 15.4 & 18.2 & 2 & 3 css101009:005825 + 283304 & css0058 + 28 & 1 & 14.1 & 19.5 & & 3 css091009:010412@xmath29031341 & css0104@xmath2903 & 1 & 17.2 & 18.8 & & 3 css080921:010522 + 110253 & css0105 + 11 & 4 & 16.4 & 20.0 & & 3 css091016:010550 + 190317 & css0105 + 19 & 4 & 16.3 & 19.6 & & 3 css080922:011307 + 215250 & css0113 + 21 & 4 & 15.1 & 20.2 & & 3 css081220:011614 + 092216 & css0116 + 09 & 3 & 16.1 & 19.2 & & 3 css091029:015021@xmath29124425 & css0150@xmath2912 & 6 & 16.3 & 19.1 & & 3 css101207:020804 + 373217 & css0208 + 37 & 3 & 15.5 & 17.9 & 2 & 5 css090928:032812 + 280631 & css0328 + 28 & 6 & 16.2 & 18.2 & 5 & 4 css081030:035035 + 353247 & css0350 + 35 & 9 & 16.5 & 19.2 & 2 & 2 css081024:040141 + 080008 & css0401 + 08 & 5 & 15.4 & 18.7 & 2 & css081118:041139 + 232220 & css0411 + 23 & 3 & 15.5 & 18.3 & 2 & css090219:044027 + 023301 & css0440 + 02 & 4 & 16.2 & 18.5 & 2 & css081201:044725 + 092439 & css0447 + 09 & 5 & 15.4 & 18.0 & 3 & css091024:050124 + 203818 & css0501 + 20 & 9 & 16.4 & 17.8 & 52 & css100321:050527 + 083415 & css0505 + 08 & 1 & 16.1 & 18.0 & 2 & 1 css101128:051458 + 083503 & css0514 + 08 & 4 & 14.3 & 18.5 & 3 & css090219:051815@xmath29024503 & css0518@xmath2902 & 15 & 15.6 & 17.5 & 2 & css111118:051923 + 155435 & css0519 + 15 & 8 & 14.5 & 17.7 & 47 & css111003:054558 + 022106 & css0545 + 02 & 3 & 14.5 & 18.7 & 2 & 5 css100114:055843 + 000626 & css0558 + 00 & 3 & 15.7 & 19.0 & 1 & 5 css091029:064729 + 495027 & css0647 + 49 & 1 & 13.3 & 16.3 & 33 & 4 css110413:065037 + 413053 & css0650 + 41 & 6 & 16.7 & 1.0 & & 5 css080409:081419@xmath29005022 & css0814@xmath2900 & 2 & 14.8 & 19.0 & 31 & 5 css090224:082124 + 454135 & css0821 + 45 & 6 & 16.8 & 19.8 & & 7 css090201:090210@xmath29113032 & css0902@xmath2911 & 5 & 13.0 & 17.5 & 28 & css091022:090516 + 120451 & css0905 + 12 & 4 & 16.0 & 19.7 & 2 & css110114:091246@xmath29034916 & css0912@xmath2903 & 7 & 15.3 & 18.0 & 34 & css080130:105550 + 095621 & css1055 + 09 & 6 & 15.3 & 18.5 & 5 & css081117:113951 + 455818 & css1139 + 45 & 2 & 15.2 & 19.5 & 2 & css110205:121119@xmath29083957 & css1211@xmath2908 & 2 & 15.2 & 17.8 & 1 & css100315:121925@xmath29190024 & css1219@xmath2919 & 0 & 18.2 & 19.0 & & 8 css100531:134052 + 151341 & css1340 + 15 & 1 & 14.5 & 18.0 & & 5 css090321:155631@xmath29080440 & css1556@xmath2908 & 6 & 15.7 & 18.4 & 2 & css100408:161637@xmath29181027 & css1616@xmath2918 & 8 & 14.9 & 17.8 & 2 & css080131:163943 + 122414 & css1639 + 12 & 11 & 17.3 & 19.2 & & 4 css090416:164413 + 054158 & css1644 + 05 & 6 & 17.7 & 19.8 & & 4 css100707:164950 + 035835 & css1649 + 03 & 5 & 14.1 & 18.9 & 1 & 8 css110426:170152 + 132131 & css1701 + 13 & 10 & 18.4 & 19.3 & & 4 css090205:170610 + 143452 & css1706 + 14 & 2 & 14.8 & 17.9 & 30 & css090929:172039 + 183802 & css1720 + 18 & 5 & 16.2 & 19.8 & 1 & 4 css090929:172734 + 130513 & css1727 + 13 & 7 & 17.3 & 19.5 & 2 & 8 css100707:172952 + 220808 & css1729 + 22 & 5 & 15.1 & 17.8 & 46 & css110623:173517 + 154708 & css1735 + 15 & 1 & 14.2 & 17.3 & 2 & css110610:175253 + 292219 & css1752 + 29 & 7 & 16.4 & 18.3 & 2 & 1 css091010:202948@xmath29155437 & css2029@xmath2915 & 7 & 14.5 & 18.0 & 3 & 6 css090528:205933@xmath29091616 & css2059@xmath2909 & 11 & 15.7 & 18.0 & 2 & 1 css100404:210954 + 163052 & css2109 + 16 & 4 & 16.8 & 19.3 & & 4 css100520:214426 + 222024 & css2144 + 22 & 2 & 14.7 & 17.1 & & 4 css090622:215636 + 193242 & css2156 + 19 & 2 & 17.1 & 18.5 & & 4 css100615:215815 + 094709 & css2158 + 09 & 1 & 13.2 & 17.6 & & 4 css090917:221344 + 173252 & css2213 + 17 & 2 & 17.2 & 19.0 & 3 & 1 css090531:222724 + 284404 & css2227 + 28 & 2 & 14.5 & 18.0 & 2 & 5 css100520:223234 + 185036 & css2232 + 18 & 3 & 16.5 & 18.8 & & 4 css090929:232716 + 413149 & css2327 + 41 & 3 & 15.9 & 18.4 & & 4 [ tab : star_info ] llrrrr css0051 + 20 & 5944.57 & mn & 600 & 19.0 & 41 + css0208 + 37 & 5587.58 & mt & 1080 & 18.3 & 56 + css0328 + 28 & 5588.62 & mt & 960 & 19.0 & 48 + css0350 + 35 & 5818.93 & os & 1200 & 18.3 & 1.5 + css0401 + 08 & 5587.66 & mt & 960 & 16.5 & 4 + css0411 + 23 & 5587.68 & mt & 1080 & 18.9 & 33 + css0440 + 02 & 5589.66 & mt & 1800 & 19.3 & 130 + css0447 + 09 & 5589.64 & mt & 1800 & 18.9 & 22 + css0501 + 20 & 5591 & mt & 27300 & 18.4 & 70 + css0505 + 08 & 5820.98 & os & 1200 & 18.5 & 23 + css0514 + 08 & 5588.68 & mt & 1800 & 20.0 & 58 + css0518@xmath2902 & 5588.70 & mt & 960 & 16.9 & 4 + css0519 + 15 & 5947 & mn & 27960 & 17.4 & 38 + css0545 + 02 & 5944.86 & mn & 1200 & 18.9 & 72 + css0558 + 00 & 5946.79 & mn & 360 & 16.6 & 7 : + css0647 + 49 & 5638 & mt & 23640 & 17.2 & 20 + css0814@xmath2900 & 5591 & mt & 18360 & 18.7 & 81 + css0902@xmath2911 & 5591 & mt & 23160 & 18.4 & 13 + css0905 + 12 & 5589.78 & mt & 1440 & 20.1 & 130 + css0912@xmath2903 & 5641 & mt & 19577 & 17.9 & 61 + css1055 + 09 & 5587 + 88 & mt & 2280 & 19.4 & 90 + css1139 + 45 & 5589.04 & mt & 1200 & 19.9 & 70 : + css1211@xmath2908 & 5733.66 & mt & 600 & 17.8 & 79 + css1556@xmath2908 & 5592.99 & mt & 1200 & 19.3 & 44 + css1616@xmath2918 & 5588.06 & mt & 960 & 15.9 & 2 : + css1649 + 03 & 5733.89 & mt & 600 & 18.9 & 97 + css1706 + 14 & 5733.8 & mt & 15480 & 17.7 & 84 + css1720 + 18 & 5733.91 & mt & 720 & 20.0 & 120 + css1727 + 13 & 5734.68 & mt & 1200 & 19.0 & 48 + css1729 + 22 & 5739 & mt & 31260 & 18.7 & 60 + css1735 + 15 & 5819.65 & os & 1200 & 18.8 & 17 + css1752 + 29 & 5819.71 & os & 1200 & 17.4 & 16 + css2029@xmath2915 & 5819.76 & os & 1800 & 18.4 & 43 + css2059@xmath2909 & 5820.74 & os & 1200 & & 35 + css2213 + 17 & 5819.92 & os & 2160 & & 80 + css2227 + 28 & 5819.89 & os & 1200 & & 115 + [ tab : spectro_summary ] lrrrrr css0501 + 20 & 55589.6842 & & & @xmath111 & @xmath112 + css0501 + 20 & 55589.6914 & & & @xmath113 & @xmath114 + css0501 + 20 & 55589.7017 & & & @xmath115 & @xmath112 + css0501 + 20 & 55589.7088 & & & @xmath116 & @xmath117 + css0501 + 20 & 55589.7160 & & & @xmath118 & @xmath119 + css0501 + 20 & 55589.8673 & & & @xmath120 & @xmath121 + css0501 + 20 & 55589.8745 & & & @xmath122 & @xmath121 + css0501 + 20 & 55589.8817 & & & @xmath123 & @xmath119 + [ tab : velocities ] css0501 + 20 emn & 55590.8497(14 ) & 0.07481(17 ) & 52(6 ) & @xmath124 & 48 & 20 + css0519 + 15 emn & 55946.935(2 ) & 0.0849(3 ) & 47(7 ) & @xmath125 & 46 & 23 + css0647 + 49 absn & 55641.756(2 ) & 0.37149(2 ) & 117(4 ) & @xmath126 & 35 & 13 + css0647 + 49 emn & 55641.574(2 ) & 0.37155(4 ) & 85(4 ) & @xmath127 & 31 & 11 + css0647 + 49 comb & & 0.37150(2 ) & & & & + css0814@xmath2900 emn & 55591.6527(12 ) & 0.07485(5 ) & 61(7 ) & @xmath128 & 31 & 22 + css0902@xmath2911 absn & 55590.747(3 ) & 0.2757(4 ) & 100(6 ) & @xmath129 & 24 & 19 + css0912@xmath2903 emn & 55640.825(2 ) & 0.0653(3 ) & 65(15 ) & @xmath130 & 29 & 38 + ( alternate ) & 55640.819(2 ) & 0.0610(3 ) & 61(15 ) & @xmath131 & 29 & 40 + css1706 + 14 emn & 55733.7552(15 ) & 0.05823(15 ) & 40(6 ) & @xmath132 & 50 & 24 + css1729 + 22 emn & 55738.704(7 ) & 0.2966(14 ) & 69(10 ) & @xmath133 & 42 & 35 + ( alternate ) & 55738.758(8 ) & 0.2248(10 ) & 60(11 ) & @xmath134 & 42 & 40 + [ tab : parameters ] llrrrr css0051 + 20 & 5823.842 & 0.55(5 ) & 18.66(1 ) & & 1.49(2 ) + css0058 + 28 & 5823.806 & 0.13(9 ) & 19.38(3 ) & & 0.20(5 ) + css0104 - 03 & 5823.819 & 0.17(18 ) & 19.80(5 ) & & 0.70(7 ) + css0105 + 11 & 5823.827 & 0.44(23 ) & 20.36(7 ) & & 0.65(9 ) + css0105 + 19 & 5823.834 & -0.09(11 ) & 19.71(4 ) & & 0.07(7 ) + css0113 + 21 & 5823.850 & -0.12(22 ) & 20.60(8 ) & & 1.01(10 ) + css0116 + 09 & 5823.856 & 0.24(6 ) & 18.89(2 ) & & 0.64(3 ) + css0150 - 12 & 5823.864 & 0.02(9 ) & 19.13(3 ) & & 0.65(4 ) + css0208 + 37 & 5585.586 & 0.22(2 ) & 18.23(1 ) & 0.24(1 ) & 0.38(1 ) + css0211 + 17 & 4524.599 & 0.18(5 ) & 19.23(3 ) & & 0.90(4 ) + css0328 + 28 & 5585.597 & 0.81(3 ) & 18.55(1 ) & 0.40(1 ) & 0.98(1 ) + css0350 + 35 & 5818.924 & & 18.32(2 ) & & + & 5818.944 & & 18.28(2 ) & & + css0505 + 08 & 5820.972 & & 18.86(3 ) & & + css0545 + 02 & 5937.729 & 0.47(9 ) & 18.75(2 ) & 0.83(2 ) & 1.71(2 ) + css0558 + 00 & 5937.742 & & 20.45(10 ) & 0.94(11 ) & 2.09(11 ) + css0647 + 49 & 5585.720 & 0.96(2 ) & 16.84(0 ) & 0.66(1 ) & 1.22(1 ) + css0650 + 41 & 5940.758 & 0.72(13 ) & 20.07(3 ) & 0.65(4 ) & 1.38(5 ) + css0814 - 00 & 5585.812 & & 19.35(4 ) & 0.95(4 ) & 1.27(4 ) + & 5585.808 & & 19.14(4 ) & & 1.22(5 ) + css0821 + 45 & 5940.842 & -0.03(17 ) & 20.52(6 ) & 0.48(10 ) & 1.54(7 ) + css0845 + 03 & 4524.703 & & 19.91(1 ) & & 0.47(3 ) + css1219 - 19 & 5940.009 & & 18.02(3 ) & 0.31(5 ) & 0.36(6 ) + & 5940.020 & 0.49(7 ) & 18.02(2 ) & 0.22(4 ) & 0.20(5 ) + css1340 + 15 & 5940.045 & 0.99(12 ) & 18.32(1 ) & 0.69(2 ) & 1.15(2 ) + css1631 + 10 & 5362.767 & -0.16(3 ) & 18.51(2 ) & & 0.81(4 ) + & 5371.767 & 0.38(11 ) & 18.19(3 ) & 0.33(4 ) & 0.71(4 ) + css1639 + 12 & 5730.810 & & 20.52(16 ) & 0.09(21 ) & 0.74(20 ) + css1644 + 05 & 5731.796 & & 20.36(7 ) & 0.38(8 ) & 0.74(8 ) + css1649 + 03 & 5730.816 & 0.50(15 ) & 18.98(5 ) & 0.15(7 ) & 0.51(7 ) + & 5731.807 & 0.17(7 ) & 19.01(2 ) & 0.17(3 ) & 0.49(3 ) + css1701 + 13 & 5730.841 & & 20.85(19 ) & 0.37(23 ) & 0.66(26 ) + css1720 + 18 & 5730.833 & & 20.46(12 ) & 0.42(15 ) & 0.80(16 ) + css1727 + 13 & 5730.824 & 0.29(2 ) & 17.14(1 ) & 0.17(1 ) & 0.38(1 ) + & 5731.815 & 0.19(2 ) & 16.96(1 ) & 0.15(1 ) & 0.30(1 ) + css1752 + 29 & 5819.705 & & 17.34(1 ) & & + css2029 - 15 & 5732.917 & 0.12(10 ) & 18.73(4 ) & 0.18(5 ) & 0.56(5 ) + & 5819.750 & & 18.05(2 ) & & + & 5819.751 & & 18.19(2 ) & & + css2109 + 13 & 5732.863 & 0.21(2 ) & 16.56(1 ) & 0.30(1 ) & 0.62(1 ) + css2109 + 16 & 5732.850 & -0.22(12 ) & 19.45(6 ) & 0.53(7 ) & 0.85(7 ) + css2144 + 22 & 5732.872 & 0.50(2 ) & 17.28(1 ) & 0.38(1 ) & 0.72(1 ) + css2156 + 19 & 5732.883 & -0.96(28 ) & 20.83(16 ) & -0.29(23 ) & 0.94(18 ) + css2158 + 09 & 5732.891 & -0.04(3 ) & 17.48(1 ) & 0.13(1 ) & 0.43(2 ) + css2213 + 17 & 5819.899 & & 19.69(6 ) & & + css2227 + 28 & 5732.900 & -0.07(4 ) & 18.17(1 ) & 0.53(2 ) & 1.12(2 ) + & 5819.876 & & 18.76(3 ) & & + css2232 + 18 & 5732.908 & -0.07(18 ) & 20.07(7 ) & 0.47(8 ) & 0.71(9 ) + css2327 + 41 & 5732.974 & & 20.41(22 ) & 0.42(25 ) & 0.98(25 ) + [ tab : photometry ] lcll 001856.93 + 345444.3 & 4 & no & nl + 002603.80 - 093021.0 & 5 & no & ncv + 002728.01 - 010828.5 & 4 & no & dn + 003941.06 + 005427.5 & 4 & no & dn - w + 004335.14 - 003729.8 & 3 & no & dn - w + 005050.88 + 000912.7 & 4 & no & dn + 013132.39 - 090122.3 & 2 & no & dn - w + 013701.06 - 091234.9 & 2 & yes & dn - w + [ tab : sdssclasses ] lrrr sdss and crts & 44 & 40 & 4 + rkcat and crts & 119 & 108 & 11 + gcvs and crts & 34 & 29 & 5 + union & 134 & 123 & 11 + sdss , not crts & 227 & 138 & 89 + rkcat , not crts & 460 & 250 & 210 + gcvs , not crts & 253 & 142 & 111 + union & 602 & 348 & 254 + [ tab : samplestats ] |
we provide conditions for the existence and the uniqueness of strictly stationary solutions of the usual dynamic conditional correlation garch models ( dcc - garch ) . the proof is based on tweedie s ( 1988 ) criteria , after having rewritten dcc - garch models as nonlinear markov chains . we also study the existence of their moments and discuss the tightness of our sufficient conditions . * key words and phrases * : multivariate dynamic models , conditional correlations , stationarity , dcc . | 1405.6905 | in multivariate extensions of garch models , modelers are faced with the problem of correlations ( between asset returns , in most applications ) . the simplest idea is to assume that these correlations are constant in time , and constitute only an additional matrix of parameters . this has provided the class of constant conditional correlations models ( ccc ) , first introduced by bollerslev ( 1990 ) . since ccc models can be seen as the components of first - order markov processes , once such models are rewritten in an extended vector space , it is relatively easy to prove the existence of strictly stationary and explicit solutions , even if the latter are analytically complex : see classical textbooks , for instance francq and zakoan ( 2010 ) . it rapidly became apparent that the assumption of constant correlations is too strong . it does not correspond to economic intuition or many empirical features : see the recent paper by otranto and bauwens ( 2013 ) and the numerous references therein , for instance . therefore , engle ( 2002 ) proposed extending ccc specifications by adding particular dynamics on the ( conditional ) correlation matrices of returns , denoted here by @xmath0 . to ensure modelers are dealing with true correlation matrices , he introduced a non - linear transform : there exists a sequence of variance - covariance matrices @xmath1 such that @xmath2 , and @xmath1-dynamics are specified instead of @xmath0-dynamics directly , contrary to other authors ( tse and tsui , 2002 or pelletier , 2006 , for instance ) . this non - linear transform ensures that @xmath3 is always a correlation matrix , i.e. positive semidefinite with ones on its main diagonal . nonetheless , it considerably complicates the work of stating dcc model stationarity conditions . indeed , analytically tractable solutions to such processes no longer exist . this explains why the existence and uniqueness of dcc model stationarity solutions have not yet been established in the literature , nor have the finiteness of their moments . particularly , this implies that theoretically sound statistical inference procedures do not yet exist , as noted in caporin and mcaleer ( 2013 ) . despite their theoretical shortcomings , dcc models have been used intensively among academics and practitioners . besides numerous applied works , several extensions of the baseline dcc representation have been proposed in the literature : inclusion of asymmetries ( cappiello , engle , and sheppard , 2006 ) , volatility thresholds ( kasch and caporin , 2013 ) , macro - variables ( otranto and bauwens , 2013 ) , univariate switching regime probabilities ( pelletier , 2006 , billio and caporin , 2005 , fermanian and malongo , 2013 ) , among others . other authors have revisited the dcc parameterization itself : billio , caporin , and gobbo ( 2006 ) , franses and hafner ( 2009 ) , etc . therefore , there is an urgent need for new theoretical results concerning the seminal dcc model itself . usually in econometrics , proving the existence of stationary solutions is the first step towards developing a full asymptotic theory ( consistency / asymptotic normality of qml estimates typically , as in comte and lieberman ( 2003 ) in the case of multivariate garch models ) , because laws of large numbers and some clts are easily obtained in this case . in the garch literature , this essential task has been fulfilled by bougerol and picard ( 1992 ) for univariate garch models , by ling and mcaleer ( 2003 ) for multivariate arma - garch models , and by boussama et al . ( 2011 ) for bekk models . in the case of dcc models , a keystone is missing : a theory for inference has been proposed by engle and sheppard ( 2001 ) , but their two stage estimation procedure is contingent on the underlying dcc process being strictly stationary and ergodic ( see their assumption a.2 ) . the goal of this paper is to fill this gap . after introducing some notations , we define dcc models at the beginning of section [ dccrewriten ] . they will be rewritten as `` almost linear '' markov chains in subsection [ dccmarkov ] . the existence of strong and weak stationary solutions is stated in subsection [ mainres ] . subsection [ unicity ] exhibits sufficient conditions to get their uniqueness . we discuss the tightness of our technical conditions from a qualitative standpoint in section [ my_discussion ] . proof of the propositions and theorems are detailed in the appendices . consider an @xmath4 matrix @xmath5_{1\leq i \leq n , 1\leq j \leq m}$ ] . * @xmath6 ( resp . @xmath7 ) means that all elements of @xmath8 are non - negative ( resp . strictly positive ) , and @xmath9_{1\leq i \leq n , 1\leq j \leq m}$ ] . * if @xmath10 , let the diagonal matrix @xmath11_{1\leq i \leq m , 1\leq j \leq m}$ ] and the vector @xmath12_{1\leq i \leq m}$ ] in @xmath13 . * if @xmath10 and @xmath8 is symmetric , @xmath14 denotes the @xmath15 column vector whose components are read from @xmath8 column - wise and without redundancy . to be formal , @xmath16_{1\leq k \leq m^*}$ ] , where @xmath17 for the unique couple of indices @xmath18 in @xmath19 , @xmath20 such that @xmath21^+ + ( i - j+1 ) = k$ ] . this defines a one - to - one mapping @xmath22 between the indices @xmath23 and the pairs @xmath18 , @xmath20 , @xmath24 , i.e. @xmath25 . * @xmath26 denotes the usual kronecker product , and @xmath27 ( @xmath28 times ) . @xmath29 denotes the element - by - element product . if @xmath30 is a vector in @xmath31 , then @xmath32_{1\leq i \leq n,1\leq j\leq m}$ ] . * we will consider several matrix norms , particularly @xmath33 and the spectral norm , defined for any squared matrix by @xmath34 besides , we will consider any norm @xmath35 for vectors , that is not the euclidian norm @xmath36 . then , we can define the norm @xmath37 for matrices by setting @xmath38 note that @xmath39 when @xmath40 . * @xmath41 denotes the spectral radius of the squared matrix @xmath8 , i.e. the largest of the modulus of @xmath8 s eigenvalues . if @xmath8 is positive semi - definite , then @xmath42 and its smallest eigenvalue is denoted by @xmath43 . * for any column vector @xmath44 , we denote @xmath45 and @xmath46 . * @xmath47 denotes a vector of ones , the dimension of which will be implicit . @xmath48 ( resp . @xmath49 ) denotes the @xmath50 matrix of zeros ( resp . identity matrix ) . when the dimension of an identity matrix is not specified , it will be denoted by @xmath51 . * if @xmath8 depends on @xmath52 , then @xmath53 is the matrix @xmath54 $ ] . let us reiterate here the standard dcc model , as introduced in engle ( 2002 ) . consider a stochastic process @xmath55 in @xmath13 , typically a vector of @xmath56 asset returns . the sigma field generated by the past information of this process until up to and including time @xmath57 is denoted by @xmath58 . modeling the expected returns of financial series is a problem per se , that has generated a huge amount of literature . in this paper , our focus will be on the dynamics of the conditional variance - covariance of @xmath59 instead . therefore , following current practice , we will assume we can remove the conditional means of our returns . let @xmath60= : e_{t-1}[y_t]$ ] be the conditional mean vector of @xmath59 . it depends on a vector of parameters @xmath61 . we define a `` detrended '' series @xmath62 by @xmath63=0.\ ] ] for convenience , the conditional mean @xmath64 is assumed to be measurable with respect to @xmath65 . therefore , @xmath66 . let us denote by @xmath67 the variance - covariance matrix of the @xmath68-observations , conditionally on @xmath58 : @xmath69 as usual with dcc - type models , we split the variance - covariance matrix @xmath67 between volatility terms on one side ( in @xmath70 ) , and correlation coefficients on the other side ( in @xmath3 ) : @xmath71 where @xmath72 denotes the `` instantaneous variance '' of the return @xmath73 ( or @xmath74 , equivalently ) , conditionally on @xmath58 . we assume garch - type models on every margin , but with potential cross - effects between all these volatilities : @xmath75 for some deterministic non - negative matrices @xmath76 and @xmath77 , and for a positive vector @xmath78 in @xmath13 . we will set @xmath79_{1\leq k , l \leq m}$ ] , @xmath80 , and @xmath81_{1\leq k , l \leq m}$ ] , @xmath82 . let us introduce the vector of so - called `` standardized residuals '' @xmath83 . obviously , @xmath84=0 $ ] and @xmath85=r_t$ ] . we impose that @xmath86 is positive definite for any positive definite matrix @xmath8 . in this case , the square root of @xmath3 is uniquely defined : see serre ( 2010 ) , theorem 6.1 . this will be our convention throughout the article . the dynamics of correlations are given by the traditional dynamic conditional correlation specification : @xmath87 where the sequence of matrices @xmath88 satisfies @xmath89 for some deterministic matrices @xmath90 and @xmath91 , and for a positive definite constant matrix @xmath92 . obviously , when such a sequence @xmath93 is initialized with @xmath94 non negative definite ( possibly null ) matrices , every @xmath95 , @xmath96 , will be definite positive . in theorem [ thmoments ] , we prove that a `` doubly infinite '' stationary sequence @xmath97 of definite positive matrices exists and which satisfies ( [ qtdef ] ) . we will set @xmath98_{1\leq p , q \leq m}$ ] , @xmath99 , and @xmath100_{1\leq p , q \leq m}$ ] , @xmath101 . in practice , the positive matrix @xmath92 ( or the constant vector @xmath102 in @xmath103 equivalently ) is a parameter that has to be estimated , most often during the first stage . aielli ( 2013 ) noticed that the estimation of the unknown matrix @xmath92 is not straightforward , because it can not be deduced trivially from the unconditional correlation between the standardized residuals @xmath104 . therefore , he introduced a new variety of dcc - garch models ( called cdcc ) , where ( [ qtdef ] ) is replaced by @xmath105 under this new assumption , cdcc can be seen as a particular bekk model ( engle and kroner , 1995 ) . therefore , aielli obtained the existence of strictly and/or weakly stationary solutions , applying the conditions of boussama , fuchs , and stelzer ( 2011 ) on bekk processes . actually , aielli s model ( [ qtdefaielli ] ) is a smart but not intuitive `` ad - hoc '' specification . its main justification appears as essentially technical , to avoid the non - linear feature of engle s original dcc model ( [ qtdef ] ) . under the standard latter specification , dcc models can no longer be rewritten as bekk models and other techniques have to be found . in this paper , we obtain similar results to aielli ( 2013 ) , but by keeping the original specification of dcc models and without relying on another surrounding family of processes . actually , it is possible to rewrite the previous dcc model as a markov chain , that looks like an ar(1 ) process . this rewrite will become a crucial tool when studying stationary solutions hereafter . set @xmath106 where @xmath107 @xmath108 @xmath109 @xmath110 the dimensions of the four previous random vectors are @xmath111 , @xmath112 , @xmath113 and @xmath114 respectively . their sum , the dimension of @xmath115 , is denoted by @xmath116 . with simple block matrix calculations , random matrices @xmath117 and a vector process @xmath118 exist , such that the dynamics of @xmath115 , any solution of the dcc model , may be rewritten as @xmath119 for any @xmath68 . we will write the block matrix @xmath120_{1\leq i , j\leq 4}$ ] with convenient random matrices @xmath121 . knowing ( [ defmarkov ] ) , the underlying process @xmath122 can be seen as a vectorial autoregressive of order one , but with random matrix - coefficients @xmath117 . let us detail the ar(1 ) form of ( [ defmarkov ] ) : * set @xmath123 when @xmath124 , @xmath125,\ ; \text{and}\;\ , t_{12,t}:= \left [ \begin{matrix } b_1 & b_2 & \cdots & \cdots & b_s \\ 0_m & \cdots & \cdots & \cdots & 0_m \\ \vdots & & & & \vdots \\ \vdots & & & & \vdots \\ 0_m & \cdots & \cdots & \cdots & 0_m \end{matrix } \right].\ ] ] * we deduce from equation ( [ dtdef ] ) that @xmath126 let us set @xmath127 , @xmath128,\ ; \text{and}\ ] ] + @xmath129\ ] ] * clearly , matrices @xmath130 , @xmath99 , exist such that @xmath131 similarly , matrices @xmath132 , @xmath101 , exist such that @xmath133 + it is possible to explicitly write the previous matrices @xmath130 and @xmath132.s indeed , with the notations of subsection [ notations ] , @xmath134_{1\leq u , v \leq m^*}$ ] where @xmath135 similarly , @xmath136_{1\leq u , v \leq m^*}$ ] and @xmath137 then , set @xmath138 , @xmath139 , \ ; \text{and}\ ; t_{34,t}:= \left [ \begin{matrix } \tilde{n}_1 & \tilde{n}_2 & \cdots & \cdots & \tilde{n}_\mu \\ 0_{m^ * } & \cdots & \cdots & \cdots & 0_{m^ * } \\ \vdots & & & & \vdots \\ 0_{m^ * } & \cdots & \cdots & \cdots & 0_{m^ * } \end{matrix } \right].\ ] ] * @xmath140 , @xmath141 , and define the @xmath142 matrix @xmath143 .\ ] ] moreover , rewrite @xmath144 where , with obvious sizes , these vectors are @xmath145 @xmath146 since @xmath70 , @xmath95 and @xmath3 are @xmath147measurable , it is easy to see that the filtration induced by the observations is the natural filtration of @xmath122 : @xmath148 for all @xmath68 . from now on , we consider @xmath149 as the filtration that is generated by @xmath122 . this means that @xmath150=e[z | \ic_{t-1}]=e[z | x_{t-1},x_{t-2},\ldots]$ ] , for any random vector @xmath151 . and a process @xmath152 is said to be ( one - order , implicitly ) @xmath153-markov if the law of @xmath154 given @xmath58 is the law of @xmath154 given @xmath155 . intuitively , the sequence @xmath122 is @xmath153-markov because it is the case for the processes @xmath118 and @xmath117 themselves . to prove this formally , we need an assumption concerning the data generating process ( dgp ) of @xmath156 . let us define the @xmath68-vector of innovations by @xmath157 note that @xmath158=0 $ ] and @xmath159=i_m$ ] by construction . the definition of these innovations implies that , for every @xmath68 , @xmath160 nonetheless , we will not establish whether there are equalities between the latter filtrations . technically speaking , this would be equivalent to stating the invertibility of the underlying process . _ * assumption a0 : * _ @xmath161 possesses the markov property with respect to the filtration @xmath153 . in particular , @xmath162= e[\eta_t |x_{t-1}]$ ] for every @xmath68 . obviously , the latter assumption is satisfied if @xmath161 is a sequence of identically distributed and mutually independent random vectors , with @xmath163=0 $ ] and @xmath164=i_m$ ] . for instance , if the random vectors @xmath165 are standardized gaussian and mutually independent , the process @xmath62 is conditionally gaussian , a standard case in practice . [ prop_markov ] under a0 , the process @xmath122 is markov of order one with respect to its natural filtration . see the proof in the appendix . it would be possible to define a slightly different dgp for the dcc model above : consider a @xmath153-martingale difference ( i.i.d . , for instance ) sequence @xmath166 and set @xmath167 , @xmath168=0 $ ] , @xmath169=i_m$ ] . in the latter case , the process @xmath170 above would be defined by @xmath171 . then , it is easy to check that @xmath170 is @xmath153-markov and is a martingale difference . in other words , a0 would apply in such circumstances . the ar dynamics of @xmath115 were defined above thanks to @xmath172 and @xmath173 , which will be stochastic only through @xmath104 , i.e. through the @xmath68-innovation @xmath165 and the @xmath58-measurable matrix @xmath3 . this creates a major difficulty in proving the existence of stationary solutions . in particular , this means that @xmath172 depends on some components of @xmath115 . therefore , it will be difficult to find explicit expressions like @xmath174 for some deterministic and measurable function @xmath175 , because the link between @xmath172 and the past innovations ( or observations ) is highly non - linear . to obtain the existence of stationary solutions in the previous dcc model , we will invoke tweedie s ( 1988 ) criterion . the latter result will provide the existence of an invariant probability measure for the markov chain defined by ( [ defmarkov ] ) . sthis technique has already been used in several papers in econometrics , notably ling and mcaleer ( 2003 ) or ling ( 1999 ) . to get the stationarity conditions of @xmath156 , we have to control the magnitude of the random matrix @xmath172 , which depends on the random variables @xmath176 , @xmath177 . the mean of the latter variables is one , but they are not independent . this is in contrast with ling and mcaleer ( 2003 ) . moreover , unfortunately , the joint law of @xmath178 is a function of @xmath3 , i.e. a function of @xmath155 . that is why we need the following condition.s _ * assumption e1 : * _ for some @xmath179 , @xmath180<\infty$ ] and @xmath181 , where @xmath182.\ ] ] we reiterate that @xmath172 depends on @xmath178 , that @xmath183 , and that the components of @xmath165 are uncorrelated . as such , the coefficients of @xmath184 are finite because all of the coefficients of @xmath3 are less than one ( in absolute values ) . when there are no correlation dynamics , the matrices @xmath185 and @xmath186 are zero and we recover ccc models . in the latter case , our assumption e1 is reduced to the main assumption of ling and mcaleer ( theorem 2.2 ) that was stated for vectorial arma - garch models . _ * assumption e2 : * _ the law of @xmath165 given that @xmath187 is absolutely continuous with respect to the lebesgue measure , and its density is denoted by @xmath188 , for every @xmath189 and @xmath68 . the function @xmath190 is continuous for every @xmath191 and @xmath68 . there exists an integrable function @xmath192 s.t . @xmath193 for every @xmath191 . moreover , @xmath194 \leq \bar h(\|\xx \|)$ ] , for some function @xmath195 that satisfies @xmath196 for every @xmath197 . the latter technical assumption is trivially satisfied when @xmath170 is an i.i.d . sequence of random vectors s.t . @xmath198<+\infty$ ] . otherwise , e2 provides some constraints insofar as the law of @xmath165 depends on the past values of the dcc process . similar conditions may appear in the literature about the non - parametric estimation of conditional expectations . however , most of them relate to the boundedness of @xmath195 and/or its derivatives ( as in assumption 3 in newey 1997 , for instance ) , or to the moments of @xmath195 ( as assumption 1 in donald et al . 2003 , for instance ) . clearly , e2 is weaker than such assumptions . [ thmoments ] under the assumptions a0 and e1-e2 , the process @xmath199 as defined by equations ( [ htdef ] ) , ( [ dtdef ] ) , ( [ rtdcc ] ) and ( [ qtdef ] ) , possesses a strictly stationary solution . the latter process is measurable with respect to the @xmath200-field @xmath153 induced by the observations . moreover , the @xmath201-th moments of a solution @xmath156 are finite . * example 1 : * in practice and for the sake of parsimony , it is usual to assume diagonal - type dcc models , where all the parameter matrices are diagonal , assuming no `` cross - effects '' in terms of volatilities and/or correlations . this means the non - negative real numbers @xmath202 , @xmath203 , @xmath204 and @xmath205 , @xmath206 , exist such that @xmath207 @xmath208 the associated matrices @xmath130 and @xmath132 are also diagonal . set @xmath209 , and check that @xmath210 . now , let us specify the previous assumption e1 when @xmath211 . since @xmath212=1 $ ] for every index @xmath213 , @xmath214 is simply @xmath215 , replacing @xmath178 by one . denote by @xmath216 the characteristic polynomial of @xmath214 , i.e. @xmath217 . it can be seen easily that two polynomials @xmath218 and @xmath219 s.t . @xmath220 exist . here , @xmath218 denotes the characteristic polynomial of the block - matrix @xmath221_{1,\leq i , j\leq 2}$ ] , replacing @xmath178 by one . @xmath219 is the characteristic polynomial of the previous matrix @xmath222 . tedious , but relatively uncomplicated , algebraic calculations provide @xmath223 @xmath224 for some integers @xmath225 and @xmath226 . let @xmath227 be a non - zero root of @xmath216 . if @xmath227 is a root of @xmath218 then there exists an index @xmath228 such that @xmath229 if @xmath230 , this implies @xmath231 on the other side and similarly , if @xmath227 is a root of @xmath219 and if @xmath230 , then there exists @xmath232 s.t . @xmath233 in other words , a sufficient condition to fulfill assumption e1 is @xmath234 nonetheless , to apply theorem [ thmoments ] in the general case , it may be hard to check the condition on the spectral radius of @xmath214 . this is due to the analytical complexity of @xmath235 , @xmath236 , or to the calculation of its eigenvalues , even when @xmath211 . in the next theorem , we provide more explicit conditions in the case @xmath211 , i.e. so that the second - order moments of @xmath156 are finite . these conditions ensure that e1 will be satisfied . in other words , the conditions will be stronger than e1 , but they may be more practical . indeed , it is often important to obtain sufficient conditions that can be written explicitly in terms of the model parameters , for instance for inference purposes ( e.g. the optimization stage to get qml estimates ) . let us consider @xmath35 ( resp . @xmath237 ) an arbitrary norm for vectors in @xmath13 ( resp . @xmath103 ) . denote by @xmath238 and @xmath239 the associated norms for matrices . if [ thsimple ] @xmath240 @xmath241 then assumption e1 is satisfied with @xmath211 . note that the conditions of theorem [ thsimple ] do not depend on the matrices @xmath186 , @xmath101 , which is a relatively unexpected result . once they are satisfied , and under a0 and e2 , theorem [ thmoments ] applies . by choosing @xmath35 as the maximum norm for vectors , it can easily be checked that @xmath242 , and similarly with the matrices @xmath243 . alternatively , we can choose @xmath244 , that induces the spectral norm @xmath245 . obviously , we can choose these norms for @xmath237 and the matrices @xmath130 . it is often of value to assume that the markov chain is initialized at @xmath246 by drawing @xmath247 following its stationary law . introducing the filtration @xmath248 , we can easily see that the dcc solution is now measurable with respect to the @xmath200-field induced by the innovations and the initial value , because , whenever @xmath249 , @xmath250 * example 1 ( continued ) : * consider a diagonal - type dcc model and maximum norms for vectors . in this case , the condition ( [ cond1thsimple ] ) becomes @xmath251 and the condition ( [ cond2thsimple ] ) is @xmath252 these two conditions are stronger than ( [ th1diagdcccond ] ) , as expected . * example 2 : * to reduce the number of free parameters even further , scalar - dcc models are often introduced . in this case , all the unknown matrices are simply products of a scalar and an identity matrix : @xmath253 @xmath254 such models are very popular , because they allow the number of free parameters to be drastically reduced . with obvious notations , the conditions of theorem [ thmoments ] and [ thsimple ] are the same as above : @xmath255 in passing , we recover the usual ( second - order and strict ) conditions of stationarity for garch - type models : @xmath256 even if stationary solutions of the dcc model do exist , we are not initially sure _ a priori _ that they are unique . besides its theoretical interest , this problem has practical implications . for instance , for any process , the convergence of simulated trajectories towards the same stationary law , independently of the initialization stage , is a desirable feature . moreover , the uniqueness of invariant measures of a markov process implies the ergodicity of the stationary solution ( see douc et al . 2014 , corollary 7.17 ) . this is particularly important for inference purposes . indeed , the estimation of dcc models is typically based on m - estimates ( quasi maximum likelihood , for instance ) . these techniques rely heavily on uniform laws of large numbers , that are most often deduced from the ergodicity of the process . the conditions for identifiability and consistency rely on some expectations with respect to the underlying invariant measure of the given stationary process . if , for a given set of parameters , several invariant measures exist , then it becomes difficult to check such conditions . finally , with several underlying invariant measures , we can not exclude the possibility of switches from one stationary trajectory to another , disturbing the econometric analysis ( stationarity tests , statistical uncertainty around estimates , etc ) . unfortunately , this uniqueness is not given `` for free '' by tweedie s lemma [ tweedielemma ] . moreover , the usual arguments concerning the uniqueness of stationary garch - type solutions do not apply here . indeed , under the markov - chain specification given by equation ( [ defmarkov ] ) , the matrix @xmath172 is itself a function of the random vector @xmath115 through the @xmath178 factors . this is a major difference with the ccc case , and we need to find another strategy . in this section , we provide some uniqueness results under some more or less restrictive assumptions . now , we will consider only stationary solutions of the dcc model , as given in section [ mainres ] . we know that such solutions exist under the ( sufficient ) conditions of theorem [ thmoments ] or [ thsimple ] , but it is not necessary to impose such conditions from the outset . obviously , we will need other technical assumptions . _ * assumption u0 : * _ the sequence of innovations @xmath161 is highly stationary and ergodic . _ * assumption u1 : * _ @xmath257 . the matrix @xmath258 has been introduced in subsection [ dccmarkov ] , under the name @xmath259 . since @xmath259 does not depend on time , we have removed the index @xmath68 here . thanks to the latter assumptions , we will be able to bound @xmath260 from above by a stationary process @xmath261 , and from below by a constant . moreover , @xmath262 will be bounded from below . these tools will be crucial in proving the uniqueness of stationary dcc solutions . to do so , let us introduce some intermediate quantities . * the process @xmath261 , defined by @xmath263 where @xmath264 * the constants @xmath265 and @xmath266 . * the constants @xmath267 * @xmath268 and , for every @xmath269 , set @xmath270 let @xmath271 be the @xmath272-squared random matrix @xmath273.\ ] ] note that the sequences @xmath274 , @xmath261 and @xmath275 are stationary and ergodic because any @xmath276 , @xmath277 or @xmath271 is a measurable function of the innovations @xmath170 that are stationary and ergodic under assumption u0 . _ * assumption u2 : * _ @xmath278<\infty$ ] and the top lyapunov exponent of the sequence @xmath275 , defined by @xmath279 $ ] , is strictly negative . such conditions are standard in the garch literature ( see francq and zakoan , 2010 , section 2.2.2 . for instance ) . note that @xmath280 $ ] for any norm @xmath281 . actually , the technical assumptions u1-u2 above will ensure the uniqueness of @xmath282 , @xmath1 and @xmath0 only . to get the uniqueness of @xmath283 and then of @xmath156 itself , we need a last assumption : with the notations of subsection [ dccmarkov ] , set @xmath284,\ ; \text{and}\ ; \bar{t}^ * = e[\bar t_t].\ ] ] note that @xmath285 does not depend on any particular sequence @xmath282 nor @xmath68 , because @xmath286=1 $ ] for every @xmath213 . _ * assumption u3 : * _ @xmath287 [ thunicity ] under a0 and u0-u3 , a strictly stationary solution of the dcc model is unique and ergodic , given a sequence @xmath170 . the latter result can be strengthened in the following particular case , which is commonly encountered in the literature . * assumption u4 : * _ the underlying dcc model is `` partially '' scalar , i.e. scalars @xmath288 exist such that @xmath289 for all @xmath99 . moreover , @xmath290 by setting @xmath291.\ ] ] obviously , u4 is not mandatory to get our uniqueness result , even if it allows the technical condition u2 to be weakened most often , by lowering the @xmath292 terms . in every case , this `` partially '' scalar case encompasses the common practice of scalar dcc ( or scalar multivariate garch ) models . [ thunicitypl ] under a0 and u0-u4 , a strictly stationary solution of the dcc model is unique and ergodic , given a sequence @xmath170 , replacing @xmath293 ( resp . @xmath294 ) by @xmath295 ( resp . @xmath296 ) in u2 . * example 2 ( continued ) : * in the case of scalar dcc models of order one , it is easy to specify the conditions above . here , @xmath297 , @xmath298 assumptions u1 and u4 are equivalent and mean @xmath299 . assumption u4 is fulfilled if @xmath300<0 $ ] , or if @xmath301 < 0.\ ] ] this expectation could be easily evaluated by simulation , by noting that @xmath165 and @xmath277 are independent . finally , @xmath302 \otimes i_m .\ ] ] through elementary algebra , it can checked that the characteristic function of @xmath303 is the function @xmath304 . then assumption u3 means @xmath305 . therefore , as expected , the conditions required for stationary dcc solutions to be unique are more demanding than for them to simply exist , due to u2 . generally , the latter condition will be fulfilled more easily if @xmath306 is `` a lot smaller '' than one , if @xmath56 is not `` too large '' , and if the tails of @xmath165 are not `` too heavy '' . now , let us discuss the sufficient conditions to obtain the existence of stationary dcc solutions , as given in theorems [ thmoments ] and [ thsimple ] . the most explicit ones are ( [ cond1thsimple ] ) and ( [ cond2thsimple ] ) . since scalar dcc models are by far the most commonly used models in the literature , we focus on the conditions of the previous examples 1 and 2 above , particularly ( [ scalardcccond ] ) . in the latter case , the conditions on the coefficients of the volatility process are those generally applied in the univariate garch literature . it can be proved they are necessary and sufficient for the existence of second - order and strictly stationary garch solutions ( see francq and zakoan , 2010 , theorem 2.6 and remark 2.6 ) . more interestingly , we can check empirically how tight the ( now ) new conditions on the coefficients of the @xmath1 process are . in other words , in example 2 , is the constraint @xmath307 close to a necessary and sufficient condition for generating stationary trajectories of @xmath156 , @xmath0 and/or @xmath1s ? for illustrative purposes , we have considered a very simple bivariate scalar dcc model of order one , given by @xmath308 @xmath309 @xmath310 @xmath311',\ ] ] with our notations . the latter process is generated by i.i.d . innovations @xmath170 that are independent standard bivariate gaussian vectors . we initialize the process at @xmath246 with @xmath312 and @xmath313 , @xmath314 . in the paper , we have stated theoretically that the values of the coefficients of the matrices @xmath186 , @xmath315 ( or the coefficients @xmath316 in the scalar case ) do not matter to obtain the existence of stationary dcc solutions . we have verified this fairly counter - intuitive fact empirically : with the model above , different coefficients @xmath317 do not seem to modify the shape of the trajectories we generate , independently of the other parameters . therefore , our experiments will lead with a fixed value @xmath318 . note that this value is larger than one and this could be seen `` naively '' as a source of non - stationarity . as expected , the value one is key for @xmath319 . when the latter is very close to one but less than one ( @xmath320 in our case ) , we check that the simulated trajectories of @xmath156 , @xmath1 and @xmath0 look stationary , once the influences of the starting values have been forgotten ( broadly speaking when @xmath321 ) : see figure [ qtfig ] ( solid lines ) . on the other side , when this auto - regressive parameter is larger than one , even by a small amount ( @xmath322 in our case ) , we observe that the @xmath1 trajectories explode : see figure [ qtfig ] ( dashed lines ) . apparently , this is not the case for the @xmath0 correlation coefficients . they tend towards some constant levels ( figure [ rtfig ] ) , but differ from one experiment to another one . this phenomenon is a consequence of the normalization stage ( [ rtdcc ] ) , but it seems difficult to maintain that feature will happen for almost every trajectory and any dcc model . indeed , by managing increasingly ( very ) high numbers with the @xmath1 process , we can not exclude the possibility of spurious unexpected @xmath0 behaviors . moreover , we have checked that the @xmath156 trajectories that we generated in this case do not seem to exhibit non - stationary patterns visually . this is logical because our dcc model tends towards a ccc model in such situations . therefore , it is likely that modelers will be able to manage ( i.e. evaluate and simulate numerically ) dcc trajectories in practice , even if ( [ cond2thsimple ] ) or its generalizations are not satisfied . actually , this task remains feasible as long as the numerical values of @xmath1 are manageable ( i.e. not too large ) by our software . this is the case when the number of dates is not too large . we have replaced gaussian innovations @xmath170 by fat - tailed random vectors , to check to what extent this may be a source of instability . the @xmath323 components , @xmath314 and @xmath324 , have been drawn following mutually independent standardized student laws with @xmath94 degrees of freedom , @xmath325 . we reiterate here that the @xmath165-moments of order @xmath94 or higher do not exist . as long as @xmath326 and @xmath327 , we do not observe explosive patterns for @xmath156 . the components of @xmath3 appear to become stationary , even if the decrease in initial value effects is very slow when @xmath94 is close to two . on the contrary , when @xmath328 , the processes @xmath1 and @xmath0 are highly unstable . the former exhibit very spiky trajectories , while the latter often tend to be attracted by the @xmath329 or @xmath330 area . besides , some @xmath156 trajectories reach very high and unrealistic values ( @xmath331 for instance ) . in every case , when @xmath332 , the @xmath1 trajectories explode and the @xmath0 ones tend to constant values that depend on each experiment . in such cases , the return process @xmath156 may reach huge values , but only when @xmath333 , apparently . this analysis illustrates the necessity of considering innovations with finite second - order moments , and the fact that higher - order moments are not mandatory to obtain stationary solutions . concerning the sufficient conditions that guarantee that stationary dcc solutions are unique , it is more difficult to evaluate their tightness because they are more intricate and involve too many model characteristics . nonetheless , with the simple scalar dcc model used in example @xmath334 , we observe that the key condition ( [ ex2unicity ] ) will be more demanding when the number @xmath56 of underlyings increases . on the contrary , smaller values @xmath335 would help . and the partially scalar case induces significantly less demanding conditions than the general case , because the values of the constants @xmath336 and @xmath296 in the denominator are a lot higher than @xmath293 and @xmath294 respectively . the effect of @xmath337 is ambiguous because it appears in several quantities , especially @xmath258 and @xmath295 . we thank c. francq and j .- m . zakoan for their valuable remarks and discussions . moreover , we are grateful to eric renault and two anonymous referees , who have proposed a number of ways of improving the article . finally , the authors thank the labex `` ecodec '' for its support . we recall tweedie s criterion , a key tool to prove the existence of an invariant probability measure for a markov chain . this result has a remarkable advantage : contrary to more commonly used techniques ( based on some lyapunov - foster conditions , for example ) , it is not necessary to state the irreducibility of the underlying markov chain , to obtain the existence of stationary solutions . technically speaking , proving the irreducibility of such a non - linear markov chain is a very challenging task in general . let @xmath338 be a temporally homogeneous markov chain with a locally compact separable metric state space @xmath339 . the transition probability is @xmath340 , where @xmath341 and @xmath342 . tweedie s ( 1988 ) theorem 2 provides : [ tweedielemma ] suppose that @xmath122 is a feller chain , i.e. for each bounded continuous function @xmath343 on @xmath344 , the function of @xmath345 given by @xmath346 $ ] is also continuous . 1 . if there exists , for some compact set @xmath342 , a non - negative function @xmath347 and @xmath348 satisfying @xmath349 then there exists a @xmath350finite invariant measure @xmath351 for @xmath352 with @xmath353 . 2 . furthermore , if @xmath354<\infty,\ ] ] then @xmath351 is finite and hence @xmath355 is an invariant probability measure . 3 . furthermore , if @xmath356 then @xmath351 admits a finite @xmath175-moment , that is @xmath357 the following lemma is our version of lemma a.2 in ling and mcaleer ( 2003 ) . therefore , its proof is omitted . [ lemmam ] for a given squared matrix @xmath358 , if @xmath359 , then there exists a vector @xmath7 such that @xmath360 . note that @xmath104 ( or @xmath178 , or even @xmath361 ) is a function of the couple @xmath362 only . due to ( [ rtdcc ] ) and ( [ qtdef ] ) , @xmath3 is a deterministic function of @xmath155 . since @xmath165 is markov with respect to @xmath153 , the law of @xmath104 knowing @xmath58 is the law of @xmath104 knowing @xmath155 merely . the same assertion applies with @xmath172 , @xmath173 , or @xmath115 itself , instead of @xmath104 . in other words , the non - linearity of the dcc model comes mainly from @xmath178 in @xmath172 . but there exist constant matrices ( of zeros and ones ) @xmath363 and @xmath364 such that ( [ defmarkov ] ) can be rewritten @xmath365 where @xmath366 ( resp . @xmath367 ) is the @xmath172 matrix ( resp . @xmath173 vector ) when @xmath368 , @xmath369'$ ] and @xmath370'$ ] . since @xmath371 and since @xmath3 is a measurable function of @xmath155 , then @xmath115 is clearly a function of @xmath155 and of the innovation @xmath165 only , that are markov . these arguments prove the markovian structure of the @xmath122 process under a0 . @xmath372 first , let us check that @xmath122 is a feller chain in a convenient space , to be able to apply lemma [ tweedielemma ] afterwards . let @xmath343 be a bounded and continuous function on @xmath373 . clearly , @xmath374= e[h(t_t\xx + \zeta_t)\,|\ , x_{t-1}=\xx ] } \\ & = & e[h ( \psi_1(\eps_t\eps_t')\xx + \psi_2(\eps_t\eps_t'))\,|\ , x_{t-1}=\xx],\end{aligned}\ ] ] for some continuous transforms @xmath375 and @xmath376 . note that @xmath371 and that @xmath377 is a continuous function of @xmath155 . indeed , @xmath378 is continuous ( see proposition 6.3 in serre ( 2010 ) , e.g. ) , and @xmath379 is continuous by construction . then , @xmath380= e[h\circ \tilde\psi(\xx , \eta_t ) \,| \ , x_{t-1}=\xx]= \int h\circ \tilde\psi(\xx , \eta ) \ , f_{\eta_t}(\eta | \xx ) \ , d\eta , \ ] ] for some continuous transform @xmath381 . now , consider a sequence of vectors @xmath382 that tends to @xmath345 when @xmath383 . since @xmath343 is bounded and since the sequence @xmath384 is convergent for every @xmath385 , we can apply the dominated convergence theorem under e2 . we deduce that @xmath386 $ ] is continuous and then @xmath122 is feller . note that the vector @xmath115 belongs to the metric space @xmath373 , endowed with the usual topology . since we impose that the matrices @xmath95 will be positive definite , @xmath115 will live in a subspace of @xmath373 , where @xmath387 will gather only the components of definite positive matrices . it is easy to check that this subspace is separable and locally compact . therefore , the assumptions of lemma [ tweedielemma ] are satisfied . second , set @xmath388 , for an arbitrary positive vector @xmath8 , that will be chosen after . let us check that the latter function can be invoked as in lemma [ tweedielemma ] . clearly , @xmath389=1 + e\left [ | ( t_t\xx + \zeta_t)^{\otimes p}|'\,|\,x_{t-1}=\xx\right]m.\ ] ] by expanding the kronecker products , we can check that @xmath390 with @xmath391 for some positive constant @xmath392 and any multiplicative matrix norm @xmath393 . note that @xmath394 . recall that @xmath172 is a function of @xmath178 , i.e. of @xmath104 . then , its conditional law depends on @xmath395 , i.e. it is a function of @xmath155 . we deduce @xmath396'm \leq & \leq & & \leq & |\xx^{\otimes p}|'(t^ * ) ' m.\end{aligned}\ ] ] now , choose @xmath8 as provided by lemma [ lemmam ] , when the matrix @xmath358 in this lemma is replaced by @xmath214 . moreover , @xmath371 , and the ( positive definite ) matrix @xmath377 can be chosen so that all its coefficients are less than @xmath397 ( diagonalize this matrix on an orthonormal basis and invoke cauchy - schwartz inequality ) . this implies that constants @xmath398 exist such that @xmath399 when @xmath400 . since @xmath401<\bar h ( \|\xx \|)$ ] by assumption , some constants @xmath402 such that @xmath403<c_{k , l } \bar h(\|\xx \|)^{(k+l)/p}$ ] for any couple @xmath404 , @xmath405 exist . we deduce the boundedness of @xmath406 $ ] , @xmath407 , and @xmath408 \leq c_1\left ( \bar h ( \|\xx \|)^{1/p } \|\xx^{\otimes ( p-1)}\|+\ldots + \bar h ( \|\xx \|)^{(p-1)/p } \| \xx\| + \bar h ( \|\xx \|)^p \right),\ ] ] for some positive constant @xmath409 . applying e2 , we have obtained @xmath410 \leq 1 + |\xx^{\otimes p}|'(t^*)'m + o\left(\sum_{k=0}^{p-1 } \| \xx^{\otimes k } \|\cdot \|\xx\|^{a } \right ) \nonumber } \\ & \leq & g(\xx ) -|\xx^{\otimes p}| ' \left ( i d - ( t^*)'\right ) m + o\left(\sum_{k=0}^{p-1 } \| \xx^{\otimes k } \|\cdot\|\xx\|^a \right ) , \label{et-1g}\end{aligned}\ ] ] for every constant @xmath411 . by lemma [ lemmam ] , @xmath412 is strictly positive . then , a positive constant @xmath413 exists such that @xmath414 for every @xmath116-dimensional vector @xmath345 . set @xmath415 . by a similar reasoning , a positive constant @xmath416 exists such that @xmath417 for every @xmath418 . moreover , by applying hlder s inequality , we have @xmath419 for every @xmath407 . then , a positive constant @xmath420 exists such that * @xmath421 , and * every `` residual '' term @xmath422 is bounded above by ( a scalar times ) @xmath423 , when @xmath424 . therefore , this provides @xmath410 \leq g(\xx ) \left [ 1 -c_0 \frac{n(\xx)}{g(\xx ) } + o\left ( \sup_{k=0,\ldots , p-1 } \frac{n(\xx)^{(k+a)/p}}{g(\xx ) } \right)\right ] } \\ & \leq & g(\xx ) \left [ 1 - \frac{c_0 n(\xx)}{1+c_2 n(\xx ) } + o\left ( \sup_{k=0,\ldots , p-1 } \frac{n(\xx)^{(k+a)/p}}{c_1 n(\xx ) } \right)\right].\end{aligned}\ ] ] let us define the set @xmath425 , for some @xmath426 . when @xmath427 is sufficiently large , we obtain , for any @xmath428 and a power @xmath429 s.t . @xmath431 \leq g(\xx ) \left [ 1 - \frac{c_0}{2c_2 } + o\left ( \frac{\delta^{a-1/p}}{c_1 } \right)\right ] < g(\xx ) \left [ 1 - \frac{c_0}{3c_2 } \right ] . \label{et-1gg}\ ] ] since @xmath432 , it follows that @xmath433 \leq g(\xx ) -\eps$ ] for some @xmath348 . this proves equation ( [ tw1 ] ) in lemma [ tweedielemma ] . therefore , there exists a @xmath200-finite invariant measure @xmath351 for the markov chain @xmath122 , and @xmath434 . for any @xmath52 , equation ( [ et-1 g ] ) provides @xmath389 \leq g(\xx ) + o\left(\sum_{k=0}^{p-1 } \| \xx^{\otimes k } \|\cdot\|\xx\|^a \right)\leq c_3\delta^{(1+a / p)}\ ] ] for some constant @xmath435 that does not depend on @xmath345 . then , @xmath436\leq \int_a \mu ( dx ) e[g(x_t ) \,|\ , x_{t-1}=\xx ] \leq c \delta^{1+a / p } \mu(a ) < \infty.\ ] ] we deduce that @xmath351 is finite and hence @xmath437 is an invariant probability measure of @xmath122 . this implies that a strictly stationary solution satisfying ( [ defmarkov ] ) exists , still denoted by @xmath122 . third , by invoking equation ( [ et-1gg ] ) , we get ( [ tw3 ] ) in lemma [ tweedielemma ] with @xmath438 , for some @xmath439 . since @xmath440 , we obtain @xmath441<\infty . \label{momentsx}\ ] ] in particular , invoking hlder s inequality , this implies that @xmath442<\infty$ ] , for every @xmath443 and every @xmath407 . @xmath372 equation ( [ momentsx ] ) provides a lot more than only the finiteness of @xmath444 s moments . overall , it means that @xmath445<\infty,\;\ ; e_\pi\left [ \sum_{i=1}^m z_{it}^{2p } \right]<\infty,\ ] ] @xmath446<\infty,\;\text{and}\ ; e_\pi\left [ \sum_{i=1}^{m } |\eps_{it}|^{2p } \right]<\infty.\ ] ] let us consider @xmath447 , a non zero eigenvalue of @xmath214 , when @xmath211 . we can easily check that this matrix is simply @xmath172 , by replacing @xmath178 by one , and replacing the coefficients of the matrices @xmath130 and @xmath132 by their absolute values ( remember that the matrices @xmath448 and @xmath243 are already non - negative ) . let @xmath449 be the associated eigenvector , where the dimensions of the subvectors @xmath450 , @xmath451 are consistent with those of @xmath115 in ( [ defxt ] ) . we can further split the latter subvectors , so that they comply with the matrices @xmath448 , @xmath243 , @xmath130 and @xmath132 . with obvious vector sizes , we will denote @xmath452 , @xmath453 , @xmath454 and @xmath455 . by simple block - matrix calculations , the relation @xmath456 implies @xmath457 @xmath458 note that @xmath459 and @xmath460 for every @xmath461 and @xmath462 . moreover , @xmath463 and @xmath464 for every @xmath213 and @xmath465 . if @xmath466 , then @xmath467 similarly , @xmath468 since @xmath469 , we obtain @xmath470 this proves the result . suppose that two strongly stationary solutions @xmath122 and @xmath471 exist . since both satisfy equation ( [ defmarkov ] ) , with obvious notations , we can write for every @xmath68 @xmath472 note that the difference between @xmath172 and @xmath473 is only due to the ( a priori different ) factors @xmath104 and @xmath474 . we want to prove that , for every @xmath68 , almost certainly @xmath475 . the problem will be solved if we prove the uniqueness of the process @xmath476 , given by subvectors of @xmath122 . for the moment , assume it has been proved . recall that @xmath109 @xmath110 @xmath0 is therefore unique , due to ( [ rtdcc ] ) . moreover , the sequence of random matrices @xmath117 and of noises @xmath118 are also unique , similarly to the ccc case . now , let us prove the uniqueness of @xmath477 , knowing @xmath170 . this would imply the uniqueness of the instantaneous volatility process @xmath283 and of the return process @xmath156 themselves . with our notations , we have @xmath478 for every @xmath68 , by setting @xmath479 the arguments are then standard : for instance , see theorem 2.4 s proof in francq and zakoan ( 2010 ) . we recall the reasoning briefly , to get the uniqueness of @xmath480 . note that @xmath481 whenever @xmath236 . since the sequences @xmath480 and @xmath482 are stationary , it is sufficient to prove that @xmath483 tends to zero a.e . when @xmath28 tends to the infinity , for any matrix norm . this is the case under assumption u3 because , for every sequence @xmath282 , @xmath484 \leq \ln e [ \|\bar t_t \bar t_{t-1}\ldots \bar t_1 \|_{1 } ] = \ln \|\left ( \bar t_t^ * \right)^t \|_{1},\ ] ] by invoking jensen s inequality , the stationarity of @xmath485 , and by noting that all the coefficients of the matrices @xmath486 are non - negative . it is well - known that @xmath487 for any squared matrix @xmath488 . apply this result with @xmath489 . therefore @xmath490 $ ] , the top lyapunov exponent of the sequence of random matrices @xmath485 , is strictly negative under assumption u3 . since the sequence of matrices @xmath485 is strictly stationary under u0 , we get @xmath491 with probability one ( theorem 2.3 in francq and zakoan , 2010 ) . this provides the uniqueness of the processes @xmath283 and @xmath156 , once we assume the uniqueness of the processes @xmath1 and @xmath282 . now , let us prove the uniqueness of @xmath476 or , in other terms , of @xmath492 . this task is clearly more tricky , because we will have to deal with the nonlinear feature of the dcc specification . here , the convenient matrix norm will be the spectral norm @xmath493 . consider two stationary solutions @xmath492 and @xmath494 . since the spectral norm is sub - multiplicative , we deduce from ( [ qtdef ] ) that @xmath495 the key point will be to bound from above the terms @xmath496 by a function of @xmath497 . to lighten the indices , we assume @xmath498 . clearly , we have @xmath499 since the rank of @xmath500 is one , @xmath501 . moreover , @xmath502 we deduce @xmath503 since the spectral norm is unitarily invariant , theorem 6.2 in hingham ( 2008 ) provides @xmath504 note that @xmath509_{1\leq i , j\leq m}$ ] and @xmath510 ( cauchy - schwartz ) . since @xmath511 , we get @xmath512 and @xmath513 . similarly , @xmath514 . by lemma [ lambdabounded ] , we obtain @xmath515 globally , we get @xmath516 everywhere . recalling ( [ inegepseps ] ) , ( [ rootrtineg ] ) ( [ boundlambda2 ] ) and ( [ rtglobal ] ) , we deduce @xmath517 setting @xmath520'$ ] , we get @xmath521 for any positive integer @xmath28 . by the stationarity of the @xmath1 and @xmath522 trajectories , the norm of @xmath523 is bounded by a constant that is independent of @xmath68 . moreover , under the assumptions u0 and u2 , @xmath524 tends to zero a.e . when @xmath525 and for any fixed @xmath68 ( see francq and zakoan 2010 , theorem 2.3 ) . we deduce @xmath526 a.e . when @xmath527 , because @xmath528 can be initialized arbitrarily far in the past . this implies that @xmath529 a.e . therefore , @xmath530 a.e . and @xmath531 a.e . , knowing @xmath170 . this concludes the proof of uniqueness . the ergodicity of the ( now unique ) dcc solution is a consequence of corollary 7.17 in douc et al . @xmath372 * proof of lemma [ qtbounded ] : * for any @xmath68 , @xmath536 moreover , since @xmath537 for any vector @xmath345 and @xmath538 for any matrix @xmath488 ( ltkepohl , 1996 , p. 111 ) , we get @xmath539 this proves the inequality @xmath540 for every @xmath68 and @xmath465 . [ lambdabounded ] under assumption u1 , for almost every trajectory of a solution @xmath1 of the dcc model , we have @xmath545 and @xmath546 , where @xmath547 and @xmath266 . in addition , if we assume u4 , then @xmath548 and @xmath549 , with @xmath550 * proof of lemma [ lambdabounded ] : * is it known that , for any two positive definite matrices @xmath488 and @xmath551 , @xmath552 ( weyl s theorem . see ltkepohl , 1996 , p. 75 ) . in our case , we deduce that @xmath553 everywhere , due to equation ( [ qtdef ] ) . we can improve this lower bound in the particular case of `` partially '' scalar dcc models . indeed , in this case , we have @xmath554 introduce the random vector @xmath555 and @xmath556 . because of ( [ ineglambda ] ) , we have @xmath557 for every @xmath68 . under assumption u4 , it is easy to check that @xmath558 is absolutely convergent and that @xmath559 for every @xmath68 . obviously , @xmath560 due to the definition of @xmath561 , this implies that all the components of @xmath562 are the same , i.e. a real number @xmath563 exists such that @xmath564 , @xmath565 . taking the first component of the vectorial equation @xmath566 provides @xmath567 . this states the lower bound of @xmath262 under u4 . consider a fixed index @xmath443 . the reasoning for the sequence @xmath568 is similar , because @xmath569 for all @xmath68 , as this inequality is playing the same role as ( [ ineglambda ] ) . this implies the desired result . |
we present vlt - isaac near infrared ( nir ) spectro - photometric observations of 16 post - starburst galaxies aimed at constraining the debated influence of tp - agb stars on the spectral energy distribution ( sed ) of galaxies with stellar ages between 0.5 and 2 gyr , hence critical for high - redshift studies . post - starburst galaxies are characterised by negligible on - going star formation and a sed dominated by the stellar population formed in a recent ( gyr ) burst . by selecting post - starburst galaxies with mean luminosity - weighted ages between 0.5 and 1.5 gyr and a broad range of metallicities ( based on sdss optical spectroscopy ) , we explore the parameter space over which the relative energy output of tp - agb stars peaks . a key feature of the present study is that we target galaxies at , so that two main spectral features of tp - agb stars ( c - molecule band - head drops at 1.41 and 1.77 m , blended with strong telluric absorption features , hence hardly observable from the ground , for targets at ) move inside the h and k atmospheric windows and can be constrained for the first time to high accuracy . our observations provide key constraints to stellar population synthesis models . our main results are : _ i ) _ the nir regions around 1.41 and 1.77 m ( rest - frame ) are featureless for all galaxies in our sample over the whole range of relevant ages and metallicities at variance with the maraston ( 2005 ) `` tp - agb heavy '' models , which exhibit marked drops there ; _ ii ) _ no flux boosting is observed in the nir . the optical - nir seds of most of our post - starburst galaxies can be consistently reproduced with bruzual & charlot ( 2003 ) models , using either simple stellar populations ( ssp ) of corresponding light - weighted ages and metallicities , or a more realistic burst plus an underlying old population containing up to approximately 60% of the total stellar mass . in contrast , all combinations of this kind based on the maraston ( 2005 ) models are unable to simultaneously reproduce the smoothness of the nir spectra and the relatively blue optical - nir colours in the observations . the data collected in this study appear to disfavour `` tp - agb heavy '' models with respect to `` tp - agb light '' ones . [ firstpage ] galaxies : stellar content stars : agb and post - agb infrared : stars infrared : galaxies galaxies : general galaxies : photometry | 1205.4717 | the modelling of the spectral energy distribution ( sed ) of galaxies and its interpretation in terms of stellar population parameters ( e.g. age , star formation history , metallicity ) via stellar population synthesis ( sps ) is a fundamental tool to understand the properties and evolution of galaxies . yet , we are far from a complete comprehension and a reliable modelling of some stellar evolutionary phases which strongly affect the energy output of stellar populations . among them , the so - called thermally pulsing - asymptotic giant branch ( tp - agb ) phase has been the focus of debate among modellers for several years ( e.g. @xcite ) . this late stage of the agb phase for low and intermediate mass stars ( @xmath4 57 m@xmath5 ) culminates in stellar populations of ages between 0.5 and 1.5 gyr . tp - agb stars emit mainly in the near infrared ( nir ) spectral range given the low temperatures of these stars . quantitative predictions of the lifetimes , luminosities and spectral shapes and features of these stars as well as of their impact on the integrated spectra of even `` simple '' stellar populations , are still very uncertain and strongly model - dependent ( e.g. * ? ? ? one further complication derives from the fact that tp - agb stars are likely to be embedded in envelopes of dust produced by themselves ( e.g. * ? ? ? * ; * ? ? ? * ) , which attenuates the emerging optical and nir radiation by a substantial fraction @xcite . the seminal work of ( * ? * ma05 hereafter ) , who adopted the `` fuel consumption theorem '' approach and calibrated the flux contribution of this phase against optical and nir photometry of magellanic clouds globular clusters , dramatically pointed out how large the impact of tp - agb stars on galaxy seds may possibly be . the ma05 models make two clear and testable predictions in contrast to `` classical '' models ( such as * ? ? ? * bc03 hereafter ) , which rely on standard , less extreme prescriptions for the tp - agb phase : in particular , _ i ) _ the nir flux of stellar populations of ages between @xmath60.5 and @xmath7 gyr is enhanced by a factor up to 3 ( as a function of age and metallicity ) ; in addition , _ ii ) _ strong , sharp absorption features appear in the spectrum , especially at nir wavelengths ( and most notably at 1.1 , 1.41 and 1.77 @xmath2 m ) . these features correspond to the band - heads of carbon composite molecules and should depend on metallicity . they represent the undebatable fingerprints of tp - agb stars as predicted by ma05 models . this can be clearly seen in fig . [ fig : ma05vsbc03 ] , where we confront the optical - nir spectra of simple stellar populations ( ssp ) of three ages ( 0.5 , 1 and 1.5 gyr ) at three different metallicities ( 0.5 , 1 and 22.5 @xmath8 ) as derived from the ma05 models ( black lines ) and the bc03 models ( red lines ) respectively . observationally , detections of near - infrared cn absorption bands have been reported in the integrated spectra of only a few seyfert galaxies @xcite and globular clusters dominated by the stochastic presence of bright carbon stars @xcite . in fact , from a theoretical point of view , carbon absorption features are not a required signature of agb - star populations . sophisticated models of agb stellar evolution predict that convective dredge - up of carbon produced by nuclear fusion during the envelope thermal pulses will increase the abundance of carbon relative to oxygen near the surface , which can eventually lead to the production of a carbon star ( e.g. * ? ? ? * ) . while the efficiency of this process is expected to increase as the metallicity decreases , the integrated spectrum of most intermediate - age stellar populations could well be dominated oxygen - rich agb stars without strong carbon absorption features . the largely different nir flux predictions for tp - agb stars by different models have strong implications in terms of sed interpretation , leading to substantially lower stellar mass and age estimates ( by up to a factor 2 or more ) when rest - frame nir bands are included in the analysis of galaxies with ages around 1 gyr ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the popularity of nir passbands as tracers of stellar mass , both at low and high redshift , thanks to their ( generally ) low sensitivity to age , metallicity and dust in terms of mass - to - light ratio ( m / l ) ( e.g. * ? ? ? * ; * ? ? ? * ) , clearly calls for a proper assessment of the models in this respect ( see also * ? ? ? verifying the reliability of sps models in the regime of ages around 1 gyr is therefore of paramount relevance for studies of galaxy properties and their evolution . since 2005 , a number of contrasting results and claims have been produced about the appropriateness of the ma05 predictions . @xcite showed that the ma05 models fit the seds of high - z ( @xmath9 ) galaxies better than bc03 models . however the same work shows that the performance of the two models differ only marginally in terms of goodness of fit if one allows for dust reddening , although the inferred stellar mass is substantially different . a better agreement of ma05 models with the observed optical - nir colours is also reported by @xcite , who used combined optical spectroscopy and optical - nir photometry to constrain the star formation history of two galaxies . @xcite also report detections of the 1.77 @xmath2 m feature in two magellanic clouds globular clusters , in agreement with ma05 predictions . on the other hand , studies of magellanic clouds globular clusters found significant discrepancies between observations and ma05 predictions . @xcite found that the co line strength ( in the nir k band ) is much weaker than predicted by ma05 in clusters of age @xmath6 1 gyr . @xcite studied the colours of magellanic clouds globular clusters as a function of their age and showed that ma05 models predict optical and nir colours that are too red . moreover the predicted age dependence of these colours does not agree with the data . as opposed , a substantial agreement with models based on much less extreme tp - agb prescriptions is found . as already pointed out by @xcite , post - starburst galaxies ( psb hereafter ) are the best galaxies to look at in order to find the fingerprints of ( tp-)agb stars on their sed . psbs have a star formation history dominated by a burst in the exact age range when tp - agb stars are expected to have the strongest impact , i.e. between 0.5 and 1.5 gyr prior to the epoch of observation . their optical spectrum is characterised by a well developed balmer / d4000 break , the lack of strong emission lines ( which testify a substantial cessation of star formation activity ) and by the very strong balmer absorption lines typical of a stars ( hence the alternative denomination of e+a or k+a galaxies ) . @xcite compared the optical - nir colours of the sample of psbs spectroscopically selected by @xcite to the colours of different sps models . they showed ( see their fig . 16 ) that @xmath10 , @xmath11 colours are not reproduced by ma05 models : not only would the blue portion of this colour plane be only explained by unrealistically low metallicities , but also the age sequences in ma05 models produce trends that are orthogonal to the observed age trends . @xcite performed another photometric test , in which they fitted the stacked ( rest - frame ) uv - optical - nir sed of 62 psbs at @xmath12 , colour selected from the _ newfirm _ medium band survey ( nmbs ) to have ages around 1 gyr , according to both bc03 and ma05 models . also in this case it is found that bc03 models can reproduce the observed seds much better , in particular the observed blue optical - nir colours , which are not reproduced by ma05 . in this paper we present the results of the experiment that we designed to test for the two main predictions of ma05 models , namely the presence of sharp nir spectral features and the boosted nir flux , on a sample of psbs carefully selected from optical spectroscopy and newly observed in nir spectroscopy ( h and k band ) with isaac at the eso - vlt . the selection criteria and properties of the sample are fully detailed in section [ sec : sample_obs ] . here we would like to stress the uniqueness of this work . _ i ) _ for the first time we cover both the optical and the nir range of psbs with flux calibrated spectra to look directly for the nir spectral features at 1.41 and 1.77 @xmath2 m predicted by the ma05 models and resulting from the large impact of tp - agb c - rich stars . these features are unaccessible or extremely difficult to measure from the ground as they roughly coincide with the atmospheric gaps between j and h , and between h and k bands , respectively , for nearby objects . therefore we observed psbs in the redshift range 0.150.25 so that the c - features move well into the observable windows and are minimally affected by uncertain telluric corrections , as shown in fig . [ fig : ma05vsbc03 ] . _ ii ) _ contrary to previous works based either on colour selection @xcite or on a purely phenomenological spectroscopic selection @xcite , the selection of our sample is based on both phenomenological criteria ( high balmer line strength and low emission lines , * ? ? ? * ) and bayesian estimates of light weighted ages from optical spectral indices as in @xcite . this allows us to focus our observations and analysis on the _ precise _ range of stellar ages where tp - agb stars are expected to contribute the most , for a variety of stellar metallicities . the paper is organised as follows . in section [ sec : sample_obs ] we present the sample , the new isaac nir spectroscopic observations and the data reduction , including spectrophotometric calibrations and integration with optical data . in section [ sec : analysis ] we present the results and discuss possible spurious effects and contaminations that might affect our conclusions . a comparison with the analysis of @xcite is also presented . finally , in section [ sec : summary ] we summarise our findings and propose our conclusions and future developments . we draw our psb targets from the spectroscopic sample of the sloan digital sky survey ( sdss , * ? ? ? * ; * ? ? ? * ) data release 7 ( dr7 , * ? ? ? the initial pre - selection was based on the line equivalent width ( ew ) criteria defined in @xcite , namely : ew(h@xmath13)@xmath14 , ew(h@xmath15)@xmath16 , ew([oii])@xmath17 . emission , the sum of the doublet is considered . ] in order to ensure that the two strong nir features corresponding to the band - heads of cn ( 1.41 @xmath2 m ) and c@xmath18 ( 1.77 @xmath2 m ) due to carbon rich agb stars ( cf . * ) are observable from the ground , we require the galaxies to be at @xmath19 , so that 1.41 @xmath2 m and 1.77 @xmath2 m ( at rest ) move into the h- and k - band windows , respectively . this selection is based on the sdss - dr7 casjobs catalog . we further restrict the sample by applying the visibility cuts in order for targets to be observable from paranal during the winter semester : dec@xmath20 deg and ra between 0 and 230 deg or ra@xmath21 deg . finally we require that the stellar light - weighted age of the galaxies is in the range 0.5 to 1.5 gyr , i.e. at the maximum of the predicted tp - agb luminosity contribution , based on the estimates computed as in @xcite and available in the mpa - jhu catalog . given the relative rarity of the post starburst phase and the restrictive age , coordinate and redshift cuts ( the latter particularly relevant as it is almost at the extreme boundary of the sdss main galaxy sample ) , the final sample is composed of sixteen galaxies only . the properties of the sample are summarised in table [ tab : sample ] and in fig . [ fig : sample_diagn ] . this figure illustrates the distribution of our galaxies in the classical h@xmath22 vs @xmath23 plane . in this plane the main sequence of galaxies with smooth continuum star formation history is displayed by the grey intensity levels , which represent the distribution of sdss - dr4 galaxies of increasing ages from top - left to bottom right . galaxies lying above this sequence , such as our galaxies , represented by the red dots with error bars , are characterised by relatively recent bursts of star formation with ages between a few hundred myr to a couple of gyr , which dominate the light of the galaxy . for comparison , tracks of simple stellar populations ( ssp ) with different metallicities from bc03 are shown ; the typical bell - shape of these ssp tracks peaks in h@xmath22 at ages of 0.30.5 gyr ( depending on metallicity ) and decreases to h@xmath24 35 for ages of a few gyr . by design , our sample populates the intermediate range of burst ages in this diagram . the statistical estimates of light - weighted stellar ages , metallicity and mass computed from optical spectral absorption indices ( @xmath23 , @xmath25 , @xmath26 , @xmath27^\prime}$ ] , @xmath28}$ ] ) with the bayesian method of @xcite and used in the selection are reported in tab . [ tab : sample ] ( columns 6 , 7 and 8) , along with the i d ( 1 ) , coordinates ( 2 , 3 ) , redshift ( 4 ) and apparent @xmath29-band petrosian magnitudes ( 5 ) of the galaxies from the sdss . the sample comprises intermediate mass galaxies , between @xmath30 and @xmath31 , with ages in the range @xmath32 gyr and metallicities between 0.2 and 2.3 solar . most galaxies have indeed @xmath33 , as expected given their stellar mass . despite the paucity of galaxies at low age ( @xmath34 gyr ) _ and _ low metallicity ( @xmath35 ) , the sample provides a good coverage of the relevant ages for the tp - agb phase and of the metallicities most typical for galaxies around the characteristic luminosity @xmath36 . [ cols= " < , > , > , > , > , > , > , > " , ] -d4000@xmath37 diagram , a proxy for burstiness vs. age of the stellar population . red points with error bars are the sixteen galaxies selected for the present study ; the underlying grey - scale image represents the density distribution of a complete sample of galaxies from sdss - dr4 . as one can see , our sample selects galaxies dominated by a burst in a range of young to intermediate ages . note that the estimates of h@xmath38 used in this plot are not exactly the same as the e.w . measurements used in the sample selection ( therefore values @xmath39 are not inconsistent with the selection criteria reported in the text ) . ] we note that the estimates of the stellar population parameters of @xcite rely on the comparison with synthetic spectral libraries based on bc03 models , therefore they could be biased with respect to estimates based on other synthesis models , most notably the ma05 ones . however , optical spectral indices short - wards of 5500 are generally well agreed upon among different modellers and , most importantly , are only weakly affected by tp - agb stars , which mainly influence the nir spectral region . thus age and metallicity estimates from optical absorption features are robust against different tp - agb prescriptions and the use of different models . this is even more the case for spectra dominated by a stars , where balmer absorptions are very strong . furthermore , in sec . [ sec : analysis ] ( [ sec : sspcomp ] and [ sec : cspcomp ] in particular ) we thoroughly explore if and how different ( combinations of ) ages and metallicities can possibly explain the observed properties of our galaxies , irrespective of their original selection . based on the sdss optical images and , to minor extent , on our isaac nir images , we note that the majority of galaxies display smooth and centrally concentrated light distributions , typical of early type galaxies , with the only exception of psbj0227@xmath400015 , which has an extended and irregular `` halo '' . however , other four galaxies ( psbj1039@xmath410604 , psbj1046@xmath410714 , psbj1119@xmath411313 and psbj1206@xmath410918 ) have close and possibly interacting companions ; in many cases , faint structures ( like shells or streamers ) can be tentatively seen . this suggests a relatively recent merger or interaction as a possible origin of the ceased starburst in these galaxies . the sixteen galaxies of the sample were observed in the programme i d 086.b-0733 ( p.i . : s. zibetti ) at the ut3 melipal of the eso - vlt at the paranal observatory . the isaac instrument @xcite has been used in two different spectroscopic setups to obtain low resolution spectra in the h and k bands . 1.5 arcsec slit and sws1-lr grism were used in both setups , with central wavelengths of 1.65 and 2.2 @xmath2 m respectively . this gives a spectral resolution of @xmath42 ( 50 ) and @xmath43 ( 70 ) in h and k respectively . observations were carried out in service mode with clear ( not necessarily photometric ) sky and maximum seeing of 1.4 arcsec to minimise slit losses and match the fibre size and seeing of the sdss optical spectra as well as possible . standard abba on - slit nodding with 70 arcsec offsets and 20 arcsec jittering were adopted to allow accurate sky subtraction and bad pixels and cosmic ray rejection . eight exposures of 180 sec each were taken in h band and six exposures of 150 sec each were taken in k , for total 24 minutes of integration in h and 15 minutes in k per target . this eventually yields typical signal - to - noise ratio ( snr ) of @xmath44 and 25 per resolution element in h and k respectively . spectro - photometric standard stars were observed with identical setup immediately after or before the science observations to provide a reliable estimate of the spectral sensitivity function and compute the corrections for telluric absorption . standard data reduction procedures have been implemented in iraf . wavelength calibration has been performed based on the sky oh lines as measured on a median combined image of the scientific frames . this approach turned out to be much more stable and reliable than using the arc lamp calibrations taken in day time . we have tried different approaches to optimally correct the spectra for telluric absorption . as our best solution we adopted the following strategy : the empirical high - resolution transmissivity function of the average atmosphere above paranal ( provided by eso ) is finely tweaked by scaling and changing the resolution until we can exactly compensate for the observed absorption at various wavelengths on the spectrum of the spectrophotometric standard ; the coadded spectrum of each galaxy is then corrected for the same absorption pattern as for the closest ( in time and airmass ) standard star . we have applied the telluric correction to both standard stars and galaxies _ before _ computing the sensitivity function and applying the flux calibration . in order to assemble a sed including the visible range and the h and k nir bands , absolute flux calibration is required . to this goal we have obtained high snr images in @xmath45 and @xmath46 bands in photometric conditions with isaac as part of the same program . `` total '' integrated magnitude derived from these images are used to rescale the spectra . by `` total '' magnitude we adopt the magnitude integrated within the petrosian aperture , as defined by the sdss photometric pipeline on the sdss @xmath29-band images in data release 8 ( dr8 , * ? ? ? dr8 is chosen as it fixes a number of issues especially with sky subtraction . @xmath45 and @xmath46 magnitudes are used to rescale the respective isaac spectra in the nir , while the optical spectrum from the sdss is normalised according to the @xmath29-band petrosian magnitude0205 and psbj0151@xmath400056 ) we find fainter magnitude by more than 0.03 mag ( + 0.040 and + 0.033 respectively ) . these small discrepancies are well within the errors adopted in sec . [ sec : sspcomp ] and [ sec : cspcomp ] . ] . conversions between nir magnitudes natively calibrated in the vega system for the isaac filters into the corresponding flux densities for our spectra are computed based on the spectrum of vega of @xcite , distributed by bc03 . typical total photometric uncertainties for isaac petrosian magnitudes are estimated @xmath47 mag and include photometric zero point fluctuations ( @xmath48 mag , derived from repeated observations of standard stars during the nights ) and background subtraction uncertainties . for the sdss we assume errors of 0.02 mag , which is the maximum reported for our objects by sdss - dr8 . foreground galactic extinction correction is finally applied using the optical depth derived from @xcite and assuming the standard extinction laws of @xcite and @xcite . the spectra of our galaxies , both sdss and isaac , are taken in relatively small apertures or slits centred on the brightness peak : 3-arcsec fibres are used in the sdss with a typical seeing of @xmath49 arcsec while 1.5-arcsec long - slits with seeing of 1.4 arcsec or better are used with isaac . relative aperture size and seeing are designed to conspire and collect a similar fraction of the galaxy light . by rescaling the spectra according to the `` total '' petrosian magnitudes we implicitly assume that stellar population gradients are of limited relevance and the central region of the galaxies are representative of the whole . @xcite have analysed a small sample of post - starburst galaxies with integral field spectroscopy and showed that , in reality , a large fraction of them have cores that are younger than the rest of the galaxy , hence our assumption does not strictly hold . unfortunately , because of different seeing conditions and geometry of the apertures , a precise and homogeneous rescaling of the spectra is very hard to achieve . however , by comparing fluxes and colours that are obtained in the petrosian and in the fibre apertures we can show that the aperture effects for our sample are small enough not to affect our conclusions . first of all , we note that at the relatively high redshift of our sample the apparent size of the galaxies is typically 2.5 arcsec ( median petrosian radius , maximum 3.1 arcsec ) , so that the 3-arcsec sdss fibre typically collects already half of the total flux . the isaac spectra roughly collect the same fraction of flux . the long slit geometry in this case integrates into the spectra not only the central part , but also the outskirts of the galaxy , so that the isaac spectra are indeed even more representative of the whole galaxy . in second place , we have checked by how much `` total '' petrosian colours differ from fibre colours in the optical . fibre colours are computed in such a way that they correspond as closely as possible to the sed portion effectively captured in the sdss spectra . the strongest effects are expected and observed in @xmath50 , since this colour roughly corresponds to @xmath51 at @xmath1 and therefore is the most sensitive to stellar population properties by bracketing the d4000 break . we find that , on average , fibre @xmath50 is bluer by 0.027 mag with respect to the corresponding total petrosian colours , with a rms of 0.037 . only four galaxies ( out of 16 ) have a fibre colour which is bluer than total by more than 0.05 mag : psbj1046@xmath410714 , psbj1015@xmath410103 , psbj0227@xmath400015 and psbj0328@xmath410045 . we dub these galaxies `` blue core '' in the following . for the remainder , differences in colour can be considered as not significant , although the data hint at starbursts preferentially occurring in the central regions of galaxies rather than in the outskirts , consistently with @xcite . looking at @xmath52 , which roughly corresponds to the rest - frame optical colour @xmath53 that will be discussed in the following , we find even smaller differences of 0.012 mag on average ( 0.021 mag rms ) . these are well within the spectrophotometric calibration errors . finally , we estimate the relative aperture corrections between the optical and nir bands . specifically , we consider the aperture corrections in @xmath29 and @xmath45-band , defined by @xmath54 . for the @xmath29-band we use the quantities provided by the sdss database . for the @xmath45-band we use the total petrosian magnitude as derived above , while to obtain fibre magnitudes we first convolve the isaac images to a common psf of fwhm@xmath552 arcsec and then integrate the flux within a circular aperture of 3 arcsecond in diameter , as done in the sdss pipeline . since for most of the isaac images we lack a sufficient number of high snr stars to compute an accurate psf , the final fwhm is approximated within 0.1 arcsec . we find an average difference of aperture corrections in @xmath29 and @xmath45-band @xmath56 mag with rms of 0.1 mag , indicating that on average our galaxies are slightly more concentrated in @xmath29 than in @xmath45 , although with very marginal significance . this is consistent with the existence of few blue cores and , in general , of very small systematic sed variations between the spectral and the total apertures . this analysis of the aperture effects justifies our choice to use the total petrosian magnitude to compute the relative normalisation of the optical and nir spectra . total petrosian magnitudes are much more stable than magnitudes integrated in small apertures comparable with the seeing , which therefore rely on uncertain psf convolution . on the other hand , we have shown that they do not introduce significant biases between different spectral regions . the systematic aperture bias of 0.026 mag between @xmath29 and @xmath45 is in fact well within the photometric uncertainties estimated in the previous section . in the following analysis we will adopt 0.1 mag as standard error on the optical - nir colours and 0.05 mag on the optical colours , to take into account both photometric errors and aperture effects . the composite optical - nir spectra obtained as explained in the previous section are presented in fig . [ fig : specs_full ] , in the rest frame and normalised to the flux density at 5500 , in order to ease a prompt comparison between all galaxies . the actual normalisation refers to the median flux in the range 54805520 , in order to minimise the effect of noise on the individual spectral pixels . the sdss portion of the spectra covers approximately the rest - frame range 30007500 , isaac h 1.251.50 @xmath2 m , and isaac k 1.652.00 @xmath2 m . for each galaxy , three panels display the full optical - nir spectral range ( top left plot ) , a zoom into the optical region that includes the most significant diagnostic lines ( top right plot ) , and the nir h k range ( bottom plot ) . green lines are used to plot the spectra at high resolution : the original @xmath57 and @xmath58 resolution at @xmath59 and @xmath60 per pixel in the rest frame is used for the h and k spectra respectively , while for the sdss spectra ( original resolution @xmath61 at @xmath62 per pixel in the rest frame ) a smoothing over 3 pixels is applied to make the plot more readable . black thick lines are used to ( over)plot heavily smoothed versions of the spectra , in order to ease the identification of possible spectral features , especially in the nir bands : smoothing over 15 pixels is adopted both in the optical and the nir spectra , resulting in rest frame resolution of @xmath63 , 60 and 90 in optical , h and k ranges , respectively . vertical dotted lines mark the position of the main ( expected ) spectral features : the balmer series ( up to h@xmath64 9 - 2 ) and the two features of cn ( 1.41 @xmath2 m ) and c@xmath18 ( 1.77 @xmath2 m ) due to carbon rich agb stars @xcite . for each galaxy in the bottom panel we report also the light - weighted stellar age and metallicity and corresponding uncertainties derived from optical absorption features as in @xcite . all galaxies display typical e+a or post - starburst spectra with very strong balmer absorptions , well developed d4000 break and blue continuum . the actual slope of the optical continuum generally shows the well known dependence on age and metallicity , although dust attenuation may affect the slope as well . we do not attempt to put any constraint on this parameter , but discuss its possible impact on our conclusions below . three galaxies , psbj0227@xmath400015 , psbj0328@xmath400045 , and psbj1150@xmath400107 show significant h@xmath15 emission ( and some higher ionisation lines for psbj0328@xmath400045 ) , indicating some residual ongoing star formation , which is also confirmed by ( marginal ) detections at 22 @xmath2 m ( w4 ) in the wise all sky survey @xcite , measured with snr of 4.4 and 2.8 , and 2.0 , respectively . however , the equivalent width in all cases matches the @xcite criteria , being @xmath65 , @xmath66 , and @xmath67 , respectively , all larger than @xmath68 . given the observed strength of all balmer absorption lines , the contribution of newly formed stars ( or agn ) can be largely neglected in comparison to the radiation by the intermediate - age population we aim at studying here . it is worth noting that two of these galaxies ( psbj0227@xmath400015 and psbj0328@xmath400045 ) also have a blue core , supporting the idea that blue cores might be related to some residual star formation extending close to the present time . the absence of emission lines and blue cores from the majority of the galaxies in our sample further confirms their nature of genuine post - starburst galaxies . what is most important for the scope of this work is the lack of any evidence for sharp features in the nir in any of the 16 galaxies . most of them display completely featureless continua with approximately uniform slope within the two nir windows . some tentative evidence for a wavelength - dependent slope is seen in a few objects : psbj1039@xmath410604 , psbj1046@xmath410714 and especially psbj1314@xmath410430 appear to change their slope around 1.41 @xmath2 m , while in psbj1119@xmath411313 and psbj1133@xmath410205 a change of slope can be tentatively seen around 1.77 @xmath2 m . similar mild changes of slope , most likely due to molecular bands in the atmospheres of cool stars , are compatible with bc03 but are in fact substantially dissimilar from the sharp drops by a few 10% in flux predicted by ma05 models ( see fig . [ fig : ma05vsbc03 ] ) . to further support these conclusions , we stacked all 16 spectra in h and k separately , after normalising them in a 1000 - wide region around 1.41 and 1.77 @xmath2 m , respectively . the results are displayed in fig . [ fig : stack_hk ] , with the black lines showing the average spectrum and the shaded green area indicating the 16 - 84% inter - percentile range ( approximately @xmath69 range ) . this figure shows that _ on average _ carbon molecules in the atmospheres of evolved stars produce band - head drops that are definitely less than 5% of the continuum . stellar absorption indices in the optical range are extremely powerful diagnostics of stellar population properties . decades of development and tests have led to substantial agreement on their interpretation in terms of ages and global metallicity . in this section we exploit part of the wealth of information that we can extract from the optical sdss spectra of our galaxies for two main purposes : _ i ) _ probe the ability of `` tp - agb light '' models ( represented by bc03 ) and `` tp - agb heavy '' models ( ma05 ) to reproduce the observed trends and distributions in the parameter space of optical absorption indices and optical - nir colours ; _ ii ) _ investigate whether the lack of nir spectral features can be due to selection biases such that we misinterpreted optical diagnostics ( e.g. due to the use of bc03-based model libraries ) and ended up missing the relevant phase for tp - agb stars . we start considering only age - sensitive absorption indices and optical - nir colours and build the diagnostic plots presented in fig . [ fig : diag_optnir_ssp ] for bc03 and ma05 in the _ top row _ and _ bottom row _ , respectively : coloured symbols and lines show the predicted optical - nir colour @xmath70 as a function of @xmath25 , @xmath23 , and the optical colour @xmath53 ( from _ left _ to the _ right _ ) . , @xmath71 and @xmath72 denote the ab magnitudes corresponding to the flux density around 0.55 , 0.70 and 1.40 @xmath2 m ( rest ) , respectively . ] for bc03 , ages range from 50 myr ( lower ages are not relevant for the present analysis as they would yield much bluer colours than observed ) to 13.6 gyr . metallicities range from 1/50 solar to 2.5 times solar , and are represented with different symbols and colours from blue to red ( solar in orange , see figure caption ) . symbols of increasing size are used to mark discrete increasing ages , with the crucial age of 1 gyr identified by a star . the spectral indices @xmath25 and @xmath23 are measured after convolving the model spectra to match the resolution and average velocity dispersion of the galaxies ( @xmath73 ) . the ma05 ssp spectra are distributed over the full optical - nir range only at low resolution ( 20 sampling ) , therefore can not be used to compute reliable indices , @xmath25 in particular . by up to @xmath74 and to overestimate @xmath23 by up to @xmath75 . ] @xcite delivered high resolution spectra for a subsample of the ma05 age - metallicity grid , over the optical range : only three metallicities are available ( 0.5 , 1.0 and 2.0 solar ) , with the lowest ages being 200 , 30 , 400 myr , respectively . we use these models ( in their stelib version ) , convolved to the effective resolution of our data , to compute the @xmath25 and @xmath23 plotted in the first two panels of the _ bottom row _ of fig . [ fig : diag_optnir_ssp ] ; optical - nir colours are computed from the original ma05 seds of corresponding age and metallicity . for the colour - colour plot the full grid of ma05 is used instead , which allows extending the metallicity down to 1/20 solar and the age down to 50 myr as for bc03 . observational data points derived from our composite optical - nir spectra are plotted as error bars . and the optical - nir colour @xmath70 respectively , following the discussion in sec . [ subsec : aper_eff ] . errors on indices are derived from the mpa - jhu catalog and , in the case of @xmath23 , incremented _ a posteriori _ to take into account flux calibration errors , as statistically derived by comparing duplicate observations of a large sample of sdss galaxies . ] as shown by the _ top _ row of fig . [ fig : diag_optnir_ssp ] , bc03 ssps provide a good match to the observed spectral diagnostics , with the only exception of a couple of galaxies , psbj1150@xmath400107 and psbj1206@xmath410918 , which appear to have a too blue @xmath70 nir - optical colour for the given @xmath53 optical colour ( or , reversely , a too red @xmath53 for the given @xmath70 ) and will be discussed in sec . [ subsec : deviant_objs ] . in particular , galaxies cover the range in age and metallicity corresponding to light - weighted ages and metallicities listed in table [ tab : sample ] , which are mostly relevant for the tp - agb phase . as far as the comparison with the ma05 models is concerned , we note that the observed age - sensitive indices approximately match the range of ages given by our light - weighted estimates . most of the observed @xmath25 and @xmath23 values correspond to ssp ages between 0.5 and 1 ( 1.5 ) gyr for the solar ( sub - solar ) metallicity track , while ages smaller than 0.5 gyr would be inferred from the 2 times solar metallicity track . for the given indices , however , the predicted @xmath70 nir - optical colours are on average 0.5 mag redder than the observed ones . this is consistent with the ma05 `` tp - agb heavy '' models having the nir very much boosted with respect to bc03 in the age range around 1 gyr implied by the balmer indices . in the rightmost panel of fig . [ fig : diag_optnir_ssp ] , _ bottom _ row , we see that , unless a metallicity as low as 1/20 solar is representative of most galaxies and ages are significantly larger than 1 gyr , the ma05 models predict @xmath70 nir - optical colours that are on average 0.5 mag redder than observed , for the observed optical @xmath53 colours . we note that the inclusion of dust with a standard effective attenuation curve in the models does not alleviate the tension between the ma05 models and observations . we illustrate this in fig . [ fig : diag_optnir_ssp ] by drawing arrows which represent the offset to apply to each model when an optical extinction of @xmath76 mag and an effective attenuation law @xmath77 ( following * ? ? ? * for an ssp ) are applied . as expected , the effect of dust is to move model tracks to redder colours : this would worsen the disagreement in the index - colour planes . the reddening vector is approximately parallel to the model tracks in the colour - colour plane : accounting for dust attenuation in the models would not substantially change the ( dis)agreement with the observations in terms of colours , although it would require even lower metallicities and ages . vs [ mgfe]@xmath78 diagnostic plot , showing that our galaxies need to have metallicity not lower than @xmath6 40% solar in order to be consistent with the observed strength of metal sensitive absorption lines . the four evolutionary tracks in colour are from bc03 models , colour coded and marked as in fig . [ fig : diag_optnir_ssp ] . ] the hypothesis of a metallicity as low as 1/20 solar is not only unlikely to apply to a full sample of relatively massive galaxies with @xmath79 , but is also inconsistent with direct metallicity constraints which can be derived from absorption indices . in table [ tab : sample ] we already showed metallicity estimates derived from the vast library ( based on bc03 ) and bayesian method of @xcite : these estimates range from 0.2 @xmath8 minimum to 2.3 @xmath8 maximum , roughly uniformly distributed . in fig . [ fig : hbeta_mgfep ] we show that metallicities as low as 1/20 solar are indeed ruled out already from a simple comparison of standard absorption line indices on which different models are in substantial agreement , namely [ mgfe]@xmath78 and h@xmath80 . [ mgfe]@xmath78 in particular is chosen as optimal metallicity indicator following @xcite who demonstrated its insensitivity to @xmath81fe abundance ratio ; h@xmath80 serves to constrain the age of the stellar population and hence alleviate the well known age - metallicity degeneracy . [ fig : hbeta_mgfep ] shows the indices measured in our galaxies as crosses with error bars , while tracks of different colours are derived from bc03 ssps in the age range 50 myr 13.6 gyr ( symbols are the same as in fig . [ fig : diag_optnir_ssp ] ) . four different metallicities , 1/50 , 0.4 , 1 and 2.5 solar are plotted in purple , blue , green orange and red , respectively . it is apparent from this plot that metallicities significantly lower than 0.4 solar are inconsistent with the data - points . we have checked that the same conclusions are obtained if indices based on bc03 models are replaced with those computed by thomas and collaborators @xcite : the only difference between the two sets of models is that thomas models shift the lower limit allowed by the data to slightly higher metallicities . a possible explanation why the strong tp - agb features predicted by ma05 models are not seen in our spectra is that in the nir there is a substantial contribution by old stars , which are almost completely outshined by the younger population in the optical bands . in fact , it is quite unlikely that the entire stellar population of galaxies as massive as several @xmath82 is formed in a burst @xmath62 gyr old and a more realistic scenario is that the burst shines atop an old component . to explore this hypothesis we have repeated the test presented in fig . [ fig : diag_optnir_ssp ] using composite stellar populations ( csp ) instead of ssps . in a first experiment , each csp is composed of two components : a fixed 10-gyr old , solar metallicity ssp ( which we will call `` old component '' in the following ) and an ssp of variable age , from 50 myr to 13.6 gyr , and metallicity , from 1/20 to 22.5 solar ( which we will call `` young '' component or `` burst '' , and corresponds to the ssps of figure [ fig : diag_optnir_ssp ] ) . in the following we refer to these models as csp - s , where `` s '' indicates that the metallicity of the old component is solar . this choice appears reasonable considering that @xmath83 is typical for the old component of massive galaxies with @xmath84 ( e.g. * ? ? ? * ; * ? ? ? we have varied the relative fraction of ( initial ) mass in the old component and in the burst from 0:1 to 0.9:0.1 . [ fig : diag_optnir_csp ] shows the same diagnostic plots we already used to analyse pure ssps ( fig . [ fig : diag_optnir_ssp ] ) , but now plotting csp - s models for the extreme `` contamination '' , i.e. for 90% of the formed mass in the old component and 10% in the burst ( as was also assumed by * ? ? ? symbols and colours are the same as in fig . [ fig : diag_optnir_ssp ] and refer to age and metallicity of the variable burst component alone . csp - s models , either bc03 or ma05 , can easily reproduce the distribution of data - points in @xmath25 and @xmath23 : the contamination by the old component is compensated by a younger burst component ( @xmath85 myr ) with respect to those that provided the best match in the pure burst ( ssp ) models . such a combination would easily explain why we do not see the sharp nir features due to tp - agb stars predicted by ma05 : not only part of the nir emission is provided by the old and tp - agb free stellar population , but also the young component is younger that the peak of the tp - agb phase . however , problems arise when optical - nir colours are considered . the addition of an old component at solar metallicity to a young burst yields significantly redder optical - nir colours with respect to a pure burst . as a consequence , bc03 models , which provided a decent match in the case of a pure ssp , are now shifted to inconsistently red colours with respect to the observations . in the case of ma05 models , which already had troubles being too red in optical - nir colours at almost any age , they become inconsistent with the observed colours of almost all galaxies . mixtures with an intermediate mass ratio between the old population and the burst produce also intermediate results in terms of indices and colours between the pure burst models ( fig . [ fig : diag_optnir_ssp ] ) and the csp - s ones presented in fig . [ fig : diag_optnir_csp ] . in fact , as shown in fig . [ fig : diag_optnir_csp_20 ] , a more moderate ratio between `` old '' and `` burst '' stars of 2:1 yields a decent match between the bc03 models and the observations . as for the ma05 models , however , the fact that already ssps are too red in optical - nir colours results in no combination of `` old '' vs `` burst '' stars being able to reproduce the observed colours . we repeated the experiment with the 9:1 ratio , but changing the metallicity of the old component from @xmath86 fixed to a variable metallicity that matches the one of the `` young '' component . we tag these models as csp - y , where `` y '' indicates that the metallicity of the old component is the same as in the young component . the results are shown in fig . [ fig : diag_optnir_csp_samez ] . bc03 models ( fig . [ fig : diag_optnir_csp_samez ] _ top _ ) are able to cover most of the parameter space occupied by the data - points ; however a significant number of galaxies are too blue in @xmath70 with respect to models at given @xmath23 . moreover , the metallicities implied by colours are systematically lower ( and to large extent inconsistent ) with those implied by the absorption indices alone . as one can argue in the light of the results of pure ssp and 2:1 ratio models , less extreme fractions of the old component can provide a much better match to the data . as for the ma05 models , fig . [ fig : diag_optnir_csp_samez ] _ bottom _ shows that not even with this kind of mix it is possible to match the distribution of @xmath70 vs @xmath23 , the optical - nir colours of the models being too red with respect to the data . we reiterate that any dust with the most commonly adopted attenuation curves would only worsen this disagreement . in summary , adding a substantial fraction of old stellar populations to reduce the impact of tp - agb stars predicted by ma05 in the nir may indeed explain the lack of observed nir features , while remaining consistent with the observed optical spectral indices . however , such a combination fails to reproduce the observed optical - nir colours at given @xmath25 , @xmath23 and optical colours , no matter if the metallicity of the `` old '' component is the same as the one of the `` young '' component or is fixed at the solar value ( as it is probably more realistic for an old stellar population in a system of a few times @xmath87 ) . we can not exclude that more complicated and _ ad hoc _ star formation histories and metallicity mixtures might be able to produce a closer match between observations and ma05 , but it appears unlikely that such a fine tuning actually applies to all 16 galaxies in the sample . on the other hand , bc03 models appear to reproduce the observations much more easily by assuming ssps or csps with even massive `` old '' components of variable sub - solar metallicity ; in order for models with an `` old '' component at solar metallicity to match the observations , `` old '' fractions less than 0.9 are required in most cases : it is sufficient to lower such fraction to 0.6 - 0.7 in order to obtain a generally good agreement . we further illustrate these points by directly comparing ssp and csp model spectra with an actual spectrum of our psb sample in fig . [ fig : example_csp_j1006 ] , where psbj1006@xmath411308 is used as a mere example which optimally serves to this goal , as it is broadly representative of the full sample . for both bc03 ( shown in the left - hand panel ) and ma05 ( in the right - hand panel ) , we pick the best fitting models from each of the three grids of ssp , csp - s and csp - y ( assuming old : young=9:1 ) . they are overplotted to the observed spectrum ( solid black line ) , in blue for the ssp , in solid red for the csp - s and dashed green for the csp - y . the `` best fitting '' models are those yielding the minimum @xmath88 over the optimal set of optical absorption indices defined in @xcite ( @xmath23 , @xmath25 , @xmath26 , @xmath27^\prime}$ ] , @xmath28}$ ] ) . due to the coarse grid in metallicity , the use of fixed old fractions and the neglect of dust attenuation , we warn the reader not to over - interpret these fits in terms of derived physical parameters . nonetheless , the following key observations can be derived from fig . [ fig : example_csp_j1006 ] . _ i ) _ all best fit models can reproduce the observed optical features extremely well ( see insets ) . the slope of the optical continuum is also generally well reproduced , although for bc03 the ssp fit is too blue in the optical range , which might be due either to a small mismatch in metallicity because of the coarse grid or to small amounts of dust , which are not included in our simple models . _ ii ) _ the nir spectrum of the best fitting ma05 ssp , with an age of 0.6 gyr and metallicity 2 times solar , has the typical nir boosting due to the ma05 implementation of the tp - agb phase , clearly recognisable from the sharp features around 1.41 and 1.77 @xmath2 m . however , no evidence of such features appears in the observed spectrum ; moreover , the observed spectrum is almost a factor of two dimmer than the ma05 ssp in the nir . in contrast , the bc03 ssp provides a very good fit in the nir both in terms of ( lack of ) features and of overall flux . _ iii ) _ the use of csps with a substantial fraction ( 90% ) of 10-gyr old stars yields a relative enhancement in the nir over optical for bc03 models , for both csp - s and csp - y . the plot suggests that a smaller old fraction and lower metallicity than the best fitting pure ssp might provide a very close match to the nir - to - optical ratio . csps for ma05 ( which by coincidence converge to the same solution for both csp - s and csp - y , i.e. the solid red line and the dashed green one overlap ) do actually yield much weaker nir features and also lower nir flux relative to optical . however , this stays significantly higher ( @xmath44% ) than observed . psbj1150@xmath400107 and psbj1206@xmath410918 are characterised by extremely blue optical - nir colours ( @xmath89 ) and intermediate / red optical colours ( @xmath90 ) . when plotted on the colour - colour diagrams of fig . [ fig : diag_optnir_ssp ] , [ fig : diag_optnir_csp ] , [ fig : diag_optnir_csp_20 ] and [ fig : diag_optnir_csp_samez ] , they significantly depart from the sequences both of observed data - points and model tracks . none of them is able to reproduce such a combination of intermediate / red optical colours and very blue optical - nir ones : for the given @xmath70 , an offset of @xmath91 to @xmath92 mag in @xmath53 would be required to bring them back in agreement with the models , or , for the given @xmath53 , an offset of @xmath93 mag in @xmath70 . from the arrows plotted in the colour - colour diagrams , it is immediate to see that a standard effective attenuation la @xcite is unable to move the models closer to these two objects : in fact , one would need an effective attenuation curve which is extremely flat , thus causing similar attenuation at 0.70 and 1.40 @xmath2 m . with standard dust grains , for which the absorption cross section decreases with wavelength , this can be only obtained by invoking a peculiar geometry in which back scattering plays a major role . significant residual star formation is witnessed in psbj1150@xmath400107 by ew(h@xmath15)@xmath94 and a marginal detection at 22 @xmath2 m in the wise all - sky survey . psbj1206@xmath410918 , however , does not display any sign of star formation , with ew(h@xmath15)@xmath95 and no detection at 22 @xmath2 m . this suggests that residual star formation or processes related to nuclear activity are not the cause of the anomaly common to these two galaxies . we note that the shape of the spectra of these two galaxies is peculiar in two main aspects : _ i ) _ the continuum appears to bend down , i.e. the slope becomes steeper , going from @xmath96 to 0.8 @xmath2 m , whereas all other galaxies display an approximate constant slope or even an up - bending shape in this region ; _ ii ) _ a significant and featureless drop long - ward of @xmath97 @xmath2 m is observed instead of the fe and mg absorption complexes , which is probably due to the low metallicity of these two galaxies , among the lowest of the entire sample . these two features appear to be responsible for the anomalous colours , by making the spectral slope between 0.7 and 1.4 @xmath2 m too steep relative to the one in the optical range 0.550.70 @xmath2 m . the fact that models fail to reproduce these colours may point to problems in the low metallicity regime , below @xmath98 . we note that , according to the ma05 models , the largest enhancement and the most dramatic features in the nir are expected for @xmath99 . instead these two low - metallicity galaxies exhibit bluer optical - nir colours than even `` tp - agb light '' models ( e.g. , bc03 ) . as already mentioned in sec . [ sec : intro ] , @xcite have obtained a detailed ( rest - frame ) uv - optical - nir sed for a sample of 62 psb galaxies observed in the _ newfirm _ medium band survey ( nmbs ) in the redshift range 0.72 . their composite sed is essentially a stack of the sed of their sample . they applied maximum likelihood fitting using both bc03 and ma05 on different wavelength ranges ( optical only or optical plus nir ) and , similarly to the results we have presented in this work , they found that ma05 can not reproduce the blue optical - nir colours observed , whereas bc03 provide an almost perfect match . in fig . [ fig : comp_kriek ] we compare the sed published by @xcite , represented by the red points with error bars , with the stack of our 16 spectra . both seds are normalised at 5500 . the black line is the median of our 16 spectra at each wavelength , while the shaded area displays the 1684 inter - percentile range . the similarity of the stacked seds from the two works is striking , typically consistent within 11.5 @xmath100 . our spectra are slightly bluer and steeper in the range 5000 to @xmath101 m , which might derive from some relative bias in the respective sample selections . , with our objects being somehow more extreme ( e.g. in terms of burst fraction , see also sec . [ sec : summary ] ) . in fact , @xcite define their sample based on k - corrected rest - frame colour cuts which are hardly comparable to our spectroscopic criteria . alternatively , this comparison might indicate that nir - bright evolved ( tp - agb ) stars have a slightly larger influence on the sed of psb galaxies at @xmath102 than at @xmath103 , consistently with expectations , e.g. , of @xcite . apart from this minor colour difference , our detailed spectroscopic study of individual psb galaxies confirms and reinforces the previous results by @xcite , based on stacking and medium band photometry and at a very different age of the universe . we have presented the first spectroscopic study of post - starburst galaxies covering both the optical ( sdss ) and nir ( our new vlt - isaac observations ) wavelength ranges , which was specifically designed to detect the two main signatures of the presence of tp - agb stars predicted by the @xcite models in contrast to `` tp - agb light '' models ( such as bc03 ) : _ i ) _ the strong nir absorption features caused by carbon composite molecular band - heads ; and _ ii ) _ the boosting of the nir flux . to this goal , we have carefully selected a sample of 16 galaxies with post - burst ages corresponding to the peak contribution by tp - agb stars to the nir flux ( 0.51.5 gyr ) , in a broad range of stellar metallicities ( @xmath1042.5 @xmath8 ) . these galaxies are at high enough redshift ( 0.150.25 ) to allow clean measurements of the nir features at rest wavelengths 1.41 and 1.77 @xmath2 m through the atmospheric h and k windows . despite these ideal conditions for the detection of the characteristic signatures of tp - agb stars predicted by the ma05 models , the 16 galaxies in our sample all display smooth , featureless nir spectra and blue optical - nir colours . interestingly , these observed spectral properties are generally well reproduced using the bc03 models assuming simple ( i.e. single - age ) stellar populations . we have investigated whether the observations could be reproduced with ma05 models as well , by combining stellar populations of different ages . we find that diluting the light of a young ( i.e. a few 100 myr old ) starburst with that of a massive old component allows the ma05 models to reproduce the typical `` e+a '' optical spectra of post - starburst galaxies characterised by strong balmer absorption lines and a well developed d4000 break . at the same time , such models exhibit very weak nir features , in agreement with the observations , because tp - agb stars from the young component have not developed yet , while the nir emission from old stellar populations is smoother . however , these models still can not account for the data , because the predicted ratio of nir to optical flux is significantly larger than observed at the metallicity of our galaxies . hence , the observed lack of deep nir absorption features and the modest ratio of nir to optical flux in our data can not be simultaneously accounted for by a simple `` light dilution '' effect . we note that the observation of metal absorption lines at optical wavelength does not support the idea that the faint nir continuum ( and the consequent dilution and/or absence of strong nir absorption features ) could arise from extremely low metallicities . the ma05 models predict too red optical - nir colours for any choice of ratio between the old and the young component . on the contrary , models based on bc03 fail for `` old '' mass fractions as high as 90% , but produce colours that are consistent with ( most of ) the observations for `` old '' mass fractions up to 6070% , which leaves enough space to accommodate more reasonable sfhs than a ( possibly unrealistic ) single burst . we note that the work of @xcite indicates old fractions close to 90% for their seven psb galaxies , which would be inconsistent with our seds . however , this estimate might be biased by their assumption of fixed solar metallicity . moreover , and probably more importantly , it is possible that our objects are somehow more extreme in terms of burst fraction , because , due to our redshift constraints , we select objects close to the faintest limit of the spectroscopic sample of the sdss . in this way we would select the intrinsically most luminous psb galaxies at @xmath103 , hence those with the largest burst fraction ( i.e. with the lowest `` old '' fraction ) : in fact , for a given total mass , a higher burst fraction yields higher luminosity . the lack of strong colour gradients in most of our psb galaxies lends support to this hypothesis , suggesting that the starburst substantially affected the full extent of the galaxy , rather than the centre only . finally , @xcite point out that their sample is less extreme than typical samples of psb galaxies in the literature due to the small physical size of the spectroscopic fiber ( few hundred parsecs ) , which bias their selection towards central rather than extended starbursts . in conclusion , the consistency of our seds with bc03-based csps and `` old '' mass fractions up to 6070% does not raise any problem for the physical interpretation of our psb galaxies using the bc03 models . we also find that attenuation by interstellar dust , using standard prescriptions , can only worsen the disagreement ( when present ) between spectral evolution models and the observed optical - nir colours of post - starburst galaxies in our sample . this is particularly true for the ma05 models , which predict systematically redder colours than observed . we note that heavier - than - implemented attenuation by _ circumstellar _ dust of the most luminous c stars about to eject their envelopes could potentially reduce the contribution by these stars to the integrated spectrum and perhaps improve the agreement of the ma05 models with observations . a more accurate theoretical and empirical assessment of the effects of circumstellar dust absorption is definitely key to make progress in the sed modelling of the stellar population affected by tp - agb stars ( e.g. * ? ? ? * ; * ? ? ? the results presented in this work confirm the previous claim by @xcite that bc03 , and more generally `` tp - agb light '' models , provide a better agreement with the spectral properties of galaxies hosting a stellar population at the peak of the tp - agb phase than ma05 models . the important novelty of our work is to base these conclusions on a higher - resolution , high - quality spectrophotometric dataset and an optimal spectroscopic sample selection . somehow in contrast with these studies and our work in particular , @xcite have detected the 1.77 @xmath2 m feature in a couple of magellanic cloud s globular clusters and found them in agreement with ma05 predictions . this result calls for more observations in order to obtain a representative sample of globular clusters over a broader range of metallicities ( the two in which the detection was obtained are both at metallicity between 0.2 and 0.6 solar ) and beat down stochasticity effects : in fact , a single ( tp-)agb star can produce a substantial fraction of the total cluster nir luminosity in the mass range explored . one further possibility to reduce the apparent disagreement between the @xcite results and ours might reside in a technical aspect of their challenging observations : the addition of targeted agb stars outside of the central region to integrate the central spectrum and obtain full spatial coverage of the globular clusters might lead to agb stars being over - represented with respect to the stars that contribute the bulk of the diffuse emission . in the light of this complex and partly controversial observational framework , the apparent agreement of the bc03 models ( of either simple or composite stellar populations ) with the observations presented in this paper should not be necessarily interpreted as an indication that the description of the tp - agb phase in these models is adequate . new models incorporating more recent developments in the theory and observation of tp - agb stars ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) are being assembled ( charlot & bruzual , in preparation ) . by providing a better account of observations , new models of this type will allow deeper insight into the physical properties of galaxies from spectral interpretations , especially at high redshift , where a large number of galaxies is going through the evolutionary phase affected by tp - agb stars . in this perspective , the present work provides key and very stringent constraints for the new generation of stellar population synthesis models . we thank the anonymous referee for useful comments and suggestions that significantly improved the clarity of the presentation . + the dark cosmology centre is funded by the danish national research foundation . + we also thank paolo cassata , harald kuntschner and mariya lyubenova for useful comments . s.z . acknowledges the generous hospitality of the dark cosmology centre and stimulating discussions with jens hjorth and sune toft . d.p . acknowledges the kind hospitality at the max - planck - institut fr extraterrestrische physik . g. , 2007 , in vallenari a. , tantalo r. , portinari l. , moretti a. , eds , from stars to galaxies : building the pieces to build up the universe vol . 374 of astronomical society of the pacific conference series , stellar populations : high spectral resolution libraries . improved tp - agb treatment . |
we present the first representation of the general term of the rayleigh - schrdinger series for quasidegenerate systems . each term of the series is represented by a tree and there is a straightforward relation between the tree and the analytical expression of the corresponding term . the combinatorial and graphical techniques used in the proof of the series expansion allow us to derive various resummation formulas of the series . the relation with several combinatorial objects used for special cases ( degenerate or non - degenerate systems ) is established . | 1011.1751 | rayleigh - schrdinger ( rs ) perturbation theory is a venerable technique to calculate the eigenvalues and eigenvectors of @xmath0 from those of @xmath1 . it was created in 1894 by lord rayleigh to describe the vibrations of a string @xcite and adapted to quantum mechanics by schrdinger in 1926 @xcite . rs perturbation theory has been used in all fields of quantum physics ( particle , atomic , molecular , solid - state physics ) and quantum chemistry . in most textbooks , the rs series deals with the perturbation of a single nondegenerate initial state . however , in many practical applications , the initial state is either degenerate or quasidegenerate ( i.e. several states are close in energy ) and the basic rs approach breaks down . quasidegenerate perturbation theory is widely used to set up effective hamiltonians @xcite , for example the spin hamiltonians of molecular chemistry and solid state physics @xcite , or to deal with the quantum electrodynamics of atoms @xcite . the terms of the rs series are notoriously complex . even mathematical physicists of the stature of reed and simon admit that the terms of the rs series are `` quite complicated '' and that `` the higher order rs coefficients are hard to compute '' ( see ref . , p. 8 and 18 ) . computer programs are available to build the terms of the rs series @xcite , but their results are intricate expressions that do not exhibit any obvious structure and that can hardly be used to carry out resummations of the series . in this paper , combinatorial physics is used to provide the first representation of the general term of the rayleigh - schrdinger ( rs ) perturbation theory for quasidegenerate systems . each term is written as a tree that faithfully reflects its algebraic structure . in particular , trees illustrate the recursive structure of rs terms , that is used to prove properties of the rs series . the purpose of this paper is to describe the general term of the rs series and to illustrate the power of our combinatorial approach by deriving a number of possible resummations of the rs series . our aim is to set up the tools for performing such resummations and not to discuss when some resummations are more convenient than others . as a consequence , no numerical example is given . when the system is not quasi - degenerate ( i.e. when it is either fully degenerate or non - degenerate ) , several graphical representations of the terms of the rs series have been proposed . we give the bijection between these representations and our trees . in the most general setting , we consider a model space @xmath2 spanned by @xmath3 eigenstates @xmath4 of @xmath1 , with energies @xmath5 . we assume that all @xmath5 of the model space are separated from the rest of the spectrum of @xmath1 by a finite gap . the projector onto the model space @xmath2 is @xmath6 , where @xmath7 runs over the basis states of @xmath2 . the wave operator @xmath8 transforms @xmath3 states @xmath9 of @xmath2 into eigenstates @xmath10 of @xmath11 : @xmath12 . we recently described @xcite the way to choose the states @xmath9 . we assume that @xmath13 and @xmath14 . then , the eigenvalues of the effective hamiltonian @xmath15 are the eigenvalues @xmath16 of @xmath11 . in other words , eigenvalues of @xmath11 can be obtained by diagonalizing the @xmath17 matrix @xmath18 . lindgren and kvasnika showed @xcite that @xmath19 } & = & v\omega - \omega v \omega . \label{lindgren}\end{aligned}\ ] ] the recursive solution of this equation gives a series expansion for @xmath8 which is the rs series for the wavefunction . however , it is not obvious that eq . ( [ lindgren ] ) has a recursive solution . indeed , we must be able to solve the equation @xmath20 } = c$ ] for @xmath21 . in general , the operator sylvester equation @xmath22 , where @xmath23 and @xmath24 are self - adjoint , has a unique solution if and only if @xmath23 and @xmath24 have no common eigenvalue @xcite . clearly , this is not the case if @xmath25 , so we recast the equation by defining @xmath26 , where @xmath27 , with @xmath28 . thus , eq . ( [ lindgren ] ) becomes an equation for @xmath29 : @xmath30 } & = & qvp + qv\chi -\chi v p - \chi v \chi . \label{lindgrenchi}\end{aligned}\ ] ] this equation is a matrix riccati equation , for which various solution methods have been proposed in the literature @xcite . solutions exist also for its time - dependent form @xcite . now , @xmath31 } = \chi h_0 - h_0 \chi = \chi p h_0 - q h_0 \chi$ ] , because @xmath32 and @xmath33 commute with @xmath1 . we obtain a sylvester equation with @xmath34 and @xmath35 and its solution is unique because we assumed that there is a finite gap between the states of the model space and the rest of the spectrum . because of its importance , the basic rs series has been dealt with in hundreds of papers . as a starting point to discover the proper combinatorial structure of the rs series , we enumerate the number of its terms : it has one term of order 0 , one term of order 1 , then 2 , 5 , 14 , 42 , 132 , 429 terms of order 2 , 3 , 4 , 5 , 6 and 7 , respectively . the number of terms of order @xmath36 is the catalan number @xmath37 . the catalan sequence is well - know in combinatorics : `` the catalan sequence is probably the most frequently encountered sequence that is still obscure enough to cause mathematicians [ ] to expend inordinate amounts of energy re - discovering formulas that were worked out long ago . '' an entire book is devoted to catalan numbers @xcite . they enumerate over 200 combinatorial objects @xcite and in the physics literature , several of these objects were used to represent the general term of the rs series : dyck paths @xcite , non - crossing partitions @xcite , bracketings @xcite , labelled diagrams @xcite and sequences of integers @xcite . however , these approaches could not be generalized to quasi - degenerate systems because they dealt with denominators @xmath38 with @xmath39 , whereas more complex denominators occur in the quasi - degenerate case . moreover , they could not be used to discover efficient resummations because they do not capture the recursive structure of the rs series . to find out which combinatorial object is best suited to deal with the rs series , we had to use a rather elaborate algebraic analysis of the time - dependent perturbation theory @xcite . it turned out in the end that planar binary trees provide the most faithful representation . as we shall see , the relation beween a tree and the corresponding term of rs series is straightforward and the recursive structure of the terms is transparent . the faithfulness is demonstrated by the easiness of the proofs and by the relation between geometrical properties of the trees and analytical properties of the terms , for instance for resummation . we first give examples of the planar binary trees that we use . we denote by @xmath40 the set of planar binary trees with @xmath36 inner vertices . a tree is called _ binary _ if each vertex has either zero or two children . it is called _ planar _ if two trees are different when they can be deduced one from the other by moving one edge over another one . for example , the two planar trees @xmath41 and @xmath42 are different . there is a single tree with zero inner vertex : @xmath43 . there is one tree with one inner vertex @xmath44 , two trees with two inner vertices @xmath45 , five trees with three inner vertices @xmath46 and fourteen with four inner vertices @xmath47 . the vertical line of a tree is called the _ root_. the trees of @xmath40 ( with @xmath48 ) can be built from smaller trees by the following relation @xmath49 , where @xmath50 is the grafting of trees @xmath51 and @xmath52 , by which the roots of @xmath51 and @xmath52 are brought together and a new root is grown from their juncture . pictorially : @xmath53 for example , @xmath54 , @xmath55 , @xmath56 . except in figure [ figarbre ] , the vertices are not drawn explicitly for notational convenience . the inner vertices of a tree are the vertices to which three edges ( or two edges and the root ) are incident , its leaves are its vertices to which a single edge is incident . a leaf is oriented either to the left or to the right . each tree of @xmath40 has @xmath36 inner vertices and @xmath57 leaves . the order @xmath58 of a tree @xmath59 is the number of its inner vertices . if @xmath60 denotes the number of elements of @xmath40 , the recursive definition of @xmath40 implies that @xmath61 and @xmath62 , so that @xmath60 is a catalan number . the recursive definition of planar binary trees make them very easy to generate with a computer . in ref . , we showed that the wave operator can be written as the sum @xmath63 where @xmath64 with @xmath65 an eigenstate of @xmath1 in the model space and @xmath4 out of it ( i.e. @xmath66 and @xmath67 ) . the scalars @xmath68 are defined by an exceedingly simple recursive relation : for any tree @xmath59 different from @xmath69 , there are two trees @xmath51 and @xmath52 such that @xmath70 . then , @xmath71 with the special cases @xmath72 if @xmath73 and @xmath74 if @xmath75 . it is clear from the definition that @xmath76 is built from a product of @xmath58 matrix elements of @xmath77 . note that , when @xmath73 , both @xmath7 and @xmath78 in @xmath79 correspond to eigenstates outside the model space , while when @xmath75 , both @xmath7 and @xmath78 in @xmath79 correspond to eigenstates in the model space . we now prove that the expansion @xmath80 is a solution of eq . ( [ lindgrenchi ] ) . first notice that taking the commutator @xmath81}$ ] amounts to multiply @xmath82 by @xmath83 . let @xmath84 . then , @xmath81}$ ] is equal to @xmath85 , or @xmath86 or @xmath87 or @xmath88 , respectively , according to whether @xmath89 , or @xmath90 and @xmath91 , or @xmath92 and @xmath93 , or @xmath92 and @xmath91 , respectively . the result follows by summing over all trees @xmath51 and @xmath52 . this very simple proof illustrates the fact that non trivial results can be easily obtained because trees encapsulate the recursive structure of the rs series . a more explicit version of the recursive relation ( [ recursive ] ) will sometimes be needed . for @xmath70 , we define equivalently @xmath76 by @xmath94 where the exponent of @xmath95 and @xmath96 means that state @xmath65 is a basis state of @xmath2 and @xmath4 a basis state of the complementary of @xmath2 . as examples , we give all @xmath76 for @xmath58=2 and 3 , the value of @xmath76 for @xmath97 ( i.e. @xmath98 ) being given by eq . ( [ rec11 ] ) ) . @xmath99 by looking at these examples , one may wonder whether some denominators could take the value zero . we show now by induction that this never happens . in other words , @xmath76 is never singular . this is true for @xmath98 because , in eq . ( [ rec11 ] ) , @xmath4 belongs to @xmath2 and @xmath65 does not belong to @xmath2 . thus , @xmath100 is greater than the gap between the model space and the rest of the spectrum . assume now that no @xmath76 is singular for trees with @xmath101 . take a tree @xmath59 with @xmath102 . then , @xmath70 , with @xmath103 and @xmath104 . assume that neither @xmath51 nor @xmath52 is equal to the root @xmath69 . then , eq . ( [ rectt ] ) and the previous remark about @xmath100 shows that @xmath76 is not singular . the same is true if @xmath51 or @xmath52 is equal to the root . thus , the denominator @xmath76 can never take the value 0 . again , the proof is made very simple by the recursive structure of @xmath76 . the recursive relation ( [ recursive ] ) is very useful to derive proofs , but we also need a non - recursive expression for the terms of the rs series . we now give an explicit relation between @xmath59 and @xmath76 @xcite . this construction can be followed in figure [ figarbre ] for @xmath105 . consider a tree @xmath59 with @xmath106 and number its leaves from 1 for the leftmost leaf to @xmath57 for the rightmost one . the numerators in the expansion of @xmath76 are @xmath107 . the state @xmath108 belongs to @xmath2 ( i.e. is @xmath109 ) if leaf @xmath78 is oriented to the right and does not belong to @xmath2 ( i.e. is @xmath110 ) if leaf @xmath78 is oriented to the left . in the example of fig . [ figarbre ] , the denominator is @xmath111 then , for each inner vertex @xmath112 of @xmath59 , take the subtree @xmath113 for which @xmath112 is just above the root . in other words , @xmath113 is obtained by chopping the edge below @xmath112 and considering the edge dangling from @xmath112 as the root of @xmath113 . in the example of fig . [ figarbre ] , we have four inner vertices labelled @xmath114 , @xmath115 , @xmath116 , @xmath117 . for vertex @xmath114 , the subtree @xmath118 is the full tree @xmath59 . for the other inner vertices the subtrees are proper ( i.e. different from @xmath59 ) and are given in fig . [ figarbre ] . for each tree @xmath113 , denote by @xmath119 the index of its leftmost leaf and by @xmath120 that of its rightmost leaf . then , the denominator is the product of @xmath121 , where @xmath112 runs over the @xmath36 inner vertices of the tree . in the example of fig . [ figarbre ] , @xmath118 gives the denominator @xmath122 , @xmath123 gives @xmath124 , @xmath125 gives @xmath126 and @xmath127 gives @xmath128 . finally , the whole fraction is summed over @xmath129 and multiplied by @xmath130 , where @xmath117 is the number of leaves pointing to the right . it can be easily proved that the above construction satisfies the recursive equation ( [ recursive ] ) . , @xmath115 , @xmath116 and @xmath117 . the couples @xmath131 are @xmath132 , @xmath133 , @xmath134 and @xmath135 for @xmath136 and @xmath117 , respectively . [ figarbre],width=302 ] if we take the example of the tree @xmath59 of fig . [ figarbre ] and consider for example the inner vertex labelled by @xmath115 , we get a contribution @xmath124 to the denominator . the total term is @xmath137 as far as we know , this tree representation provides the first description of the general term of the rs series for quasidegenerate systems . the tree representation is useful , not only to prove properties of the rs series , but also to derive new resummations of it . indeed , any way to write the set of trees as a composition of subtrees gives rise to a resummation of the rs series . for any tree @xmath59 , we define the sequence @xmath138 by @xmath139 , @xmath140 . for example , if @xmath141 , we have @xmath142 , @xmath143 . for any integer @xmath36 , @xmath138 is obtained by grafting @xmath59 on the leftmost leaf of a tree that has a single leaf oriented to the left and @xmath36 leaves oriented to the right ( such a tree is called a left comb whereas @xmath138 , @xmath48 is called a left comb grafting . notice the uniqueness rewriting property : an arbitray tree @xmath144 can be rewritten uniquely as @xmath138 , where @xmath145 or @xmath146 . the sum over graftings on left combs is made by defining @xmath147 we can now calculate @xmath148 in terms of @xmath76 . we first define @xmath149 then , we use kvasnika s trick and denote by @xmath150 the projector onto the eigenspace of @xmath1 with eigenvalue @xmath5 outside the model space , so that @xmath151 and eq . ( [ rect1 ] ) can be rewritten @xmath152 . by repeating this argument we obtain @xmath153 the map @xmath154 goes from the model space to itself and the inverse is taken within the model space . the dimension of the model space is generally small and the numerical calculation of the inverse is fast . we transform @xmath155 where @xmath156 and the inverse is again from @xmath2 to @xmath2 . to conclude , we assume that @xmath157 , where @xmath144 and @xmath112 are different from @xmath69 . then , @xmath158 where @xmath159 and @xmath160 are the eigenstates and eigenvalues of @xmath161 , so that @xmath162 . in other words , summing over all left combs amounts to replacing the eigenstates of @xmath163 by the eigenstates of @xmath161 . the formula we obtained for @xmath164 suggests to look for another expansion of the rs series involving only the eigenstates and eigenvalues of @xmath165 . the uniqueness rewriting property implies that @xmath166 , where the sum is over all trees @xmath59 with two or more inner vertices and such that @xmath167 , but this resummation can still be improved . we rewrite the kvasnika - lindgren equation . @xmath30 } & = & \sum_i q_i \chi h_0 - h_0 \sum_i q_i \chi = \sum_i q_i \chi ( h_0 - e_i).\end{aligned}\ ] ] thus , eq . ( [ lindgrenchi ] ) becomes @xmath168 that is , @xmath169 or , for all the eigenvalues @xmath5 of @xmath1 outside the model space : @xmath170 where @xmath171 this equation is similar to the kvasnika - lindgren equation and can be solved graphically and recursively by the same process . notice that the leading term of the recursive expansion of @xmath29 is now @xmath172 . let us call a tree @xmath59 right - normalized if there is no edge in @xmath59 such that the tree @xmath173 obtained by pruning the tree @xmath59 at that edge can be written @xmath174 with @xmath175 . the only left comb possibly contained in a right - normalized tree is @xmath176 . the number of right - normalized trees is considerably smaller than the number of trees . for @xmath36=1 , 2 , 3 and 4 the number of right - normalized trees is 1 ( @xmath177 , 1 ( @xmath42 ) , 2 ( @xmath178 , @xmath179 ) and 4 ( @xmath180 , @xmath181 , @xmath182 , @xmath183 ) instead of 1 , 2 , 5 , 14 . the right - normalized trees are enumerated by the motzkin numbers @xcite . the solution of eq . ( [ lindgrenchip ] ) is then @xmath184 , where @xmath59 runs over right - normalized trees and @xmath185 can be defined recursively by @xmath186 once again , this approach reduces the number . it can be expected that this resummation considerably accelerates the convergence of the series . indeed , when the model space is well chosen , the matrix elements of @xmath85 are smaller than those of @xmath187 . in this expression , all terms involving powers of @xmath187 have been resummed . a similar resummation , that we omit , can be obtained by summing over right combs . however , it is not so practical as the previous ones because it is usually not easy to invert @xmath188 . if this inversion is possible , then we can simultaneously sum over right and left combs , at least if we assume that a finite gap exists between the eigenvalues of @xmath189 and @xmath161 . we make this assumption , proceed as before and transform the kvasnika - lindgren equation . we define @xmath190 the projector onto the eigenspace of @xmath161 with energy @xmath160 and use similar notations for the projectors and energy levels of @xmath189 . thus , @xmath191 and we can rewrite the lhs of eq . ( [ lindgrenchip ] ) as @xmath192 thus eq . ( [ lindgrenchip ] ) becomes @xmath193 , or @xmath194 for all eigenvalues @xmath195 of @xmath161 , this gives us @xmath196 where we set : @xmath197 we can write directly the wave operator as a sum parametrized by trees : @xmath198 , where @xmath199 is defined recursively for all trees by @xmath200 as for the one involving a direct resummation over right combs , this resummation is not as practical as the previous one because the eigenvalues of @xmath189 are generally not known . however , this algorithm provides a way to express the eigenvectors of a matrix @xmath201 in terms of the eigenvalues and eigenvectors of @xmath23 and @xmath24 . we use now another representation of trees in terms of trees . since any tree @xmath59 decomposes uniquely as @xmath70 , a recursion on the left hand side of the decomposition shows that there is a unique integer @xmath78 and @xmath78 unique trees @xmath202 ( with @xmath203 ) such that @xmath204 where @xmath205 in words , @xmath206 is obtained by taking a left comb with @xmath78 right leaves and grafting @xmath207 on the lowest right leaf , @xmath208 on the next one , up to @xmath209 on the highest right leaf . if we sum over @xmath78 and over all trees @xmath210 we find @xmath211 where @xmath212 stands for the formal sum of all trees . we can plug the expansion of the wave operator of section [ secondsect ] in this equation and get for @xmath213 : @xmath214 therefore , eq . ( [ eqf ] ) gives us the non - perturbative equation @xmath215 if we use instead the first expansion eq ( [ omegat ] ) of the wave operator and plug the corresponding recursive formulas in @xmath206 , we get @xmath216 where @xmath217 and @xmath218 , so that : @xmath219 we believe that these relations are new . similar expansions would follow by considering the symmetric expansion of trees ( using the grafting operation of a tree on the right - most leaf of another tree instead of the grafting on the left - most tree ) . the previous works on quasidegenerate perturbation theory correspond to a summation which is symmetric to our first alternative expansion ( in the sense that they focus on the operators @xmath220 ) . by following a line suggested by kvasnika and lindgren @xcite , several authors transformed the kvasnika - lindgren equation into @xmath221 a resummation ( similar , but slightly more involved than the one leading to eq . ( [ chilk ] ) ) gives the equation @xcite @xmath222 which is a generalization of the degenerate case @xcite . suzuki and okamoto further solved this for @xmath29 but their rather complex result was not applied to concrete problems , as far as we know . up to a left / right symmetry , their result is similar to our first alternative expansion , excepted for the fact that @xmath187 and @xmath223 are grouped in a single term . the main practical difference between the result obtained by suzuki and okamoto and our second alternative expansion is the fact that we do not consider the term @xmath187 as a perturbation , we treat it exactly . this is important because , when the model space is well chosen , @xmath187 is larger than the non - diagonal terms @xmath224 . in this section , we discuss a question related to the green function of degenerate systems . consider a hamiltonian @xmath1 with a degenerate energy @xmath225 . the eigenstates of @xmath1 with energy @xmath225 span a vector space @xmath2 . the projector onto the model space @xmath2 is denoted by @xmath32 . in a series of recent papers @xcite , we proved by non - perturbative methods that there are eigenstates @xmath4 of @xmath1 , called the _ parent states _ , such that the usual gell - mann and low wavefunction has a well - defined limit when the adiabatic parameter @xmath226 goes to zero : @xmath227 where @xmath228 is the evolution operator in the interaction picture . morover , we showed that the parent states @xmath4 are eigenstates of @xmath1 ( with energy @xmath225 ) and are also eigenstates of @xmath187 . as a consequence , @xmath229 for @xmath230 if @xmath4 and @xmath65 are parent states . since the parent states solve the problem in a non - perturbative approach , one might be tempted to use them in the perturbative one . in other words , we pick up a parent state , say @xmath231 , and we calculate the rayleigh - schrdinger series corresponding to the projector @xmath232 . however , as noticed by tth @xcite , a problem appears in the perturbative expansion . this problem can be illustrated by a simple example : for @xmath233 , we have @xcite @xmath234 where @xmath226 is the adiabatic switching operator , which tends to zero at the end of the calculation . now , if @xmath4 belongs to the model space , then @xmath235 ( because @xmath236 ) and the expression converges although @xmath237 could have brought a problem . in other words , using a basis of parent states for @xmath2 has made the expression convergent . however , this trick does not always work . indeed , for @xmath141 , we have @xcite @xmath238 if @xmath4 belongs to the model space and @xmath65 is out of it , then we have @xmath237 and the limit @xmath239 is not defined because nothing insures that @xmath229 or @xmath240 . in other words , the convergence problem was solved at the non - perturbative level but remains at the perturbative one , so that the perturbative expansion has to be resummed in a proper way . now we show how to solve this problem by using a trick related to the hamiltonian shift proposed by silverstone @xcite . we assume that @xmath241 are nondegenerate . then , we rewrite @xmath242 , where @xmath243 we build the rs series for @xmath244 and @xmath245 with the one - dimensional model space @xmath2 spanned by @xmath231 . thus , @xmath246 . this gives us @xmath247 . as a consequence , @xmath248 if @xmath249 with @xmath250 . we can write @xmath251 , where @xmath33 is the projector corresponding to the initial problem and @xmath252 is the projector onto the basis states of @xmath2 different from @xmath231 . then , we have @xmath253 and @xmath254 . this gives us @xmath255 , which simplifies the evaluation of @xmath76 for @xmath98 . similarly , @xmath256 simplifies the evaluation of @xmath76 for @xmath70 . finally , the identity @xmath257 gives us @xmath258 for the evaluation of @xmath76 with @xmath259 . this gives us the following recursive expression for @xmath70 . @xmath260 the recursive expression shows that all terms are well defined if all @xmath241 are different and if @xmath261 is different from the energies @xmath262 out of the model space . we discuss here some continued - fraction resummation of the rs series . such a ( generalized ) continued fraction formula was found to be very efficient for calculating nuclear properties @xcite . the combinatorial structure of continued fractions was studied in detail by flajolet @xcite . on the other hand , lee and suzuki derived a continued fraction expression for @xmath29 and the effective hamiltonian for a degenerate system @xcite . other implementations of continued fractions for pertubation theory can be found in the literature @xcite . suzuki and lee @xcite start from the kvasnicka - lindgren equation @xmath30 } & = & qvp + qv\chi -\chi v p -\chi v \chi.\end{aligned}\ ] ] for a degenerate system @xmath263 , so that @xmath264 and @xmath265 . this equation is then reordered into @xmath266 . they consider the iterative equation ( see @xcite , eq . ( 3.27 ) p. 2102 ) . @xmath267 with the boundary condition @xmath268 . in other words @xmath269 the suzuki - lee formula has two drawbacks : it is restricted to degenerate systems @xcite and it requires the inversion of @xmath270 , which is usually infinite dimensional . to solve these two problems , we start from eq . ( [ lindgrenchip ] ) and , by using @xmath271 , we rewrite it @xmath272 . we transform this equation into @xmath273 thus , we define the system of recursive equations @xmath274 with the boundary condition @xmath268 . this generalized continued fraction has convergence properties similar to that of lee and suzuki , it is well - defined for quasi - degenerate systems and the inverse is computationally easier because it is done within the model space . many other combinatorial objects have been used to represent the terms of the rs series in non - degenerate or degenerate cases . each of these representations is useful for specific applications . it is therefore important to describe the relation between the most important of them ( bloch sequences , dyck paths , braketings and non - crossing partitions ) and the trees . most of these representations are valid for degenerate systems . thus , we start by presenting the first terms of the rs series of degenerate systems , where @xmath225 is the energy of the states of the model space . we also give an operator representation of these terms , with @xmath275 . @xmath276 partition & @xcite + 1 & ( 2,.5)(3,4 ) ( 1.4,1.3)@xmath277 ( 1,0)(1,2 ) ( 1,2)(0,3 ) ( 1,2)(2,3 ) & ( 1 ) & ( 2,0)(2,2 ) ( 0,0)(1,1)(2,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 0,0)(2,0 ) & @xmath278 & @xmath279 & 1 + 2 & ( 2,.5)(5,5 ) ( 3.4,1.3)@xmath277 ( 1.0,2.3)@xmath280 ( 3,0)(3,2 ) ( 3,2)(1,4 ) ( 3,2)(4,3 ) ( 2,3)(3,4 ) & ( 20 ) & ( 3,0)(4,2 ) ( 0,0)(1,1)(2,2)(3,1)(4,0 ) ( 0,0)(2,2 ) ( 2,2)(4,0 ) ( 0,0)(4,0 ) & @xmath281 & @xmath282 & 2 + 2 & ( 2,.5)(6,5 ) ( 3.4,1.3)@xmath277 ( 4.5,2.3)@xmath280 ( 3,0)(3,2 ) ( 3,2)(5,4 ) ( 3,2)(2,3 ) ( 4,3)(3,4 ) & ( 11 ) & ( 3,0)(4,2 ) ( 0,0)(1,1)(2,0)(3,1)(4,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 2,0)(3,1 ) ( 3,1)(4,0 ) ( 0,0)(4,0 ) & @xmath283 & @xmath284 & 1 + 3 & ( 2,.5)(5,6 ) ( 3.4,1.3)@xmath277 ( 1.0,2.2)@xmath280 ( 0.0,3.3)@xmath285 ( 3,0)(3,2 ) ( 3,2)(0,5 ) ( 3,2)(4,3 ) ( 2,3)(3,4 ) ( 1,4)(2,5 ) & ( 300 ) & ( 3,0)(6,3 ) ( 0,0)(1,1)(2,2)(3,3)(4,2)(5,1)(6,0 ) ( 0,0)(3,3 ) ( 3,3)(6,0 ) ( 0,0)(6,0 ) & @xmath286 & @xmath287 & 5 + 3 & ( 2,.5)(5,6 ) ( 3.4,1.3)@xmath277 ( 1.0,2.2)@xmath280 ( 3.4,3.5)@xmath285 ( 3,0)(3,2 ) ( 3,2)(1,4 ) ( 3,2)(4,3 ) ( 2,3)(4,5 ) ( 3,4)(2,5 ) & ( 210 ) & ( 3,0)(6,3 ) ( 0,0)(1,1)(2,2)(3,1)(4,2)(5,1)(6,0 ) ( 0,0)(2,2 ) ( 2,2)(3,1 ) ( 3,1)(4,2 ) ( 4,2)(6,0 ) ( 0,0)(6,0 ) & @xmath288 & @xmath289 & 2 + 3 & ( 2,.5)(6,6 ) ( 3.4,1.3)@xmath277 ( 4.6,2.4)@xmath280 ( 0.9,2.4)@xmath285 ( 3,0)(3,2 ) ( 3,2)(0.5,4.5 ) ( 1.5,3.5)(2.5,4.5 ) ( 3,2)(5.5,4.5 ) ( 4.5,3.5)(3.5,4.5 ) & ( 201 ) & ( 3,0)(6,2 ) ( 0,0)(1,1)(2,2)(3,1)(4,0)(5,1)(6,0 ) ( 0,0)(2,2 ) ( 2,2)(3,1 ) ( 3,1)(4,0 ) ( 4,0)(5,1 ) ( 5,1)(6,0 ) ( 0,0)(6,0 ) & @xmath290 & @xmath291 & 4 + 3 & ( 3,.5)(6,6 ) ( 3.4,1.3)@xmath277 ( 4.6,2.4)@xmath280 ( 1.8,3.4)@xmath285 ( 3,0)(3,2 ) ( 3,2)(5,4 ) ( 3,2)(2,3 ) ( 4,3)(2,5 ) ( 3,4)(4,5 ) & ( 120 ) & ( 3,0)(6,2 ) ( 0,0)(1,1)(2,0)(3,1)(4,2)(5,1)(6,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 2,0)(4,2 ) ( 4,2)(6,0 ) ( 0,0)(6,0 ) & @xmath292 & @xmath293 & 3 + 3 & ( 3,.5)(6,6 ) ( 3.4,1.3)@xmath277 ( 4.4,2.3)@xmath280 ( 5.4,3.3)@xmath285 ( 3,0)(3,2 ) ( 3,2)(6,5 ) ( 3,2)(2,3 ) ( 4,3)(3,4 ) ( 5,4)(4,5 ) & ( 111 ) & ( 3,0)(6,1 ) ( 0,0)(1,1)(2,0)(3,1)(4,0)(5,1)(6,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 2,0)(3,1 ) ( 3,1)(4,0 ) ( 4,0)(5,1 ) ( 5,1)(6,0 ) ( 0,0)(6,0 ) & @xmath294 & @xmath295 & 1 + ( 3,.5)(5,6.5 ) ( 4.4,1.3)@xmath277 ( 2.0,2.2)@xmath280 ( 1.0,3.2)@xmath285 ( 0.0,4.2)@xmath296 ( 4,0)(4,2 ) ( 4,2)(0,6 ) ( 4,2)(5,3 ) ( 3,3)(4,4 ) ( 2,4)(3,5 ) ( 1,5)(2,6 ) & ( 4000 ) & ( 4,0)(8,4 ) ( 0,0)(1,1)(2,2)(3,3)(4,4)(5,3)(6,2)(7,1)(8,0 ) ( 0,0)(4,4 ) ( 4,4)(8,0 ) ( 0,0)(8,0 ) & @xmath297 & @xmath298 & 5 + ( 4,.5)(5,6.5 ) ( 4.4,1.3)@xmath277 ( 2.0,2.2)@xmath280 ( 1.0,3.2)@xmath285 ( 3.1,4.4)@xmath296 ( 4,0)(4,2 ) ( 4,2)(1,5 ) ( 4,2)(5,3 ) ( 3,3)(4,4 ) ( 2,4)(4,6 ) ( 3,5)(2,6 ) & ( 3100 ) & ( 4,0)(8,3 ) ( 0,0)(1,1)(2,2)(3,3)(4,2)(5,3)(6,2)(7,1)(8,0 ) ( 0,0)(3,3 ) ( 3,3)(4,2 ) ( 4,2)(5,3 ) ( 5,3)(8,0 ) ( 0,0)(8,0 ) & @xmath299 & @xmath300 & 6 + ( 4,.5)(6,6.5 ) ( 4.4,1.3)@xmath277 ( 2.0,2.2)@xmath280 ( 4.7,3.7)@xmath285 ( 0.5,3.7)@xmath296 ( 4,0)(4,2 ) ( 4,2)(0.5,5.5 ) ( 4,2)(5,3 ) ( 3,3)(5.5,5.5 ) ( 1.5,4.5)(2.5,5.5 ) ( 4.5,4.5)(3.5,5.5 ) & ( 3010 ) & ( 4,0)(8,3 ) ( 0,0)(1,1)(2,2)(3,3)(4,2)(5,1)(6,2)(7,1)(8,0 ) ( 0,0)(3,3 ) ( 3,3)(5,1 ) ( 5,1)(6,2 ) ( 6,2)(8,0 ) ( 0,0)(8,0 ) & @xmath301 & @xmath302 & 14 + ( 3,.5)(6,6.5 ) ( 3.4,1.3)@xmath277 ( 1.0,2.2)@xmath280 ( 3.1,3.2)@xmath285 ( 1.1,4.2)@xmath296 ( 3,0)(3,2 ) ( 3,2)(1,4 ) ( 3,2)(4,3 ) ( 2,3)(4,5 ) ( 3,4)(1,6 ) ( 2,5)(3,6 ) & ( 2200 ) & ( 4,0)(8,3 ) ( 0,0)(1,1)(2,2)(3,1)(4,2)(5,3)(6,2)(7,1)(8,0 ) ( 0,0)(2,2 ) ( 2,2)(3,1 ) ( 3,1)(5,3 ) ( 5,3)(8,0 ) ( 0,0)(8,0 ) & @xmath303 & @xmath304 & 7 + ( 3,.5)(5,6.5 ) ( 3.4,1.3)@xmath277 ( 1.0,2.2)@xmath280 ( 3.1,3.2)@xmath285 ( 4.0,4.2)@xmath296 ( 3,0)(3,2 ) ( 3,2)(1,4 ) ( 3,2)(4,3 ) ( 2,3)(5,6 ) ( 3,4)(2,5 ) ( 4,5)(3,6 ) & ( 2110 ) & ( 4,0)(8,3 ) ( 0,0)(1,1)(2,2)(3,1)(4,2)(5,1)(6,2)(7,1)(8,0 ) ( 0,0)(2,2 ) ( 2,2)(3,1 ) ( 3,1)(4,2 ) ( 4,2)(5,1 ) ( 5,1)(6,2 ) ( 6,2)(8,0 ) ( 0,0)(8,0 ) & @xmath305 & @xmath306 & 4 + ( 3,.5)(6,6.5 ) ( 3.9,1.3)@xmath277 ( 5.2,2.7)@xmath280 ( 1.2,2.7)@xmath285 ( 0.2,3.7)@xmath296 ( 3.5,0)(3.5,2 ) ( 3.5,2)(0,5.5 ) ( 3.5,2)(6,4.5 ) ( 5,3.5)(4,4.5 ) ( 2,3.5)(3,4.5 ) ( 1,4.5)(2,5.5 ) & ( 3001 ) & ( 4,0)(8,3 ) ( 0,0)(1,1)(2,2)(3,3)(4,2)(5,1)(6,0)(7,1)(8,0 ) ( 0,0)(3,3 ) ( 3,3)(6,0 ) ( 6,0)(7,1 ) ( 7,1)(8,0 ) ( 0,0)(8,0 ) & @xmath307 & @xmath308 & 13 + ( 2,.5)(5,6.5 ) ( 2.9,1.3)@xmath277 ( 4.2,2.7)@xmath280 ( 0.2,2.7)@xmath285 ( 1.8,3.7)@xmath296 ( 2.5,0)(2.5,2 ) ( 2.5,2)(0,4.5 ) ( 2.5,2)(5,4.5 ) ( 4,3.5)(3,4.5 ) ( 1,3.5)(3,5.5 ) ( 2,4.5)(1,5.5 ) & ( 2101 ) & ( 4,0)(8,2 ) ( 0,0)(1,1)(2,2)(3,1)(4,2)(5,1)(6,0)(7,1)(8,0 ) ( 0,0)(2,2 ) ( 2,2)(3,1 ) ( 3,1)(4,2 ) ( 4,2)(6,0 ) ( 6,0)(7,1 ) ( 7,1)(8,0 ) ( 0,0)(8,0 ) & @xmath309 & @xmath310 & 11 + ( 2,.5)(5,6.5 ) ( 2.9,1.3)@xmath277 ( 4.2,2.7)@xmath280 ( 2.2,3.7)@xmath285 ( 0.2,2.7)@xmath296 ( 2.5,0)(2.5,2 ) ( 2.5,2)(5,4.5 ) ( 2.5,2)(0,4.5 ) ( 1,3.5)(2,4.5 ) ( 4,3.5)(2,5.5 ) ( 3,4.5)(4,5.5 ) & ( 2020 ) & ( 4,0)(8,2 ) ( 0,0)(1,1)(2,2)(3,1)(4,0)(5,1)(6,2)(7,1)(8,0 ) ( 0,0)(2,2 ) ( 2,2)(4,0 ) ( 4,0)(6,2 ) ( 6,2)(8,0 ) ( 0,0)(8,0 ) & @xmath311 & @xmath312 & 10 + ( 3,.5)(6,6.5 ) ( 2.9,1.3)@xmath277 ( 4.2,2.7)@xmath280 ( 5.2,3.7)@xmath285 ( 0.2,2.7)@xmath296 ( 2.5,0)(2.5,2 ) ( 2.5,2)(6,5.5 ) ( 2.5,2)(0,4.5 ) ( 1,3.5)(2,4.5 ) ( 4,3.5)(3,4.5 ) ( 5,4.5)(4,5.5 ) & ( 2011 ) & ( 4,0)(8,2 ) ( 0,0)(1,1)(2,2)(3,1)(4,0)(5,1)(6,0)(7,1)(8,0 ) ( 0,0)(2,2 ) ( 2,2)(4,0 ) ( 4,0)(5,1 ) ( 5,1)(6,0 ) ( 6,0)(7,1 ) ( 7,1)(8,0 ) ( 0,0)(8,0 ) & @xmath313 & @xmath314 & 9 + ( 3,.5)(5,6.5 ) ( 3.4,1.3)@xmath277 ( 4.4,2.3)@xmath280 ( 2.1,3.2)@xmath285 ( 1.1,4.2)@xmath296 ( 3,0)(3,2 ) ( 3,2)(5,4 ) ( 3,2)(2,3 ) ( 4,3)(1,6 ) ( 3,4)(4,5 ) ( 2,5)(3,6 ) & ( 1300 ) & ( 4,0)(8,3 ) ( 0,0)(1,1)(2,0)(3,1)(4,2)(5,3)(6,2)(7,1)(8,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 2,0)(5,3 ) ( 5,3)(8,0 ) ( 0,0)(8,0 ) & @xmath315 & @xmath316 & 8 + ( 3,.5)(5,6.5 ) ( 2.4,1.3)@xmath277 ( 3.4,2.3)@xmath280 ( 1.1,3.2)@xmath285 ( 3.1,4.3)@xmath296 ( 2,0)(2,2 ) ( 2,2)(4,4 ) ( 2,2)(1,3 ) ( 3,3)(1,5 ) ( 2,4)(4,6 ) ( 3,5)(2,6 ) & ( 1210 ) & ( 4,0)(8,2 ) ( 0,0)(1,1)(2,0)(3,1)(4,2)(5,1)(6,2)(7,1)(8,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 2,0)(4,2 ) ( 4,2)(5,1 ) ( 5,1)(6,2 ) ( 6,2)(8,0 ) ( 0,0)(8,0 ) & @xmath317 & @xmath318 & 3 + ( 3,.5)(6,6.5 ) ( 2.4,1.3)@xmath277 ( 3.4,2.3)@xmath280 ( 4.7,3.7)@xmath285 ( 0.5,3.7)@xmath296 ( 2,0)(2,2 ) ( 2,2)(5.5,5.5 ) ( 2,2)(1,3 ) ( 3,3)(0.5,5.5 ) ( 4.5,4.5)(3.5,5.5 ) ( 1.5,4.5)(2.5,5.5 ) & ( 1201 ) & ( 4,0)(8,2 ) ( 0,0)(1,1)(2,0)(3,1)(4,2)(5,1)(6,0)(7,1)(8,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 2,0)(4,2 ) ( 4,2)(6,0 ) ( 6,0)(7,1 ) ( 7,1)(8,0 ) ( 0,0)(8,0 ) & @xmath319 & @xmath320 & 12 + ( 3,.5)(6,6.5 ) ( 2.4,1.3)@xmath277 ( 3.4,2.3)@xmath280 ( 4.4,3.3)@xmath285 ( 2.1,4.3)@xmath296 ( 2,0)(2,2 ) ( 2,2)(5,5 ) ( 2,2)(1,3 ) ( 3,3)(2,4 ) ( 4,4)(2,6 ) ( 3,5)(4,6 ) & ( 1120 ) & ( 4,0)(8,2 ) ( 0,0)(1,1)(2,0)(3,1)(4,0)(5,1)(6,2)(7,1)(8,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 2,0)(3,1 ) ( 3,1)(4,0 ) ( 4,0)(6,2 ) ( 6,2)(8,0 ) ( 0,0)(8,0 ) & @xmath321 & @xmath322 & 2 + ( 3,.5)(6,6.5 ) ( 2.4,1.3)@xmath277 ( 3.4,2.3)@xmath280 ( 4.4,3.3)@xmath285 ( 5.4,4.3)@xmath296 ( 2,0)(2,2 ) ( 2,2)(6,6 ) ( 2,2)(1,3 ) ( 3,3)(2,4 ) ( 4,4)(3,5 ) ( 5,5)(4,6 ) & ( 1111 ) & ( 4,0)(8,1 ) ( 0,0)(1,1)(2,0)(3,1)(4,0)(5,1)(6,0)(7,1)(8,0 ) ( 0,0)(1,1 ) ( 1,1)(2,0 ) ( 2,0)(3,1 ) ( 3,1)(4,0 ) ( 4,0)(5,1 ) ( 5,1)(6,0 ) ( 6,0)(7,1 ) ( 7,1)(8,0 ) ( 0,0)(8,0 ) & @xmath323 & @xmath324 & 1 + bloch @xcite was the first to write the general term of @xmath8 for degenerate systems . to describe his result , we consider the wave operator @xmath325 of the hamiltonian @xmath326 . we have the series expansion @xmath327 with @xmath328 where @xmath329 , @xmath330 for @xmath331 , and where @xmath332 run over the bloch sequences . a bloch sequence is an @xmath36-tuple @xmath333 of non - negative integers , such that @xmath334 for @xmath335 and @xmath336 . the bloch sequences for @xmath36=1 , 2 and 3 are given in table [ tab123 ] and for @xmath36=4 in table [ tab4 ] . these combinatorial objects are enumerated by catalan numbers ( see example 6.24 p. 180 of ref . @xcite and item @xmath337 in stanley s catalan addendum ) . since bloch s publication @xcite , they were widely used @xcite to represent the general term of the rs series for degenerate systems . the bijection between trees and sequences is defined by @xmath338 , @xmath339 , @xmath340 and @xmath341 where @xmath342 , @xmath343 , where the product on the right hand sides is the concatenation of sequences ( i.e. @xmath344 ) , and where the operation @xmath345 acts by @xmath346 . notice that @xmath347 is clearly one to one ( injective ) . the fact that it is actually a bijection follows e.g. from the fact that trees and bloch sequences are both enumerated by catalan numbers . we prove now its compatibility with the tree and bloch expansions by induction . we write @xmath70 and we consider the usual four cases . the first case is @xmath98 , @xmath348 . this is the starting point of the inductive proof . we have @xmath349 , where we used the fact that the only bloch sequence for @xmath350 is @xmath351 . let us denote by @xmath352 the contribution of @xmath353 in eq . ( [ defomegan ] ) corresponding to the bloch sequence @xmath354 . thus , we showed that , for @xmath98 and @xmath355 , we have @xmath356 . now , assume that @xmath357 for all trees @xmath59 with @xmath101 and choose a tree @xmath59 such that @xmath102 . then , @xmath70 with @xmath103 and @xmath104 . we consider the three remaining cases . if @xmath250 and @xmath75 , then @xmath358 is obtained from eqs . ( [ defomegat ] ) and ( [ rect1 ] ) @xmath359 . by the induction hypothesis , we have @xmath360 for @xmath361 . therefore , @xmath362 . if @xmath363 , then @xmath356 with @xmath364 and the property is proved . the second case is @xmath250 and @xmath365 . then eqs . ( [ defomegat ] ) and ( [ rectt ] ) give us @xmath366 . if @xmath367 and @xmath368 , then @xmath356 with @xmath369 . thus , @xmath370 . the last case is @xmath73 and @xmath365 . equations ( [ defomegat ] ) and ( [ rec1 t ] ) give us @xmath371 . if @xmath372 , then @xmath356 with @xmath373 . in ref . @xcite , bloch also defined geometrical objects that are dyck paths rotated by @xmath374 . we prefer to use dyck paths because they are thoroughly studied in the combinatorial literature ( see , for example item i , p. 221 of ref . @xcite or example 6.2 , p. 151 of ref @xcite , where they are called _ mountain ranges _ ) . the bijection between dyck paths and bloch sequences used in ref . @xcite is well known in the combinatorial literature ( * ? ? ? * and 181 ) . to build the dyck path corresponding to the the bloch sequence @xmath333 , start from the origin and make @xmath375 steps in the north - east direction , then make one step in the south - east direction , then @xmath376 steps in the north - east direction , then one step in the south - east direction , and so on this bijection is illustrated in tables [ tab123 ] and [ tab4 ] . the terms of the rs series of degenerate systems can also be described by non - crossing partitions . this correspondence was studied by olszewski @xcite because it leads to useful factorizations of the rs terms for non - degenerate systems . consider the examples at the beginning of section [ bijection - sect ] . if we just look at the denominators , we see that each of them can be deduced from the right comb by saying that some indices are equal . for the five trees with @xmath377 , the term corresponding to @xmath378 can be obtained from that of @xmath379 by stating that @xmath380 . the other terms follow from @xmath381 , @xmath382 and @xmath383 . more generally , for a given tree @xmath59 , we say that two indices @xmath384 and @xmath78 are equivalent if and only if @xmath385 in the denominator of @xmath76 . the sets of equivalent indices form a partition of @xmath386 . we recall that a partition @xmath387 of @xmath388 is a set of disjoint subsets @xmath389 of @xmath388 whose union is @xmath388 . each subset @xmath389 is called a block of the partition . the partitions corresponding to the rs terms of order 1 to 3 are given in table [ tab123 ] and for order 4 in table [ tab4 ] . it will be shown that these partitions are _ non - crossing_. two blocks @xmath23 and @xmath24 of a partition are said to be crossing if there are @xmath390 in @xmath23 and @xmath391 in @xmath24 such that @xmath392 or @xmath393 . a partition is called non - crossing if no two of its blocks cross . we build by induction a partition from a tree . note first that for tree @xmath59 of order @xmath36 we deal with partitions of @xmath386 . a tree of order @xmath36 has @xmath57 leaves . thus , the index of the rightmost leaf is not used in the partition . for a given tree @xmath70 , we call @xmath32 ( @xmath394 , @xmath395 , respectively ) the partition corresponding to @xmath59 ( @xmath51 , @xmath52 , respectively ) . we consider the usual four cases . ( i ) for @xmath348 we associate the partition @xmath396 . ( ii ) for @xmath250 and @xmath75 , eq . ( [ rect1 ] ) gives us the additional denominator @xmath397 . therefore , an additional index is equivalent to 1 . this index is that of the rightmost leaf of @xmath51 , which was not used in @xmath394 . therefore , the partition @xmath32 of @xmath59 is obtained from @xmath394 by adding to the block of @xmath394 containing 1 the index of the rightmost leaf of @xmath51 , which is @xmath398 . ( iii ) for @xmath73 and @xmath365 , eq . ( [ rec1 t ] ) gives us the additional denominator @xmath397 . however , the leaf denoted by 1 is new and no block of @xmath395 should contain it . therefore , the partition @xmath32 of @xmath59 is obtained from @xmath395 by increasing all numbers of @xmath395 by 1 and by adding the block @xmath279 . ( iv ) for @xmath250 and @xmath365 , we compose the previous cases . we build @xmath32 by adding the index @xmath398 to the block of @xmath394 containing the index 1 and we increase all indices of @xmath395 by @xmath398 . we check that , in all cases , the resulting partition is non - crossing . note that olszewski @xcite does not explicitly use non - crossing partitions . he describes the general term of the rs series for non - degenerate systems by drawing a circle with @xmath36 points and pinching some of these points . the relation with non - crossing partitions is straightforward : all points that are pinched together belong to the same block of the partition . however , his correspondence between partitions and terms of the rs series is not the same as ours . following a suggestion by brueckner @xcite , huby and tong @xcite proposed , for the energy of nondegenerate systems , a solution in terms of bracketing , that is isomorphic with stanley s problem @xmath399 in his `` catalan addendum '' . consider the example of @xmath233 . we have @xmath400 . for a nondegenerate system @xmath401 and @xmath402 . the rule of the game is now to insert expectation values to disjoin powers of @xmath403 . thus , @xmath404 . we use tong s pictorial representation , where @xmath403 is replaced by @xmath405 , @xmath77 by @xmath406 , @xmath407 by @xmath408 and @xmath231 by @xmath409 , so that @xmath410 . the bracketings for @xmath58=1,2 and 3 are given in table [ tab123 ] and for order 4 in table [ tab4 ] . brueckner s bracketing is a powerful way to simplify the rs series , for instance by including the vacuum expectation value @xmath411 into @xmath1 , so that all rs terms involving it cancel . more details and examples can be found in some textbooks @xcite or review papers @xcite . we now describe the connection between trees and bracketings . denote by @xmath412 the bracketing corresponding to @xmath59 in tong s representation . assume that @xmath412 is known for all trees @xmath59 with @xmath101 and take a tree @xmath70 of degree @xmath36 . it @xmath73 and @xmath365 , then @xmath413 , so that @xmath414 . it @xmath250 and @xmath75 , then @xmath415 , so that @xmath416 . it @xmath250 and @xmath365 , then @xmath417 , so that @xmath418 . note that this bijection is different from the one used by tong @xcite . an equivalent bijection is obtained by numbering the inner vertices of the trees . the operation @xmath419 that associates to each tree @xmath59 its numbered tree is defined as follows . we denote by @xmath420 $ ] the numbered tree obtained from @xmath421 by adding @xmath78 to all the vertex numbers . then , @xmath421 can be defined recursively . if @xmath422 , then @xmath423 ( no number ) . if @xmath98 , then @xmath421 is obtained by assigning the number 1 to the root . if @xmath424 , then the root has number 1 , the inner vertices of @xmath52 ( if any ) are numbered as @xmath425 $ ] and the inner vertices of @xmath51 ( if any ) are numbered as @xmath426 $ ] . the numbered trees for @xmath427 are given in table table [ tab123 ] and for @xmath428 in table [ tab4 ] . for any tree @xmath59 , we can build the sets of numbers belonging to the same line oriented to the left . for example , if @xmath429 , the numbering is @xmath430{0}(2.4,1.3){$\scriptscriptstyle{1}$ } \rput[bl]{0}(3.4,2.3){$\scriptscriptstyle{2}$ } \rput[bl]{0}(4.7,3.7){$\scriptscriptstyle{3}$ } \rput[bl]{0}(0.5,3.7){$\scriptscriptstyle{4}$ } \psline(2,0)(2,2 ) \psline(2,2)(5.5,5.5 ) \psline(2,2)(1,3 ) \psline(3,3)(0.5,5.5 ) \psline(4.5,4.5)(3.5,5.5 ) \psline(1.5,4.5)(2.5,5.5 ) \end{pspicture}}$ ] and the sets are @xmath318 . these sets form a non - crossing partition which is the same as that used by olszewski @xcite . to obtain @xmath412 , first number the @xmath58 stars of @xmath412 from 1 to @xmath58 from the left to the right . then , consider a block @xmath431 of the partition . if @xmath432 , do nothing , if @xmath433 , then write a @xmath408 after star number @xmath375 , a @xmath409 before star number @xmath434 and replace star number @xmath435 with @xmath436 ( if any ) by @xmath437 . combinatorial physics is an emerging field that uses modern tools of algebraic combinatorics to solve physical problems . it was born with the investigation of the algebraic structure of renormalization in quantum field theory @xcite and showed its ability to deal with many - body problems @xcite . we showed that combinatorial physics is able to solve such long - standing problems as time - independent perturbation theory . the rs series is at the heart of many applications of quantum mechanics . it is also equivalent to more sophisticated methods such as feynman diagrams @xcite . it is even related to wilson s renormalization group @xcite . our combinatorial methods provided easy resummations of the rs series . it remains now to test their convergence properties . note that arnold also used trees in perturbation theory @xcite . however , his trees are essentially different from ours because they describe the successive degeneracy splitting due to higher order terms @xcite . g. s. t. was partially supported by the grant otka k60040 . i. de p. r. moreira , n. suaud , n. guihry , j. p. malrieu , r. caballol , j. m. bofill , and f. illas . derivation of spin hamiltonians from the exact hamiltonian : application to systems with two unpaired electrons per magnetic site . , 66:134430 , 2002 . s. fritzsche , b. k. mani , and d. angom . a computer - algebraic approach to the derivation of feyman - goldstone perturbation expansions for open - shell atoms and molecules . in j. r. sabin , p .- o . lwdin , and e. j. brandas , editors , _ adv . _ , volume 53 , pages 177215 , new york , 2008 . academic press . v. kostrykin and k. a. makarov . the adiabatic theorem of quantum mechanics and the riccati equation . in j zambrini , editor , _ 14th international congress on mathematical physics _ , pages 60510 , hackensack , 2005 . world scientific . |
the aim of this paper is to classify simply connected-dimensional torus manifolds with vanishing odd degree cohomology . it is shown that there is a one - to - one correspondence between equivariant diffeomorphism types of these manifolds and-valent labelled graphs , called torus graphs introduced by maeda - masuda - panov . using this correspondence and combinatorial arguments , we prove that a simply connected-dimensional torus manifold with is equivariantly diffeomorphic to the-dimensional sphere or an equivariant connected sum of copies of-dimensional quasitoric manifolds or-bundles over . | 1305.3174 | let @xmath6 be a @xmath7-dimensional closed , connected , oriented manifold with an effective @xmath8-dimensional ( i.e. , half - dimensional ) torus @xmath9-action . we call @xmath6 ( or @xmath10 ) a _ torus manifold _ if @xmath11 ( see @xcite ) , where @xmath12 is the set of fixed points . a toric manifold ( i.e. , a non - singular , complete toric variety viewed as a complex analytic space ) with the restricted @xmath9-action is a typical example of torus manifolds . recall that a toric manifold is a complex @xmath13-manifold with the dense orbit ( see @xcite , @xcite ) , and @xmath9 is the maximal compact subgroup of @xmath13 . a fundamental result of toric geometry tells us that there is a one - to - one correspondence between toric manifolds and combinatorial objects called fans . thus , topological ( more precisely , geometric ) invariants of toric manifolds can be described in terms of combinatorial invarinats of fans , such as equivariant cohomology rings , equivariant characteristic classes and other topological invariants . in 2003 , hattori - masuda introduced a torus manifold as the topological generalization of a toric manifold in @xcite . they also introduced the combinatorial objects , called multi - fans ( see @xcite , @xcite ) , and computed topological invariants ( such as equivariant characteristic classes or todd genus for unitary torus manifolds ) in terms of multi - fans . however , unlike toric geometry , a multi - fan does not contain enough information to determine some topological invariants of torus manifolds ( e.g. equivariant cohomology ) . so , in 2007 , maeda - masuda - panov introduced another combinatorial objects , called torus graphs motivated by gkm graphs introduced by guillemin - zara in @xcite . the combinatorial information of torus graphs can completely determine the equivariant cohomology rings of torus manifolds with vanishing odd degree cohomology , i.e. , @xmath14 ( in this paper , we only consider the integer coefficient ) , see @xcite , @xcite ( and also see section [ sect3 ] in this paper about torus graphs ) . however , in general , there is no one - to - one correspondence between torus manifolds with @xmath2 and torus graphs . so , we are naturally led to ask the following two questions : ( 1 ) which subclasses of torus manifolds are completely determined by combinatorial objects ( like multi - fans or torus graphs ) ; ( 2 ) if we find such a subclass of torus manifolds , how we can classify such torus manifolds . several mathematicians have answered to the 1st question ; for example , davis - januszkiewicz @xcite for the subclass called _ quasitoric manifolds _ ( see @xcite or section [ sect4.3 ] in this paper ) , ishida - fukukawa - masuda @xcite for the subclass called _ topological toric manifolds _ , and wiemeler for the class of simply connected @xmath0-dimensional torus manifolds with @xmath2 in @xcite ( see theorem [ wi2 ] ) . the aim of this paper is to answer to the 2nd question for the class of simply connected @xmath0-dimensional torus manifolds with @xmath2 by using torus graphs . let us briefly recall the classification results for torus manifolds with lower dimensions . if @xmath15 acts on a compact @xmath16-dimensional manifold @xmath6 , then @xmath6 is the @xmath16-dimensional sphere @xmath5 , the @xmath16-dimensional real projectiove space @xmath17 or the @xmath16-dimensional torus @xmath18 . because @xmath11 and @xmath6 is oriented , @xmath6 must be equivariantly diffeomorphic to @xmath5 with @xmath15-action ( also see @xcite ) . when @xmath19 . by orlik - raymond s theorem in @xcite , we have the following fact : [ or - thm ] let @xmath6 be a @xmath20-dimensional simply connected torus manifold . then , @xmath6 is equivariantly diffeomorphic to the @xmath20-sphere @xmath21 or an equivariant connected sum of copies of complex projective spaces @xmath22 , @xmath23 ( reversed orientation ) or hirzebruch surfaces @xmath24 . here , in theorem [ or - thm ] , a hirzebruch surface @xmath24 ( @xmath25 ) is a manifold which is defined by the projectivization of the complex @xmath16-dimensional vector bundle @xmath26 over @xmath27 , where @xmath28 is the tautological and @xmath29 is the trivial complex line bundles over @xmath27 . in this paper , we prove that an orlik - raymond type theorem in theorem [ or - thm ] also holds for simply connected @xmath0-dimensional torus manifolds with @xmath2 . before we state our main results , we introduce the result for non - simply connected torus manifolds . as one of the consequences of masuda - panov s theorem ( see theorem [ 2-equiv ] in section [ sect2.2 ] ) , we have the following proposition ( also see @xcite ) : [ wiemeler ] let @xmath30 be a @xmath0-dimensional torus manifold with @xmath31 ( might not be simply connected ) . then , there are a simply connected , @xmath0-dimensional torus manifold @xmath6 with @xmath2 and a homology @xmath1-sphere @xmath32 such that @xmath33 up to equivariantly diffeomorphism . here , in proposition [ wiemeler ] , the product manifold @xmath34 is the product of @xmath32 and the @xmath1-dimensional torus @xmath35 with the free @xmath36-action on the @xmath16nd factor , and the symbol @xmath37 represents the equivariant gluing along two free orbits of @xmath6 and @xmath34 . because the fundamental groups @xmath38 and @xmath39 are isomorphic in proposition [ wiemeler ] , @xmath30 is simply connected if and only if @xmath32 is simply connected , i.e. , the standard sphere . our main theorem is a classification of simply connected torus manifolds appeared in proposition [ wiemeler ] . [ main1 ] let @xmath6 be a simply connected @xmath0-dimensional torus manifold with @xmath2 . then , @xmath6 is equivariantly diffeomorphic to the @xmath0-sphere @xmath40 or obtained by an equivariant connected sum of copies of @xmath0-dimensional quasitoric manifolds or @xmath21-bundles over @xmath41 equipped with the structure of a torus manifold . this type of classification , i.e. , classification by equivariant connected sum , may be regarded as the @xmath0-dimensional analogue of orlik - raymond s classification in theorem [ or - thm ] . so , in this paper , we call this theorem an _ orlik - raymond type classification _ ( also see the papers @xcite and @xcite ) . in the paper @xcite , izmestiev proves an orlik - raymond type classification for some class of @xmath1-dimensional small covers ( i.e. , the real analogue of quasitoric manifolds , see section [ sect4.2 ] ) , called a linear model . the organization of this paper is as follows . in section [ sect2 ] , we recall the basic facts about torus manifolds . in section [ sect3 ] , we recall a torus graph . in particular , corollary [ key2 ] is the key fact to prove theorem [ main1 ] . in section [ sect4 ] , we introduce the torus graphs of @xmath3 , quasitoric manifolds and @xmath4-bundles over @xmath5 . these torus graphs will be the basic graphs to classify simply connected @xmath0-dimensional torus manifolds with @xmath2 . in section [ sect5 ] , we introduce the `` oriented '' torus graphs and translate the equivariant connected sum around fixed points of torus manifolds to the connected sum around vertices of oriented torus graphs . in section [ sect6 ] and [ sect7 ] , we prove theorem [ main1 ] . the brief outline of the proof is as follows . due to corollary [ key2 ] , there is a one - to - one corresponds between @xmath0-dimensional simply connected torus manifolds with @xmath2 and @xmath1-valent torus graphs . therefore , in order to prove theorem [ main1 ] , it is enough to prove that an ( oriented ) torus graph can be decomposed into basic torus graphs in section [ sect4 ] by the connected sum . we prove this by using combinatorial arguments . in this section , we recall some basic facts about torus manifolds ( see @xcite or @xcite for details ) . a @xmath7-dimensional torus manifold @xmath6 is said to be _ locally standard _ if every point in @xmath6 has a @xmath42-invariant open neighborhood @xmath43 which is weakly equivariantly homeomorphic to an open subset @xmath44 invariant under the standard @xmath9-action on @xmath45 , where two group actions @xmath46 and @xmath47 are said to be _ weakly equivariantly homeomorphic _ if there is an equivariant homeomorphism from @xmath43 to @xmath48 up to an automorphism on @xmath9 ( see e.g. ( * ? ? ? * section 2.1 ) for details ) . let @xmath49 , @xmath50 , be a codimension - two torus submanifold in a @xmath7-dimensional torus manifold @xmath6 which is fixed by some circle subgroup @xmath51 in @xmath42 . such @xmath49 is a @xmath52-dimensional torus manifold with @xmath53-action , called a _ characteristic submanifold_. because a torus manifold @xmath6 is compact , the cardinality of all characteristic submanifolds in @xmath6 is finite . if @xmath6 is locally standard , each characteristic submanifold is also locally standard . an _ omniorientation _ @xmath54 of @xmath6 is a choice of orientation for the torus manifold @xmath6 as well as for each characteristic submanifold . if there are just @xmath55 characteristic submanifolds in @xmath6 , there are exactly @xmath56 omniorientations ( see @xcite , @xcite ) . if @xmath6 has a @xmath42-invariant almost complex structure @xmath57 ( in this case , @xmath6 is automatically locally standard ) , then there exists the canonical omniorientation @xmath58 determined by @xmath57 . we call the torus manifold @xmath6 with a fixed omniorientation @xmath54 an _ omnioriented torus manifold _ and denote it by @xmath59 . the orbit space @xmath60 of a locally standard torus manifold @xmath6 naturally admits the structure of a `` topological '' manifold with corners . we next recall the basic facts of a topological manifold with corners ( cf . the definition of a smooth manifold with corners in @xcite ) and introduce the structure on @xmath60 . the following notations will be often used : @xmath61=\{0,\ 1,\ \ldots,\ n\},\end{aligned}\ ] ] and @xmath62 let @xmath63 be an @xmath8-dimensional topological manifold with boundary . chart with corners _ for @xmath63 is a pair @xmath64 , where @xmath65 is an open subset of @xmath63 and @xmath66 is homeomorphic from @xmath65 to a ( relatively ) open subset @xmath67 . two charts with corners @xmath68 are said to be _ ( topologically ) compatible _ if the composition of functions @xmath69 is a strata - preserving homeomorphism . this implies that if @xmath70 contains exactly @xmath71 zero coordinates then @xmath72 also contains exactly @xmath71 zero coordinates for @xmath73 . we call the collection of compatible charts with corners @xmath74 whose domains cover @xmath63 an _ atlas_. then , its maximal atlas is called a _ structure with corners _ of @xmath63 . a topological manifold with boundary together with a structure with corners is called a _ ( topological ) manifold with corners_. let @xmath75 be a point of an @xmath8-dimensional manifold with corners @xmath63 . a chart @xmath64 with corners such that @xmath76 defines the number @xmath77 $ ] by the number of zero - coordinates of @xmath72 . by the compatibility of charts , this number is independent on the choice of a chart with corners which contains @xmath78 . therefore , the map @xmath79 $ ] is well - defined . the number @xmath80 is called a _ depth _ of @xmath78 . we call the closure of a connected component of @xmath81 ( @xmath73 ) a _ codimension-@xmath71 face_. in particular , the codimension-@xmath82 face is @xmath63 itself . moreover , codimension-@xmath83 , @xmath84 and @xmath8 faces are called _ facet _ , _ edge _ and _ vertex _ , respectively . the set of all edges and vertices is called a _ one - skeleton of @xmath63 _ ( or a _ graph of @xmath63 _ ) . by restricting the structure with corners on @xmath63 to faces , we may regard each codimension-@xmath71 face as an @xmath85-dimensional ( sub)manifold with corners . an @xmath8-dimensional manifold with corners @xmath86 is said to be a _ manifold with faces _ ( or a _ nice manifold with corners _ ) if @xmath86 satisfies the following conditions : 1 . for every @xmath87 $ ] , there exists a codimension-@xmath71 face ; 2 . for each codimension-@xmath71 face @xmath88 , there are exactly @xmath71 facets @xmath89 such that @xmath88 is a connected component of @xmath90 ; moreover , @xmath91 for any facet @xmath92 ( @xmath93 ) . let @xmath10 be a torus manifold . when @xmath10 is locally standard , by using the differentiable slice theorem , the orbit space @xmath60 has the structure of an @xmath8-dimensional manifold with faces . on the other hand , when @xmath6 satisfies @xmath2 , its orbit space @xmath60 satisfies a stronger condition by the following masuda - panov theorem ( see ( * ? ? ? * lemma 2.1 ) and ( * ? ? ? * theorem 2 ) ) : [ 2-equiv ] let @xmath6 be a @xmath7-dimensional torus manifold . then , the following two conditions are equivalent : 1 . @xmath2 ; 2 . the @xmath42-action on @xmath6 is locally standard and its orbit space @xmath60 has the structure of an @xmath8-dimensional face acyclic manifold with corners . here , in theorem [ 2-equiv ] , an @xmath8-dimensional _ face acyclic _ manifold with corners @xmath86 is an @xmath8-dimensional manifold with faces such that all faces @xmath94 ( include @xmath86 ) of @xmath86 are acyclic , i.e. , @xmath95 . for example , if @xmath86 is a simply connected @xmath1-dimensional face acyclic manifold with corners , then it is easy to check that the boundary of @xmath86 is homeomorphic to the @xmath16-sphere @xmath5 . moreover , in this case , we can also check that @xmath86 itself is homeomorphic to the @xmath1-dimensional disk @xmath96 . therefore , as one of the conclusions of theorem [ 2-equiv ] , we have the following corollary : [ disk ] let @xmath6 be a simply connected @xmath0-dimensional torus manifold with @xmath2 . then , its orbit space @xmath60 is homeomorphic to the @xmath1-dimensional disk . by the definition of a manifold with faces @xmath86 , we can define a simplicial poset ( partially ordered set ) @xmath97 , called a _ face poset _ of @xmath86 ( see @xcite ) , by the set of faces in @xmath86 with the empty - set @xmath98 ordered by inclusion , where the empty set @xmath98 is the smallest element by this order , say @xmath99 . we often denote the face poset structure of @xmath86 as @xmath100 . let @xmath101 and @xmath102 be @xmath8-dimensional manifolds with faces . we call @xmath101 and @xmath102 are _ combinatorially equivalent _ if their face posets @xmath103 and @xmath104 are isomorphic as a poset ( i.e. , there is an order - preserving bijection between them ) . we denote them by @xmath105 . by definition of weakly equivariant homeomorphism , if two locally standard torus manifolds @xmath106 and @xmath107 are weakly equivariantly homeomorphic then @xmath108 . let @xmath6 be a @xmath7-dimensional locally standard torus manifold . by the argument demonstrated in section [ sect2.2 ] , the orbit map @xmath109 may be regarded as the projection onto some manifold with faces @xmath86 . let @xmath110 be the set of all facets in @xmath86 . by the definition of facet @xmath111 , its preimage @xmath112 is a characteristic submanifold @xmath49 . then , there exists the circle subgroup @xmath113 fixing @xmath114 ( recall that @xmath115 ) . recall that @xmath51 is determined by a primitive element in @xmath116 ( the lattice of lie algebra of @xmath42 ) . therefore , by taking this primitive element ( up to sign ) in @xmath117 , we can define the following map : @xmath118 where @xmath119 represents the quotient of @xmath117 by signs . we call @xmath120 a _ characteristic function_. now the choice of omniorientation @xmath54 of @xmath6 determines the sign of @xmath120 as follows . fix an omniorientation @xmath54 of @xmath6 . namely , we fix the orientation of the tangent bundle of @xmath6 ( resp . @xmath49 ) , say @xmath121 ( resp . @xmath122 ) . restricting @xmath121 to the submanifold @xmath49 , say @xmath123 , we obtain the @xmath9-equivariant decomposition @xmath124 , where @xmath125 is the @xmath51-equivariant normal bundle of @xmath49 . therefore , because we fix the orientation of @xmath123 ( induced from the orientation of @xmath121 ) and that of @xmath122 , we may choose an orientation of @xmath125 such that the orientation of @xmath123 coincides with that of @xmath126 ( thus , we may regard @xmath125 as the complex line bundle over @xmath49 ) . because @xmath51 acts on @xmath125 , we may choose an orientation of @xmath51 such that the @xmath51-action preserves the orientation of @xmath125 . this orientation of @xmath51 determines the sign of @xmath127 for @xmath50 . by this way , we have the following function : @xmath128 this is called an _ omnioriented characteristic function _ ( of @xmath59 ) , in this paper . [ rem - chfct ] the characteristic function defined in @xcite may be regarded as the characteristic function @xmath120 as above . on the other hand , the characteristic function defined in @xcite may be regarded as the characteristic function @xmath129 as above by taking an appropriate omniorientation ( also see ( * ? ? ? * section 5.2 ) ) . let @xmath130 . we define the subset @xmath131 $ ] as follows : @xmath132\ |\ p\in m_{i}\}.\ ] ] by the differentiable slice theorem around @xmath130 , we have that its cardinality @xmath133 for every @xmath130 . put @xmath134 . because the @xmath42-action on @xmath6 is effective , @xmath135 spans @xmath136 , i.e. , the determinant of the induced @xmath137-matrix @xmath138 satisfies that @xmath139 similarly , we have @xmath140 for each @xmath8 facets such that @xmath141 for some vertex @xmath142 ( called _ the facets around a vertex _ ) . motivated by the observations as above , we may abstractly define the characteristic function on a manifold with faces as follows ( see @xcite , @xcite for simple polytopes and @xcite , @xcite for manifold with faces ) : let @xmath86 be an @xmath8-dimensional manifold with faces and @xmath143 be the set of its facets . let @xmath117 be the lattice of lie algebra of @xmath9 and @xmath119 be its quotient by @xmath144 . then , a function @xmath145 is said to be a _ characteristic function _ if @xmath120 satisfies the relation for the facets around every vertex , and a function @xmath146 is said to be an _ omnioriented characteristic function _ if @xmath129 satisfies the relation for the facets around every vertex . we denote an @xmath8-dimensional manifold with faces @xmath86 with its characteristic function @xmath120 ( resp . omnioriented characteristic function @xmath129 ) by @xmath147 ( resp . @xmath148 ) . let @xmath101 and @xmath102 be manifolds with faces and @xmath149 ( resp . @xmath150 ) and @xmath151 ( resp . @xmath152 ) be their ( resp . omnioriented ) characteristic functions , respectively . assume that @xmath153 and it is induced by the bijective map @xmath154 . denote its restriction onto the set of facets as @xmath155 then , we call that @xmath156 and @xmath157 are _ combinatorially equivalent _ if the following diagram commutes : @xmath158^{f } \ar[r]^{\lambda_{1 } } & { \mathfrak{t}}_{{{\mathbb{z}}}}/\{\pm 1\ } \ar[d]^{id } \\ & \mathcal{f}(q_{2 } ) \ar[r]^{\lambda_{2 } } & { \mathfrak{t}}_{{{\mathbb{z}}}}/\{\pm 1\ } } \ ] ] similarly , @xmath159 and @xmath160 are _ combinatorially equivalent _ if the following diagram commutes : @xmath158^{f } \ar[r]^{\quad \lambda_{\mathcal{o}_{1 } } } & { \mathfrak{t}}_{{{\mathbb{z } } } } \ar[d]^{id } \\ & \mathcal{f}(q_{2 } ) \ar[r]^{\quad \lambda_{\mathcal{o}_{2 } } } & { \mathfrak{t}}_{{{\mathbb{z } } } } } \ ] ] note that the characteristic function @xmath120 can be obtained by ignoring sings from the omnioriented characteristic function @xmath129 ; we call such @xmath120 an _ induced characteristic function _ from @xmath129 . on the other hand , by choosing a sign for each facet , we can obtain an omnioriented characteristic function @xmath129 from the characteristic function @xmath120 ; we call such @xmath129 an _ induced oriented characteristic function _ from @xmath120 . therefore , we have the following lemma : [ omni->nonomni ] if @xmath159 and @xmath160 are combinatorially equivalent , then their induced @xmath156 and @xmath157 are also combinatorially equivalent . if @xmath156 and @xmath157 are combinatorially equivalent , then there are induced omnioriented characteristic functions @xmath150 and @xmath152 such that @xmath159 and @xmath160 are combinatorially equivalent . now we may introduce one of the key facts to prove our main theorem ( see ( * ? ? ? * theorem 1.3 , theorem 6.1 ) ) : [ wi2 ] let @xmath106 and @xmath107 be a @xmath0-dimensional simply connected torus manifold with @xmath2 and @xmath156 and @xmath157 be their orbit spaces with characteristic functions . then , the following three statements are equivalent : 1 . @xmath156 and @xmath157 are combinatorially equivalent ; 2 . @xmath106 and @xmath107 are equivariantly homeomorphic ; 3 . @xmath106 and @xmath107 are equivariantly diffeomorphic . therefore , by corollary [ disk ] and theorem [ wi2 ] , in order to classify all @xmath0-dimensional simply connected torus manifolds with @xmath2 , it is enough to classify all @xmath147 s up to combinatorial equivalent , where @xmath86 is a @xmath1-dimensional disk equipped with the structure of a manifold with faces . let @xmath59 be an omnioriented locally standard @xmath7-dimensional torus manifold and @xmath148 be its orbit space with an omnioriented characteristic function . from the one - skeleton of @xmath148 , we can define a labelled graph , called a _ torus graph_. one of the key arguments to prove the main theorem is to classify all possible torus graphs ( see section [ sect7 ] ) . to do that , in this section , we recall a torus graph defined by maeda - masuda - panov @xcite . let @xmath161 be the graph of @xmath86 . let @xmath162 be its vertices and @xmath163 be its oriented edges , i.e. , we distinguish two edges @xmath164 and @xmath165 . for @xmath166 , we denote the set of outgoing edges from @xmath78 as @xmath167 . because @xmath86 is an @xmath8-dimensional manifold with faces , @xmath168 and each edge @xmath169 is a connected component of @xmath170 , for some @xmath171 . moreover , for @xmath166 which is one of two vertices on @xmath172 , there is another facet @xmath173 such that @xmath174 is a connected component of @xmath175 . in other words , @xmath176 may be regarded as a normal facet of @xmath169 on @xmath166 . put @xmath177 . then , there exists unique @xmath178 such that @xmath179 where @xmath180 represents the pairing of @xmath181 and @xmath182 . therefore , by this way , we can define a map @xmath183 from the omnioriented characteristic function @xmath129 . this map @xmath184 is called an _ axial function _ on @xmath161 . we call the labelled graph @xmath185 a _ torus graph _ induced from @xmath148 ( or equivalently @xmath59 ) . we denote such a torus graph as @xmath186 ( or @xmath187 ) . we can easily check the following proposition by definition of torus graphs ( also see @xcite ) : [ connection ] let @xmath185 be a torus graph induced from @xmath148 . then , @xmath161 is an @xmath8-valent regular graph , i.e. , @xmath168 for all @xmath166 , and @xmath185 satisfies the following conditions : 1 . @xmath188 , where @xmath189 is the orientation reversed edge of @xmath172 ; 2 . @xmath190 spans @xmath191 for all vertices @xmath166 ; 3 . there is a bijection @xmath192 for all edges whose initial vertex is @xmath78 and terminal vertex is @xmath193 such that the following conditions hold : 1 . @xmath194 ; 2 . @xmath195 ; 3 . @xmath196 for all @xmath197 . we call @xmath198 a _ connection _ on @xmath185 . the original torus graph ( induced from an omnioriented torus manifold ) is defined by using the _ tangential representations _ , see @xcite , @xcite . the definition of torus graph as above is essentially same with this definition . in @xcite , motivated by the gkm graph introduced by guillemin - zara in @xcite , an @xmath8-valent graph @xmath161 with a label @xmath183 which satisfies three conditions in proposition [ connection ] is called an _ ( abstract ) torus graph _ ( i.e. , there might be no geometric objects which define @xmath185 ) . we next define the equivalence relation between two torus graphs . we call the map @xmath199 a _ graph isomorphism _ , if the restricted map @xmath200 and @xmath201 are bijective and the following map commutes : @xmath202^{\pi_{v_{1 } } } \ar[r]^{f|_{e } } & e(\gamma_{2 } ) \ar[d]^{\pi_{v_{2 } } } \\ & v(\gamma_{1 } ) \ar[r]^{f|_{v } } & v(\gamma_{2 } ) } \ ] ] where @xmath203 is the map projecting onto the initial vertex , i.e. , @xmath204 . in other words , the bijection @xmath205 preserves the edges . now we may define the equivalence relation . [ def_torus - graph ] let @xmath206 and @xmath207 be torus graphs . we say @xmath206 and @xmath207 are _ equivalent _ if there is a graph isomorphism @xmath208 such that the following diagram commutes : @xmath202^{f|_{e } } \ar[r]^{\quad \mathcal{a}_{1 } } & { \mathfrak{t}}_{{{\mathbb{z}}}}^ { * } \ar[d]^{id } \\ & e(\gamma_{2 } ) \ar[r]^{\quad \mathcal{a}_{2 } } & { \mathfrak{t}}_{{{\mathbb{z}}}}^ { * } } \ ] ] assume @xmath209 . let @xmath210 be the set of @xmath71-valent torus subgraphs in @xmath185 , i.e. , @xmath71-valent subgraphs in @xmath161 closed under the connection @xmath211 , where @xmath212 and we define @xmath213 . then , the set @xmath214 admits the structure of a simplicial poset by inclusion ( see @xcite ) . we denote this structure by @xmath215 . let @xmath97 be the face poset of @xmath86 ( see section [ sect2.2 ] ) and @xmath216 be the set of all @xmath71-dimensional faces , where @xmath212 and @xmath217 . then , each element of @xmath210 is nothing but the graph of an element in @xmath216 . this implies that the poset @xmath215 is equivalent to the poset @xmath218 . therefore , we have the following lemma : [ equiv<->equiv ] the following two statements are equivalent : 1 . two manifolds with faces with omnioriented characteristic functions @xmath159 and @xmath160 are combinatorially equivalent ; 2 . their induced torus graphs @xmath219 and @xmath220 are equivalent . by lemma [ omni->nonomni ] , theorem [ wi2 ] and lemma [ equiv<->equiv ] , we have the following corollary : [ key2 ] let @xmath221 and @xmath222 be @xmath0-dimensional simply connected torus manifolds with vanishing odd degree cohomology . then , the following statements are equivalent : 1 . @xmath221 and @xmath222 are equivariantly diffeomorphic ; 2 . their orbit spaces , i.e. , @xmath1-dimensional disks with the structures of manifolds with faces , with characteristic functions @xmath223 and @xmath224 are combinatorially equivalent ; 3 . there are omnioriented characteristic functions @xmath150 and @xmath152 such that their induced @xmath1-valent torus graphs @xmath225 and @xmath226 are equivalent . therefore , in order to prove our main theorem ( theorem [ main-1 ] ) , it is enough to classify all @xmath1-valent torus graphs @xmath185 , induced from @xmath59 , up to equivalent . let @xmath10 be a simply connected , @xmath0-dimensional torus manifold with @xmath2 , and @xmath227 be its torus graph induced by some omniorientation . as a preliminary to prove the main theorem ( theorem [ main-1 ] ) , in this section , we will introduce some of basic torus graphs @xmath185 and their corresponding @xmath0-dimensional torus manifolds @xmath10 . because the induced torus graphs from @xmath10 are @xmath1-valent , if there is a @xmath1-multiple edge , i.e. , three edges that are incident to the same two vertices , then it follows from proposition [ connection ] that such torus graph must be the torus graph in figure [ tg - s^6 ] . ( we denote the torus graph in figure [ tg - s^6 ] as @xmath228 . ) , where @xmath229 are @xmath230-basis . ] put @xmath231 , @xmath232 and @xmath233 , by the standard basis @xmath234 in @xmath235 . then , the following equation holds : @xmath236 let @xmath237 be the unit sphere , i.e. , the set @xmath238 such that @xmath239 . define the @xmath36-action on the first three complex coordinates in @xmath3 by @xmath240 where @xmath241 and @xmath242 , @xmath243 , is a one - dimensional complex representation defined by @xmath244 then , by choosing an appropriate omniorientation on @xmath3 , we have that its induced torus graph is equivalent to @xmath228 . therefore , by using corollary [ key2 ] , we have the following lemma : [ s^6 ] let @xmath59 be an omnioriented @xmath0-dimensional simply connected torus manifold with @xmath2 . if its induced torus graph is @xmath228 , then @xmath10 is equivariantly diffeomorphic to one of @xmath245 defined by . assume that a @xmath1-valent torus graph @xmath185 does not have @xmath1-multiple edges but have multiple edges , i.e. , two edges that are incident to the same two vertices . in this section , we classify the easiest case of such torus graphs . because @xmath161 is a one - skeleton of @xmath1-dimensional manifold with faces @xmath86 , the number of vertices @xmath246 . assume that @xmath247 . then , we can easily check that such torus manifold is the one - skeleton of the @xmath1-simplex ( see figure [ tg - cp^3 ] in section [ sect4.3 ] ) or the graph drawn in figure [ tg - bundles ] , say @xmath248 . as is well known that the torus manifold whose torus graph is the one - skeleton of the @xmath1-simplex is equivariantly diffeomorphic to the complex projective space with some @xmath42-action ( see e.g. @xcite , and also see figure [ tg - cp^3 ] in section [ sect4.3 ] ) . so , we only study the torus manifold which induces the graph @xmath248 . because @xmath86 is homeomorphic to @xmath96 , we may regard @xmath86 whose one - skeleton is @xmath248 as the product of @xmath249 , where @xmath250 is the @xmath16-dimensional disk and @xmath251 is the interval . by considering all functions on facets of @xmath86 which satisfies , we can classify all omnioriented characteristic functions @xmath129 on @xmath86 . then , by the way to induce the axial function @xmath252 from @xmath148 demonstrated in section [ sect3 ] , we can obtain all possible axial functions on @xmath248 as shown in figure [ tg - bundles ] . , where @xmath253 and @xmath254 and @xmath255 are @xmath230-basis of @xmath191 . ] the torus graph @xmath256 in figure [ tg - bundles ] can be induced from an @xmath21-bundle over @xmath41 as follows . first , by choosing @xmath253 , we may define two free @xmath15-actions on @xmath257 as follows : @xmath258 we denote @xmath259 with the above @xmath260-action by @xmath261 . note that @xmath262 is diffeomorphic to the @xmath16-sphere @xmath41 , and a complex line bundle over @xmath41 can be denoted by @xmath263 where @xmath264 is the complex @xmath83-dimensional @xmath260-representation space by @xmath71-times rotation for some @xmath25 . let @xmath265 be the trivial real line bundle over @xmath41 . take the unit sphere bundle of the following whitney sum of three vector bundles for @xmath254 : @xmath266 then , we obtain the @xmath4-bundle over @xmath5 denoted by @xmath267 for @xmath253 , @xmath254 . namely , we can identify elements in @xmath268 by @xmath269=[(t^{-1}w , t^{\epsilon}z),(t^{a}x , t^{b}y , r)]\ ] ] for any @xmath270 such that @xmath271 and @xmath272 . define a @xmath35-action on @xmath268 by @xmath269\mapsto [ ( t_{1}w , z),(t_{2}x , t_{3}y , r)],\ ] ] where @xmath273 . fix an omniorientation on @xmath268 by the induced orientations from @xmath274 . then , considering the tangential representations around each fixed point , it is easy to check that the induced torus graph is @xmath275 , where @xmath234 are the standard basis of @xmath276 . therefore , by taking the appropriate automorphism of @xmath35 , we can construct each torus graph @xmath256 in figure [ tg - bundles ] from @xmath268 . note that if @xmath277 and @xmath278 , then this is nothing but one of the torus manifolds which appeared in the classifications of torus manifolds with codimension one extended actions in @xcite . by the argument as above and corollary [ key2 ] , we establish the following lemma : [ s^4-s^2(geom ) ] let @xmath59 be an omnioriented @xmath0-dimensional simply connected torus manifold with @xmath2 . if its induced torus graph has @xmath20-vertices , then @xmath10 is equivariantly diffeomorphic to one of the followings : 1 . @xmath279 with the standard @xmath35-action up to automorphism of @xmath35 ; 2 . @xmath268 for some @xmath253 and @xmath254 . assume that there are no multiple edges in a @xmath1-valent torus graph @xmath185 , i.e. , there are no two edges that are incident to the same two vertices . a graph @xmath161 is called a _ simple _ if @xmath161 does not have both of multiple edges and loops . in this and section [ sect5 ] , we study simple torus graphs which can be realized as the one - skeleton of manifold with faces homeomorphic to @xmath96 . the typical example of such torus manifolds whose torus graphs are simple is a quasitoric manifold ( introduced by davis - januszkiewicz in @xcite ( also see @xcite ) ) . a _ quasitoric manifold _ is defined by a torus manifold whose orbit space is a _ simple convex polytope _ , i.e. , a convex polytope admitting the structure of a manifold with faces . for example , the complex projective space @xmath280 with the standard @xmath281-action is the quasitoric manifold whose orbit space is the @xmath8-dimensional simplex . the figure [ tg - cp^3 ] shows the torus graph induced from @xmath282 , i.e. , the omniorientation @xmath283 induced from the standard complex structure on @xmath284 and the standard @xmath42-action on @xmath284 . . ] we next characterize when torus graphs are induced from simple convex polytopes , i.e. , induced from quasitoric manifolds . the steinitz theorem ( see ( * ? ? ? * chapter 4 ) ) tells us that a graph @xmath161 is the one skeleton of a @xmath1-dimensional convex polytope if and only if @xmath161 is a simple , planner and @xmath1-connected graph , where @xmath161 is called a _ @xmath1-connected _ graph if it remains connected whenever fewer than @xmath1 vertices are removed . it easily follows from the steinitz theorem that we have the following lemma : [ polytope ] let @xmath86 be a manifold with faces and @xmath161 be its graph . assume that @xmath86 is homeomorphic to the @xmath1-disk @xmath96 and there are no multiple edges . then , the following two statements are equivalent : 1 . @xmath86 is combinatorially equivalent to a @xmath1-dimensional simple convex polytope @xmath285 ; 2 . @xmath161 is a @xmath1-connected graph . therefore , together with corollary [ key2 ] , we have the following fact : [ quasi - toru ] let @xmath59 be an omnioriented @xmath0-dimensional simply connected torus manifold with @xmath2 . then , the following two statements are equivalent : 1 . @xmath10 is equivariantly diffeomorphic to a quasitoric manifold ; 2 . its induced torus graph @xmath161 is a @xmath1-connected graph with no multiple edges . by the arguments in section [ sect4 ] , only the following case remains : the simply connected @xmath0-dimensional torus manifolds with @xmath2 whose induced torus graphs are simple but not @xmath1-connected . such torus manifolds can be constructed by using the connected sum of `` oriented '' torus graphs . the purpose of this section is to introduce oriented torus graphs and their connected sum . we first recall the equivariant connected sum of torus manifolds . let @xmath106 , @xmath107 be @xmath7-dimensional torus manifolds and @xmath286 , @xmath287 be fixed points . by using the slice theorem , we may take @xmath42-invariant open neighborhoods @xmath288 of @xmath78 and @xmath289 of @xmath193 . assume that @xmath290 and @xmath291 are equivariantly diffeomorphic . then , @xmath292 and @xmath293 are equivariantly diffeomorphic to @xmath294 , where @xmath295 with some effective @xmath9-action and @xmath296 with the trivial @xmath9-action for some @xmath297 . then , we glues these two neighborhood by @xmath298 defined by the identity on @xmath299 and the map @xmath300 on @xmath251 for @xmath301 . namely , we can glue @xmath302 and @xmath303 by the following identification : @xmath304 the @xmath9-manifold obtained by this way is denoted by @xmath305 or @xmath306 ( if we emphasize fixed points @xmath286 and @xmath287 ) . because each torus manifold has more than two fixed points , @xmath305 is again a torus manifold . we call this operation an _ equivariant connected sum_. the following lemma holds : [ lem - easy ] if two torus manifolds @xmath106 and @xmath107 are simply connected and @xmath307 , then @xmath305 is also simply connected and @xmath308 . it is easy to check the statement by using van - kampen s theorem and the mayer - vietoris exact sequence . assume that @xmath309 , @xmath310 are @xmath0-dimensional omnioriented simply connected torus manifolds with @xmath307 . let @xmath206 and @xmath207 be their induced @xmath1-valent torus graphs . assume that we can glue @xmath311 and @xmath287 by the connected sum . then , by considering the restriction of @xmath298 in onto @xmath295 , i.e. , the identity map , the axial functions around @xmath312 and @xmath313 must satisfy @xmath314 however , at this stage , torus graphs @xmath206 and @xmath207 do not have the information of orientations of @xmath106 and @xmath107 . note that , to do connected sum , we need the orientations around @xmath286 and @xmath287 . to encode the orientations around fixed points , we need the following notion : let @xmath185 be a torus graph . we call a triple @xmath315 with a map @xmath316 an _ oriented torus graph _ , if @xmath316 satisfies the following condition for all @xmath169 : @xmath317 where @xmath318 is the initial vertex of @xmath169 , i.e. , for @xmath319 , @xmath320 and @xmath321 . we call a map @xmath322 an _ orientation _ of @xmath185 . let @xmath59 be an omnioriented torus manifold . the oriented torus graph @xmath315 of @xmath59 is defined as follows . let @xmath130 . then , there exist exactly @xmath8 characteristic submanifolds @xmath323 such that @xmath78 is a connected component of @xmath324 . now the fixed orientations of @xmath325 determine the decomposition of the tangential representation , i.e. , @xmath326 is determined by fixing the orientations of @xmath325 . on the other hand , the orientation of @xmath6 determines the orientation of @xmath327 . so , we define the map @xmath328 by @xmath329 let @xmath330 and @xmath331 be the induced oriented torus graphs from @xmath309 and @xmath310 . if we can glue @xmath311 and @xmath287 by the connected sum , then both of the relation and the relation @xmath332 hold ( corresponds to that the orientations on @xmath333 and @xmath334 are different ) . note the induced ( oriented ) torus graph by @xmath335 is nothing but the one - skeleton of the connected sum @xmath336 of manifolds with faces , where @xmath337 is the orbit space of @xmath49 , @xmath338 ( also see ( * ? ? ? * definition 3 ) , ( * ? ? ? * section 3.1 ) for details about the connected sum of polytopes ) . therefore , conversely , if @xmath312 and @xmath313 satisfy and , then we can do the _ connected sum of ( oriented ) torus graphs _ between @xmath330 and @xmath331 , say @xmath339 or @xmath340 ( if we emphasize the vertices @xmath312 and @xmath313 ) . more precisely , @xmath339 is defined as follows ( also see figure [ operations1 ] ) . 1 . @xmath341 ; 2 . @xmath342 , where @xmath343 for @xmath243 ; 3 . @xmath344 such that @xmath345 and @xmath346 for @xmath347 and @xmath348 , and @xmath349 and @xmath350 ; 4 . @xmath351 is defined by @xmath352 for @xmath353 and @xmath354 for @xmath355 . of torus manifold ( from left to right ) and its inverse @xmath356 ( from right to left ) , where @xmath357 . here , @xmath358 are @xmath230-basis of @xmath359 and @xmath360 . ] then , we can easily check that @xmath315 is an oriented torus graph . using corollary [ key2 ] and the arguments as above , we have the following lemma : [ con - sum canonical model ] let @xmath106 , @xmath107 be @xmath0-dimensional simply connected torus manifolds with @xmath307 , and @xmath330 , @xmath331 be their induced oriented torus graphs from some omniorientations , respectively . if @xmath361 , then @xmath315 is the oriented torus graph induced from @xmath362 with some omniorientation . by using the connected sum , we can construct the torus manifolds which are not appeared in section [ sect4 ] . for example , the following torus manifold is one of such examples : @xmath363 where @xmath364 is the reversed orientation of @xmath279 . the figure [ others ] shows the torus graph induced from @xmath365 ( see the axial functions on figure [ tg - bundles ] , [ tg - cp^3 ] for details ) . we can easily check that this graph is @xmath1-valent , simple and planner but not @xmath1-connected ; therefore , due to lemma [ quasi - toru ] , this manifold is not a quasitoric manifold . ) induced from @xmath365 . ] to prove the main theorem ( theorem [ main-1 ] ) , in this section , we show the following two lemmas . [ one - pt ] let @xmath86 be a @xmath1-dimensional manifold with faces which is homeomorphic to @xmath96 and @xmath161 be its graph . then , @xmath366 is connected , for all vertices @xmath166 . because @xmath86 is homeomorphic to the @xmath1-disk @xmath96 , @xmath161 may be regarded as a planner graph by the stereographic projection of @xmath367 . assume @xmath366 is not connected . because @xmath86 is a @xmath1-dimensional manifold with faces , there are exactly three out - going edges from @xmath78 , say @xmath368 , @xmath369 and @xmath370 . therefore , we may assume that there exists a connected component @xmath371 in @xmath366 such that @xmath372 but @xmath373 ( see figure [ one - pt - con ] ) . . the facet @xmath94 has the self - intersection on the edge @xmath374 . ] since @xmath371 is also a planner @xmath1-valent graph excepts on the vertex @xmath375 ( because @xmath376 ) , there is a @xmath16-valent subgraph in @xmath371 , say @xmath377 , such that @xmath377 splits @xmath367 into two connected components @xmath378 and @xmath379 , where @xmath380 but @xmath381 . this implies that there is a facet @xmath94 in @xmath86 such that @xmath382 contains @xmath377 and @xmath374 . however , in this case , @xmath374 must be a self - intersection edge of @xmath94 ( see figure [ one - pt - con ] ) . this gives a contradiction to that @xmath86 is a manifold with faces . this establishes the statement . by lemma [ one - pt ] , if @xmath161 is not @xmath1-connected , then there are two vertices @xmath383 such that @xmath384 is not connected but both of @xmath366 and @xmath385 are connected . more precisely , we have the following lemma : [ separate - points ] let @xmath86 be a @xmath1-dimensional manifold with faces which is homeomorphic to @xmath96 and @xmath161 be its graph . assume that there are two vertices @xmath383 such that @xmath386 for any facets @xmath94 , i.e , @xmath78 and @xmath193 are not on the same facet @xmath94 . then , @xmath384 is connected . assume that @xmath387 are not on the same facet of @xmath86 . because @xmath86 is a manifold with faces , there are mutually distinct facets @xmath388 such that @xmath174 is a component of @xmath389 and @xmath390 is that of @xmath391 and we can take vertices @xmath392 and @xmath393 such that @xmath394 and @xmath395 are all outgoing edges from @xmath78 and @xmath193 , for @xmath243 . take two vertices @xmath396 from @xmath397 . by lemma [ one - pt ] , @xmath398 is connected . so there is a path @xmath28 from @xmath399 to @xmath400 in @xmath398 . if @xmath28 does not go through @xmath78 , then @xmath399 and @xmath400 are connected in @xmath397 . assume that this path @xmath28 goes through @xmath78 . then @xmath28 goes through exactly two vertices @xmath401 , @xmath402 ( we may assume @xmath375 and @xmath403 ) . moreover , one of the facets @xmath404 , say @xmath405 , contains both of @xmath375 , @xmath403 . note that @xmath405 corresponds to the @xmath16-valent subgraph in @xmath161 . therefore , we can take the path @xmath406 connecting @xmath375 , @xmath403 on @xmath405 which is not the path @xmath407 . because @xmath387 are not on the same facet , in particular @xmath408 , the path @xmath406 does not contain @xmath193 . hence , the connected subgraph @xmath409 contains both of @xmath410 but does not contain both of @xmath411 . thus , we can take the path @xmath412 from @xmath399 to @xmath400 in @xmath413 . this establishes that @xmath397 is connected . in summary , by lemma [ one - pt ] and [ separate - points ] , we have the following fact . [ key3 ] let @xmath161 be a one - skeleton of @xmath1-dimensional manifold with faces @xmath86 . then , for all @xmath166 , @xmath366 is connected . furthermore , if @xmath384 is not connected , then @xmath411 is on the same facet . the main theorem of this paper can be stated as follows : [ main-1 ] let @xmath6 be a simply connected , @xmath0-dimensional torus manifold with @xmath2 . then , @xmath6 is equivariantly diffeomorphic to one of the following manifolds : 1 . @xmath414 with a torus action induced from a ( faithful ) representation of @xmath35 on @xmath415 ; 2 . @xmath0-dimensional quasitoric manifold @xmath416 ; 3 . @xmath21-bundle over @xmath41 which is equivariantly diffeomorphic to @xmath268 for some @xmath417 , or otherwise , there are some @xmath0-dimensional quasitoric manifolds @xmath418 , for some @xmath419 , and some @xmath21-bundles over @xmath41 , say @xmath420 ( for some @xmath421 , @xmath422 and @xmath423 ) , such that @xmath6 is equivariantly diffeomorphic to @xmath424 where @xmath425 represents the equivariant connected sum around fixed points , @xmath426 for @xmath427 , and the case @xmath428 means that there is no @xmath418 factor . in this final section , we prove theorem [ main-1 ] . let @xmath6 be a simply connected , @xmath0-dimensional torus manifold with @xmath2 , @xmath86 be its orbit space which is homeomorphic to @xmath96 and @xmath187 be its induced oriented torus graph ( we omit the orientation ) . because @xmath429 is a one - skeleton of a manifold with faces which is homeomorphic to @xmath96 , it is easy to check that @xmath430 . if @xmath431 , by lemma [ s^6 ] , we have that @xmath6 is equivariantly diffeomorphic to @xmath40 , i.e. , the statement ( 1 ) . if @xmath432 , it follows from lemma [ s^4-s^2(geom ) ] that @xmath6 is equivariantly diffeomorphic to a quasitoric manifold @xmath284 or @xmath268 for some @xmath253 , @xmath254 , i.e. , the statement ( 2 ) or ( 3 ) occurs . so we may only prove the case when @xmath433 . we first claim the following lemma : [ 1-claim ] assume that @xmath433 and there is a multiple edge in @xmath429 . then , @xmath187 can be decomposed into the connected sum of the following torus graphs : @xmath434 or @xmath435 where @xmath436 is a torus graph without multiple edges and @xmath420 for @xmath437 . assume two vertices @xmath78 , @xmath193 are connected by a multiple edge , i.e. , two edges ( see the bottom graph in figure [ essential discussion to prove this theorem ] ) . as any generators in @xmath359 and @xmath438 and @xmath439 . here , the bottom graph is @xmath187 and the left upper graph is @xmath440 and the right upper graph is @xmath441 . note that if we fix the orientation of @xmath187 then the orientations of @xmath440 and @xmath441 are automatically determined . ] then , by the connection of the torus graph ( see proposition [ connection ] ) , it is easy to check that the axial functions around the vertex @xmath399 of the bottom graph in figure [ essential discussion to prove this theorem ] satisfy the axial functions expressed in figure [ essential discussion to prove this theorem ] , where we can take @xmath442 , @xmath443 , @xmath28 as @xmath230-basis @xmath359 . in this case , we can do ( inverse ) connected sum such as expressed in figure [ essential discussion to prove this theorem ] ( from the bottom to the top in figure [ essential discussion to prove this theorem ] ) . then , the induced torus graph @xmath187 is decomposed into two induced torus graphs @xmath440 and @xmath441 , where @xmath444 is some simply connected @xmath0-dimensional torus manifold with @xmath445 by lemma [ lem - easy ] . namely , we have @xmath446 if there is no multiple edges in @xmath447 , then we may put @xmath448 . assume that there is a multiple edge in @xmath447 . if there are only @xmath20 vertices in @xmath447 , then we may put @xmath444 as @xmath449 by lemma [ s^4-s^2(geom ) ] . when there are more than @xmath20 vertices in @xmath447 , with iterating the arguments as above , finally we establish the statement of this lemma . therefore , to prove theorem [ main-1 ] , it is enough to show the following lemma : [ 2-claim ] assume that @xmath433 and there are no multiple edges in @xmath429 . then , @xmath187 can be decomposed into the connected sum of the following torus graphs : @xmath450 where @xmath451 for @xmath419 is the torus graph induced from a quasitoric manifold @xmath418 , and @xmath420 for @xmath452 . if @xmath453 is @xmath1-connected , then it follows from lemma [ quasi - toru ] that the statement holds , i.e. , @xmath454 . therefore , we may assume @xmath161 is not @xmath1-connected . in this case , by corollary [ key3 ] , there is a @xmath16-valent torus subgraph @xmath455 such that some @xmath456 satisfy that @xmath397 is not connected . if @xmath94 is a triangle ( i.e. , @xmath457 ) , with the method similar to that demonstrated in the proof of lemma [ one - pt ] , we have that there is a face in @xmath86 which has the self - intersection edge . this gives a contradiction to that @xmath86 is a manifold with faces . therefore , we may assume @xmath458 . we first assume that @xmath164 is an edge of @xmath94 . then , there are two graphs @xmath371 and @xmath459 which are the connected components of @xmath397 expressed in figure [ edge - con ] . and @xmath193 from @xmath161 instead of @xmath78 and @xmath193 , the graph is also disconnected . ] if we remove the two vertices @xmath399 and @xmath193 from @xmath161 instead of @xmath78 , @xmath193 , where @xmath460 such that @xmath461 is an edge , then @xmath462 is also not connected ( see figure [ edge - con ] ) . therefore , from now , we may assume the followings : 1 . @xmath463 satisfy @xmath397 is not connected ; 2 . @xmath464 there is a @xmath16-valent torus subgraph ( facet ) @xmath94 with @xmath458 in @xmath161 such that @xmath456 . we call such facet @xmath94 a _ singular facet_. let @xmath94 be a singular facet . assume @xmath465 . when @xmath465 , where @xmath466 and @xmath467 . here , @xmath468 is a facet which intersects with @xmath94 on @xmath461 and @xmath469 . ] in this case , by the similar argument just before , we may take @xmath387 as in the position of figure [ 6-vertices ] , i.e. , @xmath78 and @xmath193 are on two separated edges @xmath470 and @xmath471 which are common edges of two facets @xmath94 and @xmath468 in figure [ 6-vertices ] ( note that @xmath399 and @xmath400 might be connected by an edge ) . moreover , by considering the omnioriented characteristic functions of the facets @xmath94 and @xmath468 , we may take the axial functions around the facet @xmath94 as in figure [ 6-vertices ] . by taking an appropriate orientation , we can do the connected sum as in figure [ 6-vertices2 ] ; here we denote the ( oriented ) torus graph containing @xmath371 as @xmath472 and that containing @xmath459 as @xmath473 . in figure [ 6-vertices ] splits into two torus graphs @xmath472 ( upper ) and @xmath473 ( lower ) . here , we omit the axial functions around the vertices @xmath474 because they are exactly same with that of figure [ 6-vertices ] . ] the torus graph obtained by this connected sum is nothing but the torus graph @xmath185 in figure [ 6-vertices ] . note that @xmath475 is simple and planner , and @xmath476 is just planner . with the method similar to that demonstrated in figure [ essential discussion to prove this theorem ] , @xmath473 can be obtained from the connected sum of @xmath256 and the simple , planner graph @xmath477 ( containing @xmath459 ) , where @xmath256 is one of the torus graphs ( by taking the appropriate axial functions ) in figure [ tg - bundles ] . namely , the torus graph in figure [ 6-vertices ] can be obtained from the following connected sum : @xmath478 here , it is easy to check that @xmath479 consists of @xmath480 and the other two facets , say @xmath481 and @xmath482 ( induced from @xmath94 and @xmath468 in @xmath161 ) . because of figure [ 6-vertices2 ] , the number of vertices of other two facets @xmath481 and @xmath482 are reduced ; in particular , the number of vertices of the facet @xmath481 induced from the singular facet @xmath94 is strictly less than @xmath0 . if both of @xmath472 and @xmath477 are @xmath1-connected , then these torus graphs are induced from quasitoric manifolds , i.e , the statements of lemma [ 2-claim ] hold . assume that @xmath472 is not @xmath1-connected . then , by the arguments before , there is a singular facet @xmath94 in @xmath472 . if @xmath465 , then @xmath472 also decomposes into @xmath483 by using the similar arguments explained in figure [ 6-vertices2 ] . iterating this arguments , we may reduce all singular facets with @xmath465 . more precisely , we may decompose @xmath185 in figure [ 6-vertices ] into @xmath484 where @xmath485 , for @xmath423 , is a torus graph in figure [ tg - bundles ] and @xmath486 , for @xmath487 , is a @xmath1-valent simple and planner torus graph which satisfies one of the followings : * @xmath1-connected ( in this case , @xmath486 is induced from a quasitoric manifold ) ; * otherwise , all singular facets @xmath94 satisfy @xmath488 or @xmath489 . assume that the number of vertices in every singular facet of the torus graph @xmath185 is less than or equal to @xmath489 . then , such torus graph is one of torus graphs expressed in figure [ 45-vertices ] . with @xmath490 or @xmath20 . here , @xmath468 is a facet which intersects with @xmath94 on @xmath461 and @xmath469 . ] however , because @xmath161 is the one - skeleton of a manifold with faces and not @xmath1-connected , it is easy to check that there exists the singular facet @xmath468 such that @xmath491 and @xmath492 . this gives a contradiction . hence , this case does not occur . this establishes lemma [ 2-claim ] . consequently , by lemma [ con - sum canonical model ] , [ 1-claim ] and [ 2-claim ] , we have the statement of theorem [ main-1 ] finally , by theorem [ main-1 ] and the mayer - vietoris exact sequence , we also have the following well - known result . [ 2nd deg ] let @xmath6 be a simply connected @xmath0-dimensional torus manifold whose cohomology ring is generated by the @xmath16nd degree cohomology . then , @xmath6 is a quasitoric manifold . the author would like to thank professors yael karshon and takashi tsuboi for providing him excellent circumstances to do research . |
we present a simple formalism describing evolution of a qubit in the process of its measurement in a circuit qed setup . when a phase - sensitive amplifier is used , the evolution depends on only one output quadrature , and the formalism is the same as for a broadband setup . when a phase - preserving amplifier is used , the qubit evolution depends on two output quadratures . in both cases a perfect monitoring of the qubit state and therefore a perfect quantum feedback is possible . | 1111.4016 | the goal of this lecture is to present a physical picture of the process of continuous quantum measurement of a qubit in the circuit quantum electrodynamics ( cqed ) setup @xcite ( fig . 1 ) , extending or reformulating the previous theoretical descriptions @xcite . understanding of the qubit evolution in the process of measurement is important for developing an intuition , which is useful in many cases , in particular in designing various schemes of the quantum feedback @xcite . when a quantum measurement is discussed @xcite , there are usually two different types of questions to answer : we can either focus on obtaining information on the initial state ( before measurement ) or focus on the quantum state after the measurement ( i.e. evolution in the process of measurement ) . let us emphasize that we consider the latter problem here and essentially extend the collapse postulate by describing continuous evolution `` inside '' the collapse timescale . in the cqed setup ( fig . 1 ) a qubit interacts with a ghz - range microwave resonator , whose frequency slightly changes depending on whether the qubit is in the state @xmath0 or @xmath1 @xcite . in turn , this frequency shift affects the phase ( and in general amplitude ) of a probing microwave , which is transmitted through the resonator ( in another setup the microwave is reflected from the resonator , but the difference is not important ) . the outgoing microwave is amplified , and after that the rf signal is downconverted by mixing it with the original microwave tone , so that the low - frequency ( @xmath2 mhz ) output of the iq mixer provides information on the qubit state . the output noise is mainly determined by the first amplifying stage , the pre - amplifier . with recent development of nearly quantum - limited superconducting parametric amplifiers @xcite , it is natural to use them as pre - amplifiers @xcite instead of cryogenic high - electron - mobility transistors ( hemts ) @xcite , which usually have noise temperature above 3 k. is transmitted through ( or reflected from ) the resonator of frequency @xmath3 , which slightly changes , @xmath4 , depending on the qubit state . after amplification the microwave is sent to the iq mixer , which produces two quadrature signals : @xmath5 and @xmath6 . for a phase - preserving amplifier we define @xmath5 as the quadrature carrying information on the qubit state , while for a phase - sensitive amplifier we define @xmath5 as corresponding to the amplified quadrature.,width=321 ] continuous quantum measurement in the cqed setup in some sense falls in between a qubit measurement by a quantum point contact ( qpc ) or a single - electron transistor ( set ) , the theory of which has been developed over a decade ago @xcite , and continuous quantum measurement in optics , for which the theory of quantum trajectories has been developed even earlier @xcite . nevertheless , the cqed setup differs from both these cases , and this is probably the reason why there is still a confusion about the proper physical description of the measurement process . the measurement by the qpc or set is of the broadband type , meaning that the monitored frequency band starts from zero . in contrast , the cqed setup is of the narrowband type : we deal only with a relatively narrow band around the probing microwave frequency @xmath7 . this necessarily involves two orthogonal quadratures @xcite : we work with rf signals of the type @xmath8 , and there are essentially two signals @xmath9 and @xmath10 instead of only one in the broadband case . in this sense the cqed setup is similar to the optical ( especially cavity qed ) setup @xcite ; however , there is an important difference : in the cqed case the outgoing microwave is amplified ( fig . 1 ) before being mixed with the original microwave , while there is no amplification stage in the standard optical setup . the operation will obviously depend on whether a phase - sensitive or a phase - preserving amplifier is used , since a phase - preserving amplifier necessarily adds the half quantum of noise into any quadrature @xcite . notice that the quantum trajectory theory for the cqed setup was developed in @xcite ; however , the amplifier stage was essentially missing in the analyzed model . in this lecture we consider the simplest cqed case , assuming dispersive regime @xcite , exactly resonant microwave frequency , absence of the rabi drive , and sufficiently wide resonator and amplifier bandwidths for the markov approximation . some generalizations are rather straightforward ; however , our goal is a simple picture in a simple case . a description of continuous qubit measurement is essentially a description of the quantum back - action . following the same quantum bayesian framework as for the measurement by qpc / set @xcite ( see @xcite for review ) , we will discuss two kinds of measurement back - action onto the qubit , which we name here as `` spooky '' and `` realistic '' . _ the `` spooky '' ( or `` quantum '' , `` informational '' , `` non - unitary '' ) back - action does not have a physical mechanism _ and therefore can not be described by the schrdinger equation ( in contrast to what people often think , trying to find a mechanism for the quantum collapse ) ; however , _ it is a common - sense consequence of acquiring information _ on the qubit state in the process of measurement . this is essentially the same back - action which is discussed in the epr paradox @xcite and bell inequality violation @xcite ; the only difference is that in our case the information is incomplete and therefore we have to use the quantum bayes rule @xcite instead of the projective collapse rule . in contrast , the `` realistic '' ( or `` classical '' , `` unitary '' ) back - action has a physical mechanism : in the cqed case it is a fluctuation of the number of photons in the resonator , which affects the phase of the qubit state . the `` realistic '' back - action is usually discussed in the standard theories of the cqed measurement @xcite . actually , there is a certain spookiness even in the `` realistic '' back - action ( it may be affected by a delayed choice , as discussed in conclusion ) ; however , we do not want to emphasize it to keep the picture simple . when we measure the @xmath11-coordinate of the qubit state on the bloch sphere ( the basis states @xmath1 and @xmath0 correspond to the north and south poles ) , then the `` spooky '' back - action changes the @xmath11-coordinate and leads to the state evolution along the meridian lines , while the `` realistic '' back - action leads to the evolution around the @xmath11-axis , i.e. along the parallels . it is important to notice that when the probing microwave leaves the resonator after interaction with the qubit , one quadrature of the microwave carries information about the qubit state , while the orthogonal quadrature carries information on the fluctuating number of photons in the resonator @xcite . therefore , if a phase - preserving amplifier is used , then the `` spooky '' and `` realistic '' back - actions are fully separated and correspond to two orthogonal quadratures @xmath5 and @xmath6 measured after the mixer ( it is trivial to choose the proper linear combinations of the i / q mixer outputs ) . the signals @xmath5 and @xmath6 are necessarily noisy , and the measurement back - actions are stochastic ; however , there is a correlation ( full correlation in the ideal case ) between the output noise and the back - action noise in both channels . as a result ( derived later ) , for a quantum - limited phase - preserving amplifier and absence of extra decoherence , the measured quadratures @xmath5 and @xmath6 give us _ full information about the back - action _ , so that a _ random evolution of the qubit wavefunction can be monitored precisely _ ( a useful analogy is with a brownian particle under a microscope : we can not predict its motion , but we can monitor it ) . this is what is needed , in particular , for arranging a perfect quantum feedback control of the qubit state . it is interesting to notice that for an ensemble - averaged evolution ( in which the random but monitorable qubit evolution is replaced by dephasing ) , exactly one half of the ensemble dephasing @xmath12 comes from the `` spooky '' back - action , and the other half comes from the `` realistic '' back - action . in the case of a phase - sensitive amplifier it is sufficient to measure after the mixer only the quadrature which was amplified ; let us still denote it as @xmath5 , though now its phase is determined by the amplifier instead of the microwave - qubit interaction . in this case the `` spooky '' and `` realistic '' back - actions are in general mixed ( not separated ) , because there is only one output signal @xmath5 . this situation exactly corresponds to the broadband measurement by the qpc / set with a correlation between the output and `` realistic '' back - action noises @xcite . the situation simplifies when the amplified quadrature is the one which carries information about the qubit state ( @xmath11-coordinate ) . then in the quantum - limited case the `` realistic '' back - action is fully absent : we can not measure the photon number fluctuation and correspondingly it does not fluctuate ( in the imperfect case the effect of the remaining `` realistic '' back - action can be described by an extra dephasing ) . so we are left with only the `` spooky '' back - action , and the quantum measurement description coincides with the simpler theory of measurement by a symmetric qpc @xcite , which does not produce the `` realistic '' back - action . in contrast , in the case when the photon - number quadrature is amplified , we do not obtain any information on the qubit @xmath11-coordinate , and therefore there is no `` spooky '' back - action , but only the `` realistic '' one . in a general case , when the amplified quadrature makes an arbitrary angle @xmath13 with the qubit - information quadrature , both types of the back - action are present , and their strength depends on @xmath13 . it is important to mention that the ensemble dephasing rate @xmath12 does not depend on @xmath13 , as required by causality . in particular , in the quantum - limited case the contribution @xmath14 comes from the `` spooky '' back - action , while @xmath15 comes from the `` realistic '' back - action . let us emphasize that both the phase - sensitive and phase - preserving amplifiers permit exact monitoring of the qubit state and therefore a perfect quantum feedback . the necessary condition in both cases is that the detection system is quantum - limited . in the following sections a formal description of the above discussed results is presented . we start with reviewing the bayesian approach for the broadband qubit measurement , then briefly discuss the difference between phase - preserving and phase - sensitive amplifiers , and then present the formalism of the narrowband continuous measurement of a qubit in the cqed setup . in conclusion we briefly discuss generalizations of the formalism , quantum feedback , and the causality principle . we note that our approach can be converted into the formal language of the positive - operator - valued - measure(povm)-type generalized quantum measurement @xcite ( then separation of the `` spooky '' and `` realistic '' back - actions corresponds to the decomposition of the measurement operator into diagonal and unitary parts see later ) , and our results for the case of a phase - sensitive amplifier are very similar to the results of ref . in this section we review the bayesian formalism @xcite for the broadband measurement of a qubit , considering only the simple case without additional evolution , and thus emphasizing the main physical idea of the formalism . we start with the broadband formalism because it is simpler than for the narrowband ( cqed ) measurement and it can be used as a natural step in understanding the cqed setup . for definiteness let us assume that the qubit is a double quantum dot populated with one electron ( fig . 2 ) , and the states @xmath0 and @xmath1 correspond to the electron localized in one or the other dot . the qubit is measured by a small - transparency tunnel junction ( model of qpc ) , whose barrier height depends on the electron location , so that the two qubit states correspond to different average currents @xmath16 and @xmath17 through the qpc . the voltage across the qpc is sufficiently large to make the detector output classical ( markov approximation ) , and @xmath18 , where @xmath19 is the response and @xmath20 is the mean value ; this weak response assumption allows us to consider the qpc current @xmath5 as a quasicontinuous noisy signal ( see @xcite for the detailed discussion of required assumptions ; the formalism needs only a minor change if @xmath21 ) . then the output signal of the detector is i(t)=i_c + ( i/2 ) z(t ) + ( t ) , s _ ( ) = s , where @xmath22 is the @xmath11-component of the bloch sphere representation of the qubit density matrix @xmath23 , and @xmath24 is the white shot noise with spectral density @xmath25 ( we use the single - sided definition for the spectral density , in which the signal variance ( `` power '' ) corresponds to @xmath26 ; the definition of @xmath27 is twice smaller in ref . @xcite and @xmath28 times smaller in refs.@xcite ) . we emphasize that the detector signal @xmath5 is classical , and the qubit state @xmath23 is practically unentangled from the detector , but obviously depends on @xmath5 . is the qpc current . , width=188 ] the detector hamiltonian and the qubit - detector interaction hamiltonian are given in refs . @xcite , they are not really important for our discussion here . for simplicity let us assume that the qubit hamiltonian is zero , @xmath29 , so that the qubit evolution is due to the measurement only . in this case the qubit evolution during time @xmath30 happens to be determined only by the time - averaged value of the measured detector output _ m ( t ) = _ 0^t i(t ) dt , which would contain full information for a classical measurement . because of the correspondence principle , the evolution of the diagonal elements of the qubit density matrix @xmath31 ( @xmath32 ) should correspond to the classical evolution of probabilities , which are given by the classical bayes rule . the bayes rule says that an updated ( a posteriori ) probability of a system state is proportional to the initial ( a priori ) probability and the probability ( likelihood ) of the obtained measurement result assuming this particular state . in our case @xmath33 is the measurement result , and its probability for the qubit in the basis state @xmath34 has the gaussian distribution p_|j(_m)= , d= , [ gaussian ] where @xmath35 is the variance , which decreases with the measurement time @xmath30 . therefore the correspondence principle demands the bayesian evolution = , [ rho - diag]which in our terminology is due to the `` spooky '' back - action ; it can not be described by the schrdinger equation , but follows from common sense . if the phase of qubit state is not affected by the measurement process ( no `` realistic '' back - action ) , then an arbitrary initial wavefunction @xmath36 ; therefore for an arbitrary mixed state we get _ 01(t)= _ 01(0 ) . [ rho - off]equations ( [ rho - diag ] ) and ( [ rho - off ] ) describe the `` spooky '' back - action . now assume that due to the qubit - detector interaction ( e.g.coulomb interaction ) , each electron passing through the detector rotates the qubit phase @xmath37 by a small amount @xmath38 . from the measured result @xmath33 we know exactly how many electrons passed through [ @xmath39 with @xmath40 being the electron charge ] , and can easily introduce the corresponding phase factor into eq . ( [ rho - off ] ) : _ 01(t)= _ 01(0 ) , [ rho - off-2]where @xmath41 . the non - stochastic factor @xmath42 can be obviously ascribed to the qubit hamiltonian ; however , this is not important here . the factor @xmath43 $ ] in eq.([rho - off-2 ] ) is the effect of the `` realistic '' back - action . it may or may not be present in a particular physical situation ; for example , @xmath44 for measurement by a symmetric qpc , while @xmath45 in an asymmetric qpc or set case . finally , if there is an extra pure dephasing of a qubit with rate @xmath46 , then eq . ( [ rho - off-2 ] ) becomes _ 01(t)= _ 01(0 ) e^-t . [ rho - off-3]equations ( [ rho - diag ] ) and ( [ rho - off-3 ] ) is _ the main starting point of the bayesian formalism _ it is then easy to include non - zero qubit hamiltonian @xmath47 by differentiating eqs . ( [ rho - diag ] ) and ( [ rho - off-3 ] ) over time ( paying attention to whether the stratonovich or it definition of the derivative is used ) and adding terms due to @xmath47 . energy relaxation and other mechanisms of the qubit state evolution can be included in the same way . actually , there are many ways to derive the bayesian equations ( [ rho - diag ] ) and ( [ rho - off-3 ] ) @xcite , but we focus here only on their meaning , not on their derivation . notice that averaging of eqs . ( [ rho - diag ] ) and ( [ rho - off-3 ] ) over the measurement result @xmath48 ( i.e. ensemble averaging ) with the probability distribution p(_m)=_00(0)p_|0(_m)+ _ 11(0)p_|1(_m ) [ p(i ) ] gives the same evolution as for a pure dephasing : the diagonal matrix elements of @xmath31 do not evolve , while the off - diagonal element @xmath49 decays as @xmath50 with the ensemble dephasing rate @xcite = + + , [ gamma]which has clear contributions from the `` spooky '' back - action , `` realistic '' back - action , and additional dephasing . in the case @xmath51 an initially pure qubit state remains pure ; in other words we can monitor evolution of a qubit wavefunction . this property can be used as the definition of a quantum - limited detector @xcite . the quantum efficiency @xmath52 can then be naturally defined as @xmath53 . if by some reason the `` realistic '' back - action is considered as dephasing ( i.e. only in the averaged way ) , then the quantum efficiency can be defined as @xmath54 ( here the definitions of @xmath55 and @xmath52 are exchanged compared with the definitions in @xcite ) . in other words , @xmath56 is the relative contribution of only the `` spooky '' back - action in the ensemble dephasing @xmath12 . in particular , this definition is relevant to the peak - to - pedestal ratio of the rabi spectral peak @xcite , which is equal to @xmath57 . as an example , if @xmath51 and contributions in eq . ( [ gamma ] ) from the `` spooky '' and `` realistic '' back - actions are equal to each other ( as in the cqed setup with a phase - preserving amplifier ) , then @xmath58 but @xmath59 . a non - ideal detector ( @xmath60 ) can be modeled in two equivalent ways @xcite : we either add an extra dephasing @xmath46 to the qubit or we add an extra noise to the output of the ideal detector . only the total dephasing @xmath12 , response @xmath61 , total output noise @xmath27 , and correlation factor @xmath62 are the physical ( i.e. experimentally measurable ) parameters , while distribution of the non - ideality between the extra dephasing and additional output noise is a matter of convenience [ here @xmath63 , and notation @xmath64 reminds about averaging over additional classical noise at the output ] . we emphasize that the bayesian formalism deals only with the experimentally measurable parameters @xmath61 , @xmath27 , @xmath65 , @xmath12 , and the output signal @xmath5 . in the ideal case ( @xmath58 ) the evolution equations ( [ rho - diag ] ) and ( [ rho - off-2 ] ) can be translated into the language of povm - type generalized measurement . in this approach the effect of measurement is described as @xcite |(t)= , ( t)= , where @xmath66 is the so - called measurement ( kraus ) operator , corresponding to the result @xmath67 . the probability of the result @xmath67 is @xmath68 using wavefunctions or @xmath69 $ ] using density matrices ; therefore the povm elements @xmath70 should satisfy the completeness condition @xmath71 . the relation between this approach and the quantum bayesian approach can be understood via the operator decomposition m_r = u_r , where @xmath72 is unitary and the square root of the positive operator @xmath70 is defined in the natural way in the diagonalizing basis . it is easy to see that @xmath73 is essentially the quantum bayes rule ( in the diagonalizing basis ) ; in our terminology it corresponds to the `` spooky '' back - action , while @xmath72 corresponds to the `` realistic '' back - action . for the discussed setup the result @xmath67 is @xmath33 , the `` spooky '' back - action @xmath74^{1/2}$ ] should be determined by the probabilities @xmath75 given by eq . ( [ gaussian ] ) , and the `` realistic '' back - action @xmath72 is given by the phase factor in eq . ( [ rho - off-2 ] ) . therefore the corresponding measurement operator is @xmath76 where @xmath77 is the pauli matrix . before discussing microwave amplifiers , let us consider a measurement of an oscillator , for example , a mechanical resonator with frequency @xmath3 and mass @xmath78 . this is a very well studied problem @xcite , so we will only discuss a way to understand the results . a classical resonator position @xmath79 oscillates as @xmath80 ( in this section @xmath79 stands for the usual spatial coordinate , not for the bloch sphere coordinate ) . the corresponding quantum state is called the `` coherent state '' in the optical language ; it is represented by the wavefunction @xmath81\exp(i p_c x/\hbar)$ ] , where @xmath82 is the ground state and @xmath83 is the classical momentum . so the coherent state is essentially the ground state with oscillating center position . notice that continuous quantum measurement of a resonator position can be described in the same bayesian way @xcite as in the previous section ; for example the `` spooky '' back - action gives the evolution @xmath84/{\rm norm}$ ] , where @xmath85 is the average detector signal for the resonator position @xmath79 , and @xmath86 is normalization [ see eqs.([rho - diag ] ) and ( [ rho - off ] ) ] . the time step @xmath30 in this case should be chosen much shorter than @xmath87 so that the unitary evolution and evolution due to measurement may be simply added . let us consider the following game : charlie prepares an oscillator in a coherent state with quadratures @xmath88 and @xmath89 , gives it to david , and david s goal is to find @xmath88 as accurately as possible . an optimal strategy is rather obvious : david should make a projective measurement of @xmath79 at time @xmath90 with any integer @xmath91 ( to avoid contribution from the @xmath89-term ) , and the measurement result is the best estimate of @xmath88 [ if the measurement is done at @xmath92 , then the result should be multiplied by @xmath93 . even though the strategy is optimal , the inaccuracy of david s result is obviously the width ( standard deviation ) @xmath94 of the ground state shape @xmath95 ; in energy units this inaccuracy corresponds to one half of the energy quantum . now assume that david can not make projective measurements , but only `` finite - strength '' ( i.e. imprecise ) measurements . the best accuracy @xmath96 can still be achieved if the measurement is done in the simple but very clever `` quantum non - demolition '' ( qnd ) way : many finite - strength measurements are made at times @xmath97 ; this is called `` stroboscopic '' measurement @xcite . since the oscillator returns to the same state after the period @xmath98 , the unitary evolution is not important , and many finite - strength measurements ( described by the bayesian equation above ) are `` stacked '' to produce a strong , essentially projective measurement . more generally , the necessary condition to have the best accuracy @xmath96 for @xmath88 is that the measurement is not sensitive to the quadrature @xmath89 . now assume that david is only allowed to make a continuous measurement with unmodulated weak strength ( so that the inaccuracy achieved after @xmath87 is much larger than @xmath96 ) . then the `` spooky '' back - action gets mixed with the unitary evolution , essentially adding noise into the monitored evolution , so that after a while the resonator state becomes mostly determined by the back - action and almost not dependent on the initial state . as the result , the best accuracy for measurement of @xmath88 becomes @xmath99 , which in energy units corresponds to two half - quanta @xcite , that is twice worse than for the projective or stroboscopic measurement . however , the continuous monitoring gives us an information about @xmath89 in the same way as for @xmath88 , so the accuracy of @xmath89-measurement is also @xmath100 . therefore , in some sense continuous phase - insensitive measurement brings the same total information as the phase - sensitive ( e.g. stroboscopic ) measurement ; however , in our game only half of this information is useful for the david s goal . after discussing measurement of the resonator state it is easy to understand the quantum limits for the high - gain microwave amplifiers . now suppose charlie prepares a coherent state of a microwave resonator with quadratures @xmath88 and @xmath89 , gives it to david to find @xmath88 , and david uses an amplifier for amplification of the microwave field , which slowly leaks from the resonator until it is empty . there is only classical signal processing after the amplifier , so amplification is essentially the quantum measurement . the results are the same as above @xcite : a phase - sensitive amplifier , which amplifies only @xmath88-quadrature and `` de - amplifies '' ( attenuates ) @xmath89-quadrature can measure @xmath88 with accuracy @xmath96 , while a phase - preserving amplifier can measure @xmath88 only with accuracy @xmath99 ( and also measures @xmath89 with the same accuracy @xmath99 ) . technically , the accuracy is limited by the noise at the amplifier output , so this noise should forbid measuring of @xmath88 with accuracy better than @xmath96 by a phase - sensitive amplifier , and better than @xmath99 by a phase - preserving amplifier . therefore in the quantum - limited case the output noise power of a phase - preserving amplifier ( per quadrature ) is twice larger than for a phase - sensitive amplifier with the same gain ; this is often called an `` additional noise '' , corresponding to one half of the energy quantum ( one more half - quantum is present in both cases ) @xcite . it may be somewhat confusing why this result does not depend on the rate with which the microwave leaks from the resonator . so let us check the scaling : for @xmath101 times slower leakage ( @xmath101 times larger @xmath102-factor ) the microwave amplitude is @xmath103 times smaller , but accumulation time is @xmath101 times longer , therefore the measured signal for the quadrature @xmath88 is @xmath103 times larger , which is the same factor as for the noise accumulation . therefore , the signal - to - noise ratio which determines the @xmath88-accuracy does not depend on the leakage rate ( resonator bandwidth ) . using the above discussion of microwave amplifiers , it is easy to extend the bayesian approach for a broadband quantum measurement to the narrowband cqed setup . we consider the standard cqed setup @xcite , in which a qubit interacts with a microwave resonator , and assume the dispersive regime with the hamiltonian @xmath104 where @xmath105 is the lamb - shifted qubit frequency with no photons in the resonator , @xmath106 is the effective coupling with @xmath107 being the jaynes - cummings coupling , @xmath3 is the bare resonator frequency , pauli operator @xmath77 acts on the qubit state in the energy basis @xmath108 , and the resonator creation / annihilation operators are @xmath109 and @xmath110 . notice that the resonator frequency increases by @xmath111 when the qubit state changes from @xmath0 to @xmath1 ; conversely , the qubit frequency increases by @xmath111 per each additional photon in the resonator . to measure the qubit state , a microwave field with frequency @xmath7 is either transmitted through or reflected from the resonator , then amplified and sent to the iq mixer , which measures both quadratures relative to the original microwave tone ( fig . 1 ) . the qubit state affects the resonator frequency and therefore affects the phase ( and in general amplitude ) of the outgoing microwave . an elementary fabry - prot analysis gives the classical ( complex ) microwave field @xmath112 inside the resonator : f_r = , [ fabry - perot]where @xmath113 is the applied incident field , @xmath114 is the transmission amplitude of the barrier from the incident side , @xmath115 is the resonator bandwidth due to the microwave leakage from the both sides ( the @xmath102-factor is @xmath116 ) , and the round - trip time is @xmath117 for a half - wavelength resonator and @xmath118 for a quarter - wavelength resonator . a similar formula with the same denominator describes a lumped resonator . in presence of the qubit , the resonator frequency @xmath3 in this formula is substituted by @xmath119 , depending on the qubit state . notice that for the quantum measurement analysis _ there is no difference between the cases of transmission and reflection _ for the same @xmath112 and @xmath115 , because the field leaking from the resonator is determined only by @xmath112 and @xmath115 . ( the reflection case has a technical advantage of dealing with a twice smaller outgoing microwave field for the same measured signal . ) however , an important parameter is the collected fraction @xmath120 of the leaking microwave power ; we will often assume the ideal case @xmath121 ( for the transmission setup this requires strongly asymmetric coupling , @xmath122 ) . for simplicity we assume the resonant case , @xmath123 , then the ensemble qubit dephasing due to measurement is @xcite = 8 ^ 2|n/ , [ gamma - cqed]where @xmath124 is the average number of photons in the resonator . it is easy to include rabi oscillations into the model ; however , we do not do it for simplicity and also for more transparent analogy with sec . [ broadband ] , in which we considered a qubit with zero hamiltonian , evolving only due to measurement ; this case exactly corresponds to the cqed hamiltonian ( [ h - dispersive ] ) in the rotating frame . we will need several assumptions to describe the qubit state evolution in the process of measurement . first , for the validity of the dispersive approximation ( [ h - dispersive ] ) we need sufficiently large qubit - resonator detuning , @xmath125 , and not too many photons in the resonator , @xmath126 ( we do not consider the recently discovered nonlinear regime @xcite ) . second , to use the markov approximation for the evolution we need the so - called `` bad cavity '' assumption : @xmath127 ( if the qubit evolves due to rabi oscillations with frequency @xmath128 , we also need @xmath129 ) . this assumption means that the photons leave the resonator much faster than evolution of the qubit state , and therefore there is practically no entanglement between the qubit and unmeasured microwave field . this assumption also implies that the two resonator states for the qubit states @xmath0 and @xmath1 are almost indistinguishable , @xmath130 . third , we use the `` weak response '' assumption , which requires a small phase difference between the two resonator states , @xmath131 . this means that each outgoing photon carries only a little information about the qubit state . notice that for @xmath132 the previous assumption @xmath133 automatically implies the weak response , and even for @xmath134 the weak response assumption is not always needed . fourth , we will neglect the qubit energy relaxation due to measurement @xcite , which can be added later . a coherent state in the resonator with average @xmath124 photons and zero average phase corresponds to the oscillation of the field expectation value @xmath135 , where @xmath96 is the ground state width ( rms uncertainty ) and we assume @xmath123 . ( notice that the amplitude @xmath96 corresponds to 1/4 photon . ) interaction with the qubit slightly changes the phase , @xmath136 , depending on the qubit state , so that @xmath137 where @xmath11 is the qubit bloch coordinate . thus the small @xmath89-quadrature carries information about the qubit state , while larger @xmath88-quadrature may give us information on the fluctuations of the photon number in the resonator . in the optical representation ( fig . 3 ) with axes @xmath138 and @xmath139 , the two resonator states for the qubit states @xmath0 and @xmath1 are shown as two `` error circles '' @xcite with rms uncertainty 1 along any direction and distance @xmath140 between the origin and circle centers ( if axes @xmath141 and @xmath142 are used , then the distance is @xmath143 , while the uncertainty is 1/2 ) . in the resonator is shown @xcite as an `` error circle '' with radius 1 , shifted by @xmath140 from the origin . axes are normalized by the standard deviation @xmath96 of the ground state . the @xmath89-quadrature carries information about the qubit state , the @xmath88-quadrature corresponds to the number of photons in the resonator.,width=188 ] let us start with the case when a phase - sensitive amplifier is used in the cqed setup . also , we first assume the most ideal case : the amplifier is quantum - limited , it amplifies the optimal @xmath89-quadrature , there is no microwave collection loss ( @xmath144 ) , and there is no extra noise or dephasing . then , as discussed in the previous section , measuring once the microwave contents of the resonator ( by fully emptying it ) we can measure the @xmath89-quadrature with imprecision @xmath96 . therefore , in the continuous measurement for time @xmath30 the @xmath89-quadrature is measured with imprecision @xmath145 , which converts into the imprecision @xmath146 of the qubit @xmath11-coordinate . following the language of sec . [ broadband ] , let us discuss the signal and noise at the output of the setup . there are two outputs of the iq mixer ; however , only the amplified quadrature carries an information , so let us denote the corresponding output of the mixer ( or their linear combination ) as @xmath5 . then the response @xmath147 corresponds to @xmath148 and @xmath149 . for measurement during time @xmath30 the above variance @xmath150 of the @xmath11-coordinate converts into the variance @xmath151 of the measured output @xmath152 . equating it with @xmath153 , we find the ( single - sided ) spectral density of the @xmath5 noise : s_min=(i_max)^2/(32 ^2 |n ) , [ s]where we replaced @xmath27 with @xmath154 and @xmath61 with @xmath155 to remind that we consider the quantum - limited case , and the response is maximized by amplifying the optimal quadrature . notice that since @xmath156 , the noise @xmath154 does not depend on the qubit or resonator properties ; it is essentially the amplified vacuum noise and depends only on the amplifier gain . and @xmath89 are for the microwave field in the resonator . ( a ) for a phase - sensitive amplifier the amplified quadrature makes an angle @xmath13 with the informational @xmath89-quadrature . the corresponding quadrature at the mixer output is defined as @xmath5 [ the output @xmath6 is then useless ] . ( b ) for a phase - preserving amplifier we define output quadratures @xmath5 and @xmath6 as corresponding to the resonator quadratures @xmath89 and @xmath88.,width=264 ] obtaining information on the qubit @xmath11-coordinate via the signal @xmath5 with response @xmath157 and noise @xmath154 , we necessarily cause the `` spooky '' back - action described by eqs . ( [ rho - diag ] ) and ( [ rho - off ] ) . as discussed in sec . [ broadband ] , this is the consequence of the corresponding principle or just the common sense . now averaging the @xmath49 evolution in eq.([rho - diag ] ) over the measurement result @xmath158 with its probability distribution ( [ p(i ) ] ) and ( [ gaussian ] ) , we see that the `` spooky '' back - action dephases an ensemble of qubits with the rate [ see eq . ( [ gamma ] ) ] @xmath159 . this rate coincides with the total ensemble dephasing ( [ gamma - cqed ] ) , and therefore the qubit state can not additionally fluctuate due to any other reason . thus we derived an important result : _ in the ideal case with phase - sensitive amplifier there is only the `` spooky '' back - action and no `` realistic '' back - action_. this means that the _ number of photons in the resonator does not fluctuate _ ( otherwise there would be an additional dephasing ) , which makes sense since we can not measure the @xmath88-quadrature , carrying information on the photon number . notice that it is also easy to prove this result when @xmath160 . then from eq . ( [ fabry - perot ] ) we obtain that the informational quadrature amplitude is multiplied by the factor @xmath161^{-1/2}$ ] compared with eq . ( [ a - b - quadratures ] ) . the response @xmath157 is multiplied by same factor , while the noise @xmath154 does not change . therefore , the `` spooky '' back - action contribution into the ensemble dephasing is multiplied by the factor @xmath161^{-1}$ ] , which again coincides with the result @xcite for the total ensemble dephasing @xmath12 . this proves the absence of the `` realistic '' back - action for the non - resonant case @xmath160 as well . now let us consider the case when an ideal phase - sensitive amplifier amplifies the @xmath88-quadrature ( we again assume @xmath123 for simplicity ) . then we do not get any information on the qubit @xmath11-coordinate , and therefore there is no `` spooky '' back - action , but there is the `` realistic '' back - action due to fluctuating number of photons . the description of evolution in this case is essentially the standard description @xcite . let us still denote with @xmath5 the output signal from the mixer , corresponding to the amplified quadrature . for measurement during time @xmath30 we measure @xmath88-quadrature with imprecision @xmath145 . this is consistent with fluctuation of the number @xmath162 of emitted photons : @xmath163 , @xmath164 . the correlation function of photon number in the resonator depends on time as @xmath165 @xcite , which means that each extra photon inferred from @xmath5 fluctuation spends ( on average ) time @xmath166 in the resonator and therefore changes the qubit phase @xmath37 by @xmath167 ( the correlation time @xmath166 is essentially the lifetime of the field , not power @xcite ) . then @xmath37-variance is @xmath168 , and the corresponding ensemble dephasing is @xmath169 . as expected , this reproduces the standard result ( [ gamma - cqed ] ) for the ensemble dephasing , while for individual qubit evolution we have the above discussed correlation : each additional photon inferred from @xmath5 fluctuation changes @xmath37 by @xmath167 . for the same amplifier gain and noise as for measuring @xmath89-quadrature , we get @xmath170 , and therefore the correlation is @xmath171 . it is easy to check that @xmath172 [ see eq . ( [ gamma ] ) ] coincides with ensemble dephasing @xmath12 from eq.([gamma - cqed ] ) , as expected for the presence of only `` realistic '' back - action . finally , assume that the phase - sensitive amplifier amplifies the quadrature , which makes angle @xmath13 with the optimal @xmath89-quadrature and angle @xmath173 with the @xmath88-quadrature . the measured signal @xmath5 still denotes the output of the iq mixer , corresponding to the amplified quadrature ; now it gives information about both @xmath89 and @xmath88 quadratures , with the factors @xmath174 and @xmath175 , respectively . combining the `` spooky '' and `` realistic '' back - actions , we get the same formulas as for the broadband detection of sec . [ broadband ] : @xmath176 , \,\,\ , d=\frac{s}{2 t } , \qquad \label{ph - sens - diag}\\ & & \frac{\rho_{01}(t)}{\rho_{01}(0)}= \frac{\sqrt{\rho_{00}(t)\ , \rho_{11}(t ) } } { \sqrt{\rho_{00}(0)\ , \rho_{11}(0)}}\ , \exp [ i k \tilde{i}_m(t)\ , t ] , \label{ph - sens - off}\\ & & \delta i= i_1-i_0= \delta i_{max}\cos\varphi , \,\,\ , k = k_{max } \sin\varphi , \qquad \label{ph - sens - deltai}\\ & & k_{max } = \delta i_{max}/s , \label{ph - sens - kmax}\\ & & \tilde i_m(t ) = \frac{1}{t } \int_0^\infty i(t')\ , dt ' - \frac{i_0+i_1}{2 } , \label{ph - sens - im}\\ & & i(t)=\frac{i_0+i_1}{2 } + \frac{\delta i}{2 } \ , z ( t ) + \xi(t ) , \,\,\ , s_\xi(\omega ) = s . \label{ph - sens - i(t)}\end{aligned}\ ] ] here we introduced @xmath177 by subtracting the constant @xmath178 from @xmath48 and did a simple algebra to convert eq . ( [ rho - diag ] ) into eq . ( [ ph - sens - diag ] ) ; the qubit rotating frame corresponds to @xmath124 photons , @xmath179 is the above discussed correlation for @xmath88-quadrature amplification , @xmath157 is the response for @xmath89-quadrature amplification , and @xmath180 . notice that the total ensemble dephasing ( [ gamma ] ) does not depend on @xmath13 : ( i_max)^2/4s_min+ ( k_max)^2 s_min/4= . so far we discussed only the ideal case . there are several mechanisms for non - ideality . first , the qubit may have additional environmental dephasing @xmath181 . this will lead to the extra factor @xmath182 in eq . ( [ ph - sens - off ] ) and increase the ensemble dephasing @xmath12 by @xmath181 . following the definitions in sec . [ broadband ] , the corresponding quantum efficiency is @xmath183 . second , not all microwave power leaking from the resonator may be collected and amplified . this can be characterized by the collection efficiency @xmath120 and multiplies the response @xmath184 and correlation @xmath65 by the factor @xmath185 , while not affecting the output noise @xmath27 . third , if the phase - preserving amplifier is not quantum - limited , it introduces additional noise @xmath186 compared with the quantum limit @xmath154 [ given by eq.([s ] ) when @xmath121 ] . the corresponding amplifier efficiency is @xmath188 . this does not affect @xmath61 but multiplies @xmath65 by @xmath189 ( because for uncorrelated gaussian - distributed random numbers @xmath190 and @xmath191 , the averaging of @xmath190 for a fixed sum @xmath192 gives the correlation @xmath193 $ ] ) . if all three mechanisms of the non - ideality are present , then the evolution can still be described @xcite by eqs.([ph - sens - diag])([ph - sens - i(t ) ] ) , but @xmath27 is now the total ( experimental ) output noise , @xmath157 is the experimental response for @xmath194 , so that @xmath195 , the correlation @xmath196 is still given by eqs . ( [ ph - sens - deltai])([ph - sens - kmax ] ) , and the only change is the extra factor in eq . ( [ ph - sens - off ] ) : @xmath197 \ , e^{-\gamma t } , \qquad \label{ph - sens - off-2}\\ & & \gamma = \gamma -(\delta i_{max})^2/4s , \label{ph - sens - gamma}\end{aligned}\ ] ] where the ensemble dephasing is now @xmath198 . we emphasize that the qubit evolution depends only on the experimentally measurable parameters @xmath157 , @xmath27 , @xmath12 , @xmath13 , and the output signal @xmath5 . the quantum efficiencies can be expressed as = = _ amp_col_env , = ^2 , where as in sec . [ broadband ] , @xmath52 is the relative contribution to @xmath12 from both the `` spooky '' and `` realistic '' back - actions , while @xmath55 is the relative contribution from only the `` spooky '' back - action . the definition of @xmath55 corresponds to replacing the `` realistic '' back - action factor @xmath199 in eq . ( [ ph - sens - off-2 ] ) with the corresponding ensemble dephasing @xmath200 . as mentioned in sec . [ broadband ] , the peak - to - pedestal ratio of the spectral peak of contiuous rabi oscillations is @xmath201 . now assume that a phase - preserving amplifier is used ( this includes parametric amplifier , hemt , etc . ) . now both the @xmath88-quadrature and @xmath89-quadrature of eq . ( [ a - b - quadratures ] ) are amplified independently with the same gain . correspondingly , both quadratures at the iq mixer output carry physical information instead of only one quadrature in the case of a phase - sensitive amplifier . let us denote with @xmath5 the output of the iq mixer , corresponding to the @xmath89-quadrature ; thus @xmath5 provides information on the qubit @xmath11-coordinate . the output signal for the orthogonal quadrature is denoted @xmath6 ; it corresponds to the @xmath88-quadrature in the resonator and provides information on the fluctuating number of photons . the main difference from the case of a phase - sensitive amplifier is that now the `` spooky '' and `` realistic '' back - actions are related to two different output signals : @xmath5 and @xmath6 . let us start with the quantum - limited case and assume an amplifier with the same gain as in the phase - sensitive case , so that the @xmath5-channel response is the same as the optimal phase - sensitive response , @xmath202 . the `` spooky '' back - action is always described by the quantum bayes formulas ( [ rho - diag])([rho - off ] ) , but now the noise @xmath27 of the output @xmath5 is twice larger than the value ( [ s ] ) for the phase - sensitive amplifier , @xmath203 ( see discussion in sec.[amplifiers ] ) ; therefore the `` spooky '' evolution is twice slower than in the phase - sensitive case with @xmath194 . the signal @xmath6 has the same noise @xmath203 ; it is again twice larger than for the phase - sensitive case with @xmath204 ; therefore the correlation factor @xmath205 $ ] for the `` realistic '' back - action is twice smaller : @xmath206 ( this reduction is similar to the effect of a non - ideal amplifier discussed above ) . we see that @xmath207 , and the ensemble dephasing is at least @xmath208 . this again coincides with @xmath209 , and therefore there can be no additional evolution of the qubit besides these `` spooky '' and `` realistic '' back - actions . thus in the ideal case the qubit evolution is @xmath176 , \,\,\ , d=\frac{s}{2 t } , \qquad \label{ph - pres - diag}\\ & & \frac{\rho_{01}(t)}{\rho_{01}(0)}= \frac{\sqrt{\rho_{00}(t)\ , \rho_{11}(t ) } } { \sqrt{\rho_{00}(0)\ , \rho_{11}(0)}}\ , \exp [ i k \tilde{q}_m(t)\ , t ] , \label{ph - pres - off}\\ & & \delta i= i_1-i_0 , \,\,\ , k = \delta i / s , \label{ph - pres - k}\\ & & \tilde q_m(t ) = \frac{1}{t } \int_0^\infty q(t')\ , dt ' - \langle q\rangle , \,\,\ , s_q = s_i = s , \label{ph - pres - q}\end{aligned}\ ] ] where @xmath210 is the average value of @xmath6 ( which depends on @xmath124 ) , @xmath211 is defined by eq.([ph - sens - im ] ) , and the channels @xmath5 and @xmath6 both have the same ( uncorrelated ) noise @xmath212 . notice that @xmath213 , and therefore _ in the phase - preserving case the ensemble dephasing @xmath12 contains equal contributions @xmath214 from the `` spooky '' and `` realistic '' back - actions_. we emphasize that eqs.([ph - pres - diag])([ph - pres - off ] ) still _ allow us to monitor a qubit wavefunction _ if the initial qubit state is pure . a non - ideal case can be analyzed in the same way as for the phase - sensitive amplifier . an extra dephasing @xmath181 of the qubit is described by @xmath215 , imperfect collection efficiency is described by @xmath120 , and the amplifier efficiency is @xmath189 . we define @xmath216 for a phase - preserving amplifier by comparing its output noise @xmath27 ( per quadrature ) with the quantum limit for a phase - preserving amplifier : @xmath217 , so that @xmath218 in the quantum - limited case . we also define @xmath219 by comparison with a phase - sensitive amplifier having the same gain , so that @xmath220 and obviously @xmath221 . similarly to the phase - preserving case , incomplete microwave collection multiplies the response @xmath61 and correlation @xmath65 by the factor @xmath185 but does not change the noise @xmath27 ; the extra noise in the amplifier multiplies @xmath65 by @xmath189 but does not change @xmath61 . the qubit evolution can still be described by eqs.([ph - pres - diag])([ph - pres - q ] ) with the only change in eq.([ph - pres - off ] ) : @xmath222 \ , e^{-\gamma t } , \qquad \label{ph - pres - off-2}\\ & & \gamma=\gamma-2(\delta i)^2/4s , \label{ph - pres - gamma}\end{aligned}\ ] ] where now @xmath27 is the total ( experimental ) noise per quadrature , @xmath61 is the experimental response , and @xmath198 is the total ensemble dephasing . the qubit evolution is determined by the parameters @xmath61 , @xmath27 , @xmath12 , and output signals @xmath5 and @xmath6 . the quantum efficiencies are = _ amp_col_env=(i)^2/2s , = /2 . here @xmath52 is the fraction of @xmath12 due to the contribution from both the `` spooky '' and `` realistic '' back - actions . the efficiency @xmath223 is the fraction from only the `` spooky '' contribution ; it corresponds to replacing the term @xmath224 in eq . ( [ ph - pres - off-2 ] ) with the dephasing term @xmath225 . in particular , the peak - to - pedestal ratio of the rabi spectral peak for the signal @xmath5 is @xmath226 . let us mention that eqs.([ph - pres - diag])([ph - pres - k ] ) for the ideal phase - preserving case can also be obtained from eqs.([ph - sens - diag])([ph - sens - kmax ] ) for the phase - sensitive case in the following way . let us think about a phase - preserving amplifier as a phase - sensitive amplifier , in which the angle @xmath13 rapidly changes with time , and we have to average over @xmath13 . when coefficients @xmath174 and @xmath175 in eq.([ph - sens - deltai ] ) are substituted into eqs.([ph - sens - diag ] ) and ( [ ph - sens - off ] ) , we see a natural formation of the quadratures @xmath227 and @xmath228 of the phase - preserving setup . then the exponential factor in eq.([ph - sens - diag ] ) becomes @xmath229 , and the exponential factor in eq . ( [ ph - sens - off ] ) becomes @xmath230 . now let us take into account that the average response is @xmath231 , and the phase - sensitive amplifier noise @xmath27 splits equally between the @xmath5 and @xmath6 quadratures ( the orthogonal , de - amplified quadrature is noiseless ) . the mutual cancellation of these two factors of 2 leads to the same form of eq . ( [ ph - pres - diag ] ) as in eq . ( [ ph - sens - diag ] ) and the relation @xmath207 in eq . ( [ ph - pres - k ] ) . one more way to understand the relation between the ideal phase - sensitive and phase - preserving cases is the following . instead of using a phase - preserving amplifier , let us split the outgoing microwave into two equal parts and use phase - sensitive amplifiers with @xmath194 and @xmath204 in the two channels . to keep the same noise @xmath27 per channel , we increase the gain by the factor @xmath232 , that also compensates the signal loss at the splitter . then the channel @xmath194 produces the `` spooky '' back - action ( [ ph - pres - diag ] ) , while the channel @xmath204 informs us of the `` realistic '' back - action ( [ ph - pres - off ] ) , and the relation ( [ ph - pres - k ] ) between @xmath65 and @xmath61 is the same as between @xmath179 and @xmath157 in ( [ ph - sens - kmax ] ) . in the ideal case ( @xmath58 ) the qubit evolution description can be translated into the language of the povm - type measurement . in the same way as in sec . [ broadband ] , eqs . ( [ ph - pres - diag ] ) and ( [ ph - pres - off ] ) can be converted into the measurement operator @xmath233 where the probabilities @xmath234 are given by eq.([gaussian ] ) . similarly , eqs . ( [ ph - sens - diag ] ) and ( [ ph - sens - off ] ) for the case of a phase - sensitive amplifier can be converted into the same measurement operator ( [ meas - op - qed ] ) , in which @xmath235 is replaced with @xmath177 . we have presented a simple physical picture of the qubit evolution due to its measurement in the circuit qed setup . the `` spooky '' back - action is universal , it is caused by gradual extraction of information about the qubit state . the `` realistic '' back - action is due to a specific mechanism : fluctuation of the photon number in the resonator . for a phase - sensitive amplifier the qubit evolution is described by eqs . ( [ ph - sens - diag ] ) and ( [ ph - sens - off-2 ] ) ; it is determined by the output signal @xmath5 , which corresponds to the amplified quadrature . for a phase - preserving amplifier the evolution is described by eqs.([ph - pres - diag ] ) and ( [ ph - pres - off-2 ] ) ; it is determined by two output signals : @xmath5 and @xmath6 , where @xmath5 now corresponds to the quadrature , which provides information about the qubit state ( @xmath89-quadrature in the resonator ) and @xmath6 corresponds to the orthogonal @xmath88-quadrature , which gives us a record of the photon number fluctuations in the resonator . while the cqed setup significantly differs from both the broadband quantum measurement setup @xcite and the standard optical setup @xcite , we see that the description of the qubit evolution is exactly the same as in both these cases if a phase - sensitive amplifier is used . the description is only slightly different when a phase - preserving amplifier is used : we should assign the `` spooky '' and `` realistic '' back - actions to the separate output signals @xmath5 and @xmath6 instead of only one signal . it is also useful to think about the phase - preserving case via the model in which we split the outgoing microwave ( the quantum signal ) into two equal parts and then use 90-degree - shifted phase - sensitive amplifiers for these two channels . we intentionally considered only the simplest case , because most of the further steps and generalizations are quite straightforward @xcite . in particular , it it very simple to include rabi oscillations and energy relaxation of the qubit state . for that we have to take time - derivative of the evolution equations and add the terms due to rabi oscillations and energy relaxation . if the stratonovich definition of the derivative is used , we get the equations of the bayesian formalism @xcite ; if the it derivative is used , we get the equations of the quantum trajectory formalism @xcite . generalization to measurement of several entangled qubits is also straightforward @xcite . we have considered only the resonant case , @xmath123 ; however , generalization to the case @xmath236 is quite simple ( see @xcite ) : we just need a different definition of the informational @xmath89-quadrature and photon - fluctuation @xmath88-quadrature . in our formalism we implicitly assumed sufficiently wide bandwidth of the amplifier ( much larger than the ensemble dephasing @xmath12 and rabi frequency @xmath128 ) . if this is not the case , the formalism should change significantly . however , we believe that in most of the practical cases we can take this effect into account by adding a classical narrowband filter to the classical signal at the amplifier output ; this will correspond to passing the signals @xmath5 and @xmath6 through the low - pass filters . a much more serious change of the theory is required when the resonator bandwidth @xmath115 is comparable to @xmath12 or @xmath128 ; this still has to be done . understanding the difference between the `` spooky '' and `` realistic '' back - actions is important for designing the quantum feedback control of the rabi oscillations @xcite . the simplest case is when a phase - sensitive amplifier amplifies the informational @xmath89-quadrature . then there is no `` realistic '' back - action , and the feedback loop should only modulate the amplitude of the rabi drive ( i.e. the rabi frequency @xmath128 ) ; this case was well studied for the broadband setup @xcite . the situation is different for a phase - preserving amplifier . then we need two feedback channels : the first ( usual ) channel should modulate the rabi frequency @xmath128 to compensate the `` spooky '' back - action , while the second channel should compensate the `` realistic '' back - action by modulating the qubit frequency @xmath237 or the the frequency of the rabi drive @xmath238 . the controller for the second feedback channel is quite simple : it should compensate the contribution @xmath239 to the qubit phase derivative @xmath240 due to the @xmath65-term in eq . ( [ ph - pres - off-2 ] ) . therefore the controller is ( _ qb-_r)= -k[q(t)-q ] , i.e. we should directly apply the signal @xmath6 to modulate @xmath237 or @xmath238 . the second feedback channel essentially eliminates the @xmath65-term in eq . ( [ ph - pres - off-2 ] ) and decreases the ensemble dephasing @xmath12 by @xmath241 . correspondingly , in the absence of the first ( main ) feedback channel the peak - to - pedestal ratio of the rabi peak increases from 2 to 4 in the quantum - limited case . the first feedback channel should be the same as for the broadband setup ; it depends on the signal @xmath5 and can be realized using various ideas for the controller ( `` direct '' , bayesian , `` simple '' , etc . @xcite ) . notice that _ without the second channel the feedback performance is determined by the quantum efficiency @xmath55 , while with the second channel it is determined by @xmath242 _ ( this is one more combination of the terms in eq . ( [ gamma ] ) , which can be used for the definition @xcite of quantum efficiency ) . the case of a phase - sensitive amplifier , which amplifies a non - optimal quadrature ( @xmath243 , @xmath244 ) is similar to the case of a phase - preserving amplifier , but both feedback channels should start with the same signal @xmath5 . _ in both the phase - sensitive and phase - preserving setups a perfect feedback control is possible in the quantum - limited case @xmath58 . _ discussion of the `` spooky '' and `` realistic '' back - actions in the cqed setup necessarily raises the question of causality . when the microwave leaves the resonator , it does not yet `` know '' in which way it will be measured ( phase - preserving or phase - sensitive , which angle @xmath13 , etc . ) . moreover , when a circulator is used for the outgoing microwave , the field in the resonator and the qubit can never `` know '' in a realistic way which method of measurement is used . nevertheless , the qubit evolution strongly depends on the measurement method . as we discussed , the `` spooky '' evolution moves the qubit state along the meridians of the bloch sphere , the `` realistic '' back - action moves the state along the parallels , and the measurement method determines whether the qubit experiences the `` spooky '' or `` realistic '' back - action ( or their combination ) . in this sense the `` realistic '' back - action is not fully realistic : it has the physical mechanism , but whether this mechanism works or not is determined in a spooky way . the causality requires that we can not pass a `` useful '' information to the qubit by choosing the measurement method . this means that _ the ensemble - averaged evolution of the qubit can not depend on the measurement method ( this is the general requirement of causality in quantum mechanics)_. it is surely satisfied in our cqed setup . the author thanks michel devoret , konstantin likharev , patrice bertet , and farid khalili for useful discussions . the work was supported by aro muri grant w911nf-11 - 1 - 0268 and by nsa / iarpa / aro grant w911nf-10 - 1 - 0334 . h. m. wiseman and g. j. milburn , phys . a * 47 * , 642 ( 1993 ) ; h. j. carmichael , _ an open system approach to quantum optics _ ( springer , berlin , 1993 ) ; h. m. wiseman and g. j. milburn , _ quantum measurement and control _ ( cambridge univ . press , 2009 ) . |
the contributions to the heat kernel coefficients generated by the corners of the boundary are studied . for this purpose the internal and external sectors of a wedge and a cone are considered . these sectors are obtained by introducing , inside the wedge , a cylindrical boundary . transition to a cone is accomplished by identification of the wedge sides . the basic result of the paper is the calculation of the individual contributions to the heat kernel coefficients generated by the boundary singularities . in the course of this analysis certain patterns , that are followed by these contributions , are revealed . the implications of the obtained results in calculations of the vacuum energy for regions with nonsmooth boundary are discussed . the rules for obtaining all the heat kernel coefficients for the minus laplace operator defined on a polygon or in its cylindrical generalization are formulated . | hep-th0207038 | the asymptotic expansion of the heat kernel proves to be important in a series of physical applications . its coefficients specify the divergences and conformal anomalies taking place in a concrete field theory model @xcite , the high temperature behaviour of the thermodynamic functions @xcite and so on @xcite . for a flat manifold with a smooth boundary the heat kernel coefficients are determined by local characteristics of the boundary @xcite . if the boundary has discontinuities ( for example , it is piecewise smooth ) , then the latter give additional contributions to the heat kernel coefficients @xcite . usually in physical applications one assumes that the boundary is smooth . however there is a series of problems where such assumption is certainly not acceptable . a typical example here is supplied by fields defined between two plates which cross at a given angle , i.e. , inside a dihedral angle . such a configuration is considered when calculating the casimir effect for a conducting wedge @xcite . if the fields inside a dihedral are subjected to periodicity condition with respect to the angular variable then one is concerned in fact with the fields on a cone , and the point where the boundary has discontinuity becomes an internal point of the cone surface . however the origin of this singularity is the same as in the case of fields inside the wedge . the conical singularity proved to be very important in many areas of mathematical physics @xcite and lately it is investigated in connection with studies of quantum fields on the background of black holes @xcite and cosmic strings @xcite . the general consideration of the boundary nonsmothness in terms of the heat kernel expansion lacks till now @xcite . such contributions to the heat kernel coefficients @xmath0 and @xmath1 have been investigated in papers @xcite . for a plane domain the contribution to @xmath2 generated by an edge of the boundary is known @xcite and by its limiting configuration , by cusp . it is interesting to note that a cusp pointing outward , with respect to the domain under study , leads to change of the power of the asymptotic variable ( time ) in the heat kernel expansion @xcite in comparison with the standard case . in references @xcite the asymptotic expansion of the heat kernel with allowance for the boundary nonsmothness has been built by calculating the relevant green function of the heat conduction equation . the present paper seeks to show the effectiveness of applying the spectral zeta functions for the calculation of the contributions to the heat kernel coefficients caused by such boundary discontinuities as the corners . for this goal we shall use the technique for constructing the zeta functions developed in @xcite and extended in @xcite this method proves to be very effective for calculating the heat kernel coefficients for different boundary conditions given on a sphere and cylinder @xcite . a close approach has been used in @xcite . we shall consider internal and external parts of a plane sector formed by two infinite radial rays emerging from the center of a circle of radius @xmath3 at angle @xmath4 to one another ( see figure 1 , where i is the internal circular sector and ii is the external circular sector ) . the choice of such domains with nonsmooth boundary is caused by the possibility of constructing for them the global zeta functions . the latter can not be done , for example , for an open angle ( figure 1 without circular arc inside the angle ) . this point will be discussed in detail in section ii . in both the sectors the laplace operator is defined acting on scalar real functions subjected to the dirichlet or neumann boundary conditions . if we substitute in figure 1 the radial rays by crossed planes and the circular arc 1 2 by an appropriate part of a cylinder surface we arrive at the boundary value problems which have the same heat kernel coefficients as in the plane case . identifying the points of the boundary with the polar coordinates @xmath5 and @xmath6 , i.e. , imposing the periodicity condition with respect to the angular variable @xmath7 with a period @xmath4 , we arrive obviously at the spectral problem for the laplace operator on two parts of lateral surface of a cone @xmath8 : internal part @xmath9 and external part @xmath10 . at @xmath11 we can , as before , impose the dirichlet or neumann conditions . in the present paper , six coefficients of the asymptotic expansion of the heat kernel for the boundary value problems specified above will be calculated by making use of the relevant zeta functions . it will be shown that each of these coefficients , starting from the third one , is the sum of contributions generated by the corners of the boundary and by the curvature of the arc 12 . analysis of the obtained results enables one to reveal some regularities obeyed by the contributions of the boundary nonsmoothness to the heat kernel coefficients , namely : i ) the corner contribution substantially depends on whether the corner is made up by crossing two straight lines or two lines with nonzero curvature ; ii ) contributions of adjacent angles to the heat kernel coefficients @xmath12 and @xmath13 have opposite signs in the same way as the contributions of a circular arc to the coefficients @xmath14 and @xmath15 for the internal and external regions . in the case of a polygon or its cylindrical generalization the rules are formulated for obtaining all the heat kernel coefficients for the minus laplace operator . the implications of the obtained results in the casimir energy calculations with employment of the zeta function technique are also considered . in particular , it is shown that nonsmoothness of the boundary does not make always worse the situation with calculation of the vacuum energy in the framework of the zeta regularization method . for example , the corner contributions can simply be mutually cancelled in the same way as the contributions of the curved boundary are cancelled when taking into account the internal and external regions . besides , the analysis conducted enables one to infer that it is very unlikely to get a finite and unique result for the vacuum energy by smoothing the boundary singularities , e.g. , by taking into account the atomic structure of the boundary or quantum fluctuations of the boundary . the paper is organized as follows . section 2 is devoted to the detailed discussion of the choice of the domains with piece - wise smooth boundary ( internal and external circular sectors ) , for which the complete spectral zeta functions can be constructed . in section 3 the heat kernel coefficients are calculated for internal circular sector by making use of the counter integral representation for the corresponding zeta functions @xcite . in section 4 the contributions to the heat kernel coefficients of the individual boundary discontinuities ( corners of different angles ) are identified . further ( in section 5 ) the heat kernel coefficients for the union of both the sectors are calculated and the coefficients for the external sector are obtained as the regarding differences . technically it turns out to be simpler in comparison with calculation of the heat kernel coefficients for the external sector alone . the asymptotic expansion of the heat kernel for the union of internal and external sectors is constructed by differentiation of the logarithm of the function specifying the frequency equation @xcite . in section 6 the identification of the individual contributions of boundary nonsmoothness to heat kernel coefficients for external sector is carried out . some general rules for these contributions are revealed here for both sectors . in section 7 we conclude with a few summarizing remarks . when choosing the domain for investigating the contribution to the heat kernel coefficients of the boundary discontinuities we pursue two goals : the boundary of a domain should have a sufficient number of discontinuities and at the same time for this domain one can construct the spectral zeta function . in order to investigate the corner singularities of the boundary one should at first sight take the most simple configuration , namely , the angle on plane formed by two radial unrestricted rays or dihedral in space . however , for these domains the global zeta functions can not be constructed . let us explain this point in detail . we consider the laplace operator @xmath16 acting on the scalar functions defined inside the dihedral of opening angle @xmath17 ( wedge of an angle @xmath4 , @xmath18 ) and subjected to the dirichlet conditions on the wedge sides . in cylindrical coordinate system ( @xmath19 ) the eigenfunctions in this problem are @xmath20 @xmath21 where @xmath22 is the bessel function . the operator @xmath23 has the following eigenvalues @xmath24 these eigenvalues do not depend on the quantum number @xmath25 , i.e. , there is a degeneracy with respect to this number , the multiplicity of this degeneracy being infinite and @xmath26 is a nonisolated point . in fact , the spectrum is continuous according to the terminology of spectral theory of operators in hilbert space @xcite and @xmath27 are not eigenvalues as the eigenfunctions are not square integrable . ] @xmath28 the global spectral zeta function of the operator @xmath29 , @xmath30 , is determined as the trace of the operator @xmath31 @xmath32 where @xmath33 is the @xmath34th eigenvalue of the operator @xmath29 , and @xmath35 is the degeneracy of this eigenvalue . obviously , this definition does not work when @xmath35 is infinite . in order to remove this degeneracy we put inside a dihedral a cylindrical boundary as it is shown in figure 1 ( the arc 12 ) . on the internal and external sides of this boundary the scalar field will obey the dirichlet or neumann conditions . thus , we are considering the internal ( i ) and external ( ii ) sectors . certainly , the union of these two sectors is not identical to unrestricted dihedral ( to a wedge ) , because now the values of the field on the arc 12 are not arbitrary but they are determined by the corresponding boundary conditions . in the new configuration there appear additional discontinuities of the boundary at the points 1 and 2 . however at these points the angle , at which the involved boundary surfaces intersect , is fixed ( it is equal to @xmath36 ) . it will be shown below that the contribution of such boundary singularities to the heat kernel coefficients can be easily separated from the contribution of the corner at the origin . the latter is proportional to the difference @xmath37 , because at @xmath38 the singularity at the origin disappears . we shall concern with the standard asymptotic expansion of heat kernel @xmath39 where @xmath40 is the dimension of the manifold under study and es stands for the exponentially small corrections as @xmath41 . this definition leads to the same heat kernel coefficients @xmath42 for a dihedral and for corresponding plane problem obtained by crossing of the dihedral by a transverse plane . in fact , the eigenvalues of the operator @xmath43 in these two problems obey the relation @xmath44 hence @xmath45 taking into account the definition ( [ eq2 - 5 ] ) one easily deduce from ( [ eq2 - 7 ] ) that the heat kernel coefficients @xmath42 in two eigenvalue boundary problems , mentioned above , are equal . here one should bear in mind that the heat kernel @xmath46 and its coefficients are referred to a unite length along the oz axes . in view of this , when calculating the heat kernel coefficients we shall consider either the spectral problem on a plane @xmath47 or in the space @xmath48 , pursuing only simplicity of calculation . at first we consider only internal sector ( the region i in figure 1 ) for @xmath49 , i.e. , for internal sector of a dihedral or a wedge . we employ here the technique developed in @xcite and extended to the generalized bounded cone in @xcite ( see also book @xcite ) . in this approach the spectral zeta function @xmath50 should be constructed in the beginning , and then the relevant heat kernel coefficients are calculated through the relation @xcite @xmath51 for the dirichlet boundary conditions the eigenfunctions of the operator @xmath52 in the region i are defined by equation ( [ eq2 - 1 ] ) , with @xmath26 being the roots of the equations @xmath53 here the subscript @xmath54 numbers the nonzero roots of these equations for fixed @xmath25 . the relevant eigenvalues are defined by equation ( [ eq2 - 2 ] ) with @xmath55 @xmath56 in view of the behavior of the bessel function near zero @xmath57 it follows that the frequency equation ( [ eq3 - 2 ] ) has the zero root of ` multiplicity ' @xmath58 . such roots should be removed from the definition of the spectral zeta function ( [ eq2 - 4 ] ) . therefore instead of equation ( [ eq3 - 2 ] ) we shall use the following frequency equation @xmath59 according to the general definition ( [ eq2 - 4 ] ) the zeta function in the problem under study is given by @xmath60 as usual we substitute the sum over @xmath61 in this formula by the contour integral in the plane of a complex variable @xmath26 @xmath62\,{,}\ ] ] where the contour @xmath63 encloses counter clockwise the positive roots of equations ( [ eq3 - 4 ] ) . on deforming the contour @xmath63 in an appropriate way and integrating with respect to @xmath64 by means of the formula @xmath65 we arrive at the result @xmath66\,{,}\ ] ] where @xmath67 is the modified bessel function . the analytical continuation of ( [ eq3 - 8 ] ) into the left half - plane of the complex variable @xmath68 is accomplished by making use of the uniform asymptotic expansion of the function @xmath69 @xmath70 where @xmath71 and @xmath72 are the known polynomials in @xmath73 . their explicit form and the corresponding recurrent relations can be found in references @xcite . keeping in this expansion all the terms proportional to @xmath74 with @xmath75 one can write @xmath76 where @xmath77 in accordance with this expansion we represent the zeta function in equation ( [ eq3 - 9 ] ) as the following sum @xmath78 where @xmath79 in these formulas @xmath80 is the riemann zeta function and @xmath81 substituting ( [ eq3 - 9a ] ) into equation ( [ eq3 - 12 ] ) and taking into account the value of the integral @xmath82 we obtain for the function @xmath83 the final expression @xmath84 the product @xmath85 has simple poles at the points @xmath86 , and @xmath87 with the respective residua @xmath88 multiplying these residua by @xmath89 we obtain , according to equation ( [ eq3 - 1 ] ) , the contributions of the function @xmath90 to the heat kernel coefficients @xmath91 , and @xmath14 , respectively . by making use of the table integral @xcite @xmath92 we recast equation ( [ eq3 - 13 ] ) to the form @xmath93 the expression @xmath94 has simple poles at the points @xmath95 and @xmath96 with the residua @xmath97 which give the respective contributions to the heat kernel coefficients @xmath98 and @xmath14 ( see table 1 ) . calculation of the functions @xmath99 , defined in equation ( [ eq3 - 14 ] ) , can be carried out with hardly any trouble . their contribution to the heat kernel coefficients are given in table 1 . in order to get the complete values of these coefficients one should sum all the elements of respective rows . in this case the operator @xmath100 has the following eigenfunctions in internal circular sector i ( see figure 1 ) @xmath101 @xmath102 here @xmath103 are the roots of the equations @xmath104 taking into account the behavior of the derivative of the bessel function at the origin @xmath105 , we multiply equation([eq3 - 23 ] ) by @xmath106 in order to exclude the zero multiple roots @xmath107 by making use of the frequency equations ( [ eq3 - 24 ] ) we can write immediately the integral representations for the zeta function in the problem at hand analogous to equation([eq3 - 8 ] ) @xmath108{,}\end{aligned}\ ] ] where the coefficient @xmath109 is has been defined in equation([eq3 - 15 ] ) . an important distinction of this equation is a new term with @xmath110 , that was absent in the equation ( [ eq3 - 8 ] ) for the dirichlet boundary condition . we again use the uniform asymptotic expansion @xcite but now for the derivative of the bessel function @xmath111 @xmath112 where @xmath113 is defined in equation ( [ eq3 - 9a ] ) and the functions @xmath114 are the known polynomials @xcite in @xmath115 . again we use the approximation @xmath116 where @xmath117 instead of equation ( [ eq3 - 11 ] ) we have now @xmath118 where @xmath119\,{,}\label{eq3 - 30 } \\ \fl z_{-1}^n(s)&=&z_{-1}^d(s),\qquad z_{0}^n(s)=z_{0}^d(s)\ , { , } \label{eq3 - 30a}\\ \fl z_{j}^n(s)&=&c(s)\,p^{-2s-1+j}\zeta_r(2s+j-1)\int_{0}^{\infty}\rmd y\,y^{1 - 2s } \frac{\rmd}{\rmd y}f_j^n(t){,}\quad j=1 , 2,3,4{. } \label{eq3 - 30b}\end{aligned}\ ] ] finding the functions @xmath120 in the expansion ( [ eq3 - 29 ] ) presents no difficulty while the function @xmath121 needs more thorough treatment . by making use of the asymptotics @xmath122&\simeq&\frac{2}{y}+\frac{1}{4}\,y+{\cal o } ( y^3 ) , \quad y\to 0\ , { , } \label{eq3 - 31a } \\ \frac{\rmd}{\rmd y}\ln[yi'_0(y)]&\simeq & 1+\frac{1}{2y}+\frac{3}{8y^2}+\frac{3}{8 y^3}+\frac{63}{128y^4}+{\cal o}(y^{-5}),\quad y\to \infty \label{eq3 - 31b}\end{aligned}\ ] ] it is easy to make sure that the integral in the definition of the function @xmath121 does not converge at any values of @xmath68 . in order to overcome this drawback first the zeta function @xmath123 for a scalar field with nonzero mass @xmath61 should be constructed and on calculating the residua according to equation([eq3 - 1 ] ) with this zeta function the mass @xmath61 should be put equal to zero case for the neumann boundary conditions has been explained in detail in @xcite . ] . following this line we consider the function @xmath124\,{.}\ ] ] it is defined in the region @xmath125 , the lower ( upper ) limit in these inequalities being determined by the convergence of the integral when @xmath126 ( @xmath127 ) . in order to find , by equation ( [ eq3 - 1 ] ) , the contribution of the function @xmath128 to the coefficient @xmath129 the analytical continuation of @xmath128 to the region @xmath130 is required . in the most simple way it can be done by adding and subtracting from the integrand its asymptotics when @xmath127 . for our goals it is sufficient to take only the first term in this asymptotics @xmath131 - f_m \right \}+ f_m\int_{m}^{\infty}\!\!\rmd y\ , ( y^2-m^2)^{1/2-s}{,}\ ] ] where @xmath132\right |_{y = m}{.}\ ] ] the first term in equation ( [ eq3 - 33 ] ) is regular at the point @xmath133 , but the integral in the second term gives a simple pole at this point @xmath134 however the gamma function @xmath135 responsible for this pole is canceled by the same multiplier in the denominator of the coefficient @xmath109 ( see equation ( [ eq3 - 15 ] ) ) . as a result the function @xmath128 does not give the contribution to the heat kernel coefficient @xmath129 . for calculating the contribution of the function @xmath136 to the rest heat kernel coefficients by equation([eq3 - 1 ] ) , this function should be analytically continued to the region @xmath137 . it can again be done by adding and subtracting the asymptotics of the integrand when @xmath138 now . further the residua are found according to equation([eq3 - 1 ] ) and only after that the mass @xmath61 is put equal to zero . the corresponding results are presented in table 2 . finding the contributions of the rest functions @xmath139 to the heat kernel coefficients presents no trouble ( see table 2 ) . imposing on the eigenfunctions the periodicity condition with respect to the angular variable @xmath140 with a period @xmath17 we pass from a wedge to a cone because in this case the respective points on the radial rays @xmath141 and @xmath142 are identified ( see figure 1 ) , the circular arc 12 being converted into a circumference apparently , the boundary discontinuities at the points 1 and 2 disappear . this circumference separates two parts ( internal and external ) of the cone surface which for simplicity will be refereed to as the internal and external sectors of the cone . on the circumference separating them we impose on the eigenfunctions , as before , the dirichlet or neumann conditions . the corresponding ( unnormalized ) eigenfunctions for internal cone sector are @xmath143 where @xmath103 are the roots of the frequency equations for the dirichlet boundary conditions at @xmath11 @xmath144 or of those for the neumann conditions @xmath145 for the external sector the bessel functions in equations([eq3 - 34 ] ) ( [ eq3 - 36 ] ) are replaced by the hankel functions @xmath146 . from ( [ eq3 - 34 ] ) it follows that on the cone all the states with @xmath147 are double degenerate @xmath148 thus , in order to find the heat kernel coefficients for the internal cone sector one should put in previous calculations @xmath149 take into account the degeneracy of states ( [ eq3 - 38 ] ) , and sum up over @xmath25 starting with @xmath110 . let us consider the internal sector of a cone with the dirichlet conditions on the circumference 12 . for the corresponding zeta function @xmath150 the representation ( [ eq3 - 8 ] ) holds with @xmath151 defined in ( [ eq3 - 38 ] ) and with the summation replaced by @xmath152 , where the prime on the summation sign means that the @xmath110 term is counted with half weight . for @xmath153 in the sum ( [ eq3 - 11 ] ) there arises an additional term with @xmath110 @xmath154 where @xmath155 and the functions @xmath156 are defined by equations ( [ eq3 - 12 ] ) ( [ eq3 - 14 ] ) with @xmath157 . the asymptotics @xmath158 imply that the function @xmath159 in equation([eq3 - 40 ] ) is defined in the region @xmath160 to single out in the integral ( [ eq3 - 40 ] ) the pole contribution at the point @xmath133 we rewrite the function @xmath161 as follows @xmath162 \nonumber \\ & & + \frac{c(s)}{2}\int_{0}^{1 } \rmd y\,y^{2 - 2s}+c(s)\int_{1}^{\infty}\rmd y\ , y^{1 - 2s}\frac{\rmd}{\rmd y}i_0(y)\,{. } \label{eq3 - 44}\end{aligned}\ ] ] the integrals in the first and third terms in equation([eq3 - 44 ] ) are regular at the point @xmath133 . substituting the second term from equation ( [ eq3 - 44 ] ) into definition ( [ eq3 - 1 ] ) we get @xmath163 thus the function @xmath164 does not give any contribution to the coefficient @xmath129 . so as to find the contribution of the function @xmath165 to the heat kernel coefficients @xmath166 we again split the domain of integration in equation ( [ eq3 - 40 ] ) into two intervals @xmath167 and @xmath168 . when integrating over the second interval we add and subtract under the integral sign the asymptotics ( [ eq3 - 42 ] ) . when calculating the residua at the points @xmath169 we shall take the right - hand limits in equation ( [ eq3 - 1 ] ) it gives the following contributions to the heat kernel coefficients @xmath170 , and @xmath13 , respectively @xmath171 in order to evaluate the contributions of the functions @xmath172 to the heat kernel coefficients , in addition to ( [ eq3 - 45 ] ) , one should substitute in table 1 @xmath4 by @xmath173 and multiply all the elements of this table by 2 . summing the contributions ( [ eq3 - 45 ] ) and the data from table 1 we obtain the heat kernel coefficients for the internal sector i on the cone with the dirichlet conditions on the circle 12 @xmath174 @xmath175 for neumann boundary conditions the spectral zeta function for internal sector on a cone is given by @xmath176 where the functions @xmath177 and @xmath178 are determined in equations ( [ eq3 - 30 ] ) ( [ eq3 - 30b ] ) with @xmath4 replaced by @xmath179 . the corresponding coefficients of the heat kernel expansion are @xmath180 @xmath181 let us envisage the internal sector i ( see figure 1 ) . its boundary possesses the following peculiarities which give contribution to the heat kernel coefficients : nonzero curvature of the arc 12 ; right - angled corners at the points 1 and 2 ; corner of angle @xmath17 at the origin . the arc contribution is proportional to its length , i.e. , to @xmath17 , the contributions of the right - angled corners does not depend on @xmath4 , contribution of the corner at the origin vanishes when @xmath182 . it is sufficient to separate in each of the heat kernel coefficients the contribution due to each boundary singularity enumerated above . we demonstrate this considering the heat kernel coefficients for internal circular sector i with dirichlet conditions on its boundary ( see table 1 ) . the first coefficient @xmath183 is the area @xmath184 of the circular sector i. the second heat kernel coefficient @xmath185 where @xmath29 is the length of the sector boundary ( its perimeter ) . as concerns the coefficient @xmath186 the situation is more complicated . it is clear that @xmath36 is the contribution of two right - angled corners at the points 1 and 2 . the term @xmath187 contains the contribution of the curvature of the arc 12 ( denote it by @xmath188 ) and a part of the contribution of the corner at the origin @xmath189 ( the latter is equal to @xmath190 ) . in terms of these notations the complete contribution of the corner at the origin is @xmath191 from the condition @xmath192 it follows that @xmath193 further we find @xmath194 @xmath195 thus the corner of an angle @xmath17 on the boundary gives the contribution to @xmath14 defined by equation ( [ eq4 - 5 ] ) . at first time this contribution has been calculated in@xcite . another method to derive it is described in@xcite . in both the cases the green s function of the equation of heat conductivity was considered . in our approach it is found by making use of the spectral zeta function technique . now we are in position to check the consistency of our reasoning , namely , we can calculate the contribution to the coefficient @xmath14 due to the right - angled corners at the points 1 and 2 by making use of equation ( [ eq4 - 5 ] ) . it gives @xmath196 it is this value that has been attributed to this contribution above ( see equation ( [ eq4 - 1 ] ) ) . identification of individual contributions to the rest of heat kernel coefficients can be done in a direct way . the terms independent of the angle @xmath4 are attributed to the right - angled corners at the points 1 and 2 , while the linear in @xmath4 terms are due to the curvature of the arc 12 ( see table 3 ) . merely such a separation of individual peculiarity contributions leads to a correct value of the arc curvature contribution which is known from the heat kernel expansion for a smooth boundary ( see below ) . it is worth noting that the angle @xmath4 at the origin does not contribute to the heat kernel coefficients @xmath197 with @xmath198 even when @xmath199 . it may be explained only taking into account that the higher derivatives of the radius vector of the boundary curve behave in a different way at the origin and at the points 1 and 2 . let @xmath200 be a simply connected region of a plane bounded by a smooth curve @xmath201 . for the heat kernel of the minus laplace operator with the dirichlet conditions on @xmath201 the following asymptotic expansion holds when @xmath202 ( see , for example , references @xcite ) @xmath203+{\cal o}(t^2)\ , { , } \label{eq4 - 7}\end{aligned}\ ] ] where @xmath204 is the area of @xmath200 , @xmath29 is the length of @xmath201 , @xmath205 is the curvature of the curve @xmath201 at the point @xmath68 , @xmath206 , where @xmath207 is a parametric representation of the curve @xmath201 ; @xmath68 is the natural parameter along @xmath201 : @xmath208 ; @xmath209 . for convex portions of @xmath201 @xmath205 is considered to be positive , and for concave parts of @xmath201 @xmath205 is assumed to be negative . in the expansion ( [ eq4 - 7 ] ) the numerical coefficients of @xmath210 , and @xmath211 are derived from the contributions , proportional to @xmath4 , to the heat kernel coefficients @xmath212 , and @xmath213 , respectively ( see table 3 ) . here it should be taken into account that in the problem under study @xmath214 and the coefficients @xmath197 enter the heat kernel expansion ( [ eq2 - 5 ] ) with the multiplier @xmath215 . if we go from a wedge to a cone by identifying the radial rays @xmath141 and @xmath142 the corners at the points 1 and 2 disappear . the sole singular point remains the origin @xmath189 which becomes an internal point of the cone surface . however the contribution of this singularity to the coefficient @xmath14 has the same nature as in the case of wedge . the heat kernel coefficients for internal sector i on the cone with dirichlet condition on the boundary 12 are listen in equation ( [ eq3 - 45a ] ) . the coefficients @xmath129 and @xmath98 obey general equations ( [ eq4-a ] ) and ( [ eq4-b ] ) with corresponding values for @xmath184 and @xmath29 . separation of the contributions to @xmath14 generated by singularity at the origin and by the curvature of the circle 12 can be conducted in the same way as for a wedge . let @xmath216 be the contribution of the boundary 12 and @xmath217 be the contribution due to the singularity at the origin . when @xmath218 the surface of a cone becomes a plane and the singularity at the origin @xmath189 disappear . therefore , @xmath219 . it gives @xmath220 i.e. , for this quantity we have the same value as in the case of a wedge ( see ( [ eq4 - 4 ] ) ) . with ( [ eq4 - 12 ] ) allowed for , one deduces from the equation ( [ eq4 - 11 ] ) @xmath221 where @xmath194 is defined in equation ( [ eq4 - 5 ] ) . identification of the individual contributions to the heat kernel coefficients for the neumann boundary conditions is conducted in the same way ( see table 3 ) . the first coefficient @xmath129 is , as before , the area @xmath184 of the sector i. for the second coefficient @xmath98 wehave equation ( [ eq4-b ] ) with oposite sign in the right - hand side @xmath222 the contribution to the coefficient @xmath14 generated by the singularity at the origin proves to be the same as for the dirichlet conditions , @xmath194 . for neumann boundary conditions we failed to find in the literature the formula analogous to equation ( [ eq4 - 7 ] ) , i.e. , the asymptotic expansion of the heat kernel for the operator @xmath23 defined in the region of a plane with smooth boundary curve . however , as was noted earlier , the spectral problem on the plane , envisaged by us , and its cylindrical generalization have the same coefficients @xmath197 . therefore , for verification of our results , concerning the contributions to @xmath197 due to the smooth parts of the boundary , we used the expansion of the heat kernel for robin conditions with smooth boundary that has the dimension greater than 1 ( see references @xcite ) . here we shall use the technique applied in @xcite . it is close to the method developed in @xcite and employed in preceding subsections . let us consider the spectral zeta function depending on a parameter @xmath223 @xmath224 it may be regarded as an extension to the general spectral problem of the epstein - hurwitz zeta function @xmath225 it turns out that the heat kernel coefficients @xmath197 can be found from the expansion of the function @xmath226 in terms of inverse powers of @xmath227 developed for a certain value of @xmath68 . it is convenient to chose this value to be equal to @xmath228 . in fact , from the definition of the gamma function it follows that @xmath229 for @xmath230 equation ( [ eq3 - 48 ] ) gives @xmath231 on substituting the asymptotic expansion ( [ eq2 - 5 ] ) in equation ( [ eq3 - 49 ] ) we obtain @xmath232{. } \label{eq3 - 50}\end{aligned}\ ] ] let @xmath233 be the frequency equation which determines the spectrum @xmath234 in the problem under consideration . we also suppose that the function @xmath235 allows one to rewrite this equation in the form @xmath236 taking into account the equality @xmath237 we recast the left - hand side of equation ( [ eq3 - 50 ] ) to the form @xmath238 obviously formula ( [ eq3 - 54 ] ) is applicable only to the manifolds of even dimension . rather than to calculate the heat kernel coefficients @xmath239 for the external sector alone it is simpler to find first the coefficients @xmath240 for the union of the sectors i and ii . then the coefficients @xmath239 are obtained as the corresponding differences @xmath241 when calculating the heat kernel coefficients for the union of the internal ( i ) and external ( ii ) sectors on should take into account the following . the technique used by us @xcite gives the difference between the zeta function for the region i+ii and the zeta function for the corresponding part of the euclidean space . the last contribution is usually referred to as the minkowski space - time contribution @xcite . in the case under consideration the zeta function for an open angle @xmath17 ( without circular arc 1 - 2 ) is subtracted from the zeta function sought for . the heat kernel coefficients @xmath242 , corresponding to the minkowski space - time contribution , can be calculated in the following way : in the respective heat kernel coefficients for the sector i one should put @xmath243 and discard the curvature contribution of the arc 1 - 2 and of two rightangled internal corners at the points 1 and 2 . now we proceed to practical using the general formulas ( [ eq3 - 50 ] ) and ( [ eq3 - 54 ] ) for calculation of the heat kernel coefficients . we consider the scalar laplace operator on the union of internal and external circular sectors on a plane ( see figure 1 ) with dirichlet conditions on the arc 12 separating these sectors . as it was explained in section 2 the heat kernel coefficients for the corresponding boundary value problem in space ( @xmath49 ) the frequency equations are will be the same . @xmath244 further we shall concern with the product @xmath245 and use for it the uniform asymptotic expansion @xcite which depends only on @xmath223 . in this case the condition ( [ eq3 - 52 ] ) is apparently satisfied . this point can also be explained in the following way . in view of the formula@xcite @xmath246 for the function @xmath247 the representation of type ( [ eq3 - 52 ] ) holds . here @xmath248 are the nonzero roots of the function @xmath249 . the multiplier @xmath250 in equation ( [ eq3 - 57 ] ) is canceled in the product @xmath251 with the multiplier @xmath252 following from the small @xmath227 asymptotics of the function @xmath253 : @xmath254 . hence the requirement ( [ eq3 - 52 ] ) is satisfied . setting in ( [ eq3 - 54 ] ) @xmath255 and substituting in it the left - hand sides of the frequency equations ( [ eq3 - 55 ] ) and ( [ eq3 - 56 ] ) we obtain @xmath256 now we use the uniform asymptotic expansion for the product of the modified bessel functions @xcite @xmath257 where @xmath258 and @xmath259 are the polynomials in @xmath260 expressed in terms of the known functions @xmath72 . in order to calculate the first six coefficients in the expansion ( [ eq3 - 50 ] ) it is sufficient to keep two terms of the sum in ( [ eq3 - 59 ] ) . the relevant coefficients @xmath261 and @xmath262 are @xmath263 substituting in ( [ eq3 - 58 ] ) differentiation with respect to @xmath227 by differentiation with respect to @xmath260 we get @xmath264 \nonumber \\ & = & -r^4\sum_{n=1}^\infty{}\left [ \frac{t^4}{2\nu^4}+\frac{t^6}{4\nu^6}(1 - 18t^2 + 30t^4 ) \right . \nonumber \\ & & \left .+\frac{3t^8}{32\nu^8}(13 - 568t^2 + 3540t^4 - 6780t^6 + 3955t^8 ) \right ] { . } \label{eq3 - 63}\end{aligned}\ ] ] it will be recalled that @xmath260 depends on @xmath227 and @xmath25 through ( [ eq3 - 60 ] ) . all the sums in the equation ( [ eq3 - 63 ] ) are finite . hence the problem of analytic continuation does not emerge here . in order to do the summation exactly we use the formula @xcite @xmath265 when @xmath126 , the function @xmath266 affords the value of the sum on the left - hand side of equation ( [ eq3 - 64 ] ) up to exponentially small corrections . step - by - step differentiation with respect of @xmath267 of the left and right hand sides of the equation([eq3 - 64 ] ) gives the values of all the sums entering ( [ eq3 - 63 ] ) @xmath268 in order to express the zeta function @xmath269 in terms of @xmath270 explicitly the following substitutions should be accomplished in equation ( [ eq3 - 63 ] ) @xmath271 @xmath272 the change ( [ eq3 - 67 ] ) being done only in the round brackets in this equation . as a result the zeta function @xmath273 acquires the form @xmath274 \nonumber \\ \fl & & -\frac{3r^4}{32p^8}\left [ 160\ , s_4({\tilde x } ) -1992\,{\tilde x}^2\,s_5({\tilde x})+6930\ , { \tilde x}^4\,s_6({\tilde x } ) \right . \fl & & \left .-9040\,{\tilde x}^6\,s_7({\tilde x})+3955\,{\tilde x}^8\,s_8({\tilde x } ) \right ] { . } \label{eq3 - 68}\end{aligned}\ ] ] substitution of the explicit expressions for the functions @xmath275 from ( [ eq3 - 65 ] ) to ( [ eq3 - 68 ] ) gives @xmath276 comparing the expansions ( [ eq3 - 50 ] ) and ( [ eq3 - 69 ] ) we obtain the values of the first seven heat kernel coefficients in the problem under consideration @xmath277 @xmath278 where @xmath279 , and @xmath242 are the heat kernel coefficients corresponding to the open angle @xmath4 ( without the arc 1 - 2 ) @xmath280 all the rest of coefficients @xmath281 vanish @xmath282 . using equations ( [ eq3 - 70 ] ) , ( [ eq5 - 24 ] ) and the results for the internal sector i , presented in table 1 we deduce the heat kernel coefficients for the external sector ii with dirichlet boundary condition @xmath283 in the case of neumann boundary conditions on a separating arc the heat kernel coefficients for the union of internal and external sectors are calculated completely in a similar line . now the frequency equations read @xmath284 the angular part of the eigenfunctions is proportional to @xmath285 , therefore the index @xmath25 takes integer values starting with zero . for the zeta function in this eigenvalue problem the representation analogous to ( [ eq3 - 58 ] ) holds @xmath286{.}\ ] ] again we use the uniform asymptotic expansion @xcite @xmath287 \simeq -\ln 2 \nu - \ln t+ \sum_{j=1}^{2}\frac{g^n_{2j}(t)}{\nu^{2j}}\,{,}\ ] ] where @xmath288 on substituting ( [ eq3 - 75 ] ) and ( [ eq3 - 76 ] ) and differentiating in equation ( [ eq3 - 74 ] ) , the zeta function under consideration assumes the form @xmath289{.}\label{eq3 - 77}\end{aligned}\ ] ] in order to take the sum over @xmath25 in ( [ eq3 - 77 ] ) we apply the formula that follows from ( [ eq3 - 64 ] ) @xmath290 sequential differentiation of equation ( [ eq3 - 78 ] ) gives @xmath291 on making use of the change of variables ( [ eq3 - 66 ] ) and ( [ eq3 - 67 ] ) in equation ( [ eq3 - 77 ] ) the zeta function @xmath292 assumes the form @xmath293 \nonumber \\ & & + \frac{r^4}{32p^8}\left [ 512\,\bar s_4({\tilde x})-6712\,{\tilde x}^2\,\bar s_5({\tilde x})+ 24362\ , { \tilde x}^4\,\bar s_6({\tilde x } ) \right . \nonumber \\ & & \left .-32928\,{\tilde x}^6\,\bar s_7({\tilde x})+14847\,{\tilde x}^8\,\bar s_8({\tilde x})\right ] { . } \label{eq3 - 80}\end{aligned}\ ] ] substitution of the function @xmath294 with @xmath295 from ( [ eq3 - 78 ] ) and ( [ eq3 - 79 ] ) gives @xmath296 comparison of the expansions ( [ eq3 - 81 ] ) and ( [ eq3 - 50 ] ) gives the following values for the heat kernel coefficients @xmath297 for neumann boundary conditions @xmath298 @xmath299 the heat kernel coefficients corresponding to the minkowski space - time contribution in the case under consideration are derived from the respective coefficients for internal sector i by putting there @xmath243 and omitting the contribution due to the curvature of the arc 1 - 2 and contributions of the rightangled corners at the points 1 and 2 ( see tables 2 and 3 ) @xmath300 by making use of the equations ( [ eq3 - 82 ] ) , ( [ eq5 - 37 ] ) and table 2 we derive the heat kernel coefficients for the external sector ii with neumann condition @xmath301 @xmath302 in the case of a cone the eigenfunctions are defined in ( [ eq3 - 34 ] ) with @xmath103 being the roots of the frequency equations ( [ eq3 - 55 ] ) , ( [ eq3 - 56 ] ) and ( [ eq3 - 72 ] ) , ( [ eq3 - 73 ] ) where @xmath303 . for both dirichlet and neumann conditions on the circle 12 the index @xmath25 ranges from @xmath110 . the state degeneracy in both the cases is determined by equation ( [ eq3 - 37 ] ) . all this implies that in order to proceed to a cone one should put in the relevant formulas for a wedge @xmath303 and use for summation over @xmath25 the following relations ( instead of equations ( [ eq3 - 65 ] ) , ( [ eq3 - 78 ] ) , and ( [ eq3 - 79 ] ) ) @xmath304 it is essential that these summation formulas should be employed for both dirichlet and neumann boundary conditions given on the circle 12 and separating the internal and external sectors of a cone . first we consider dirichlet conditions . carrying out in ( [ eq3 - 68 ] ) the change of variables ( [ eq3 - 38 ] ) and using for summation the functions @xmath305 from equation([eq3 - 84 ] ) we obtain the expansion for the zeta function in the spectral problem at hand @xmath306 this expansion can formally be derived from the relevant zeta function for a wedge , equation ( [ eq3 - 69 ] ) , by omitting the terms with even powers of @xmath227 . comparison of the series ( [ eq3 - 85 ] ) with equation ( [ eq3 - 50 ] ) affords the values of the first seven heat kernel coefficients @xmath297 @xmath307 the heat kernel coefficients for minkowski spacetime contribution are @xmath308 we deduce from equations ( [ eq3 - 86 ] ) , ( [ eq5 - 42 ] ) , and ( [ eq3 - 45a ] ) the heat kernel coefficients for external sector of a cone with dirichlet boundary conditions on the circle 12 @xmath309 @xmath310 by analogy with this we obtain from equation ( [ eq3 - 80 ] ) for neumann boundary conditions @xmath311 this expansion can formally be derived from the relevant zeta function for a wedge equation ( [ eq3 - 81 ] ) by omitting the terms with even powers of @xmath227 . comparing the equations([eq3 - 88 ] ) and ( [ eq3 - 50 ] ) we obtain the values of the first seven heat kernel coefficients @xmath297 @xmath312 the heat kernel coefficients for minkowski spacetime contribution in this case are @xmath313 from equations ( [ eq3 - 89 ] ) , ( [ eq5 - 46 ] ) , and ( [ eq3 - 46a ] ) we derive the heat kernel coefficients for the external sector ii of the cone with neumann boundary conditions on the circle 12 @xmath314 @xmath315 the heat kernel coefficients for external sector are given in equations ( [ eq5 - 25 ] ) , ( [ eq3 - 83 ] ) , ( [ eq3 - 87 ] ) , and ( [ eq3 - 90 ] ) . identification of the individual contributions to these coefficients due to the boundary nonsmoothness does not differ basically from the analogous procedure for internal sector carried out in section 4 . the sole point that should be noted here is the following . the area @xmath184 and the perimeter @xmath29 are in this case infinite . therefore they should be treated as the limit , when @xmath316 tends to infinity , of the expressions @xmath317 the coefficients @xmath129 and @xmath98 are defined , as before , by equations ( [ eq4-a ] ) , ( [ eq4-b ] ) , and ( [ eq4 - 13a ] ) with @xmath184 and @xmath29 defined in ( [ eq6 - 1 ] ) . contributions to the rest of the heat kerenel coefficients , which are proportional to the angle @xmath17 are due to the curvature of the arc 12 and the contributions which are independent of @xmath17 are due to the rght - angled corners at the points 1 and 2 . the results of this analysis are represented in table 3 together with the heat kernel coefficients for the internal sector i. analysis of this table enables one to reveal some general rules for corner contributions . it is interesting to note that the right - angled corners at the points 1 and 2 give the contributions to the coefficients @xmath12 and @xmath13 which have opposite signs for internal and external sectors , it being valid for all configurations and boundary conditions considered . such a behaviour seems to be related with convexity ( internal sector ) or concavity ( external sector ) of the arc 12 . it is analogous to the contributions of the smooth segments of the boundary to the heat kernel coefficients @xmath14 and @xmath15 . the boundary discontinuities due to the corner at the origin @xmath189 and to the corners at the points 1 and 2 contribute to the heat kernel coefficients basically in different ways . corner at the origin gives contribution only to the coefficient @xmath14 ( even when @xmath199 ) . the corners at the points 1 and 2 contribute to all the heat kernel coefficients starting with @xmath318 . obviously the reason of this distinction is a geometrical one , the corner at the origin is formed by crossing two straight lines , while the corners at the points 1 and 2 are the result of intersection of lines one of which has a nonzero curvature . this general assertion concerning the corner contribution to the heat kernel coefficients can be illustrated by a known expression for the heat kernel expansion for a rectangle with sides @xmath319 and @xmath320 @xmath321 = \left ( \frac{a}{\sqrt{4\pi t}}-\frac{1}{2 } \right ) \left ( \frac{b}{\sqrt{4\pi t}}-\frac{1}{2 } \right ) + \mbox{es } \nonumber \\ \fl & = & \frac{ab}{4\pi t } -\frac{a+b}{4\sqrt{\pi t}}+\frac{1}{4}+ \mbox{es}\,{,}\end{aligned}\ ] ] where es denotes the exponentially small corrections as @xmath41 ( see , for example , @xcite ) . here the scalar operator @xmath100 with dirichlet boundary conditions is considered . for a rectangle the coefficient @xmath14 is obviously equal to the contributions of four right - angled corners . indeed , the third term in the expansion ( [ eq4 - 14 ] ) can be represented in the form @xmath322 where @xmath194 is given in equation ( [ eq4 - 5 ] ) . besides these corners the boundary of a rectangle has no other singularities , therefore the heat kernel coefficients @xmath197 with @xmath198 vanish in this problem . these rules for obtaining the heat kernel coefficients are directly generalized to an arbitrary polygon with the angles @xmath323 . the first two coefficients @xmath129 and @xmath98 are defined by equations ( [ eq4-a ] ) and ( [ eq4-b ] ) , respectively , where @xmath184 is the area of the polygon and @xmath29 is its perimeter . the third coefficient @xmath14 is equal to the sum of the contributions due to the angles @xmath323 @xmath324 the rest of the coefficients @xmath325 vanish . in particular , it implies that the zeta function technique should provide a finite value of the casimir energy for a polygon on a plane ( @xmath326 ) and for a cylindrical generalization of the polygon spectral problem ( @xmath327 ) . these subjects have been discussed earlier in papers @xcite . in reference @xcite the vacuum energy of massless fields including electromagnetic field was calculated for the boundary configuration shown in figure 1 with @xmath38 . both versions of this boundary value problem were considered , three - dimensional one ( a semi - circular infinite cylinder ) and two - dimensional spectral problem on the plane . in both the cases the zeta regularization did nt give a finite value of the casimir energy . as known @xcite , the reason of this is nonzero heat kernel coefficients @xmath15 for @xmath49 and @xmath12 for @xmath255 . using the results of the present paper we can elucidate the geometrical origin of this fact . let us consider electromagnetic filed in internal and external sectors of the wedge with @xmath182 . for the boundaries with cylindrical symmetry electromagnetic field reduces to two massless fields subjected to dirichlet and neumann boundary conditions @xcite . from table 3 it follows that @xmath328 the nonzero value of the coefficient @xmath15 is due to the contribution of four right - angled corners at the points 1 and 2 . it has been noted at first time in reference @xcite . in the case of the coefficient @xmath12 the contributions of the corners from internal and external sectors are mutually canceled , while the contributions of the curvature of the arc 1 2 from internal and external sectors are added . different geometrical origins of the zeta function failure to provide a finite value of the vacuum energy in the two- and three - dimensional versions of the boundary value problem in question probably imply the impossibility of obtaining a finite and unique value of this quantity by taking advantage of the atomic structure of the boundary @xcite or its quantum fluctuations @xcite . it is clear because any physical reason of the casimir energy divergences should be valid simultaneously in the two- and three - dimensional versions of the boundary configuration under consideration . the basic result of the paper is the calculation of the individual contributions to the heat kernel coefficients generated by such particularities of the boundary as the corners . in the course of this analysis certain patterns , that are followed by these contributions , have been revealed . as a by product , the rules for obtaining all the heat kernel coefficients for the minus laplace operator defined on a polygon or in its cylindrical generalization are formulated , these rules being valid both for dirichlet and neumann boundary conditions . implications of the obtained results in calculations of the vacuum energy for regions with nonsmooth boundary are discussed . our calculations comport the conventional point of view according to which the heat kernel coefficients are determined , in the case under consideration , by the local properties of the boundary . in any case , the heat kernel coefficients obtained in the present paper can be used for verification of the general methods of calculating the contributions of boundary discontinuities to the heat kernel coefficients which may be developed in the future . these general methods must in particular allow one to calculate the contribution of an arbitrary corner to the coefficients @xmath197 with @xmath198 in terms of the jumps of the derivatives of the boundary curve ( or its geometrical invariants ) at this corner , i.e. , the formulas analogous to equation ( [ eq4 - 5 ] ) should be found for @xmath197 with @xmath198 . we are greatly indebted to one of the referee who found out a substantial error in our consideration of the external sector in the first version of the paper . this study was started during the visit of vvn to nuclear physics institute of academy of sciences of the czech republic in e near prague . the work was supported by the votruba - blokhintsev program , russian foundation for basic research ( grant no . 00 - 01 - 00300 ) , intas ( project no.578 ) , and grant agency of academy of sciences of the czech republic ( grant no . iaa1048101 ) . vvn is grateful to d v fursaev for elucidating discussion of the cone singularity and its physical implication . 99 elizalde e , odintsov s d , romeo a , bytsenko a a and zerbini z 1994 _ zeta regularization technique with applications _ ( singapore : world scientific ) elizalde e 1995 _ ten physical applications of spectral zeta functions _ ( berlin : springer ) bordag m , mohideen u and mostepanenko v m 2001 _ phys . rep . _ * 353 * 1 dowker j s and kennedy g 1978 895 bordag m , nesterenko v v and pirozhenko i g 2002 _ phys . _ d * 65 * , 045011 kirsten k 2001 _ spectral functions in mathematics and physics _ ( boca raton fl : chapman & hall / crc ) stewartson k and waechter r t 1971 _ proc . soc . _ * 69 * 353 kennedy g 1978 _ j. phys . _ a * 11 * l173 elizalde e and romeo a 1990 _ int . j. mod . phys . _ a * 5 * 1653 gilkey p b 1995 _ invariance theory , the heat equation and the atiyah - singer index theorem _ ( boca raton fl : chapman & hall / crc ) branson t p , gilkey p b , kirsten k and vassilevich d v 1999 _ nucl . _ b * 563 * 603 apps j s and dowker j s 1995 _ class . quantum grav . _ * 12 * 1363 apps j s and dowker j s 1998 _ class . quantum grav . _ * 15 * 1121 brevik i and lygren 1996 157 brevik i and pettersen k 2001 267 nesterenko v v , lambiase g and scarpetta g 2002 403 sommerfeld a 1897 _ proc . london math . _ * 28 * 417 carslaw h s 1898 _ proc . london math . * 20 * 121 fedosov b v 1963 _ sov . mat . dokl . _ * 4 * 1092 fedosov b v 1964 _ sov dokl . _ * 5 * 988 dowker j s 1977 115 dowker j s 1978 d * 18 * 1856 cheeger j 1983 _ j. diff . geom . _ * 18 * 575 howls c j and trasler s a 1998 1911 cognola g , kirsten k and vanzo l 1994 d * 49 * 1029 zerbini s , cognola g and vanzo l 1996 _ phys . d _ * 54 * 2699 fursaev d v 1994 _ phys . _ b * 334 * , 53 fursaev d v 1994 1431 fursaev d v 1997 , 1059 frolov v p and serebriany e m 1987 d * 35 * 3779 dowker j s 2002 _ preprint _ hep - th/0006138 kac m 1966 _ amer . math . monthly _ * 73 * no . 4 part ii 1 mckean jr . h p and singer i m 1967 _ j. diff . . _ * 1 * 43 pleijel 1954 _ ark . mat . _ * 2 * 553 waechter r t 1972 _ proc . _ * 72 * 439 bordag m , elizalde e and kirsten k 1996 _ j. math . * 37 * 895 bordag m , kirsten k and dowker j s 1996 _ commun . phys . _ * 182 * 371 apps j s , bordag b , dowker j s and kirsten k 1996 _ class . quantum grav . _ * 13 * 2911 bordag m , kirsten k and vassilevich d v 1999 d * 59 * 085011 . m and pirozhenko i g 2001 d * 64 * , 025019 lambiase g , nesterenko v v and bordag m 1999 6254 moss i g 1989 759 moss i g and poletti s j 1994 _ phys . _ b * 333 * 326 death p d and esposito g v m 1991 d * 43 * 3234 death p d and esposito g v m 1991 d * 44 * 1713 esposito g 1994 _ quantum gravity , quantum cosmology and lorentzian geometries _ ( berlin : springer ) blank j , exner p and havlak m 1994 _ hilbert space operators in quantum physics _ ( new york : aip press , american institute of physics ) reed m and simon b 1978 _ methods of modern mathematical physics , _ vol 1 , functional analysis , vol 4 , analysis of operators ( new york : academic press ) . berezin f a and shubin m a 1983 _ schrdinger equation _ ( moscow : moscow state university press ) [ in russian ] 1972 ed m abramowitz and i stegun ( new york : dover publications , inc . ) olver f w j 1954 _ phil . royal soc . london _ a * 247 * 328 gradshteyn i s and ryzhik i m 1994 _ table of integrals , series and products _ , fifth edition ( new york : newacademic press ) cognola g , elizalde e and kirsten k 2001 _ j. phys . a _ : math . gen . * 34 * 7311 kennedy g , critchley r and dowker j s 1980 346 bordag m , elizalde e , kirsten k and leseduarte s , 1997 d * 56 * 4896 watson g n 1944 _ theory of bessel functions _ ( cambridge : cambridge university press ) brevik i and yousef t a 2000 5819 dowker j s 1994 557 dowker j s 1994 _ commun . _ * 162 * 633 nesterenko v v , lambiase g and scarpetta g 2001 1974 borgnis e and papas c h 1958 `` electromagnetic waveguides and resonators '' in _ encyclopedia of physics _ vol 16 ed s flge ( berlin : springer verlag ) p 285 stratton j a 1941 _ electromagnetic theory _ ( new york : mcgraw - hill ) candelas p 1982 241candelas p 1986 257 marachevsky v n 2002 _ int . a _ * 17 * 786 rezaeian a h and saharian a a 2002 3625 ford l h and svaiter n f 1998 _ phys _ d * 58 * 065007 @ccccccc & @xmath90&@xmath329&@xmath330&@xmath331&@xmath332&@xmath333 + + @xmath334&@xmath335 & & & & + + @xmath336&@xmath337&@xmath338 & & & & + + @xmath339&@xmath340 & @xmath341&@xmath342 & & & + + @xmath343&&&@xmath344 & @xmath345 & & + + @xmath346&&&&@xmath347 & @xmath348 & + + @xmath349&&&&&@xmath350 & @xmath351 + @cccccccc & @xmath352 & @xmath353&@xmath354&@xmath355&@xmath356&@xmath357&@xmath358 + + @xmath334&&@xmath335 & & & & + + @xmath359&@xmath360&@xmath337&@xmath361 & & & & + + @xmath339&@xmath362&@xmath363 & @xmath364&@xmath365 & & & + + @xmath343&@xmath366&&&@xmath367 & @xmath368 & & + + @xmath346&@xmath369&&&&@xmath370 & @xmath371 & + + @xmath349&@xmath372&&&&&@xmath373 & @xmath374 + @ccccc & curvature & right - angled corners & corner of angle @xmath17 + & & of the arc 12 & at the points 1 and 2 & at the origin + + @xmath375 & @xmath376 & @xmath377&@xmath378&@xmath379 + + & @xmath380&@xmath381&@xmath378&@xmath379 + + & @xmath382&@xmath381 & & @xmath383 + + & @xmath384 & @xmath385&&@xmath386 + @xmath387 & @xmath376 & @xmath388&@xmath389 & + + & @xmath380 & @xmath390&@xmath391 & + + & @xmath382&@xmath388 & & + + & @xmath384&@xmath390 & + @xmath392 & @xmath376 & @xmath393 & @xmath394 & + + & @xmath380 & @xmath395&@xmath396 & + + & @xmath382 & @xmath397 & & + + & @xmath384 & @xmath398 & & + @xmath399 & @xmath376 & @xmath400&@xmath401 & + + & @xmath380&@xmath402&@xmath403 & + + & @xmath382 & @xmath404 & & + + & @xmath384 & @xmath405 & & + |
deep learning is currently the subject of intensive study . however , fundamental concepts such as representations are not formally defined researchers `` know them when they see them '' and there is no common language for describing and analyzing algorithms . this essay proposes an abstract framework that identifies the essential features of current practice and may provide a foundation for future developments . the backbone of almost all deep learning algorithms is backpropagation , which is simply a gradient computation distributed over a neural network . the main ingredients of the framework are thus , unsurprisingly : ( i ) game theory , to formalize distributed optimization ; and ( ii ) communication protocols , to track the flow of zeroth and first - order information . the framework allows natural definitions of semantics ( as the meaning encoded in functions ) , representations ( as functions whose semantics is chosen to optimized a criterion ) and grammars ( as communication protocols equipped with first - order convergence guarantees ) . much of the essay is spent discussing examples taken from the literature . the ultimate aim is to develop a graphical language for describing the structure of deep learning algorithms that backgrounds the details of the optimization procedure and foregrounds how the components interact . inspiration is taken from probabilistic graphical models and factor graphs , which capture the essential structural features of multivariate distributions . | 1509.08627 | deep learning has achieved remarkable successes in object and voice recognition , machine translation , reinforcement learning and other tasks @xcite . from a practical standpoint the problem of supervised learning is well - understood and has largely been solved at least in the regime where both labeled data and computational power are abundant . the workhorse underlying most deep learning algorithms is error backpropagation @xcite , which is simply gradient descent distributed across a neural network via the chain rule . gradient descent and its variants are well - understood when applied to convex or nearly convex objectives @xcite . in particular , they have strong performance guarantees in the stochastic and adversarial settings @xcite . the reasons for the success of gradient descent in non - convex settings are less clear , although recent work has provided evidence that most local minima are good enough @xcite ; that modern convolutional networks are close enough to convex for many results on rates of convergence apply @xcite ; and that the rate of convergence of gradient - descent can control generalization performance , even in nonconvex settings @xcite . taking a step back , gradient - based optimization provides a well - established set of computational primitives @xcite , with theoretical backing in simple cases and empirical backing in others . first - order optimization thus falls in broadly the same category as computing an eigenvector or inverting a matrix : given sufficient data and computational resources , we have algorithms that reliably find good enough solutions for a wide range of problems . this essay proposes to abstract out the optimization algorithms used for weight updates and focus on how the components of deep learning algorithms interact . treating optimization as a computational primitive encourages a shift from low - level algorithm design to higher - level mechanism design : we can shift attention to designing architectures that are guaranteed to learn distributed representations suited to specific objectives . the goal is to introduce a language at a level of abstraction where designers can focus on formal specifications ( grammars ) that specify how plug - and - play optimization modules combine into larger learning systems . let us recall how representation learning is commonly understood . et al _ describe representation learning as `` learning transformations of the data that make it easier to extract useful information when building classifiers or other predictors '' @xcite . more specifically , `` a deep learning algorithm is a particular kind of representation learning procedure that discovers multiple levels of representation , with higher - level features representing more abstract aspects of the data '' @xcite . finally , lecun _ et al _ state that multiple levels of representations are obtained `` by composing simple but non - linear modules that each transform the representation at one level ( starting with the raw input ) into a representation at a higher , slightly more abstract level . with the composition of enough such transformations , very complex functions can be learned . for classification tasks , higher layers of representation amplify aspects of the input that are important for discrimination and suppress irrelevant variations '' @xcite . the quotes describe the operation of a successful deep learning algorithm . what is lacking is a characterization of what makes a deep learning algorithm work in the first place . what properties must an algorithm have to learn layered representations ? what does it mean for the representation learned by one layer to be useful to another ? what , exactly , is a representation ? in practice , almost all deep learning algorithms rely on error backpropagation to `` align '' the representations learned by different layers of a network . this suggests that the answers to the above questions are tightly bound up in first - order ( that is , gradient - based ) optimization methods . it is therefore unsurprisingly that the bulk of the paper is concerned with tracking the flow of first - order information . the framework is intended to facilitate the design of more general first - order algorithms than backpropagation . * * to get started , we need a theory of the meaning or semantics encoded in neural networks . since there is nothing special about neural networks , the approach taken is inclusive and minimalistic . definition [ d : meaning ] states that the meaning of _ any _ function is how it implicitly categorizes inputs by assigning them to outputs . the next step is to characterize those functions whose semantics encode knowledge , and for this we turn to optimization @xcite . * * nemirovski and yudin developed the black - box computational model to analyze the computational complexity of first - order optimization methods @xcite . the black - box model is a more abstract view on optimization than the turing machine model : it specifies a _ communication protocol _ that tracks how often an algorithm makes _ queries _ about the objective . it is useful to refine nemirovski and yudin s terminology by distinguishing between black - boxes , which _ respond _ with zeroth - order information ( the value of a function at the query - point ) , and gray - boxes , which respond with zeroth- and first - order information ( the gradient or subgradient ) . with these preliminaries in hand , definition [ d : foo ] proposes that a _ representation _ is a function that is a _ local _ solution to an optimization problem . since we do not restrict to convex problems , finding global solutions is not feasible . indeed , recent experience shows that global solutions are often not necessary practice @xcite . the local solution has similar semantics to that is , it represents the ideal solution . the ideal solution usually can not be found : due to computational limitations , since the problem is nonconvex , because we only have access to a finite sample from an unknown distribution , etc . to see how definition [ d : foo ] connects with representation learning as commonly understood , it is necessary to take a detour through distributed optimization and game theory . game theory provides tools for analyzing distributed optimization problems where a set of players aim to minimizes losses that depend not only on their actions , but also the actions of all other players in the game @xcite . game theory has traditionally focused on convex losses since they are more theoretically amenable . here , the only restriction imposed on losses is that they are differentiable almost everywhere . allowing nonconvex losses means that error - backpropagation can be reformulated as a game . interestingly , there is enormous freedom in choosing the players . they can correspond to individual units , layers , entire neural networks , and a variety of other , intermediate choices . an advantage of the game - theoretic formulation is thus that it applies at many different scales . nonconvex losses and local optima are essential to developing a _ scale - free _ formalism . even when it turns out that particular units or a particular layer of a neural network are solving a convex problem , convexity is destroyed as soon as those units or layers are combined to form larger learning systems . convexity is not a property that is preserved in general when units are combined into layers or layers into networks . it is therefore convenient to introduce the computational primitive @xmath0 to denote the output of a first - order optimization procedure , see definition [ d : foo ] . * * a potential criticism is that the formulation is too broad . very little can be said about nonconvex optimization in general ; introducing games where many players jointly optimize a set of arbitary nonconvex functions only compounds the problem . additional structure is required . a successful case study can be found in @xcite , which presents a detailed game - theoretic analysis of rectifier neural networks . the key to the analysis is that rectifier units are almost convex . the main result is that the rate of convergence of a neural network to a local optimum is controlled by the ( waking-)regret of the algorithms applied to compute weight updates in the network . whereas @xcite relied heavily on specific properties of rectifer nonlinearities , this paper considers a wide - range of deep learning architectures . nevertheless , it is possible to carve out an interesting subclass of nonconvex games by identifying the composition of simple functions as an essential feature common to deep learning architectures . compositionality is formalized via distributed communication protocols and grammars . * * neural networks are constructed by composing a series of elementary operations . the resulting feedforward computation is captured via as a computation graph @xcite . backpropagation traverses the graph in reverse and recursively computes the gradient with respect to the parameters at each node . section [ sec : comp ] maps the feedforward and feedback computations onto the queries and responses that arise in nemirovski and yudin s model of optimization . however , queries and responses are now highly structured . in the query phase , players feed parameters into a computation graph ( the query graph @xmath1 ) that performs the feedforward sweep . in the response phase , oracles reveal first - order information that is fed into a second computation graph ( the response graph @xmath2 ) . in most cases the response graph simply implements backpropagation . however , there are examples where it does not . three are highlighted here , see section [ sec : advers ] , and especially sections [ sec : pg ] and [ sec : kb ] . other algorithms where the response graphs do not simply implement backprop include difference target propagation @xcite and feedback alignment @xcite ( both discussed briefly in section [ sec : kb ] ) and truncated backpropagation through time @xcite , where a choice is made about where to cut backprop short . examples where the query and response graph differ are of particular interest , since they point towards more general classes of deep learning algorithms . a _ distributed communication protocol _ is a game with additional structure : the query and response graphs , see definition [ d : dcp ] . the graphs capture the compositional structure of the functions learned by a neural network and the compositional structure of the learning procedure respectively . it is important for our purposes that ( i ) the feedforward and feedback sweeps correspond to two distinct graphs and ( ii ) the communication protocol is kept distinct from the optimization procedure . that is , the communication protocol specifies how information flows through the networks without specifying how players make use of it . players can be treated as plug - and - play rational agents that are provided with carefully constructed and coordinated first - order information to optimize as they see fit @xcite . finally , a _ grammar _ is a distributed communication protocol equipped with a guarantee that the response graph encodes sufficient information for the players to jointly find a local optimum of an objective function . the paradigmatic example of a grammar is backpropagation . a grammar is a thus a game designed to perform a task . a representation learned by one ( p)layer is useful to another if the game is guaranteed to converge on a local solution to an objective that is , if the players interact though a grammar . it follows that the players build representations that jointly encode knowledge about the task . * * what follows is provisional . the definitions are a first attempt to capture an interesting , and perhaps useful , perspective on deep learning . the essay contains no new theorems , algorithms or experiments , see @xcite for `` real work '' based on the ideas presented here . the essay is not intended to be comprehensive . many details are left out and many important aspects are not covered : most notably , probabilistic and bayesian formulations , and various methods for unsupervised pre - training . * * in line with its provisional nature , much of the essay is spent applying the framework to worked examples : error backpropagation as a supervised model @xcite ; variational autoencoders @xcite and generative adversarial networks @xcite for unsupervised learning ; the deviator - actor - critic ( dac ) model for deep reinforcement learning @xcite ; and kickback , a biologically plausible variant of backpropagation @xcite . the examples were chosen , in part , to maximize variety and , in part , based on familiarity . the discussions are short ; the interested reader is encouraged to consult the original papers to fill in the gaps . the last two examples are particularly interesting since their response graphs differ substantially from backpropagation . the dac model constructs a zeroth - order black - box to estimate gradients rather than querying a first - order gray - box . kickback prunes backprop s response graph by replacing most of its gray - boxes with black - boxes and approximating the chain rule with ( primarily ) local computations . bottou and gallinari proposed to decompose neural networks into cooperating modules @xcite . decomposing more general algorithms or models into collections of interacting agents dates back to the shrieking demons that comprised selfridge s pandemonium @xcite and a long line of related work @xcite . the focus on components of neural networks as players , or rational agents , in their own right developed here derives from work aimed at modeling biological neurons game - theoretically , see @xcite . a related approach to semantics based on general value functions can be found in sutton _ et al _ @xcite , see remark [ rem : sutton ] . computation graphs as applied to backprop are the basis of the python library theano @xcite and provide the backbone for automatic / algorithmic differentiation @xcite . grammars are a technical term in the theory of formal languages relating to the chomsky hierarchy @xcite . there is no apparent relation between that notion of grammar and the one presented here , aside from both relating to structural rules governing composition . formal languages and deep learning are sufficiently disparate fields that there is little risk of terminological confusion . similarly , the notion of semantics introduced here is distinct from semantics in the theory of programming languages . although game theory was originally developed to model human interactions @xcite , it has been pointed out that it may be more directly applicable to interacting populations of algorithms , so - called _ machina economicus _ @xcite . this paper goes one step further to propose that games played over first - order communication protocols are a key component of the foundations of deep learning . a source of inspiration for the essay is bayesian networks and markov random fields . probabilistic graphical models and factor graphs provide simple , powerful ways to encode a multivariate distribution s independencies into a diagram @xcite . they have greatly facilitated the design and analysis of probabilistic algorithms . however , there is no comparable framework for distributed optimization and deep learning . the essay is intended as a first step in this direction . this section defines semantics and representations . in short , the semantics of a function is how it categorizes its inputs ; a function is a representation if it is selected to optimize an objective . the connection between the definition of representation below and `` representation learning '' is clarified in section [ sec : ebp ] . possible world semantics was introduced by lewis to formalize the meaning of sentences in terms of counterfactuals @xcite . let @xmath3 be a proposition about the world . its truth depends on its content and the state of the world . rather than allowing the state of the world to vary , it is convenient to introduce the set @xmath4 of all possible worlds . let us denote proposition @xmath3 applied in world @xmath5 by @xmath6 . the meaning of @xmath3 is then the mapping @xmath7 which assigns 1 or 0 to each @xmath5 according to whether or not proposition @xmath6 is true . equivalently , the meaning of the proposition is the ordered pair consisting of : all worlds , and the subset of worlds where it is true : @xmath8 for example , the meaning of @xmath9``__that _ _ is blue '' is the subset @xmath10 of possible worlds where i am pointing at a blue object . the concept of blue is rendered explicit in an exhaustive list of possible examples . a simple extension of possible world semantics from propositions to arbitrary functions is as follows @xcite : [ d : meaning]@xmath11 + given function @xmath12 , the * semantics * or * meaning * of output @xmath13 is the ordered pair of sets @xmath14 functions implicitly categorize inputs by assigning outputs to them ; the meaning of an output is the category . whereas propositions are true or false , the output of a function is neither . however , if two functions both optimize a criterion , then one can refer to how _ accurately _ one function _ represents _ the other . before we can define representations we therefore need to take a quick detour through optimization : [ d : opt]@xmath11 + an * optimization problem * is a pair @xmath15 consisting in parameter - space @xmath16 and objective @xmath17 that is differentiable almost everywhere . the * solution * to the global optimization problem is @xmath18 which is either a maximum or minimum according to the nature of the objective . the solution may not be unique ; it also may not exist unless further restrictions are imposed . such details are ignored here . next recall the black - box optimization framework introduced by nemirovski and yudin @xcite . [ d : protocol]@xmath11 + a * communication protocol * for optimizing an unknown objective @xmath17 consists in a user ( or player ) and an oracle . on each round , user presents a * query * @xmath19 . oracle can * respond * in one of two ways , depending on the nature of the protocol : * _ black - box ( zeroth - order ) protocol . _ + oracle responds with the value @xmath20 . + + at ( 0,0 ) ( player ) * player * ; at ( 3,0 ) ( block ) @xmath21 ; ( player ) edge node[above ] @xmath22 ( block ) ; at ( 8,0 ) ( player_a ) * player * ; at ( 11,0 ) ( oracle ) @xmath21 ; ( oracle ) edge node[above ] @xmath20 ( player_a ) ; + + * _ gray - box ( first - order ) protocol . _ + oracle responds with either the gradient @xmath23 or with the gradient together with the value . + + at ( 0,0 ) ( player ) * player * ; at ( 3,0 ) ( block ) @xmath21 ; ( player ) edge node[above ] @xmath22 ( block ) ; at ( 8,0 ) ( player_a ) * player * ; at ( 11,0 ) ( oracle ) * oracle@xmath24 * ; ( oracle ) edge node[above ] @xmath25 ( player_a ) ; the protocol specifies how player and oracle interact without specifying the algorithm used by player to decide which points to query . the next section introduces _ distributed communication protocols _ as a general framework that includes a variety of deep learning architectures as special cases again without specifying the precise algorithms used to perform weight updates . unlike @xcite we do not restrict to convex problems . finding a global optimum is not always feasible , and in practice often unnecessary . [ d : foo]@xmath11 + let @xmath26 be a function space and @xmath27 be a map from parameter - space to functions . further suppose that objective function @xmath28 is given . a * representation * is a local solution to the optimization problem @xmath29 corresponding to a _ local _ maximum or minimum according to whether the objective is minimized or maximized . intuitively , the objective quantifies the extent to which functions in @xmath30 categorize their inputs similarly . the operation @xmath0 applies a first - order method to find a function whose semantics resembles the optimal solution @xmath31 where @xmath32 . in short , representations are functions with useful semantics , where usefulness is quantifed using a specific objective : the lower the loss or higher the reward associated with a function , the more useful it is . the relation between definition [ d : foo ] and representations as commonly understood in the deep learning literature is discussed in section [ sec : ebp ] below . [ rem : sutton]@xmath11 + in related work , sutton _ et al _ @xcite proposed that semantics i.e. knowledge about the world can be encoded in general value functions that provide answers to specific questions about expected rewards . definition [ d : meaning ] is more general than their approach since it associates a semantics to _ any _ function . however , the function must arise from optimizing an objective for its semantics to accurately represent a phenomenon of interest . the main example of a representation arises under supervised learning . [ rep : sup]@xmath11 + let @xmath33 and @xmath34 be an input space and a set of labels and @xmath35 be a loss function . suppose that @xmath36 is a parametrized family of functions . * _ nature _ which samples labeled pairs @xmath37 i.i.d . from distribution @xmath38 , singly or in batches . * _ predictor _ chooses parameters @xmath19 . * _ objective _ is @xmath39.\ ] ] the query and responses phases can be depicted graphically as + at ( 0,0 ) ( player ) * predictor * ; at ( 5,0 ) ( block ) @xmath40}$ ] ; ( player ) edge node[above ] @xmath22 ( block ) ; at ( 10,0 ) ( player_a ) * predictor * ; at ( 15,0 ) ( oracle ) * oracle@xmath24 * ; ( oracle ) edge node[above ] @xmath25 ( player_a ) ; + the predictor @xmath41 is then a representation of the optimal predictor @xmath42 . a commonly used mapping from parameters to functions is @xmath43 where a feature map @xmath44 is fixed . the setup admits a variety of complications in practice . firstly , it is typically infeasible even to find a local optimum . instead , a solution that is within some small @xmath45 of the local optimum suffices . secondly , the distribution @xmath38 is unknown , so the expectation is replaced by a sum over a finite sample . the quality of the resulting representation has been extensively studied in statistical learning theory @xcite . finally , it is often convenient to modify the objective , for example by incorporating a regularizer . thus , a more detailed presentation would conclude that @xmath46 yields a representation @xmath47 of the solution to @xmath48 $ ] . to keep the discussion and notation simple , we do not consider any of these important details . it is instructive to unpack the protocol , by observing that the objective @xmath21 is a composite function involving @xmath49 , @xmath50 and @xmath51 $ ] : + at ( 0,0 ) ( glabel ) * query * ; at ( 4.5,0 ) ( unknown_in ) * nature * ; at ( 0,-2 ) ( player ) * predictor * ; at ( 3,-2 ) ( block1 ) @xmath52 ; at ( 6,-2 ) ( block2 ) @xmath53 ; ( unknown_in ) edge node[left ] @xmath54 ( block1 ) ; ( player ) edge node[above ] @xmath22 ( block1 ) ; ( block1 ) edge node[left ] @xmath55 ( block2 ) ; ( unknown_in ) edge node[right ] @xmath56 ( block2 ) ; at ( 9,0 ) ( glabel ) * response * ; at ( 9,-2 ) ( player_out ) * predictor * ; at ( 15,-2 ) ( oracle_l ) * oracle@xmath57 * ; at ( 12,0 ) ( oracle_f ) * oracle@xmath58 * ; at ( 12,-2 ) ( block ) @xmath59 ; ( oracle_l ) edge node[above ] @xmath60 ( block ) ; ( oracle_f ) edge node[right ] @xmath61 ( block ) ; ( block ) edge node[below ] @xmath62 ( player_out ) ; ( block ) edge node[above ] @xmath63 ( player_out ) ; the notation @xmath63 is borrowed from backpropagation . it is shorthand for the derivative of the objective with respect to parameters @xmath22 . nature is not a deterministic black - box since it is not queried directly : nature produces @xmath37 pairs stochastically , rather than in response to specific inputs . our notion of black - box can be extended to stochastic black - boxes , see e.g. @xcite . however , once again we prefer to keep the exposition as simple as possible . the second example concerns fitting a probabilistic or generative model to data . a natural approach is to find the distribution under which the observed data is most likely : @xmath11 + let @xmath33 be a data space . * _ nature _ samples points from distribution @xmath64 . * _ estimator _ chooses parameters @xmath19 . * _ operator _ @xmath65 computes a probability density on @xmath33 that depends on parameter @xmath66 . * _ operator _ @xmath67 acts as a loss . the objective is to mimimize @xmath68.\ ] ] at ( 0,0 ) ( player ) * estimator * ; at ( 5,0 ) ( block ) @xmath69}$ ] ; ( player ) edge node[above ] @xmath22 ( block ) ; at ( 10,0 ) ( player_a ) * estimator * ; at ( 15,0 ) ( oracle ) * oracle@xmath24 * ; ( oracle ) edge node[above ] @xmath25 ( player_a ) ; + the estimate @xmath70 , where @xmath71 , is a representation of the optimal solution , and can also be considered a representation of @xmath64 . the setup extends easily to maximum _ a posteriori _ estimation . as for supervised learning , the protocol can be unpacked by observing that the objective has a compositional structure : + at ( 0,0 ) ( glabel ) * query * ; at ( 3,0 ) ( unknown_in ) * nature * ; at ( 0,-2 ) ( player ) * estimator * ; at ( 3,-2 ) ( block1 ) @xmath72 ; at ( 6,-2 ) ( block2 ) @xmath73 ; ( unknown_in ) edge node[left ] @xmath54 ( block1 ) ; ( player ) edge node[above ] @xmath66 ( block1 ) ; ( block1 ) edge node[left ] @xmath55 ( block2 ) ; at ( 9,0 ) ( glabel ) * response * ; at ( 9,-2 ) ( player_out ) * estimator * ; at ( 15,-2 ) ( oracle_l ) * oracle@xmath74 * ; at ( 12,0 ) ( oracle_f ) * oracle@xmath75 * ; at ( 12,-2 ) ( block ) @xmath59 ; ( oracle_l ) edge node[above ] @xmath76 ( block ) ; ( oracle_l ) edge node[below ] @xmath77 ( block ) ; ( oracle_f ) edge node[left ] @xmath78 ( block ) ; ( block ) edge node[below ] @xmath79 ( player_out ) ; ( block ) edge node[above ] @xmath80 ( player_out ) ; the third example is taken from reinforcement learning @xcite . we will return to reinforcement learning in section [ sec : pg ] , so the example is presented in some detail . in reinforcement learning , an agent interacts with its environment , which is often modeled as a markov decision process consisting of state space @xmath81 , action space @xmath82 , initial distribution @xmath83 on states , stationary transition distribution @xmath84 and reward function @xmath85 . the agent chooses actions based on a _ policy _ : a function @xmath86 from states to actions . the goal is to find the optimal policy . actor - critic methods break up the problem into two pieces @xcite . the critic estimates the expected value of state - action pairs given the current policy , and the actor attempts to find the optimal policy using the estimates provided by the critic . the critic is typically trained via temporal difference methods @xcite . let @xmath87 denote the distribution on states @xmath88 at time @xmath89 given policy @xmath90 and initial state @xmath91 at @xmath92 and let @xmath93 . let @xmath94 be the discounted future reward . define the value of a state - action pair as @xmath95.\ ] ] unfortunately , the value - function @xmath96 can not be queried . instead , temporal difference methods take a bootstrapped approach by minimizing the bellman error : @xmath97\ ] ] where @xmath88 is the state subsequent to @xmath91 . @xmath11 + critic interacts with black - boxes actor and nature . * _ critic _ plays parameters @xmath98 . * _ operator _ @xmath99 and @xmath100 estimates the value function and compute the bellman error . in practice , it turns out to _ clone _ the value - estimate periodically and compute a slightly modified bellman error : @xmath101\ ] ] where @xmath102 is the cloned estimate . cloning improves the stability of td - learning @xcite . a nice conceptual side - effect of cloning is that td - learning reduces to gradient descent . + + at ( 0,0 ) ( glabel ) * query * ; at ( 6,0 ) ( unknown_in ) * nature * ; at ( 3,0 ) ( actor ) * actor * ; at ( 0,-2 ) ( player ) * critic * ; at ( 3,-2 ) ( block1 ) @xmath103 ; at ( 6,-2 ) ( block2 ) @xmath104 ; ( unknown_in ) edge node[left ] @xmath91 ( block1 ) ; ( player ) edge node[above ] @xmath105 ( block1 ) ; ( block1 ) edge node[left ] @xmath55 ( block2 ) ; ( unknown_in ) edge node[right ] @xmath106 ( block2 ) ; ( actor ) edge node[left ] @xmath107 ( block1 ) ; + at ( 9,0 ) ( glabel ) * response * ; at ( 9,-2 ) ( player_out ) * critic * ; at ( 15,-2 ) ( oracle_l ) * oracle@xmath108 * ; at ( 12,0 ) ( oracle_f ) * oracle@xmath109 * ; at ( 12,-2 ) ( block ) @xmath59 ; + ( oracle_l ) edge node[above ] @xmath110 ( block ) ; ( oracle_f ) edge node[right ] @xmath111 ( block ) ; ( block ) edge node[above ] @xmath112 ( player_out ) ; + + the estimate is a representation of the true value function . @xmath11 + temporal difference learning is not strictly speaking a gradient - based method @xcite . the residual gradient method performs gradient descent on the bellman error , but suffers from double sampling @xcite . projected fixpoint methods minimize the _ projected _ bellman error via gradient descent and have nice convergence properties @xcite . an interesting recent proposal is implicit td learning @xcite , which is based on implicit gradient descent @xcite . section [ sec : pg ] presents the deviator - actor - critic model which simultaneously learns a value - function estimate and a locally optimal policy . it is often useful to decompose complex problems into simpler subtasks that can handled by specialized modules . examples include variational autoencoders , generative adversarial networks and actor - critic models . neural networks are particularly well - adapted to modular designs , since units , layers and even entire networks can easily be combined analogously to bricks of lego @xcite . however , not all configurations are viable models . a methodology is required to distinguish good designs from bad . this section provides a basic language to describe how bricks are glued together that may be a useful design tool . the idea is to extend the definitions of optimization problems , protocols and representations from section [ sec : reps ] from single to multi - player optimization problems . [ d : game]@xmath11 + a * distributed optimization problem * or * game * @xmath113,\theta , { \ensuremath{\boldsymbol\ell}})$ ] is a set @xmath114=\{1,\ldots n\}$ ] of players , a parameter space @xmath115 , and loss vector @xmath116 . player @xmath117 picks moves from @xmath118 and incurs loss determined by @xmath119 . the goal of each player is to minimize its loss , which depends on the moves of the other players . the classic example is a _ finite game _ @xcite , where player @xmath117 has a menu of @xmath120-actions and chooses a distribution over actions , @xmath121 on each round . losses are specified for individual actions , and extended linearly to distributions over actions . a natural generalization of finite games is _ convex games _ where the parameter spaces are compact convex sets and each loss @xmath122 is a convex function in its @xmath123-argument @xcite . it has been shown that players implementing no - regret algorithms are guaranteed to converge to a correlated equilibrium in convex games @xcite . computation graphs are a useful tool for calculating derivatives @xcite . for simplicity , we restrict to deterministic computation graphs . more general stochastic computation graphs are studied in @xcite . * _ query phase . _ players provide inputs to the query graph _ ( @xmath1 ) _ that operators transform into outputs . * _ response phase . _ operators in @xmath1 act as oracles in the response graph _ ( @xmath2 ) _ : they input subgradients that are transformed and communicated to the players . the protocol specifies how players and oracles communicate without specifying the optimization algorithms used by the players . the addition of a response graph allows more general computations than simply backpropagating the gradients of the query phase . the additional flexibility allows the design of new algorithms , see sections [ sec : pg ] and [ sec : kb ] below . it is also sometimes necessary for computational reasons . for example , backpropagation through time on recurrent networks typically runs over a truncated response graph @xcite . suppose that we wish to optimize an objective function @xmath124 that depends on all the moves of all the players . finding a global optimum is clearly not feasible . however , we may be able to construct a protocol such that the players are jointly able to find local optima of the objective . in such cases , we refer to the protocol as a grammar : [ d : grammar]@xmath11 + a * grammar * for objective @xmath124 is a distributed communication protocol where the response graph provides _ sufficient _ first - order information to find a local optimum of @xmath125 . the guarantee ensures that the representations constructed by players in a grammar can be combined into a coherent distributed representation . that is , it ensures that the representations constructed by the players transform data in a way that is useful for optimizing the shared objective @xmath21 . the players losses need not be explicitly computed . all that is necessary is that the response phase communicate the gradient information needed for players to locally minimize their losses and that doing so yields a local optimum of the objective . * * functions can be inserted into grammars as lego - like building blocks via function composition during queries and the chain rule during responses . let @xmath126 be a function that takes inputs @xmath22 and @xmath127 , provided by a player and by upstream computations respectively . the output of @xmath128 is communicated downstream in the query phase : + at ( 0,0 ) ( glabel ) * query * ; at ( 0,-2 ) ( unknown_in ) @xmath55 ; at ( 6,-2 ) ( unknown_out ) @xmath55 ; at ( 3,0 ) ( player ) * player * ; at ( 0,-4 ) ( unknown_x ) @xmath55 ; at ( 3,-2 ) ( block ) @xmath129 ; ( unknown_in ) edge node[above ] @xmath127 ( block ) ; ( player ) edge node[left ] @xmath22 ( block ) ; ( block ) edge node[above ] @xmath128 ( unknown_out ) ; at ( 0,0 ) ( glabel ) * response * ; at ( 0,-2 ) ( unknown_in ) @xmath55 ; at ( 6,-2 ) ( unknown_out ) @xmath55 ; at ( 3,0 ) ( player ) * player * ; at ( 3,-2 ) ( block ) @xmath59 ; at ( 3,-4 ) ( oracle ) * oracle@xmath130 * ; ( block ) edge node[left ] @xmath131 ( player ) ; ( block ) edge node[right ] @xmath132 ( player ) ; ( block ) edge node[above ] @xmath133 ( unknown_in ) ; ( block ) edge node[below ] @xmath134 ( unknown_in ) ; ( unknown_out ) edge node[above ] @xmath135 ( block ) ; ( oracle ) edge node[right ] @xmath136 ( block ) ; + the chain rule is implemented in the response phase as follows . oracle@xmath137 reports the gradient @xmath138 in the response phase . operator `` @xmath139 '' computes the products @xmath140 via matrix multiplication . the projection of the product onto the first and second components '' produce two outputs , the entire vector can be reported in both direction with the irrelevant components ignored . ] are reported to player and upstream respectively . 1 . _ exact gradients . _ + under error backpropagation the response graph implements the chain rule , which guarantees that players receive the gradients of their loss functions ; see section [ sec : ebp ] . surrogate objectives . _ + the variational autoencoder uses a surrogate objective : the variational lower bound . maximizing the surrogate is guaranteed to also maximize the true objective , which is computational intractable ; see section [ sec : vae ] . learned objectives . _ + in the case of generative adversarial network and the dac - model , some of the players learn a loss that is guaranteed to align with the true objective , which is unknown ; see sections [ sec : advers ] and [ sec : pg ] . 4 . _ estimated gradient . _ + in the dac - model and kickback , gradient estimates are substituted for the true gradient ; see sections [ sec : pg ] and [ sec : kb ] . guarantees are provided on the estimates . @xmath11 + there is considerable freedom regarding the choice of players . in the examples below , players are typically chosen to be layers or entire neural networks to keep the diagrams simple . it is worth noting that zooming in , such that players correspond to individual units , has proven to be a useful tool when analyzing neural networks @xcite . the game - theoretic formulation is thus scale - free and can be coarse- or fine - grained as required . a mathematical language for tracking the structure of hierarchical systems at different scales is provided by operads , see @xcite and the references therein , which are the natural setting to study the composition of operators that receive multiple inputs . the main example of a grammar is a neural network using error backpropagation to perform supervised learning . layers in the network can be modeled as players in a game . setting each ( p)layer s objective as the network s loss , which it minimizes using gradient ascent , yields backpropagation . [ c : chain_rule]@xmath11 + an @xmath141-layer neural network can be reformulated as a game played between @xmath142 players , corresponding to _ nature _ and the _ layers _ of the network . the query graph for a 3-layer network is : + * _ nature _ plays samples datapoints @xmath37 i.i.d . from @xmath143 and acts as the zeroth player . * _ layer@xmath144 _ plays weight matrices @xmath145 . * _ operators _ compute @xmath146 for each layer , along with loss @xmath147 . the protocol can be extended to convolutional networks by replacing the matrix multiplications performed by each operator , @xmath148 , with convolutions and adding parameterless max - pooling operators @xcite . * * we are now in a position to relate the notion of representation in definition [ d : foo ] with the standard notion of representation learning in neural networks . in the terminology of section [ sec : reps ] , each player learns a representation . the representations learned by the different players form a coherent distributed representation because they jointly optimize a single objective function . if we set @xmath153 then the function @xmath154 fits the definition of representation above . moreover , the compositional structure of the network implies that @xmath155 is composed of subrepresentations corresponding to the optimizations performed by the different players in the grammar : each function @xmath156 is a local optimum where @xmath157 is optimized to transform its inputs into a form that is useful to network as a whole . * * little can be said in general about the rate of converge of the layers in a neural network since the loss is not convex . however , neural networks can be decomposed further by treating the individual units as players . when the units are linear or rectilinear , it turns out that the network is a _ circadian game_. the circadian structure provides a way to convert results about the convergence of convex optimization methods into results about the global convergence a rectifier network to a local optimum , see @xcite . the next example extends the unsupervised setting described in section [ sec : unsupervised ] . suppose that observations @xmath158 are sampled i.i.d . from a two - step stochastic process : a latent value @xmath159 is sampled from @xmath160 , after which @xmath161 is sampled from @xmath162 . the goal is to ( i ) find the maximum likelihood estimator for the observed data and ( ii ) estimate the posterior distribution on @xmath163 conditioned on an observation @xmath164 . a straightforward approach is to maximize the marginal likelihood @xmath165 and then compute the posterior @xmath166 however , the integral in eq . is typically untractable , so a more roundabout tactic is required . the approach proposed in @xcite is to construct two neural networks , a decoder @xmath167 that learns a generative model approximating @xmath168 , and an encoder @xmath169 that learns a recognition model or posterior approximating @xmath170 . it turns out to be useful to replace the encoder with a deterministic function , @xmath171 , and a noise source , @xmath172 that are compatible . here , compatible means that sampling @xmath173 is equivalent to sampling @xmath174 and computing @xmath175 . * _ environment _ plays i.i.d . samples from @xmath176 * _ noise _ plays i.i.d . samples from @xmath172 . it also communicates its density function @xmath172 , which is analogous to a gradient and the reason that _ noise _ is gray rather than black - box . * _ encoder _ and _ decoder _ play parameters @xmath177 and @xmath22 respectively . * _ @xmath178 is a neural network that encodes samples into latent variables . * _ operator _ @xmath179 is a neural network that estimates the probability of @xmath164 conditioned on @xmath163 . * the remaining operators compute the ( negative ) variational lower bound @xmath180}_{{{\mathcal l}}_2}.\ ] ] at ( 0,0 ) ( glabel ) * response * ; at ( 2,-2 ) ( oracle1 ) * oracle@xmath130 * ; at ( 13,-2 ) ( oracle2)*oracle@xmath181 * ; at ( 2,-4 ) ( oracle4 ) * oracle@xmath182 * ; at ( 13,-4 ) ( oracle3 ) * oracle@xmath183 * ; at ( 7,-4 ) ( oracle5 ) * oracle@xmath184 * ; at ( 5,-2 ) ( block1 ) @xmath185 ; at ( 9,-2 ) ( block2 ) @xmath186 ; at ( 5,-4 ) ( block3 ) @xmath186 ; at ( 9,-4 ) ( block4 ) @xmath186 ; at ( 2,0 ) ( block5 ) @xmath186 ; at ( 5,0 ) ( player1 ) * encoder * ; at ( 9,0 ) ( player2 ) * decoder * ; ( oracle1 ) edge node[left ] @xmath187 ( block5 ) ; ( oracle2 ) edge node[above ] @xmath188 ( block2 ) ; ( block4 ) edge node[right ] @xmath189 ( block2 ) ; ( oracle3 ) edge node[above ] @xmath190 ( block4 ) ; ( oracle4 ) edge node[above ] @xmath191 ( block3 ) ; ( oracle5 ) edge node[above ] @xmath192 ( block3 ) ; ( oracle5 ) edge node[above ] @xmath192 ( block4 ) ; ( block1 ) edge node[below ] @xmath135 ( block5 ) ; ( block2 ) edge node[above ] @xmath193 ( block1 ) ; ( block3 ) edge node[left ] @xmath191 ( block1 ) ; ( block5 ) edge node[above ] @xmath194 ( player1 ) ; ( block2 ) edge node[left ] @xmath195 ( player2 ) ; ( block2 ) edge node[right ] @xmath63 ( player2 ) ; 1 . maximizing the variational lower bound yields ( i ) a maximum likelihood estimator and ( ii ) an estimate of the posterior on the latent variable @xcite . the chain rule ensures that the correct gradients are communicated to encoder and decoder . a recent approach to designing generative models is to construct an adversarial game between forger and curator @xcite . forger generates samples ; curator aims to discriminate the samples produced by forger from those produced by nature . forger aims to create samples realistic enough to fool curator . if forger plays parameters @xmath22 and curator plays @xmath177 then the game is described succinctly via @xmath196 + \operatorname*{\mathbb e}_{{\ensuremath{\boldsymbol\epsilon}}\sim{{\mathbb p}}_{noise}({\ensuremath{\boldsymbol\epsilon}})}\big[\log ( 1 - d_{{{\ensuremath{\boldsymbol{\phi}}}}}(g_{{{\ensuremath{\boldsymbol{\theta}}}}}({\ensuremath{\boldsymbol\epsilon}})))\big ] \right],\ ] ] where @xmath197 is a neural network that converts noise in samples and @xmath198 classifies samples as fake or not . * _ environment _ samples images i.i.d . from @xmath176 . * _ noise _ samples i.i.d . from @xmath199 . * _ forger _ and _ curator _ play parameters @xmath22 and @xmath177 respectively . * _ operator _ @xmath197 is a neural network that produces fake image @xmath200 . * _ operator _ @xmath201 is a neural network that estimates the probability that an image is fake . * the remaining operators compute a loss that _ curator _ minimizes and _ forger _ maximizes @xmath202}_{{\ensuremath{\boldsymbol\ell}}_{disc } } + \underbrace{\operatorname*{\mathbb e}_{{\ensuremath{\boldsymbol\epsilon}}\sim { { \mathbb p}}({\ensuremath{\boldsymbol\epsilon}})}\big[\log\big(1-d_{{\ensuremath{\boldsymbol{\phi}}}}(g_{{\ensuremath{\boldsymbol{\theta}}}}({\ensuremath{\boldsymbol\epsilon}}))\big)\big]}_{{\ensuremath{\boldsymbol\ell}}_{gen}}\ ] ] note there are two copies of operator @xmath203 in the query graph . the response graph implements the chain rule , with a tweak that multiplies the gradient communicated to _ forger _ by @xmath204 to ensure that _ forger _ maximizes the loss that _ curator _ is minimizing . + at ( 0,0 ) ( glabel ) * response * ; at ( 0,-3 ) ( player1 ) * forger * ; at ( 4,0 ) ( player2 ) * curator * ; at ( 4,-3 ) ( block1 ) @xmath186 ; at ( 8,-3 ) ( block2 ) @xmath186 ; at ( 12,-3 ) ( block3 ) @xmath186 ; at ( 12,-1 ) ( block4 ) @xmath186 ; at ( 8,-1 ) ( block4a ) @xmath186 ; at ( 4,-1 ) ( block5 ) @xmath185 ; at ( 12,-4 ) ( oracle3 ) * oracle@xmath205 * ; at ( 12,0 ) ( oracle4 ) * oracle@xmath206 * ; at ( 12,-2 ) ( oracle5 ) * oracle@xmath184 * ; at ( 4,-4 ) ( oracle1 ) * oracle@xmath130 * ; at ( 8,-4 ) ( oracle2 ) * oracle@xmath207 * ; at ( 8,0 ) ( oracle2a ) * oracle@xmath207 * ; ( block1 ) edge node[below ] @xmath131 ( player1 ) ; ( block1 ) edge node[above ] @xmath63 ( player1 ) ; ( block2 ) edge node[right ] @xmath208 ( block5 ) ; ( oracle4 ) edge node[right ] @xmath209 ( block4 ) ; ( block2 ) edge node[above ] @xmath135 ( block1 ) ; ( block2 ) edge node[below ] @xmath210 ( block1 ) ; ( oracle2 ) edge node[right ] @xmath211 ( block2 ) ; ( oracle2a ) edge node[right ] @xmath212 ( block4a ) ; ( oracle1 ) edge node[left ] @xmath213 ( block1 ) ; ( oracle3 ) edge node[right ] @xmath214 ( block3 ) ; ( oracle5 ) edge node[right ] @xmath192 ( block3 ) ; ( oracle5 ) edge node[right ] @xmath192 ( block4 ) ; ( block4 ) edge node[above ] @xmath215 ( block4a ) ; ( block4a ) edge node[above ] @xmath216 ( block5 ) ; ( block4 ) edge node[below ] @xmath217 ( block4a ) ; ( block3 ) edge node[above ] @xmath218 ( block2 ) ; ( block3 ) edge node[below ] @xmath214 ( block2 ) ; ( block5 ) edge node[left ] @xmath219 ( player2 ) ; the generative - adversarial network is the first example where the response graph does not simply backpropagate gradients : the arrow labeled @xmath223 is computed as @xmath224 , whereas backpropagation would use @xmath225 . the minus sign arises due to the adversarial relationship between forger and curator they do not optimize the same objective . as discussed in section [ sec : valprox ] , actor - critic algorithms decompose the reinforcement learning problem into two components : the critic , which learns an approximate value function that predicts the total discounted future reward associated with state - action pairs , and the actor , which searches for a policy that maximizes the value appoximation provided by the critic . when the action - space is continuous , a natural approach is to follow the gradient @xcite . in @xcite , it was shown how to compute the policy gradient given the true value function . furthermore , sufficient conditions were provided for an approximate value function learned by the critic to yield an unbiased estimator of the policy gradient . more recently @xcite provided analogous results for deterministic policies . the next example of a grammar is taken from @xcite , which builds on the above work by introducing a third algorithm , deviator , that directly estimates the gradient of the value function estimated by critic . at ( 0,-1.5 ) ( glabel ) * query * ; at ( 0,0 ) ( actor ) * actor * ; at ( 0,-3 ) ( deviator ) * deviator * ; at ( 14,0 ) ( critic ) * critic * ; at ( 14,-3 ) ( noise ) * noise * ; at ( 6.5,-1.5 ) ( env ) * nature * ; at ( 3,0 ) ( block1 ) @xmath90 ; at ( 3,-3 ) ( block2 ) @xmath226 ; at ( 10,0 ) ( block3 ) @xmath103 ; at ( 10,-3 ) ( block4 ) @xmath53 ; ( actor ) edge node[above ] @xmath22 ( block1 ) ; ( deviator ) edge node[above ] @xmath105 ( block2 ) ; ( critic ) edge node[above ] @xmath227 ( block3 ) ; ( noise ) edge node[above ] @xmath228 ( block4 ) ; ( block1 ) ( block2 ) ; ( block1 ) ( block3 ) ; ( block2 ) ( block4 ) ; ( block3 ) ( block4 ) ; ( env ) edge node[above ] @xmath91 ( block1 ) ; ( env ) edge node[above ] @xmath91 ( block2 ) ; ( env ) edge node[above ] @xmath91 ( block3 ) ; ( env ) edge node[above ] @xmath106 ( block4 ) ; * _ nature _ samples states from @xmath84 and announces rewards @xmath229 that are a function of the prior state and action ; _ noise _ samples @xmath230 . * _ actor _ , _ critic _ and _ deviator _ play parameters @xmath22 , @xmath227 and @xmath105 respectively . * _ operator _ @xmath90 is a neural network that computes actions @xmath231 . * _ operator _ @xmath232 is a neural network that estimates the value of state - action pairs . * _ operator _ @xmath233 is a neural network that estimates the gradient of the value function . * the remaining _ operator _ computes the bellman gradient error ( bge ) which _ critic _ and _ deviator _ minimize @xmath234 at ( 0,-1.5 ) ( glabel ) * response * ; at ( 0,0 ) ( actor ) * actor * ; at ( 0,-3 ) ( deviator ) * deviator * ; at ( 14,0 ) ( critic ) * critic * ; at ( 5.5,0 ) ( oracle1 ) * oracle@xmath236 * ; at ( 5.5,-1.5 ) ( oracle2)*oracle@xmath237 * ; at ( 3,-1.5 ) ( op2 ) @xmath238 ; at ( 7.5,0 ) ( oracle3 ) * oracle@xmath239 * ; at ( 10,-3 ) ( oracle4 ) * oracle@xmath57 * ; at ( 3,0 ) ( block1 ) @xmath59 ; at ( 5.5,-3 ) ( block2 ) @xmath59 ; at ( 10,0 ) ( block3 ) @xmath59 ; ( block1 ) edge node[below ] @xmath240 ( actor ) ; ( block1 ) edge node[above ] @xmath241 ( actor ) ; ( block2 ) edge node[below ] @xmath242 ( deviator ) ; ( block2 ) edge node[above ] @xmath112 ( deviator ) ; ( block3 ) edge node[below ] @xmath243 ( critic ) ; ( block3 ) edge node[above ] @xmath244 ( critic ) ; ( oracle1 ) edge node[below ] @xmath245 ( block1 ) ; ( oracle4 ) edge node[below ] @xmath246 ( block2 ) ; ( oracle4 ) edge node[above ] @xmath247 ( block2 ) ; ( oracle4 ) edge node[right ] @xmath248 ( block3 ) ; ( oracle4 ) edge node[left ] @xmath249 ( block3 ) ; ( oracle2 ) edge node[left ] @xmath250 ( block2 ) ; ( op2 ) edge node[right ] @xmath226 ( block1 ) ; ( op2 ) edge node[left ] @xmath251 ( block1 ) ; ( oracle3 ) edge node[below ] @xmath252 ( block3 ) ; note that instead of backpropagating first - order information in the form of gradient @xmath253 , the response graph instead backpropagates zeroth - order information in the form of _ gradient - estimate _ @xmath226 , which is computed by the query graph during the feedforward sweep . we therefore write @xmath251 and @xmath254 ( instead of @xmath255 and @xmath256 to emphasize that the gradients communicated to actor are estimates . 1 . _ critic _ estimates the value function via td - learning @xcite with cloning for improved stability @xcite . deviator _ estimates the value gradient via td - learning and the gradient perturbation trick @xcite . _ actor _ follows the correct gradient by the policy gradient theorem @xcite . the internal workings of each neural network are guaranteed correct by the chain rule . two appealing features of the algorithm are that ( i ) actor is insulated from critic , and only interacts with deviator and ( ii ) critic and deviator learn different features adapted to representing the value function and its gradient respectively . previous work used the derivative of the value - function estimate , which is not guaranteed to have compatible function approximation , and can lead to problems when the value - function is estimated using functions such as rectifiers that are not smooth @xcite . finally we consider kickback , a biologically - motivated variant of backprop with reduced communication requirements @xcite . the problem that kickback solves is that backprop requires two distinct kinds of signals to be communicated between units feedforward and feedback whereas only one signal type spikes are produced by cortical neurons . kickback computes an estimate of the backpropagated gradient using the signals generated during the feedforward sweep . kickback also requires the gradient of the loss with respect to the ( one - dimensional ) output to be broadcast to all units , which is analogous to the role played by diffuse chemical neuromodulators @xcite . at ( 0,0 ) ( glabel ) * query * ; at ( 0,-2 ) ( player_env1 ) * nature * ; at ( 4,0 ) ( player1 ) * layer@xmath258 * ; at ( 8,0 ) ( player2 ) * layer@xmath259 * ; at ( 12,0 ) ( player3 ) * layer@xmath260 * ; at ( 16,0 ) ( player_env2 ) * nature * ; at ( 2.5,-3 ) ( out1 ) @xmath55 ; at ( 6.5,-3 ) ( out2 ) @xmath55 ; at ( 10.5,-3 ) ( out3 ) @xmath55 ; at ( 4,-2 ) ( block1 ) @xmath261 ; at ( 8,-2 ) ( block2 ) @xmath262 ; at ( 12,-2 ) ( block3 ) @xmath263 ; at ( 16,-2 ) ( block4 ) @xmath53 ; ( player_env1 ) ( block1 ) ; ( player1 ) edge node[left ] @xmath264 ( block1 ) ; ( player2 ) edge node[left ] @xmath265 ( block2 ) ; ( player3 ) edge node[left ] @xmath266 ( block3 ) ; ( player_env2 ) edge node[left ] @xmath56 ( block4 ) ; ( block1 ) edge node[above ] @xmath267 ( block2 ) ; ( player_env1 ) edge node[above ] @xmath54 ( block1 ) ; ( block2 ) edge node[above ] @xmath268 ( block3 ) ; ( block3 ) edge node[above ] @xmath269 ( block4 ) ; ( block1 ) edge node[right ] @xmath270 ( out1 ) ; ( block2 ) edge node[right ] @xmath271 ( out2 ) ; ( block3 ) edge node[right ] @xmath272 ( out3 ) ; ( block1 ) edge node[left ] @xmath273 ( out1 ) ; ( block2 ) edge node[left ] @xmath274 ( out2 ) ; ( block3 ) edge node[left ] @xmath275 ( out3 ) ; * _ nature _ samples labeled data @xmath37 from @xmath143 . * _ layers _ by weight matrices @xmath145 . the output of the neural network is required to be one - dimensional * _ operators _ for each layer compute two outputs : @xmath276 _ and _ @xmath277 where @xmath278 and 0 otherwise . * the task is regression or binary classification with loss given by the mean - squared or logistic error . it follows that the derivative of the loss with respect to the network s output @xmath279 is a _ scalar_. the response graph contains a single oracle that broadcasts the gradient of the loss with respect to the network s output ( which is a scalar ) . estimates _ for each _ layer _ are computed using a mixture of oracle and local zeroth - order information referred to as _ kicks _ : + at ( 2,0 ) ( glabel ) * response * ; at ( 4,-2 ) ( oracle1 ) * kick@xmath280 * ; at ( 8,-2 ) ( oracle2 ) * kick@xmath281 * ; at ( 12,-2 ) ( oracle3 ) * kick@xmath282 * ; at ( 10,-4 ) ( oracle_env2 ) * oracle@xmath57 * ; at ( 6,-2 ) ( block1 ) @xmath186 ; at ( 10,-2 ) ( block2)@xmath186 ; at ( 14,-2 ) ( block3 ) @xmath186 ; ( oracle_env2 ) edge node[left ] @xmath284 ( block2 ) ; ( oracle_env2 ) edge node[left ] @xmath285 ( block1 ) ; ( block1 ) edge node[right]@xmath286 ( player1 ) ; ( block2 ) edge node[right]@xmath287 ( player2 ) ; ( block3 ) edge node[right]@xmath288 ( player3 ) ; ( oracle1 ) ( block1 ) ; ( oracle2 ) ( block2 ) ; ( oracle3 ) ( block3 ) ; ( s1 ) edge node[above]@xmath295 ( block1 ) ; ( s2 ) edge node[above]@xmath296 ( block2 ) ; ( s3 ) edge node[above]@xmath297 ( block2 ) ; ( s3 ) edge node[below]@xmath298 ( block2 ) ; ( block1 ) ( zero ) ; ( block2 ) ( block1 ) ; + where @xmath299 is coordinatewise multiplication and @xmath300 is the outer product . if @xmath301 then _ nature _ is substituted for @xmath295 . if @xmath302 then @xmath303 is replaced with the scalar value @xmath192 . the loss functions for the layers are not computed in the query graph . nevertheless , the gradients communicated to the layers by the response graph are exact with respect to the layers losses , see @xcite . for our purposes it is more convenient to focus on the global objective of the neural network and treat the gradients communicated to the layers as _ estimates _ of the gradient of the global objective with respect to the layers weights . the guarantee for kickback is that , if the network is coherent , then the gradient estimate @xmath306 computed using the zeroth - order kicks has the same sign as the backpropagated error @xmath307 computed using gradients , see @xcite for details . as a result , smalls steps in the direction of the gradient estimates are guaranteed to decrease the network s loss . @xmath11 + kickback uses a single oracle , analogous to a neuromodulatory signal , in contrast to backprop which requires an oracle per layer . the rest of the oracles are replaced by kicks zeroth - order information from which gradient - estimates are constructed . importantly , the kick computation for layer @xmath117 only requires locally available information produced by its neighboring layers @xmath308 and @xmath309 during the feedforward sweep . the feedback signals @xmath297 are analogous to the signals transmitted by nmda synapses . two recent alternatives to backprop that also do not rely on backpropagating exact gradients are target propagation @xcite and feedback alignment @xcite . target propagation makes do without gradients by implementing autoencoders at each layer . unfortunately , optimization problems force the authors to introduce a correction term involving _ differences _ of targets . as a consequence , and in contrast to kickback , the information required by layers in difference target propagation can not be computed locally but instead requires recursively backpropagating differences from the output layer . feedback alignment solves a different problem : that feedback and forward weights are required to be equal in backprop ( and also in kickback ) . the authors observe that using random feedback weights can suffice . unfortunately , as for difference target propagation , feedback alignment still requires separate feedforward and recursively backpropagated training signals , so weight updates are not local . unfortunately , at a conceptual level kickback , target propagation and feedback alignment all tackle the wrong problem . the cortex performs reinforcement learning : mammals are not provided with labels , and there is no clearly defined output layer from which signals could backpropagate . a biologically - plausible deep learning algorithm should take advantage of the particularities of the reinforcement learning setting . hinton g , deng l , yu d , dahl ge , mohamed a , jaitly n , senior a , vanhoucke v , nguyen p , sainath tn , kingsbury b : * deep neural networks for acoustic modeling in speech recognition*. _ ieee signal processing magazine _ 2012 , * 29*:8297 . mnih v , kavukcuoglu k , silver d , rusu aa , veness j , bellemare mg , graves a , riedmiller m , fidjeland ak , ostrovski g , petersen s , beattie c , sadik a , antonoglou i , king h , kumaran d , wierstra d , legg s , hassabis d : * human - level control through deep reinforcement learning*. _ nature _ 2015 , * 518*(7540):529533 . bergstra j , breuleux o , bastien f , lamblin p , pascanu r , desjardins g , turian j , warde - farley d , bengio y : * theano : a cpu and gpu math expression compiler*. in _ proc . python for scientific comp . ( scipy ) _ 2010 . bastien f , lamblin p , pascanu r , bergstra j , goodfellow i , bergeron a , bouchard n , bengio y : * theano : new features and speed improvements*. in _ nips workshop : deep learning and unsupervised feature learning _ 2012 . williams rj , zipser d : * gradient - based learning algorithms for recurrent networks and their computational complexity*. in _ backpropagation : theory , architectures , and applications_. edited by chauvin y , rumelhart d , lawrence erlbaum associates 1995 . von bartheld cs , wang x , butowt r : * anterograde axonal transport , transcytosis , and recycling of neurotrophic factors : the concept of trophic currencies in neural networks*. _ molecular neurobiology _ 2001 , * 24*. sutton r , modayil j , delp m , degris t , pilarski pm , white a , precup d : * horde : a scalable real - time architecture for learning knowledge from unsupervised motor interaction*. in _ proc . 10th int . conf . on aut agents and multiagent systems ( aamas ) _ 2011 . balduzzi d : * falsification and future performance*. in _ algorithmic probability and friends : bayesian prediction and artificial intelligence _ , _ volume 7070 of _ lnai__. edited by dowe d , springer 2013:6578 . sutton r , szepesvri c , maei hr : * a convergent @xmath310 algorithm for off - policy temporal - difference learning with linear function approximation*. in _ adv in neural information processing systems ( nips ) _ 2009 . |
bio - inspired algorithms like genetic algorithms and fuzzy inference systems ( fis ) are nowadays widely adopted as hybrid techniques in commercial and industrial environment . in this paper we present an interesting application of the fuzzy - ga paradigm to smart grids . the main aim consists in performing decision making for power flow management tasks in the proposed microgrid model equipped by renewable sources and an energy storage system , taking into account the economical profit in energy trading with the main - grid . in particular this study focuses on the application of a hierarchical genetic algorithm ( hga ) for tuning the rule base ( rb ) of a fuzzy inference system ( fis ) , trying to discover a minimal fuzzy rules set in a fuzzy logic controller ( flc ) adopted to perform decision making in the microgrid . the hga rationale focuses on a particular encoding scheme , based on control genes and parametric genes applied to the optimization of the fis parameters , allowing to perform a reduction in the structural complexity of the rb . this approach will be referred in the following as fuzzy - hga . results are compared with a simpler approach based on a classic fuzzy - ga scheme , where both fis parameters and rule weights are tuned , while the number of fuzzy rules is fixed in advance . experiments shows how the fuzzy - hga approach adopted for the synthesis of the proposed controller outperforms the classic fuzzy - ga scheme , increasing the accounting profit by 67% in the considered energy trading problem yielding at the same time a simpler rb . * _ keywords : _ * microgrid , energy management system , battery energy storage , power flow optimization , storage system management , fuzzy systems , evolutionary computation , hierarchical genetic algorithms . | 1604.04789 | the world wide power grid can be considered one of the greatest masterpiece of engineering that human being has ever made . moreover , starting from the first industrial revolution , smart grids ( sgs ) are one of the most important breakthrough that science and engineering fields are carrying out . currently , the smart grid concept is founded on a paradigmatic revolution that will permeate many aspects of human life . from the power system point of view , sgs can be considered as a way to transform the electric energy infrastructures from a centralized , producer - controlled network , into a distributed and consumer - interactive system , leveraging the same concepts and technologies that enabled the emergence and spread of the internet . sgs vision promises a power grid infrastructure with increased automation of the grid operations and self - healing capabilities . sgs integrate the renewable energy production , seamlessly balancing energy supply and demand , and can be considered as a key technology for facilitating the spread of electric mobility . to achieve that vision , the current power grid has to be thought as a _ technological ecosystem _ that needs a strong _ injection _ of distributed intelligence @xcite . g. k. venayagamoorthy argues that the current power grid can be considered a spatially and temporally complex , nonlinear and non - stationary system with a lot of uncertainties @xcite . accordingly , the sg can be examined for all intents and purposes as a complex system , and computational intelligence ( ci ) and soft computing ( sc ) techniques @xcite , are widely adopted to face a plethora of applications and problems arising in the sg context . the main ci paradigms for sg related problem solutions are : neuro - fuzzy , neuro - swarm , fuzzy - pso , fuzzy - ga , neuro - genetic @xcite . in fact , in mg related tasks , such as control and flow management , the presence of uncertainty and non - linearity , for example in the power demand profile of a large amount of users or in the power produced by solar or wind sources , makes related problems extremely challenging . hence , sc techniques can help managing the complexity of problems offering reliable solutions , especially in presence of non - linearity @xcite and in presence of storage devices that increase the solution space of the unit commitment problem @xcite . consequently , since linear techniques can not be considered adequate in solving problems whose nature is nonlinear and even stochastic , sc techniques offer a suitable framework introducing learning capabilities in the design of the mg controllers , especially in presence of renewable energy sources and storage . the current research follows our previous work @xcite about an application of what we call classic fuzzy - ga paradigm to the problem of flow control optimization in a microgrid ( mg ) . the mg can be thought as a sub - network of the sg characterized by the presence of autonomous ( often renewable ) energy sources buffered by some type of battery energy storage system ( bess ) and locally controlled in order to achieve smart energy flows management . the flow control task is carried out by a flc with two fuzzy inference systems ( fiss ) of mamdani type . fiss , relaying on approximated reasoning based on fuzzy logic ( fl ) , are in charge to optimize energy flows and the overall accounting profit in energy trading operations with the main - grid . the choice of mamdani type fis is related to its simplicity in incorporating human knowledge in the rb . specifically , in the application at hand the computational overhead introduced by the defuzzification process ( absent in a sugeno type fis ) is not a real problem . furthermore , the hga paradigm with its suitable coding scheme is well suited to optimize the structure of a mandani fis . the current work is focused on two main objectives : @xmath0 improve the mg model , in particular the bess model , @xmath1 optimize the rule base ( rb ) of the flc adopted to control power flows in the mg . as concerns the first goal we move on from an ideal battery adopted in @xcite designing an energy storage device based on a real - world model with suitable efficiency parameters . for the second objective , we designed an optimization method based on a suited genetic algorithm ( ga ) that is in charge of learning the fis parameters , optimizing at the same time both the economic return in energy trading and the cardinality of the fuzzy rb . the adopted optimization algorithm is known as hierarchical genetic algorithm ( hga ) aiming to perform at the same time a fine tuning of the fuzzy membership functions ( mfs ) and the structural optimization of the fiss in the following we will refer on this paradigm as `` fuzzy - hga paradigm '' . in fact , while in our previous work the fis structure is constrained to be fixed , with a predefined number of antecedent and consequent terms and , thereby , with an immutable number of fuzzy rules , the adopted hga scheme allows to relax these constraints . moreover , the standard ga approach deals with a chromosome of fixed length whose encoding scheme leads to a lower flexibility in the rb tuning . a ga algorithm able to emulate a variable length chromosome with a suitable encoding scheme of the fis is ideal for optimizing the number of rules in the given rb . finally , the fuzzy rule optimization can lead to an improved performance of the fis , discarding pre - defined low performing rules . the idea behind hgas is based on the biological inspired gene structure of a chromosome formed in a hierarchical fashion , emulating the encoding approach of the deoxyribonucleic acid ( dna ) . in nature the genes can be classified into two different types : regulatory sequences and structural genes . one of the regulatory sequences found in dna is called the promoter " with the task of activating or deactivating structural genes . therefore the presence of active and inactive genes in the structural genes leads to the idea of a hierarchical structure formulation of the chromosome that consists of _ control genes _ and _ parametric genes_. the activation of the parametric genes is governed by the value of the control genes . accordingly , the strategy suggested by nature can be modeled for solving a number of engineering problems , such as those involving mix integer programming methods @xcite or fuzzy control applications demanding the joint optimization of both fiss parameters and rbs @xcite . in the last case the novelty of a hierarchical coding scheme is based on the definition of suitable genetic operators moving from standard ga algorithms to more advanced ones . the hierarchical encoding scheme allows to code the fiss parameters , more precisely mf parameters , as parametric genes and , at the same time , control genes can be used to activate and deactivate mfs composing a given fuzzy rb , thus tuning the overall number of fuzzy rules . the work is organized as follows . [ sec : related_works ] is a literature review about the use of flc and the fuzzy - ga paradigm in the smart grid context . in sec . [ sec : background ] we introduce the optimization problem and the mg model . [ sec : pf ] clarifies the level of abstraction of the problem , introducing the adopted notation and explaining how the fuzzy controller works . the fuzzy control scheme for a mg , together with the classic fuzzy - ga and fuzzy - hga paradigms are treated in sec . [ sec : fuzzy_c_mg ] and related subsections . in sec . [ sec : experim_ev ] , soon after the introduction of the examined mg scenarios and the algorithm settings , the main results are reported and discussed . finally , conclusions are drawn in sec . [ sec : conclusions ] , while feature works are discussed in sec . [ sec : fws ] . in the sc field , fl is a well consolidated discipline able to integrate approximate reasoning methodologies in engineering systems . sc techniques are commonly adopted in applications dealing with uncertainties @xcite . an interesting review on fuzzy logic and its hybrid approaches employed in the smart grids and microgrids context can be found in @xcite . in @xcite a small - scale mg designed to provide power to local communities , able to connect or disconnect from the main - grid , is studied . the control scheme is based on a non optimized flc considering as input the electricity prices , the renewable production , and the load demand . the rb is developed in order to ensure reliable grid operations taking into account the financial aspect to decide the load modification s level . flcs can be designed to control the amount of power that should be taken from the battery system in case of power deficiency to cover the load demand in a scheduling process @xcite . the latter problem is faced in @xcite , optimizing distribution system operations in a sg , from cost and system stability points of view . in the adopted mg model the flc was used only in the case when the instantaneous load demanded is higher than the instantaneous available power from renewable energy sources and the system is not at the peak period ( the battery is operated in discharge mode ) . however the flc does not foresee any further optimization . in @xcite is exploited a flc in a mg energy management system ( ems ) equipped by a bess in order to apply peak shaving and valley filling and to limit the oscillation of energy exchanged with the network buffering the ess . the flc aims to reduce the variation of the energy exchanged with the grid . moreover , it exploits a low bandwidth filter set before the flc . the flc is also optimized , but the adopted algorithm is not specified . in @xcite it has been considered a mg grid - connected equipped with a battery and a fuel cell . the mg consists of a wind turbine , a pv plant and a micro - turbine ; the ems must satisfy the load demand . in this work , the considered objective functions to be minimized are the energy expenses and the emissions of a mg , through multi - objective optimization . the proposed ems approach consists in the joint formulation of multi - objective optimization approach based on linear programming and battery scheduling ( working both on - line ) . the battery scheduling , as a part of an optimal online energy management , is made through a flc designed by an expert operator for the on - line scheduling of the bess . it defines whether the battery should be charged or discharged and at which convenient rates . one of the flc input is the predicted energy in the day ahead both for the load and the generation . after the scheduling of the bess , the remainder of the power flows are optimized with a multi objective linear programming algorithm . in the mg modeling field a fl - based framework is proposed in @xcite aimed at controlling a bess to achieve an high efficient management . authors propose a mg model able to operate synchronously with the main - grid ( grid connected mode ) or independently ( islanded mode ) aiming to control the amount of power delivered to / taken from the storage unit in order to improve a cost function based on purchasing / selling power to the main - grid . predetermined human reasoning - based fuzzy rules are adopted to define a flc with inputs variables such as consumer s load demand , electricity price rate and renewable energy generation rate . specifically , data are sampled every 15 minutes and the evaluated figure of merit consists of the summation of payments / revenues ( similarly to the current work ) along with system losses . however , the fuzzy rb is kept fixed , thereby any subsequent learning and optimization procedure is not considered . it is well known that fl , especially in flc of mamdani type , is well - suited in incorporating human knowledge in artificial systems . however , the synthesis of the fuzzy rules in a complex control problem can be a challenging task and in presence of many free parameters in the rb the human expertise is not sufficient to assure satisfying performances . thereby , the fuzzy - ga modeling can help to improve the inherent human approximation level in generating fuzzy rules . furthermore , an evolutionary strategy can fill the gap between a fixed rb and a flexible one introducing a suitable learning scheme . thereby , an hga is able to tackle the problem of managing many free parameters of a mamdani fis , optimizing at the same time the number of rules and selecting the ones that perform better in terms of the considered objective function . in @xcite , the fuzzy - ga approach is adopted for controlling energy flows and for managing the mg , in presence of an electric storage unit and distributed renewable energy generation . the main goal consists in optimizing a cost function defined as the net expense in electric energy purchasing from the main - grid . in @xcite authors propose , within the fuzzy - ga paradigm , an algorithm for optimizing the mg operations . in particular given the load demand , the wind power generation and day - ahead electricity prices , the algorithm is able to set up a mg generation schedule and to synthesize a suitable fuzzy expert system for allocating the energy from the storage system . moreover , two gas are used alternatively , one to determine both the microgrid scheduling and the fuzzy rules , the other to tune the mfs . the proposed expert system , responsible for setting the power withdrawn from the battery , is provided with learning capabilities so that it can optimize its knowledge base ( fuzzy rules and mfs ) according to the given scenario . a fl - based energy management , adopting a fixed mamdani fis , is studied in @xcite for house consumption applications in presence of a solar source . the main aim here is maximizing the use of solar energy and reducing the impact on the power grid , while satisfying the energy demand of house appliances , giving a fixed level of comfort . the proposed approach concerns a control scheme relating to energy flows of a mg belonging to an energy district see fig [ fig : energy_district ] and connected to the main - grid or even other mgs . the mg can perform its operations in a grid connected mode " and in islanded mode " . in the proposed model the energy production is typically provided by renewable distributed resources ( drs ) such as solar , wind , and micro - hydro generators . in addition , the mg can be equipped with a traditional energy source , such as a turbogas generator . moreover , the mg is provided by a li - ion bess ; the model is compliant with technical specifications of a specific toshiba device . the bess is controlled through a battery management system ( bms ) and can be used to compensate the lack of energy production when drs are unable to meet domestic demand or to accumulate energy in case of overproduction . the bms can be programmed also to store energy that can be resold to the main - grid when favorable economic conditions apply ( based , for example , on energy price signals ) . the main scenarios , in which the proposed controller can act , are : * the energy production within the mg meets ( or exceed ) the energy demand ; * the energy production within the mg does not meet the demand . in the former case the mg operates in islanded mode " and the bess can be used to store a fraction of the surplus of energy , selling the remainder to the main - grid . in the latter case the mg can operate in grid connected mode " and can satisfy the loads by drawing a fraction of the negative energy balance from the bess , while buying the remainder from the main - grid . the two main operations introduced above depend strictly on the mg internal state that includes the state of charge ( soc ) of the bess and thus the available stored energy , the current price of energy and the generated and demanded power . it is noted that the two main objectives of the proposed controller are : * to guarantee the continuous to power supplies of energy availability without interruptions . thus the mg must guarantee an high quality of service . * to maximize the accounting profit for possible energy trading with the main - grid or other mgs , given the collected information about the energy price and the mg internal state . in fig . [ fig : s_c_scheme ] is depicted the simplified control scheme for a given mg . the controller is immersed in an external environment where information is assumed to be originated from a centralized data - center or directly from other connected mgs , in an automatic cooperative / competitive trading scenario . input information to the controller are the aggregated energy time profile requested from the mg loads , the aggregated energy time profile produced by the drs , the energy price signal coming from the main - grid and the soc of the bess . the output of the controller is a pair of signals towards the main - grid or the bess . the decision making module adopts a fuzzy inference scheme , implementing two distinct fiss for deciding the best actions trying to maximize the accounting profit of the mg in a given time span . a hierarchical genetic algorithm is adopted to tune the set of fuzzy rules and their parameters and thus to learn their best combination exploiting a given suitable objective function . the entire control chain in a general microgrid can be considered as organized as a three level system : the field level ( data acquisition and decentralized control ) , the scada level ( supervision and control functions ) , and the planning and management level , which can be a decentralized ems @xcite . in this work , we focus on the third level . thereby , the model that we will discuss is based on a number of hypothesis that define the level of abstraction useful to correctly place the problem under analysis . the main assumptions are reported below . * the considered timestamps are discretized and monospaced . * the present study does not consider low level operations such as voltage and reactive power control . * only high level operations are evaluated , such as the control of energy flow within a mg . * in the simulation several energy prosumers are contemplated , each one has its own characteristic energy profile with a timing defined each quarter of an hour . * for each time sample the flow control system takes into account the aggregate energy production and demand . * in each time sample the control system is able to fulfill a decision making request . * the demanded energy is actually consumed . * the main - grid has an unlimited energy capacity . * the current version of the mg flow control system deals with a single bess , as an overall model of the total storage capacity within the mg ( possibly due to more than one distinct devices ) . * the current work does not cover data exchange models within the main - grid and it is supposed that the mg is equipped with a suitable data network infrastructure with low delay and high qos in exchanging data packets . hence , the adopted controller focus on the energy exchanges between portions of the mg and the main - grid considered fairly of unlimited capacity given the powers involved in small - scale mgs , as the one herein considered . furthermore a discrete data model for the measured power consumed and demanded is used with a sample time of 15 minutes , as adopted in most high level control schemes . the last assumption , about the low delay in data exchange , is reliable considering both the low delays characterizing communication protocols used in modern sgs and the small extension of the adopted mg model . for instance , cellular wireless networks such as gprs , umts , or 4 g technologies like 802.16 m and lte could be used for the interface between smart meters and the central system @xcite . furthermore , multiple communication technologies and standards could coexist in different parts of the smart grid , especially in a multi - microgrid context , and their impact deserves a deep investigation of communication devices and channel networks . in general , given a mg like the one depicted in fig . [ fig : energy_district ] , thus consisting of several energy generators and loads , the aggregate electricity production can be defined as : @xmath2 where @xmath3 is the electrical power produced by the @xmath4-the producer within a set of @xmath5 producers , at time sample @xmath6 . + similarly , we can define the aggregate electrical power demand within the given mg : @xmath7 where @xmath8 is the amount of the demanded electricity effectively consumed by the @xmath9-th user within a set of @xmath10 consumers , at time sample @xmath6 . the mg is equipped with a bess , whose detailed model will be provided in sec . [ sec : bm ] below . hence , at the abstraction level of the controller , the electricity storage can be characterized by the soc variable , defined as : @xmath11 where @xmath12 and @xmath13 represent the lower and the maximum threshold values for the soc , respectively . in other words , these values constitute the interval of admissibility for the charge level of the battery unit . + considering both the aggregate demand time profile ( [ eq : prod_agg ] ) and the aggregate production time profile ( [ eq : cons_agg ] ) within the given mg , the power balance can be defined as : @xmath14 finally , we will denote with @xmath15 and @xmath16 the price signals in buying and selling energy from / to the main - grid , respectively . the mg state at time sample @xmath6 , from the point of view of the controller , depends on the energy balance , the energy price signals and the state of charge ( soc ) of the storage unit . therefore , it is possible to define a generic mg state as : @xmath17 the energy balance value ( [ eq : bal ] ) defines three distinct sub - states in which the mg can be found , namely : 1 . if @xmath18 , the mg has a deficit of energy ( underproduction state ) ; 2 . if @xmath19 , the produced energy fulfills the internal demand ( equilibrium state ) ; 3 . if @xmath20 , the mg has a surplus of energy ( overproduction state ) . for each given sub - state we can have the following contingencies for the storage unit : 1 . if @xmath21 , the available charge has a null value ; 2 . if @xmath22 , the charge has a maximum value ; 3 . if @xmath23 , the available charge has an intermediate value . the above reported fine grained classification leads to define a set of control actions taking into account also the energy price signals . the inference process underlying the controller is performed by a set of control actions dispatched to actuators . the control actions depend on the mg state @xmath24 . it is noted that , in the following , the term _ crisp _ " @xcite relies to the distinction made in fl among _ crisp _ sets , that are ordinary sets and fuzzy sets defined by a mf , as it will discussed in the following sec . [ sec : fuzzy_c_mg ] . the set of control actions is defined as : @xmath25 hence , for the generic action we have : @xmath26 in addition , the set of actions just defined can be split in two disjoint subsets : 1 . forced " action that are crisp " action due to : @xmath0 the energy deficit in the mg and the lack of available bess charge , or @xmath1 the energy surplus in the mg and the capacity limit reached by the bess ; 2 . fuzzy " actions that depend on two real valued parameters @xmath27 $ ] encoding respectively : 1 . the quantity of energy to sell to the main - grid or used to feed the battery ; 2 . the quantity of energy to purchase from the main - grid or to draw from the battery . the disambiguation task about these cases is performed by a suitable module [ fig : fd ] able to evaluate the mg state @xmath24 , in charge of selecting the activation of the crisp decision making procedure or the fuzzy one . + for the first subset of control actions it is defined the energy exchange in input and output from / to the main - grid , as : @xmath28 \delta t \\ e_{in}(t)= \beta(t ) \ , bal(t)= \beta(t ) \ , [ p_{agg}(t)-d_{agg}(t ) ] \delta t , \label{eq : pout}\end{aligned}\ ] ] where @xmath29 , @xmath30 are real numbers in @xmath31 $ ] and @xmath32 is the time slot width . it is noted that the unitary value for both the @xmath29 and @xmath30 control parameters leads to actions that overlaps with the set of crisp " actions . however , these ones must be classified as _ deliberate _ choice leaded by the inference module , rather than forced actions " . the second subset of control actions refers to the energy exchanges with the energy storage device , that at the abstraction level of the controller , is governed with the following dynamical equation regarding the soc @xcite : @xmath33 where @xmath34 is the signed quantity of charge exchanged with the mg , as required at each time step by the flc . given the purchasing and selling price signals and the energy exchanged with the main grid [ eq : pout ] , it follows that the economical balance of the mg is : @xmath35 the last equation is the accounting profit ( or the accounting loss , depending on the sign of the first term of ( [ eq : profit ] ) for each time sample and it is the difference between the revenues and the expenses arising from the energy trading . in eq . ( [ eq : profit ] ) the revenues and the expenses are defined respectively : @xmath36 @xmath37 where @xmath16 is the selling price and @xmath15 is the purchasing price . the last two equation in ( [ eq : rev_exp]),([eq : rev_exp2 ] ) and thus the ( [ eq : profit ] ) are functions of the parameters @xmath29 and @xmath30 that are the output of the inference process based on the flc . the presence of the energy storage capacity that constitutes a degree of freedom of the control system allows , for example , in the case of surplus of the energy production to store energy if the current selling price is low and to sell the stored energy when the price is high . on the basis of the economic policies chosen for the given mg , it is possible to define an objective function that the controller has to fulfill . the objective function introduced in the current work , that is optimized by an evolutionary algorithm capable of fine tuning the flc parameters , is the overall profit on a finite temporal horizon of @xmath38 time samples . therefore , the objective function is defined as : @xmath39 at the abstraction level of the controller the energy storage capacity linked to the given mg is modeled by means of the ( [ eq : soc_eq ] ) , through the temporal dynamic of the soc parameter @xcite . in order to exploit a more realistic model of the mg , the bess model is enriched with physical real - world characteristics regarding battery devices currently used in battery - to - grid applications . precisely , the model proposed in the current section is based on a real - word li - on battery suitable for grid - tied applications @xcite . today , li - on batteries are not only used in consumer electronics but also in smart grid , as energy storage capacity for mgs or in v2 g applications @xcite . the li - on technology is characterized by both a good ratio power / weight ( about 150 wh / kg ) and an high efficiency that exceed 90% . a fundamental parameter in a battery model is the efficiency @xmath40 , defined as the energy fraction that we can recover from the battery after storing a given amount of energy . the efficiency parameter can be represented as : @xmath41 moreover , starting from the equivalent circuit of a generic battery [ fig : batt_scheme ] it is possible to outline the fundamental parameters able to increase the accuracy of the battery model . the model foresees an ideal voltage source @xmath42 that consists in the open - circuit voltage of the battery and an internal resistance @xmath43 . the @xmath44 and @xmath45 represent respectively the output current and voltage . from the circuit scheme depicted in fig . [ fig : batt_scheme ] , @xmath42 and @xmath43 depend on the soc value , hence : @xmath46 @xmath47 the @xmath48 and the @xmath49 values are respectively the charging and discharging efficiency , while @xmath50 is the capacity of the battery ( it is measured in ah ) representing the amount of electric current intensity that the battery is able to supply in a time interval of one hour . both the efficiencies @xmath48 and @xmath49 together with the internal resistance @xmath43 can take values depending on the state of the battery . therefore , @xmath42 and @xmath43 are variables that are functions of the charging state @xcite . fl is a very useful paradigm in dealing with fuzzy concepts expressed by fuzzy words ( i.e. high , low , warm , cold , etc . ) in computational and algorithmic frameworks and , as lotfi a. zadeh stated : fuzzy logic means computing with words " @xcite . moving from this general concepts , in engineering and in particular in the control systems field , researchers developed the fuzzy logic modeling ( flm ) paradigm to tackle control problems in complex systems aiming to model the underlying uncertainty @xcite . as baseline for the control scheme of the proposed mg model , in our previous work @xcite it is developed a flc of mamdani type capable to express an approximate reasoning method in order to control a real - world system , integrating knowledge coming from human experts @xcite . + as established in sec . [ sec : ca ] , the controller can manage a particular type of control rules that depend on two real - valued parameters , namely @xmath29 and @xmath30 . such parameters are currently computed by means of two fiss whose scheme is depicted in fig . [ fig::func_diag_dec_mak ] . formally , input variables , normalized in the real - valued range [ 0,1 ] , are computed separately by two functions @xmath51 and @xmath52 defined as follows : @xmath53 [ eq::param ] both fiss are based on a mamdani type fuzzy linguistic model able to introduce semi - qualitative if - than fuzzy rules . the paradigm used for our problem is the multiple input single output ( miso ) . hence , given a fixed number of inputs , a miso fis compute a single real - valued output . more specifically , given @xmath54 antecedents in a conjunctive form , the structure of the fuzzy rule can be written as : @xmath55,\quad i = 1,2, ... ,k . \end{split } \label{eq::fuzzy_r}\ ] ] where @xmath56 is a real number interpreted as a singleton mf , @xmath57 and @xmath58 are fuzzy labeled sets , i.e. , the antecedent and consequent linguistic terms , and @xmath59 are rule weights measuring the @xmath9-th rule strength among a set of @xmath5 fuzzy rules . moreover , the _ fuzzy conjunction _ of the antecedents i.e. , the and " operator , and the _ fuzzy implication _ , i.e. the then " operator , are suitable _ triangular norm _ @xcite . fuzzy rules , representing fuzzy implications or relations @xcite , in a conjunctive form decomposes the input domain into a set of fuzzy hyper - boxes , obtaining a _ grid partition _ of the fis input domain . if fis parameters are tuned by an optimization procedure the partition is said to be adaptive @xcite . nowadays , numerous fuzzy reasoning models have been developed and they are characterized by the number and the type of parameters , the shape of mfs and the defuzzification methods " . the last refers to a procedure to obtain a single scalar value from a fuzzy set @xcite . within the fuzzy - ga paradigm and in the simpler learning scheme , a genetic algorithm ( ga ) is adopted to optimize only the mf parameters . the rationale is that the ga evolves a set of fiss and returns , by means of an evolutionary procedure imitating the natural selection , the best among a population in terms of a suitable and well - defined performance measure or objective function . depending on the parameters of a flc that one decides to tune , we can fulfill different parameter learning schemes . in the first fuzzy control scheme , that is the simpler , the ga is applied to a procedure for _ knowledge learning _ and in part of _ structure learning _ the basics of the optimization scheme for the proposed control problem , obtained through a standard ga is depicted in fig . [ fig::gen_opt ] . for each of the three input in ( [ eq::param ] ) and for the output we employed three fuzzy labels , while for the shape of the input and output mfs we adopted the triangular ones . the form of a general fuzzy rule with three antecedents is : @xmath60 for the number of fuzzy rules we have chosen a maximal set , that is the maximum allowed by a regular grid partition , induced by the fuzzy rules structure defined in , namely 27 . having three inputs and working with triangular mfs the _ descriptive _ fuzzy reasoning method leads to a simple encoding scheme [ fig::enc_scheme ] for the flc parameters , namely the mf parameters and the rule weights @xmath59 . for both the fiss each gene represents a particular fis instance and each allele represents the parameter subject to the evolutionary tuning . having three parameters for each triangular mf with three input and one output for the flc , the length of the overall chromosome accounts for @xmath61 , being 9 the number of parameters that needs for describing three triangular shaped mfs . the fis composed by 27 rules is thus characterized by a chromosome consisting of @xmath62 elements . consequently , the whole chromosome , defining the two fis in the proposed flc , will account for 126 real bounded values . given an instance of a basic flc , as the one described in sec . [ sec : sfcs ] , the task is now to try to optimize together with the mf parameters and weights also the number of fuzzy rules . in other words starting from the adaptive grid partition structure generated by the antecedents of the fuzzy reasoning , we want to find a minimal set of rules . the basic flc , in fact , operates the inference with a maximal set of fuzzy rule that are encoded following a suitable scheme depicted in fig . [ fig::enc_scheme ] . in the search process , considering the number @xmath5 of fuzzy rules as a further degree of freedom , it is possible to improve the performance of the overall mechanism of inference and at the same time obtaining a simpler and more understandable structure of the fuzzy rb . this problem can be faced , within the fuzzy - ga paradigm , using a multiple chromosome scheme . the rationale is to partition the chromosome in two separate sections @xmath63 and @xmath64 , each one with a different meaning . more precisely , genes in @xmath63 can activate or deactivate genes of @xmath64 . the increasing complexity of the genetic code demands suited genetic operators . by adopting a multiple chromosome scheme it is possible to decompose the problem in small sub - problems , reflecting the decomposition of the genetic code in smaller units , each one ruled by its own genetic operator . the multi - chromosome scheme , as the entire genetic programming discipline , is inspired by nature . in fact , such methodology is widely used in erokaryotes organisms , i.e. some bacteria and plant species @xcite . this approach , whose performances are studied in @xcite , is adopted only when the problem at hand need such characteristic representation of the chromosome . in @xcite authors apply the multiple chromosome scheme to a mixed integer programming problem consisting in the control of an industrial machinery , where the need is to activate the machinery @xmath65 only when it is compliant with the overall cost reduction . the multiple chromosome has a section containing the _ control genes _ that encode the activation status of a particular machinery , while the remaining part , consisting of _ parametric genes _ , encodes for the amount of produced goods . in @xcite the multiple chromosome scheme is adopted for the solution of the multiple traveling salesman problem ( mtsp ) , that is a complex combinatorial optimization problem . + a particular approach to the multiple chromosome scheme is the _ hierarchical chromosome _ one , that leads to a family of gas , known as hierarchical genetic algorithms ( hgas ) . the multiple chromosome representation is based on partitioning it in two sections , one containing the control genes and the other the parametric genes . the control genes encode for the activation of a particular sub - section of the parametric genes . the adopted chromosome encoding scheme is based on a mixed binary and real - valued string @xcite , where the real - valued genes encode the mf parameters and the rule weights , while the binary genes , as control genes , encode the activation of mfs . in fig . [ fig::multiple_enc_scheme ] is depicted the adopted encoding scheme for both fiss . the overall chromosome can be formally written as : @xmath66 , \label{pbin - preal}\ ] ] where @xmath67 , \ ; p_{i}\in\left\lbrace0 , 1\right\rbrace , \ ; \forall i=1,\dots , n$ ] , being n the total number of the input mfs and @xmath68 , \ ; p_{j}\in r , \ ; \forall j = n+1,\dots , m$ ] , being @xmath69 the total number of mfs and fuzzy rules encoded in the chromosome . + considering as granulation of each input 2 external trapezoidal shaped mfs , described by 2 parameters each , and 3 internal triangular shaped mfs , described by 3 parameters each , the chromosome structure for one fis is the following . * the number @xmath70 of control genes is given by the product @xmath71 , where @xmath72 is the number of the input mfs and @xmath73 is the number of the fis input variables . thus , for for an input described by 5 mfs , we have @xmath74 control genes . * the number @xmath75 of parametric genes is composed by the number of input and output mf parameters @xmath76 followed by the number of rules weights , @xmath77 . thus , we have @xmath78 parametric genes . * finally , the total length of the genetic code is equal to @xmath79 genes . consequently , the whole chromosome , defining the two fis in the proposed flc , will account for 384 real bounded values . + among the several ways available for generating the initial population , in the current work it is adopted the simple random generation , compliant with given suitable constraints , with the aim to cover the overall search space for the ga . regarding the admissible values of the alleles encoding the mf parameters , we follow the method presented in @xcite . as concerns the constraints for the weights of the fuzzy rules , @xmath80 , each one is allowed to range in @xmath31 $ ] . once we define the constraints , the initial values for each allele can be generated through the extraction of a uniformly distributed random variable within the interval constraints . for the control genes , as binary strings , the domain of the random variables is the binary valued set @xmath81 and the random extraction process assigns equal probability to the two possible values . the chromosome representation given in ( [ pbin - preal ] ) is based on the distinction between the control genes and the parametric genes . this separation leads to an heavy simplification of the design process for the genetic operators . it is well - known that the crossover operator is an important genetic operator that can effectively exchange and recombine genetic material from two selected parents . the hierarchical scheme allows to a semantic separation of the genetic material leading to design a mixed crossover operator consisting of a specific simple operator for control genes and a suitable operator for the parametric genes . hence , the adopted mixed crossover is realized as follows : * for the control genes @xmath82 it is designed a simple `` one cut - point '' crossover where the cutting point is set by randomly drawing from a uniform distribution ; * for the parametric genes @xmath83 it is applied a convex crossover operator " described by the following relations @xcite : @xmath84 where @xmath85 and @xmath86 are the offspring s chromosomes , while @xmath87 and @xmath88 are the parent s chromosomes and @xmath89 is a random value , such as @xmath90 $ ] . the mutation operator is crucial for the exploration capability of the overall optimization algorithm , allowing a suitable variability in the population . more specifically , the mutation operator is designed to change one or more genes depending on a suitable mutation rate used as threshold for a given probability value for the mutation of each allele . even for the mutation operator it is adopted a mixed scheme : + * for the binary section @xmath82 of the chromosome it is applied a single point mutation operator in which the mutation point is chosen randomly . + * for the real - valued section @xmath83 of the chromosome the non - uniform " mutation operator is performed . the genetic operator takes into account the age of the given population in a way that the mutation effect vanishes with the progression of the generations . more specifically , given a chromosome @xmath83 for a selected gene @xmath91 , the resulting offspring gene is generated choosing randomly among the following relations @xcite : + + @xmath92 + where @xmath93 and @xmath94 are , respectively , the upper and lower bounds for @xmath91 , @xmath95 $ ] is a random uniformly distributed real value , _ g _ is the current generation index and _ g _ the predefined maximum number of generations . finally , @xmath96 is a parameter determining the decay speed . we remark that the main difference between the two evolutionary approaches , namely classic fuzzy - ga and fuzzy - hga , lies in a higher number of degrees of freedom available for the fis optimization . in fact , a composite encoding scheme capable of turning off / on each fuzzy set allows a more effective exploration of the search space . if we call @xmath97 the set of all fis@xmath98 depending on a set of structural parameters @xmath99 , where the subscript @xmath100 refer to the membership function parameters , the rule weights and the fuzzy rules , respectively , the fuzzy - hga scheme is able to explore a subset @xmath101 . borrowing the linear algebra jargon , the fuzzy - hga scheme can explore a subspace of the overall space which hosts the fiss . @xmath97 can contain fiss with poorly performing fuzzy rules that the ga could evaluate in order to optimize the objective function , lowering in this way the overall search efficiency . in fact , among a huge set of fuzzy rules , many of them can be useless ( for example because are centered onto input space regions corresponding to infrequent system states ) , or even harmful for the overall performances ( for example , if the parameters of the related membership functions are not well tuned ) . thus , according to the occam s razor criterion , a tuning procedure able to perform a reduction of the structural complexity of the rule base ( and thus of the fis ) is fundamental to improve the overall performances of the ems this section reports several experimental evaluations with the aim to measure the performance of the fuzzy - hga control system adopted for the energy flow optimization task in the proposed mg model . specifically the performance of the proposed flc is measured in a suitable scenario on a time - span of one year reporting the results for several configurations of the energy storage device connected to the mg and for two mutation parameters of the given ga . finally , it is provided a comparison with the basic fuzzy - ga paradigm described in [ sec : simple_scheme ] and studied in our previous work @xcite , where only the mf parameters and the rule weights are tuned . the scenario considered for experiments consists of a medium dimension mg . the basic scheme is depicted in fig . [ fig::mg_scheme ] . the mg model is equipped with a distributed resources like a photo - voltaic generator ( pvgs ) and a wind generator ( wpgs ) . the model is completed with an energy storage capacity of 24 kwh and the control system ( ems ) based on fuzzy - hga paradigm where a hierarchical evolutionary algorithm ( hga ) is in charge of learning the flc parameters and , at the same time , the number of adopted fuzzy rules . the control system manages the energy flow of the mg in input and output maximizing the accounting profit , dispatching control actions depending on : the energy balance , the soc of the battery and the price signal of the energy [ sec : pf ] . experiments are performed on real - world data set provided by acea distribuzione s.p.a . , the electric distribution utility that manages the power grid of rome in italy . the electricity demand and production time profiles regard measurements in a time - span of one year and are sampled every 15 minutes . the overall dataset has been split in two disjoint set , _ training set _ and _ test set _ , considering the odd and the even time samples , respectively . the parameters of the flc are learned on the training set and the performances of the ems are evaluated on the test set . the daily cost profiles , in generic monetary units ( mu ) , are reported in fig . [ fig : cost - prof ] . it is noted that the adopted electricity prices are multi - hours tariffs . in fact , the selling price is high during the day when the demanded energy is high , while it is lower in the night when the power is generated by only the wind source . the model of the energy storage capacity is the one described in sec . [ sec : bm ] referring to the circuital scheme of fig . [ fig : batt_scheme ] . the storage unit used for experiments is modeled on the basis on a scib battery , developed by toshiba , with storage capacity of 24 kwh . simulations are performed on the basis of two main configurations resumed in tab . [ tab : config - batt ] . the defuzzification method used is the mean of maximum ( mom ) . for the two fiss , the starting configuration is depicted in fig . [ fig : fisalfa - orig ] . the fuzzy rules optimization task is obtained by a hierarchical genetic algorithm ( hga ) as described in sec . [ sec : hga ] . the initial population is composed by 40 individuals . the genetic operators used for the proposed hga are described in sec . [ sec : mixed_g_o ] . the selection operator is the stochastic uniform " , while the crossover and the mutation operators are of mixed type , such as the one point " crossover and mutation adopted for the control genes , while for the parametric genes a convex " crossover ( [ eq : xover - conv2 ] ) and a non - uniform " mutation ( [ eq : non_uniform_mut ] ) are used . the decay speed @xmath96 is set to 1 , the crosover fraction " is 0.8 , while the mutation rate @xmath102 is evaluated for two values : 0.001 and 0.01 . finally the proposed objective function is given by expression ( [ eq : obj_f ] ) . as preliminary experiments we compare the two battery configurations resumed in tab . [ tab : config - batt ] with two scenarios characterized by two different mutation rate values : @xmath103 . the electricity prices , as input data , are those described above in sec . [ sec : scenery ] and range in the entire year . the hga searching results within the config . n 1 and @xmath104 , leaded to the mfs depicted in figs . [ fig : fisalfa - mf - ott ] , [ fig : fisbeta - mf - ott ] with the following control chromosome sections [ sec : mcr ] : @xmath105 ; \\ p_{bin}(\beta)=[0\;0\;1\;1\;1\;|\;1\;1\;1\;0\;1\;|\;1\;0\;1\;1\;1].\end{aligned}\ ] ] parameter for config . n 1 . ] parameter for config . n 1 . ] results reported in tab . [ tab : annual - test ] show how the flc obtains good accounting profit performances also using an higher mutation rate . it is noted how the config . n 2 , characterized by a higher efficiency and a lower threshold limit ( @xmath106 ) for the soc , leads to a further improvement of the accounting profit due to the lower expenses related to energy purchase . the main causes can be found in : @xmath0 a better use of the energy storage system with an high value of the efficiency that allows to store a greater amount of electric charge leading to lower losses ; @xmath1 the introduced threshold value constrains the controller to undertake crisp actions " charging the battery when the soc is lower than @xmath12 . the above described behavior can be observed in figs . [ fig : test2-perf ] , [ fig : test3-perf ] where are reported , for a suitable time interval : the soc , the power production , the power consumption , the energy purchased and the energy sold for the second and third configuration , respectively , and reported in tab . [ tab : annual - test ] . [ fig : test23-power - batt ] shows , for the same configurations , the electric power drawn from and stored in the battery , while in fig . [ fig : test23-revexp ] are depicted the expenses , the revenues and the accounting profit time profiles of one sample day . as concerns the number of rules returned by the fuzzy - hga algorithm for all the configurations evaluated , it has been found lower than the maximum value corresponding to the complete coverage of the input space using all the available mfs . the following section proposes a performance comparison between the herein studied fuzzy - hga scheme and the classic fuzzy - ga scheme reported in @xcite and resumed in sec . [ sec : simple_scheme ] . as concerns the fuzzy - hga scheme , we have the same considerations reported in sec . [ sec : scenery ] . both systems are provided with same input data and parameters as concerns the mg model . the number of ga generations is fixed to the value of 100 and the initial population , initialized randomly , consists of 40 individuals . the fitness function is the one reported in ( [ eq : obj_f ] ) . for the configuration of the energy storage system both mgs are provided with an accumulator with 24 kwh capacity whose configuration is the n 2 ( @xmath104 ) reported in tab . [ tab : config - batt ] . figs . [ fig : test - old-3-perf ] and [ fig : test - old-3-revexp ] show the time profiles of the battery soc together with the power produced and demanded , the energy sold and purchased and the overall financial accounting performances . results in tab . [ tab : test - old-3 ] show an increase of performance for the fuzzy - hga system compared with the classic fuzzy - ga approach . in fact the hierarchical structure of the ga allows to learn , in addition to the shape of mfs and the rules weights , also the number of rules searching for a suitable cardinality of the rb . this greater degree of freedom is obtained allowing more adaptation capability of the fis to the environment with the cost of a greater complexity of the whole ga algorithm . however , this increment in the learning procedure complexity has the advantage of reducing the structural complexity of the optimized fis , in terms of both the number of rules and the number of antecedents , as shown for instance in fig . [ fig : fisalfa - mf - ott ] and fig . [ fig : fisbeta - mf - ott ] . this results in simpler fiss , lowering the overall computational cost of the ems when taking decisions in real time . computational intelligence techniques are today a consolidated framework for solving engineering problems such as challenges arising in smart grid context . in this paper we study a portion of a smart grid , acting as a microgrid , characterized by the presence of renewable sources and equipped with a bess unit allowing trading operations with the main - grid related to energy exchanges . the considered mg model has been sized for small - scale applications , such as small energy grids in rural areas or small housing units in urban areas , considered as atomic elements of a larger a multi - microgrid system . the decision making task is carried on through a fuzzy logic controller optimized by a hierarchical genetic algorithm able to tune flcs parameters ( i.e. the mf parameters and the fuzzy rule weights ) minimizing , at the same time , the number of fuzzy rules . the fuzzy rules number optimization is carried out in flcs thanks to the hierarchical structure of the adopted chromosome representation consisting of two main sections having a different semantic meaning . the control section , encoded as a binary string , controls the activation of the parametric section representing the mf parameters and the weights of fuzzy rules . the specific encoding scheme of the hga offers a higher flexibility in tuning the flc in comparison with the classic scheme which foresees a fixed and maximal number of rules due to the rigidity of the chromosome length . the proposed method for ems synthesis incorporates a learning step . in fact , once fis rules are optimized by considering training set data through an off - line procedure running on a plain workstation , the trained rules can be uploaded into an embedded system ( a microcontroller ) ready to work in real - time . from this point of view , the added complexity of the hga scheme can be well managed off - line by an inexpensive workstation ( such as an intel i7 6thh generation with 32 gb of ram ) , exploiting its power in evolving suitable rules without affecting the real time operations . note that the final aim of the hga algorithm is the reduction of the structural complexity of the rule base in the fis , dropping the number of rules and the number of antecedents in each rule . consequently , with respect to a fis resulting from a learning procedure based on a plain version of a genetic algorithm , the microcontroller is in charge to handle a lower number of rules in performing decision making , in a given time interval . moreover , note that the typical dynamics characterizing a microgrid ( at the considered abstraction level ) are slow enough to set the length of the time interval in the order of minutes ( 15 minutes , in the considered case ) . consequently , even an entry level and inexpensive microcontroller ( such as an `` arduino due '' board , based on the atmel sam3x8e arm cortex - m3 cpu ) has sufficient computational power to handle a great number of fuzzy rules in such time intervals . the fuzzy - hga paradigm is thereby evaluated for some configuration of the mg parameters , battery model parameters and genetic operator parameters with the aim to study the overall behavior of the fuzzy control system . a further comparison is performed between a classic optimization scheme within the fuzzy - ga paradigm and the adopted fuzzy - hga . the classic ga scheme foresees the optimization of the mf parameters together with the fuzzy rule weights in a fixed rule base ( rb ) whose cardinality is maximal ( grid partition ) . the fuzzy - hga algorithm outperforms the classic fuzzy - ga scheme by 67% . the hierarchical encoding method of the fis parameters and the definition of suited genetic operators explain the improvement in performance reached at the cost of a more complex design of the ga and of the encoding scheme . however , this additional computational cost at design and training stages is balanced with the possibility to obtain simpler flcs in terms of number of rules , allowing the real time implementation of the complete control system on low cost embedded electronic devices . results are encouraging and suggest to work in future developments on both mg modeling and learning . future developments foresee the definition of suitable measures , such as i ) the stress of the battery in charging and discharging processes , ii ) the stress of the main - grid in terms of undesired fast changes of power flow between the mg and the main - grid . these measures can be considered as additional objectives ( that are likely conflicting with the main one based on prosumer profit ) aiming at controlling both the stress of the storage system and that of the main - grid . a mono - objective or a muti - objective ga scheme can be adopted taking into account the possibility of tuning the fuzzy controller allowing different behaviors , such as promoting the auto - consumption , lowering the battery stress or the workload on the main - grid . furthermore , an interesting future scenario consists in performing a spatial granulation of the whole considered distribution network , where each territorial granule acts as a mg , each one able to exchange energy with other spatially close mgs and/or with the main - grid , with different pricing programs . the mgs can have different characteristics in terms of aggregate demand , production , energy source type and storage systems . moreover , to increase the model reliability , it is interesting to consider power losses in energy exchange between mgs . the intrinsic heterogeneity of mgs together with the loss factors and the learning features fulfilled within the fuzzy - hga paradigm allow to study the behavior of the mg acting as agents , with possible emerging cooperative behavior in satisfying objective functions defined by a higher level controller . hence , in this augmented scenario some of the initial assumption adopted through the model can be relaxed and the mgs can be equipped with a suitable model of telecommunication network , taking into account also the capacity constraints in data delivery . moreover the mg can foresee a demand composed of schedulable loads such as smart appliances and electric vehicles , allowing to design suitable demand response ( dr ) programs able to act together with the flc aiming to optimize predefined objectives . from the learning side , it is really interesting to apply other nature inspired methods such as autoimmune or bacterial optimization algorithms for the optimization of the flcs . furthermore , a flc with a sugeno inference core can be adopted as the main fuzzy decision module and the hga could be suitably adapted in order to obtain a fair comparison with the mamdani controller . m. ansari , a. t. al - awami , m. abido , and e. sortomme . optimal charging strategies for unidirectional vehicle - to - grid using fuzzy uncertainties . in _ t d conference and exposition , 2014 ieee pes _ , pages 15 , april 2014 . doi : 10.1109/tdc.2014.6863314 . d. arcos - aviles , j. pascual , l. marroyo , p. sanchis , and f. guinjoan . fuzzy logic - based energy management system design for residential grid - connected microgrids . _ ieee transactions on smart grid _ , pp0 ( 99):0 11 , 2016 . issn 1949 - 3053 . doi : 10.1109/tsg.2016.2555245 . y. y. f. c. n. ko , t. l. lee and c. j. wu . simultaneous auto - tuning of membership functions and fuzzy control rules using genetic algorithms . in _ systems , man and cybernetics , 2006 . smc06 . ieee international conference on _ , volume 2 , pages 11021107 . ieee , 2006 . d. cartes , j. chow , d. mccaugherty , s. widergren , and g. venayagamoorthy . the ieee computer society smart grid vision project opens opportunites for computational intelligence . in _ evolving and adaptive intelligent systems ( eais ) , 2013 ieee conference on _ , pages 144150 , april 2013 . doi : 10.1109/eais.2013.6604117 . s. chakraborty , t. ito , t. senjyu , and a. saber . intelligent economic operation of smart - grid facilitating fuzzy advanced quantum evolutionary method . _ sustainable energy , ieee transactions on _ , 40 ( 4):0 905916 , oct 2013 . issn 1949 - 3029 . doi : 10.1109/tste.2013.2256377 . a. chaouachi , r. m. kamel , r. andoulsi , and k. nagasaka . multiobjective intelligent energy management for a microgrid . _ ieee transactions on industrial electronics _ , 600 ( 4):0 16881699 , april 2013 . issn 0278 - 0046 . doi : 10.1109/tie.2012.2188873 . a. chehri and h. t. mouftah . : fuzzy - based energy management system for green houses using hybrid grid solar power . _ journal of renewable energy _ , 2013:0 e785636 , 2013 . issn 2314 - 4386 . doi : 10.1155/2013/785636 . c. chen and s. duan . optimal integration of plug - in hybrid electric vehicles in microgrids . _ industrial informatics , ieee transactions on _ , 100 ( 3):0 19171926 , aug 2014 . issn 1551 - 3203 . doi : 10.1109/tii.2014.2322822 . o. cordn , f. herrera , f. hoffmann , and l. magdalena . _ genetic fuzzy systems : evolutionary tuning and learning of fuzzy knowledge bases _ , volume 19 of _ advances in fuzzy systems applications and theory_. world scientific publishing co. pte , 1-st edition , 2001 . e. de santis , a. rizzi , a. sadeghian , and f. m. frattale mascioli . genetic optimization of a fuzzy control system for energy flow management in micro - grids . in _ 2013 joint ifsa world congress and nafips annual meeting _ , pages 418423 . ieee , 2013 . e. de santis , a. rizzi , a. sadeghian , and f. m. f. mascioli . a learning intelligent system for fault detection in smart grid by a one - class classification approach . in _ neural networks ( ijcnn ) , 2015 international joint conference on_. ieee , 2015 . j. p. fossati , a. galarza , a. martn - villate , j. m. echeverra , and l. fontn . optimal scheduling of a microgrid with a fuzzy logic controlled storage system . _ international journal of electrical power and energy systems _ , 68:0 61 70 , 2015 . issn 0142 - 0615 . doi : http://dx.doi.org/10.1016/j.ijepes.2014.12.032 . h. he . toward a smart grid : integration of computational intelligence into power grid . in _ neural networks ( ijcnn ) , the 2010 international joint conference on _ , pages 16 , july 2010 . doi : 10.1109/ijcnn.2010.5596724 . x. hu , r. xiong , and b. egardt . model - based dynamic power assessment of lithium - ion batteries considering different operating conditions . _ industrial informatics , ieee transactions on _ , 100 ( 3):0 19481959 , aug 2014 . issn 1551 - 3203 . doi : 10.1109/tii.2013.2284713 . c. ioakimidis , l. oliveira , and k. genikomsakis . wind power forecasting in a residential location as part of the energy box management decision tool . _ industrial informatics , ieee transactions on _ , 100 ( 4):0 21032111 , nov 2014 . issn 1551 - 3203 . doi : 10.1109/tii.2014.2334056 . g. iyer , p. agrawal , e. monnerie , and r. s. cardozo . performance analysis of wireless mesh routing protocols for smart utility networks . in _ 2011 ieee international conference on smart grid communications ( smartgridcomm ) _ , pages 114119 , oct 2011 . doi : 10.1109/smartgridcomm.2011.6102301 . r. jang and c .- t . _ neuro - fuzzy and soft computing : a computational approach to learning and machine intelligence_. prentice - hall , inc . , upper saddle river , nj , usa , 1997 . isbn 0 - 13 - 261066 - 3 . w. v. leekwijck and e. e. kerre . defuzzification : criteria and classification . _ fuzzy sets and systems _ , 1080 ( 2):0 159 178 , 1999 . issn 0165 - 0114 . doi : http://dx.doi.org/10.1016/s0165-0114(97)00337-0 . x. li , l. xu , j. hua , x. lin , j. li , and m. ouyang . power management strategy for vehicular - applied hybrid fuel cell / battery power system . _ journal of power sources _ , 1910 ( 2):0 542 549 , 2009 . issn 0378 - 7753 . doi : http://dx.doi.org/10.1016/j.jpowsour.2009.01.092 . x. li , d. hui , x. lai , m. ouyang , and y. min . fuzzy logic based coordinated controller for wind / battery / idsms hybrid micro - grid power system . in _ computer engineering and technology ( iccet ) , 2010 2nd international conference on _ , volume 4 , pages v4263v4267 , april 2010 . doi : 10.1109/iccet.2010.5485606 . x. li , j. li , l. xu , m. ouyang , x. han , l. lu , and c. lin . online management of lithium - ion battery based on time - triggered controller area network for fuel - cell hybrid vehicle applications . _ journal of power sources _ , 1950 ( 10):0 3338 3343 , 2010 . issn 0378 - 7753 . doi : http://dx.doi.org/10.1016/j.jpowsour.2009.11.099 . x. luo , j. wang , m. dooner , and j. clarke . overview of current development in electrical energy storage technologies and the application potential in power system operation . _ applied energy _ , 1370 ( 0):0 511 536 , 2015 . issn 0306 - 2619 . doi : http://dx.doi.org/10.1016/j.apenergy.2014.09.081 . e. mamdani and s. assilian . an experiment in linguistic synthesis with a fuzzy logic controller . j. hum .- comput . _ , 510 ( 2):0 135147 , aug . 1999 . issn 1071 - 5819 . doi : 10.1006/ijhc.1973.0303 . y. manjili , r. vega , and m. jamshidi . cost - efficient environmentally - friendly control of micro - grids using intelligent decision - making for storage energy management . _ intelligent automation & soft computing _ , 190 ( 4):0 649670 , 2013 . doi : 10.1080/10798587.2013.842346 . y. s. manjili , a. rajaee , m. jamshidi , and b. t. kelley . fuzzy control of electricity storage unit for energy management of micro - grids . in _ world automation congress ( wac ) , 2012 _ , pages 16 , june 2012 . a. mohamed and o. mohammed . real - time energy management scheme for hybrid renewable energy systems in smart grid applications . _ electric power systems research _ , 96:0 133 143 , 2013 . issn 0378 - 7796 . doi : http://dx.doi.org/10.1016/j.epsr.2012.10.015 . s. m. nosratabadi , r .- a . hooshmand , and e. gholipour . a comprehensive review on microgrid and virtual power plant concepts employed for distributed energy resources scheduling in power systems . _ renewable and sustainable energy reviews _ , 67:0 341363 , 2017 . s. d. ramchurn , p. vytelingum , a. rogers , and n. r. jennings . putting the smarts into the smart grid : a grand challenge for artificial intelligence . , 550 ( 4):0 8697 , apr . issn 0001 - 0782 . doi : 10.1145/2133806.2133825 . l. roin , k. therani , y. s. manjili , and m. jamshidi . microgrid energy management system using fuzzy logic control . in _ 2014 world automation congress ( wac ) _ , pages 462467 , aug 2014 . doi : 10.1109/wac.2014.6936001 . e. d. santis , l. livi , a. sadeghian , and a. rizzi . modeling and recognition of smart grid faults by a combined approach of dissimilarity learning and one - class classification . _ neurocomputing _ , 170:0 368 383 , 2015 . issn 0925 - 2312 . doi : http://dx.doi.org/10.1016/j.neucom.2015.05.112 . m. r. souryal and n. golmie . analysis of advanced metering over a wide area cellular network . in _ 2011 ieee international conference on smart grid communications ( smartgridcomm ) _ , pages 102107 , oct 2011 . doi : 10.1109/smartgridcomm.2011.6102299 . k. b. wagholikar and a. w. deshpande . fuzzy relation based modeling for medical diagnostic decision support : case studies . j. know .- based intell . _ , 120 ( 5,6):0 319326 , dec . 2008 . issn 1327 - 2314 . p. werbos . computational intelligence for the smart grid - history , challenges , and opportunities . _ computational intelligence magazine , ieee _ , 60 ( 3):0 1421 , aug 2011 . issn 1556 - 603x . doi : 10.1109/mci.2011.941587 . z. yang , j. zhang , m. c. w. kintner - meyer , x. lu , d. choi , j. p. lemmon , and j. liu . electrochemical energy storage for green grid . _ chemical reviews _ , 1110 ( 5):0 35773613 , 2011 . doi : 10.1021/cr100290v . pmid : 21375330 . |
we use the moduli space approximation to study the time evolution of magnetically charged configurations in a theory with an su( ) gauge symmetry spontaneously broken to u(1)su()u(1 ) . we focus on configurations containing two massive and massless monopoles . the latter do not appear as distinct objects , but instead coalesce into a cloud of non - abelian field . we find that at large times the cloud and the massless particles are decoupled , with separately conserved energies . the interaction between them occurs through a scattering process in which the cloud , acting very much like a thin shell , contracts and eventually bounces off the cores of the massive monopoles . the strength of the interaction , as measured , e.g. , by the amount of energy transfer , tends to be greatest if the shell is small at the time that it overlaps the massive cores . we also discuss the corresponding behavior for the case of the su(3 ) multimonopole solutions studied by dancer . epsf + + + + + + | hep-th0105211 | there has long been an interest in the magnetic monopoles that arise as classical solutions in certain spontaneously broken gauge theories . when the theory is quantized , these give rise to magnetically charged particles that can be regarded as the counterparts of the electrically charged elementary quanta of the theory . in certain supersymmetric theories , these two classes of particles are believed to be related by an exact duality symmetry @xcite . when the gauge symmetry is spontaneously broken , the gauge bosons corresponding to the generators of the unbroken subgroup remain massless . if the unbroken subgroup contains a non - abelian factor , some of these massless gauge bosons transform nontrivially under the gauge group and thus carry an `` electric '' charge . duality suggests that these should have massless magnetically - charged counterparts . although these can not be realized as isolated classical solutions , evidence for their existence can be found by analyzing multimonopole solutions . in this paper we will investigate some of the properties of these massless monopoles by examining the role that they play in low energy scattering processes . recall that an arbitrary magnetically charged bogomolnyi - prasad - sommerfield ( bps ) @xcite solution can be naturally understood as being composed of a number of fundamental monopoles of various types @xcite . if the gauge symmetry @xmath4 is broken to an abelian subgroup u(1)@xmath5 , as happens for generic choices of the adjoint higgs vacuum expectation value , there is an integer - valued topological charge for each u(1 ) factor . associated to each of these is a fundamental monopole carrying a single unit of that topological charge . each of these fundamental monopoles can be realized as a classical solution by embedding the unit su(2 ) monopole solution in an appropriate subgroup of @xmath4 . for the special choices of the higgs vacuum expectation value that give a non - abelian unbroken subgroup , the bps mass formula implies that some of the fundamental monopoles should become massless . although the corresponding embedding solutions reduce to pure vacuum solutions in this limit , analysis of multimonopole solutions suggests that the collective coordinates of the massless fundamental monopoles survive as degrees of freedom . examining static classical solutions with magnetic charges corresponding to a sum of massive and massless monopoles , one finds that the massive monopoles are surrounded by one or more `` clouds '' of non - abelian field @xcite . the number of collective coordinates @xcite needed to describe these clouds is exactly equal to the number that one would have attributed to the massless monopoles . to make these ideas more concrete , consider the case of an su(@xmath0 ) gauge theory with an adjoint higgs field whose asymptotic value ( in some fixed direction ) can be brought into the form @xmath6 with @xmath7 . the asymptotic magnetic field in the same direction can be written as @xmath8 , where the magnetic charge @xmath9 is of the form @xmath10 the @xmath11 are all integers , and give the number of fundamental monopoles of the various types . the mass of the @xmath12th species of fundamental is proportional to @xmath13 . hence , if the @xmath14 are all distinct , so that the symmetry is broken maximally , to u(1)@xmath15 , the @xmath16 species of fundamental monopoles are all massive . if @xmath17 of the @xmath18 are equal , there is an enlarged unbroken symmetry , with @xmath19 of the u(1 ) factors being replaced by an su(@xmath20 ) . we will focus on solutions containing one of each species of monopole , with @xmath21 . the unbroken group is then u(1)@xmath1su(@xmath2)@xmath1u(1 ) , and our solutions contain two massive and @xmath3 massless monopoles . this example has two advantages for us . first , a number of analytic results about the static solutions and their moduli space are already known . second , because the two massive monopoles correspond to commuting su(2 ) subgroups of the su(@xmath0 ) , they can have no direct interactions . any interaction between them must be due to the mediation of the massless monopoles , and thus can give us insight into the properties of the cloud . explicit solutions for an su(@xmath0 ) theory with the @xmath11 all equal to unity were obtained in ref . @xcite . for either type of symmetry breaking , these solutions are described by @xmath22 collective coordinates . when the fundamental monopoles are all massive , these have a natural interpretation as the spatial coordinates , @xmath23 , and u(1 ) phases of the @xmath16 constituent monopoles ; as long as the separation between the monopole positions is large compared to the monopole core sizes , the multimonopole nature of the configuration is evident in the classical solution . as the higgs field approaches the value corresponding to the non - abelian breaking , monopoles 1 and @xmath16 , which remain massive , clearly retain their identity . has @xmath24 and corresponds to an su(2 ) embedding in the @xmath25 and @xmath26 rows and columns of the su(@xmath27 matrices . ] however , the @xmath3 monopoles that become massless are replaced by a single cloud surrounding the massive monopoles ; this cloud can be viewed as being formed by the merger of the cores of the massless monopoles . inside this cloud , the magnetic field is approximately equal to the the coulomb field appropriate to the charges of the massive monopoles ; it has both abelian and non - abelian components . outside the cloud , the coulomb component of the magnetic field is purely abelian , with the non - abelian contributions falling with a higher power of distance . the size of this cloud is characterized by a `` cloud parameter '' @xmath28 that is equal to the sum of the separations @xmath29 between the monopoles . any value @xmath30 , the separation of the two massive monopoles , is possible , and the static energy is independent of @xmath28 . all other trace of the massless monopole positions appears to be lost in this limit . in particular , the fields show no special behavior at the @xmath23 corresponding to the massless monopoles . in fact , any change in these `` positions '' that leaves @xmath28 invariant can be compensated by a gauge transformation of the solution . our goal in this paper is to gain further insight into the nature of these massless monopoles by examining the role that the cloud plays in monopole scattering . to do this we use the moduli space approximation @xcite , which reduces the dynamics of the infinite number of field degrees of freedom to that of a finite number of collective coordinates @xmath31 . the lagrangian for this reduced set of variables can be written as @xmath32 where @xmath33 is a naturally defined metric on the moduli space of solutions . the evolution of the collective coordinates is simply geodesic motion in this metric . when the monopoles are all massive , this reduction to a finite number of degrees of freedom fits quite well with the point of view that particles arising from classical soliton solutions should be seen as having a similar status to those arising as quanta of the elementary fields , and that the solutions with higher charge should be interpreted as multiparticle states . when some of the monopoles become massless , the classical solutions lose their manifest multiparticle nature . nevertheless , the massless monopole collective coordinates survive ( although with their physical interpretations modified ) and the moduli space lagrangian has a smooth limit @xcite , provided that the total magnetic charge is abelian . the moduli space approximation is usually expected to be valid in the limit of small monopole velocities . this ensures that the monopole kinetic energy is small enough that the excitation of the nonzero - frequency modes of the massive fields is energetically suppressed . although simple energetic arguments are not sufficient to rule out excitation of any massless fields that might be present , more detailed analysis @xcite shows that the approximation holds if the long - range fields are all abelian . as we will see , the situation with massless non - abelian fields is more complex . to use the moduli space approximation , we need the metric on the space of multimonopole solutions . the exact form of this metric is known for only a relatively few cases . however , these include @xcite the case where there is at most one of each species of fundamental monopole , which is precisely what we need . unfortunately , the form of this metric that was obtained in ref . @xcite was expressed in terms of the positions and u(1 ) phases of the @xmath34 individual monopoles . although these are the natural coordinates to use when the monopoles are all massive , they are much less useful in the limit we are interested in , where the massless monopole positions have little direct physical meaning . a more natural set of coordinates would include the positions and u(1 ) phases of the two massive monopoles , the cloud parameter @xmath28 , and the parameters needed to specify the global su(@xmath2 ) orientation of the cloud . thus , our first task is to change to a new set of variables . we do this in two steps . in sec . ii we transform to an intermediate set of coordinates that allows us to separate the metric into two parts , one that gives the terms in the lagrangian containing @xmath35 and @xmath36 , and an `` angular '' part that contains the entire dependence on the spatial and gauge orientation angles and phases . the natural next step would be to relate these intermediate coordinates to the spatial euler angles and their gauge counterparts . this turns out to be somewhat nontrivial . however , we can bypass this step by rephrasing the problem in terms of the corresponding angular momenta and charges . we do this in sec . iii , leaving the discussion of the euler angles to the appendix . with these results in hand , we are then ready to determine the geodesics of the metric and the behavior of the monopoles in scattering processes . this is described in sec . we summarize our results and make some concluding remarks in sec . the moduli space of solutions for a system comprising two distinct fundamental monopoles was obtained in ref . @xcite ; the extension to any number of distinct monopoles was conjectured in ref . @xcite and proven in ref . locally , this space can be written as a product of a flat 4-dimensional space , corresponding to the center - of - mass position and an overall u(1 ) variable , and a relative moduli space . for the case of interest to us , an su(@xmath0 ) theory with one monopole of each species , the moduli space is @xmath22-dimensional , and the @xmath37 coordinates for the relative moduli space can be chosen to be the relative positions @xmath38 ( @xmath39 ) and the corresponding relative u(1 ) phases @xmath40 . in ref . @xcite it was shown that when the @xmath41 symmetry is broken to u(1)@xmath1su(@xmath2)@xmath1u(1 ) this metric can be written as @xmath42 ^ 2 + \kappa \sum_a \left[\frac{1}{r_a } d{\bf r}_a^2 + r_a ( d\psi_a+\cos\theta_ad\phi_a)^2\right ] - \lambda \left[\sum_a r_a\,(d\psi_a+\cos\theta_ad\phi_a)\right]^2 \label{rametric}\ ] ] where @xmath43 , @xmath44 , and @xmath45 are the spherical coordinates of the vector @xmath46 , @xmath47 is the reduced mass of the two massive monopoles , and @xmath48 \ , \,.\end{aligned}\ ] ] here @xmath49 ( with @xmath50 the gauge coupling of the theory ) is the magnitude of the fundamental monopole magnetic charge , while @xmath51 is approximately equal to the sum of the core radii of the two massive monopoles . the difficulty with using this expression for our purposes is that , as we noted in sec . i , the positions of the massless monopoles do not have any direct physical meaning . we would therefore like to transform to a more physically meaningful set of variables . this set should include the cloud parameter @xmath28 and the spherical coordinates @xmath52 , @xmath53 , and @xmath54 of the vector @xmath55 that gives the relative separation of the two massive monopoles , but must also include the relative u(1 ) phase and a number of su(@xmath2 ) orientation angles . one might expect that @xmath56 $ ] such angles would be needed to specify the gauge orientation of the solution . for @xmath57 this would require more collective coordinates than are available . however , examination of the explicit solutions shows that there is always a u(@xmath58 ) subgroup of the unbroken su(@xmath2 ) that leaves the fields invariant [ i.e. , any of these su(@xmath0 ) solutions is essentially an embedding of an su(4 ) solution ] . hence , there are only @xmath59 $ ] global su(@xmath2 ) parameters . together with the u(1 ) phases , the cloud size @xmath28 , and the positions of the two massive monopoles , this gives a total of @xmath22 collective coordinates , as required . we begin by defining complex coordinates and @xmath60 . ] @xmath61 these satisfy @xmath62 where @xmath63 and the @xmath64 are the pauli matrices . hence , the relative position of the two massive monopoles is and related quantities should be understood as being summed over . ] @xmath65 while @xmath66 when rewritten in terms of these coordinates , the metric takes the form @xmath67 ^ 2 + 4 \kappa \ , d\bar z\ , dz + \lambda\ , [ \bar z\ , dz -d\bar z\ , z]^2 \cr & = & \mu \,d{\bf r}^2 + 4 \kappa\ , d\bar z\ , dz + \lambda\ , [ \bar z\ , dz -d\bar z\ , z]^2 \ , \ , . \label{zmetric}\end{aligned}\ ] ] next , we find an su(2 ) matrix @xmath68 satisfying @xmath69 where @xmath70 . ( thus , @xmath68 corresponds to a rigid rotation of the entire assembly of monopoles that puts the two massive monopoles on the @xmath71-axis . ) this allows us to define a new set of complex variables @xmath72 in terms of which @xmath52 and @xmath28 are given by the simple expressions @xmath73 these identities imply that the @xmath74 must be of the form @xmath75 where the @xmath76 are a pair of complex @xmath2-component vectors obeying @xmath77 for later use , we also introduce @xmath58 orthonormal vectors @xmath78 ( @xmath79 ) orthogonal to the @xmath76 . we must now rewrite the metric in terms of these new variables . it will be helpful to introduce some new notation . first , we define two unit vectors @xmath80 and @xmath81 orthogonal to @xmath82 by requiring that @xmath83 next , we define @xmath84 and @xmath85 ( @xmath86 ) by @xmath87 and @xmath88 \ , \ , . \label{vjdef}\ ] ] finally , we define @xmath89 \ , \ , . \label{v0def}\ ] ] we also need two identities . first , by differentiating both sides of eq . ( [ uequation ] ) , one is led to @xmath90 next , viewing the @xmath91 as vectors in an @xmath2-dimensional complex vector space , we write @xmath92 \ , dw_b^a \label{dwdw}\ ] ] where @xmath93 is the projection operator onto the two - dimensional subspace spanned by @xmath94 and @xmath95 ( or by @xmath96 and @xmath97 ) and @xmath98 is the projection operator onto the orthogonal @xmath99-dimensional subspace . using eqs . ( [ widentities ] ) , ( [ wform ] ) , ( [ vjdef ] ) , and ( [ v0def ] ) , we obtain the second identity , @xmath100 ^ 2 \over 8 ( b+r ) } + { [ d(b - r)]^2 \over 8 ( b - r ) } + { b \over 4 ( b^2-r^2 ) } \sum_{\nu=0}^3 v_\nu^2 - { r \over 2(b^2-r^2 ) } ( v_0 v_3 ) + d\bar w \,\pi^t \,dw \,\ , .\ ] ] we have not yet obtained explicit expressions for @xmath53 , @xmath54 , or the gauge orientation phases in terms of the @xmath101 or @xmath68 . however , it is easy to see that neither the @xmath102 and @xmath103 nor the combination @xmath104 can contain any factors of @xmath105 or @xmath106 . with this in mind , and using our two identities and the above definitions , we substitute eq . ( [ ztow ] ) into eq . ( [ zmetric ] ) and obtain @xmath107 ^ 2 \over 2 ( b+r ) } + { [ d(b - r)]^2 \over 2 ( b - r ) } \right\ } + { \cal g}_{\rm ang } \label{metricdecomp}\ ] ] where @xmath108 + \left({\kappa b \over b^2-r^2 } - \lambda \right)\ , ( v_0 + r a_3)^2 \cr & & \qquad + { \kappa b \over b^2-r^2}\ , ( v_3 + ba_3)^2 - { 2\kappa r \over b^2-r^2}\ , ( v_0 + r a_3)(v_3 + ba_3 ) + 4 \kappa \ , d\bar w \,\pi^t \,dw \label{gangav}\end{aligned}\ ] ] contains the entire dependence of the metric on the spatial angles and the various gauge orientation phases . in the next section we will rewrite @xmath109 in a form that will allow us to analyze the evolution of these angle and gauge variables . before doing so , we comment on a point that we have skipped over . equation ( [ uequation ] ) does not completely determine the matrix @xmath68 , but instead leaves the freedom to make a further transformation of the form @xmath110 this implies a corresponding transformation @xmath111 of the @xmath101 , and rotates the unit vectors @xmath112 and @xmath113 so that @xmath114 this ambiguity in the definition of @xmath68 is not simply due to our failure to impose a sufficient number of conditions . instead , it reflects the fact that these solutions possess an axial symmetry . they are invariant under rotations about the axis joining the two massive monopoles , provided that these are combined with an appropriate global gauge transformation . because of this symmetry , there is no natural way to pick out any particular pair of axes in the plane perpendicular to @xmath115 . as a result , the form of the metric should be invariant under such redefinitions of @xmath68 . to verify that it is , we note that the auxiliary quantities @xmath102 and @xmath103 transform as @xmath116 substituting these results into eq . ( [ gangav ] ) , we see that the metric is indeed unchanged , thus providing a useful consistency check on our calculations . in this section we will bring eq . ( [ gangav ] ) for @xmath117 into a form more suitable for analyzing the dynamics of the angle and phase variables . the most straightforward method for doing this would begin by defining generalized euler angles to describe the spatial and the u(@xmath2 ) phase orientations . given these , one can construct a set of 1-forms @xmath118 invariant under both the spatial so(3 ) and the internal su(@xmath2 ) symmetries of the theory , and write @xmath119 the symmetries of the system imply the existence of a conserved angular momentum and conserved u(@xmath2 ) charges . the `` body - frame '' components of these quantities , defined by @xmath120 are the variables conjugate to the invariant one - forms . writing @xmath121 and then substituting this into eq . ( [ metricdecomp ] ) gives a convenient starting point for studying the dynamics of the system . a complication in our case is that , because of the extra symmetries of these multimonopole solutions , the number of angular variables is less than might be expected . for most choices of the magnetic charges , an arbitrary configuration would require 3 spatial euler angles and @xmath122 phases specifying the u(@xmath2 ) orientation . by contrast , the relative moduli space in our case is @xmath37-dimensional so that , after extracting @xmath52 and @xmath28 , we are left with only @xmath123 angle and phase variables . the `` missing '' @xmath124 variables are explained by extra invariances of the solutions we are considering . first , we noted previously , these su(@xmath0 ) solutions are essentially equivalent to embeddings of su(4 ) solutions , so that there is always a u(@xmath58 ) subgroup that leaves the configuration invariant . ) subgroup depends on the su(@xmath2 ) orientation of the configuration , and thus changes as the various phases vary with time . ] the @xmath125 `` euler angles '' corresponding to this subgroup play no role in the dynamics , and hence do not enter the lagrangian . second , the axial symmetry discussed at the end of the previous section implies that some linear combination of a spatial euler angle and a u(@xmath2 ) phase must also be absent from the metric . , but rather an infinitely thin top with @xmath126 and @xmath127 . ] the fact that these extra symmetries involve mixtures of spatial rotations and gauge transformations complicates the explicit construction of the @xmath128 appearing in eq . ( [ gomega ] ) ; they are not simply subsets of the standard invariant one - forms for so(3 ) and u(@xmath2 ) . fortunately , it turns out that we can bypass the construction of the @xmath128 and go directly from eq . ( [ gangav ] ) to an expression of the form given in eq . ( [ gwithiinv ] ) . we will follow this procedure in this section , leaving the discussion of the generalized euler angles and the @xmath128 to the appendix . we begin by identifying the conserved quantities of the system . recall that if a metric has a killing vector of the form @xmath129 then the charge @xmath130 is a constant along any geodesic . since the classical solutions of the moduli space lagrangian are all geodesics , these @xmath131 are the desired conserved charges . the killing vectors of the metric of eq . ( [ rametric ] ) were determined in ref . @xcite . written in terms of our complex variables @xmath132 , the rotational killing vectors take the form @xmath133 \label{rotkilling}\ ] ] while the @xmath122 killing vectors @xmath134 \label{unkilling}\ ] ] satisfy the algebra of u(@xmath2 ) . using eqs.([killingdef ] ) and ( [ chargedef ] ) , we construct from these the angular momentum @xmath135 + i \kappa \ , [ \bar z \ , \bsigma\ , dz - d \bar z \ , \bsigma \ , z ] - i \lambda \,[\bar z \ , dz - d\bar z \ , z ] ( \bar z \,\bsigma \,z ) \cr & = & \mu \ , { \bf r } { \bf \times } d{\bf r } + \kappa \ , \left [ i(\bar w \ , u \bsigma u^{-1 } \ , dw - d\bar w \ , u \bsigma u^{-1 } \ , w ) + i \bar w ( { \cal a } \ , u \ , \bsigma \ , u^{-1 } + u \bsigma u^{-1 } \,{\cal a})w \right ] \cr & & \qquad - \lambda \ , { \bf r}\ , ( v_0 + ra_3 ) \label{jequation}\end{aligned}\ ] ] ( where @xmath136 ) and the u(@xmath2 ) charges @xmath137 + i \lambda\ , \bar z_b \ , z_a \,[\bar z \ , dz - d\bar z \ , z ] \cr & = & \sqrt{2}\kappa \left [ i(d\bar w_b \ , w_a - \bar w_b \ , dw_a - \bar w_b \,\sigma_j \ , w_a \ , a_j \right ] + \sqrt{2 } \lambda \ , \bar w_b \ , w_a ( v_0 + ra_3 ) \ , \ , . \cr & & \label{eabequation}\end{aligned}\ ] ] the components implied by the bold - face notation in eq . ( [ jequation ] ) and by the subscripts in eq . ( [ eabequation ] ) are those corresponding to a fixed `` space - frame '' and are conserved . what we actually need are the components corresponding to a moving `` body - frame '' . for the angular momentum , this requires identifying three axes that rotate with the monopole system ; a suitable choice is @xmath80 , @xmath81 , and @xmath138 . the body frame components ( which are themselves rotational scalars ) are then @xmath139 in obtaining this , we have used eqs . ( [ nhatdef ] ) and ( [ vjdef ] ) . to construct a `` body - frame '' for the u(@xmath2 ) charges , we need a set of @xmath2 complex basis vectors in the internal space that transform under the fundamental representation under the action of the @xmath140 . we can take these to be the @xmath76 and the @xmath78 that were introduced in sec . 2 . using the @xmath76 and the pauli matrices , we can construct three u(@xmath2)-invariant charges @xmath141_{ab } \,p_b^{(b ) } \,{\cal e}_{ab } \ , \,.\ ] ] ( the definition of these has been chosen so that they correspond to generators of an su(2 ) subgroup with the standard normalization . ) straightforward calculations give @xmath142 \ , & $ i = 1 , 2 $ \cr \cr r \lambda ( v_0 + ra_3 ) - \kappa ( v_3 + b a_3 ) \ , & $ i = 3 $ \ , \ , . } \label{bodyt}\ ] ] comparing eqs . ( [ bodyj ] ) and ( [ bodyt ] ) , we see that @xmath143 which is a reflection of the axial symmetry of the solutions . additional `` body - frame '' components can be constructed by utilizing contractions with the @xmath78 . components involving two such contractions vanish ; i.e. , eigenvalues of the moment of inertia tensor vanish because of invariance under a u(@xmath58 ) subgroup . ] @xmath144 on the other hand , the charges @xmath145 and their complex conjugates @xmath146 are in general nonzero ; they are related to the last term in eq . ( [ gangav ] ) by @xmath147 \ , \ , . \label{dbarwpidw}\ ] ] finally , we define a conserved relative u(1 ) charge @xmath148 \,{\cal e}_{ab } \,dt = { 1 \over \sqrt{2 } } \sum_{a=1}^2 p_b^{(a ) } \ , \bar p_a^{(a ) } \ , { \cal e}_{ab } \,dt \cr & = & ( b \lambda - \kappa ) ( v_0 + r a_3 ) \ , \ , . \label{qdef}\end{aligned}\ ] ] using eqs . ( [ gangav ] ) , ( [ bodyj ] ) , ( [ bodyt ] ) , ( [ dbarwpidw ] ) , and ( [ qdef ] ) , we can now rewrite @xmath117 in terms of the body - frame components , obtaining @xmath149 \right . \cr & + & \left . { 1 \over \kappa(b^2 -r^2 ) } \left [ b ( j_3 ^ 2 + q^2 ) + 2 r j_3 q \right ] + { \mu \over \kappa^2}q^2 + { 4 \over \kappa } \sum_{r=3}^n \left [ { t_{1r } \bar t_{1r } \over b+r } + { t_{2r } \bar t_{2r } \over b - r } \right ] \right\ } dt^2 \ , \ , . \label{gangfinal}\end{aligned}\ ] ] it must be remembered that the `` body frame '' components that appear in this expression are not in general conserved . while the square of the angular momentum , @xmath150 the su(@xmath2 ) casimir , @xmath151 and the u(1 ) charge @xmath152 are constant , determining the time - dependence of the individual components of @xmath153 and @xmath154 is in general rather complex . there is considerable simplification , however , if the su(@xmath2 ) charges all vanish . the only nonzero charges remaining are then @xmath155 , @xmath156 , and @xmath152 , with the former two entering @xmath157 only through @xmath158 ; using the relation between @xmath159 and @xmath102 that follow from the vanishing of the @xmath160 , one finds that this is given by @xmath161 equation ( [ gangfinal ] ) then reduces to @xmath162 q^2 \right\ } dt^2 \ , \ , . \label{gangtzero}\ ] ] we are now ready to obtain the equations of motion of our system and study the behavior of the cloud and the massive monopoles in scattering processes . we restrict ourselves here to the case of vanishing su(@xmath2 ) charges , @xmath163 , leaving a discussion of the more general case to the appendix . because the massive monopoles have well - defined positions while the massless ones do not , there may be a tendency to view the various scattering processes as resulting from interactions between the massive monopoles , with the cloud dynamics being in the background . this would be incorrect . the massive monopoles correspond to two mutually commuting su(2 ) subgroups of su(@xmath0 ) . were it not for the presence of the cloud , they would not interact at all and would move on straight lines with constant velocity . hence , when following the motion of the massive monopoles it is important to also keep track of the size and motion of the cloud at the same time . equations ( [ metricdecomp ] ) and ( [ gangtzero ] ) provide a suitable starting point for obtaining the equations of motion of our system . because the angle and phase variables enter the latter only through conserved quantities , we can view @xmath109 as defining an effective potential for @xmath52 and @xmath28 that can be combined with the first two terms in eq . ( [ metricdecomp ] ) to yield an effective lagrangian , which gives a @xmath152-dependent increase in energy but has no effect on the evolution of @xmath28 or @xmath52 . ] @xmath164 - { b { \bf j}^2\over 2 r^2(\kappa + \mu b ) } - { b q^2\over 2\kappa(b^2 -r^2 ) } \ , \ , . \label{leffective}\ ] ] we begin our analysis of the solutions of this lagrangian by examining the asymptotic solutions at large @xmath52 and @xmath28 ; because the terms containing @xmath153 and @xmath152 both fall rapidly in this regime , it is sufficient to do this asymptotic analysis with @xmath165 . we find that the cloud and the massive monopoles decouple from each other at large times . after studying the properties of these asymptotic solutions , we then turn to the behavior at finite time , where the actual interactions take place , beginning with @xmath165 . finally , we consider the effects of nonzero @xmath153 and @xmath152 . before beginning this analysis , we must address the apparent singularity in the kinetic energy at @xmath166 . this singularity is not physical , and can be eliminated by defining @xmath167 \cr\cr y = \sqrt{\kappa } \left [ \sqrt{b+r } - \sqrt{b - r}\right ] } \ , \qquad\qquad 0 \le y \le x \ , \ , . \label{firstoctant}\ ] ] in terms of these coordinates , the lagrangian takes the nonsingular form @xmath168 where @xmath169 as we have indicated in eq . ( [ firstoctant ] ) , the entire physical range @xmath170 is mapped onto the octant @xmath171 of the @xmath172-@xmath173 plane . nevertheless , it will become clear that if either @xmath153 or @xmath152 vanish some trajectories can cross the boundaries of this octant . these can be understood as follows . the @xmath172-axis corresponds to @xmath174 . a trajectory crossing this line corresponds to a motion in which the two massive monopoles approach head - on , meet ( at @xmath175 ) , and then pass through one another . to accommodate this , we extend the definition of @xmath172 and @xmath173 into the next octant by @xmath176 \cr\cr y = \sqrt{\kappa } \left [ \sqrt{b - r } - \sqrt{b+r}\right ] } \ , \qquad\qquad 0 \le -y \le x \ , \,.\ ] ] similarly , a trajectory crossing the line @xmath177 corresponds to a motion in which the cloud shrinks to its minimum size , @xmath178 , and then begins to grow again . to describe this , we define @xmath179 \cr\cr y = \sqrt{\kappa } \left [ \sqrt{b+r } + \sqrt{b - r}\right ] } \ , \qquad\qquad 0 \le x \le y \ , \,.\ ] ] extending these definitions in a similar fashion to the other octants , we obtain an eight - fold mapping of the physical @xmath52-@xmath28 space onto the @xmath172-@xmath173 plane . throughout the plane we have the relations @xmath180 so that the @xmath172- and @xmath173-axes correspond to @xmath174 while the lines @xmath181 correspond to minimum cloud size , @xmath178 . if @xmath165 the lagrangian of eq . ( [ xylag ] ) reduces to @xmath182 the resulting euler - lagrange equations are @xmath183 multiplying the first of these by @xmath172 and the second by @xmath173 , and then subtracting , we find that @xmath184 . using this to rewrite eq . ( [ eleqs ] ) , we obtain @xmath185 there are two constants of the motion . the time - independence of the lagrangian implies the conservation of the energy ( which is actually equal to the lagrangian ) @xmath186 in addition , multiplication of the first of eq . ( [ neweleqs ] ) by @xmath187 and the second by @xmath188 shows that @xmath189 is also conserved . ( we have no deep understanding as to why @xmath190 is constant ; we have not found any generalization of @xmath190 that is conserved when either @xmath153 or @xmath152 is nonzero . ) if @xmath191 , neither @xmath188 nor @xmath187 can vanish , implying that both @xmath172 and @xmath173 vary monotonically ; if @xmath192 but @xmath193 , then one of @xmath188 or @xmath187 vanishes for all @xmath194 , while the other never vanishes . hence , there are no closed orbits in the @xmath172-@xmath173 plane . it is then straightforward to show that @xmath195 must tend to infinity as @xmath196 . there are two possibilities for the large time behavior . the first is that @xmath197 tends to a nonzero finite constant . examination of eq . ( [ neweleqs ] ) shows that this is possible only if @xmath198 , which can only happen if @xmath199 . this gives solutions where @xmath200 varies linearly with time , corresponding to a solution with a minimal ( @xmath178 ) cloud surrounding two massive monopoles with constant relative velocity . the other possibility is that @xmath201 tends to either 0 or @xmath202 ; because of the eight - fold mapping of @xmath28 and @xmath52 onto the @xmath172-@xmath173 plane , the two are are equivalent . we will analyze in detail the case where @xmath203 , so that asymptotically @xmath204 . by integrating the dominant terms in the second of eq . ( [ neweleqs ] ) , we find that @xmath205 asymptotically tends to a constant . substituting this result into eq . ( [ bdef ] ) for @xmath190 , we see that @xmath188 must tend to a constant . it is then straightforward to obtain the asymptotic solution @xmath206 where @xmath207 , @xmath208 , @xmath209 , and @xmath210 are constants of integration . of these constants , @xmath208 simply corresponds to a shift of the zero of time , while @xmath210 is related to the others by @xmath211 the meanings of @xmath209 and @xmath207 are clarified by writing the asymptotic solution in terms of @xmath28 and @xmath52 : @xmath212 we see that asymptotically the massive monopoles move with a constant relative velocity @xmath209 , while @xmath28 varies quadratically with time . approaches @xmath202 ( @xmath213 ) , the cloud parameter increases ( decreases ) at a rate that exceeds the speed of light ; we will return to this point below . ] substituting this asymptotic solution back into the energy , we find that @xmath214 as @xmath215 the terms indicated by dots vanish , while the two terms shown explicitly tend to constants . these latter two terms can be interpreted as a massive monopole energy @xmath216 and a cloud energy @xmath217 , with @xmath218 because of the multiple mapping of @xmath28 and @xmath52 into the @xmath172-@xmath173 plane , for each asymptotic solution in eq . ( [ asymsolution ] ) there are seven other physically equivalent solutions that are obtained by reversing the signs of @xmath172 or @xmath173 or by the interchange of @xmath172 and @xmath173 . having found the possible asymptotic behaviors , we now want to obtain solutions of eq . ( [ neweleqs ] ) for finite times . as with the asymptotic solutions , these will depend on four constants of integration . one of these can be absorbed in a rescaling of @xmath194 ( with a corresponding rescaling of @xmath219 and @xmath190 ) and a second in a shift of the zero of time . hence , the trajectories in the @xmath172-@xmath173 plane depend nontrivially only on two constants , which we take to be @xmath220 and @xmath221 ; without loss of generality , we can choose the trajectories to begin in the first octant , so that the latter quantity is equal to @xmath222 . equations ( [ xyenergy ] ) and ( [ bdef ] ) can be combined to yield a quadratic equation for @xmath223 . the solution of this equation gives @xmath197 as a function of @xmath172 , @xmath173 , and @xmath220 . hence , two trajectories with the same value of @xmath220 but different @xmath207 can not cross . while our numerical solutions indicate that the same may be true for trajectories with the same @xmath207 and different @xmath220 , we have no proof of this statement . we used numerical integration to obtain solutions for various values of @xmath207 and @xmath220 . we will begin by describing our results for @xmath224 ( i.e. , initially equal values of @xmath216 and @xmath217 ) . in fig . 1 , we show some trajectories in the @xmath172-@xmath173 plane for @xmath225 and several values of @xmath220 . figure 2 shows @xmath52 and @xmath28 as functions of time for these trajectories . note that @xmath52 and @xmath28 have been plotted in units of @xmath226 ; recall that this is approximately equal to the sum of the core radii of the two massive monopoles . in examining these solutions , it is useful to focus on the points where @xmath178 ( corresponding to crossing one of the 45-degree lines @xmath227 ) and where @xmath174 ( corresponding to crossing either of the coordinate axes in the @xmath172-@xmath173 plane ) . when @xmath178 , the cloud is of minimal size , . in most cases @xmath228 when @xmath28 reaches its minimum , so that the smallest value of @xmath28 corresponds to a nonminimal cloud . ] and can be viewed as having shrunk to the line segment joining the massive monopoles . the field configuration is an embedding of a purely abelian @xmath229 solution , and the component of the field in the unbroken su(@xmath2 ) is a pure dipole field . in terms of massless monopole positions , this corresponds to a configuration where all of the massless monopoles are located on top of one of the massive monopoles . all trajectories pass through at least one such minimal cloud configuration . if @xmath230 , the trajectory passes through two such configurations . whether @xmath178 once or twice does not appear to have a dramatic influence on the large time behavior of the solutions . the points on a trajectory where @xmath174 are , of course , the points where the two massive monopoles pass through one another . since @xmath172 and @xmath173 both vary monotonically , it is clear that every trajectory must cross at least one coordinate axis and so must have at least one such point . thus , the two massive monopoles that are approaching each other on a straight line ( because @xmath231 ) can not stop and reverse direction before they meet . however , it is possible for them , after passing through each other once , to stop , reverse direction , pass through each other a second time , and finally emerge with their original direction of motion reversed . whether this second possibility is realized is determined , for a given @xmath207 , by @xmath220 . for @xmath232 , the monopoles reverse direction if @xmath233 ; the boundaries of this range correspond to solutions where the massive monopole velocities vanish as @xmath234 . the trajectories displayed in figs . 1 and 2 were chosen to illustrate the various possible behaviors ( @xmath178 once or twice , @xmath174 once or twice ) . several points should be noted . first , there does not appear to be any particular striking effect when @xmath174 . because @xmath52 is by definition positive , @xmath35 changes sign , but the magnitude of @xmath35 varies continuously . similarly , @xmath235 shows no unusual behavior when @xmath52 vanishes . all this is consistent with our expectations ; since the two massive monopoles have no direct interactions , they should be able to pass through each other with little effect . by contrast , each of the massive monopoles does interact directly with one of the massless monopoles comprising the cloud . this might suggest that there should be a noticeable effect when a massive and a massless monopole position coincide ; i.e. , when @xmath178 . on the other hand , the massless monopoles are hardly point particles , since the entire cloud region can be viewed as being composed of the massless cores . nevertheless , it does seem that the greatest effect on the trajectories occurs when @xmath236 . thus , the largest changes in the slope of @xmath237 coincide almost exactly with the vanishing of @xmath238 . note also that the effect of these massive - massless `` collisions '' seems to be greatest when they occur at small @xmath28 and @xmath52 . this is perhaps clearest in the trajectories where the interaction is strong enough to reverse the direction of motion of the massive monopoles , so that @xmath174 twice . in these cases , @xmath28 and @xmath52 coincide when @xmath52 is so small that the two massive cores overlap each other . even when the massive monopoles do not change direction , we can gauge the strength of their interaction with the cloud by the amount of energy transferred between the cloud and the massive monopoles . in fig . 3 we show this as a function of @xmath220 for @xmath239 . although the details of this plot vary somewhat with @xmath207 , two points are fixed : \1 ) if @xmath240 , @xmath187 vanishes identically , so @xmath173 is constant in time and there is no energy transfer . \2 ) if @xmath241 , the trajectory asymptotically approaches one of the lines @xmath227 at large time , corresponding to a final state with minimal cloud , @xmath178 . in this case , all of the initial cloud energy is transferred to the massive monopole kinetic energy . figure 3 should be compared with fig . 4 , where we we plot the minimum value of @xmath28 as a function of @xmath220 . we see that the largest energy transfers are associated with small values of @xmath28 , and that the energy transfer decreases as @xmath242 increases for @xmath243 . in fig . 4 we have also shown the value of @xmath238 at the time that @xmath28 achieves its minimum . note that this value is always less than or comparable to the monopole core radius . we have also studied solutions with @xmath244 ( i.e. , @xmath245 ) and @xmath246 ( @xmath247 ) . the general picture is very much as for the case of @xmath239 . the analogues of figs . 1 and 2 are qualitatively quite similar . we again find that @xmath238 is small when the cloud parameter @xmath28 achieves its minimum size , and the greatest amount of energy transfer between the cloud and the massive monopoles occurs when @xmath242 is small . one notable difference is found in those @xmath248 solutions where the massive monopoles reverse direction . in contrast with the @xmath239 case , the massive monopoles need not be overlapping at the time of this reversal . this is readily understood . because the cloud energy is so much greater than the massive monopole energy , relatively little energy transfer is required to reverse the sign of @xmath35 , and so @xmath28 at the time of the collision can be larger than it was in the previous case . this emphasizes that the critical factor for the @xmath239 reversal was that @xmath28 was small ; the fact that the massive cores overlapped was coincidental . the effective potential terms in eq . ( [ leffective ] ) come into play when either @xmath153 or @xmath152 is nonzero . with nonvanishing angular momentum , there is a centrifugal barrier that prevents the vanishing of @xmath52 . the trajectory in the @xmath172-@xmath173 plane can not cross either axis , and so is confined to a single quadrant . for large @xmath153 , this barrier ensures that @xmath52 , and hence @xmath28 , is always large , and thus suppresses interactions between the massive monopoles and the cloud . for small but nonzero @xmath153 , the effect of the centrifugal barrier is to make the trajectories appear to reflect off the @xmath172- and @xmath173-axes . we illustrate this in fig . 5 . except for the value of @xmath250 , the initial data for the trajectories in this figure are the same as for those shown in fig . 1 . equation ( [ jsquare ] ) for @xmath251 implies that the motion of the massive monopoles is confined to a plane . in fig . 6 , we illustrate this motion by showing the path of @xmath55 in the plane perpendicular to @xmath153 for one of the trajectories shown in fig . 5 . note once again that there is little evidence of interaction when the massive monopoles are at their closest approach . if @xmath252 , there is a barrier that forbids the vanishing of @xmath238 , or equivalently , the crossing of the lines @xmath253 . when @xmath152 is large , @xmath238 can never be small , and so there is little interaction among the monopoles . for small nonzero @xmath152 , we find behavior analogous to that for @xmath254 , with trajectories appearing to bounce off the lines @xmath253 . in this paper we have studied the properties of the magnetically charged counterparts of the electrically - charged massless particles that arise when a gauge theory is spontaneously broken to a non - abelian subgroup . previous studies of static bps solutions have shown that these massless monopoles coalesce into a cloud of non - abelian field surrounding one or more massive monopoles . in the scattering solutions that we have studied , these clouds act very much like thin shells , with a size of order @xmath28 , that contract until they collide with and bounce off the cores of the massive monopoles . at large times , when this shell is far from the massive cores , the cloud and the massive monopoles are essentially decoupled , and have separately conserved energies . their interaction occurs primarily at the time that the shell overlaps the massive cores . we find that the strength of this interaction , as measured , e.g. , by the amount of energy transfer , tends to be greatest if @xmath28 is small at the time of this overlap . further evidence in support of this picture can be obtained from another system that has been studied in some detail . when su(3 ) gauge theory is spontaneously broken to su(2)@xmath1u(1 ) there are two species of fundamental monopoles , one massive and one massless . the nahm data and moduli space metric for the bps solutions containing one massless and two massive monopoles were obtained by dancer @xcite . these solutions depend on twelve parameters . ten of these correspond to the center - of - mass position , global su(2)@xmath1u(1 ) phases , and overall spatial rotations . the remaining two , @xmath12 and @xmath255 , arise as parameters in elliptic functions that enter the nahm data . these span a geodesic submanifold which , after a nonlinear change of variables to eliminate coordinate singularities in the moduli space metric , is often illustrated as in fig . ( this figure is actually a six - fold covering of the submanifold , completely analogous to our eight - fold mapping of @xmath28 and @xmath52 onto the @xmath172-@xmath173 plane . ) the precise relations between position in this figure and cloud size and monopole separation are rather complex , with analytic expressions only known asymptotically @xcite . however , the qualitative picture is easily described . points far out along the three legs of the diagram correspond to solutions containing two well - separated massive monopoles , with the separation increasing with distance along the leg , while points in the central part of the figure correspond to solutions in which the massive monopole cores overlap and deform each other . the boundaries of the figure correspond to embedded su(2 ) two - monopole solutions ; in our language , these are solutions with infinite cloud parameter . ( these boundaries are infinitely far , in metric distance , from any interior point . ) the lines that bisect the legs correspond to special axially symmetric solutions @xcite . the solid portions of these lines in fig . 7 correspond to `` hyperbolic solutions '' that can be viewed as solutions with minimal cloud ( the analogue of our @xmath178 solutions ) and with the separation between the massive monopoles increasing with distance from the center . the dashed portions of these lines correspond to `` trigonometric solutions '' that are composed of two overlapping massive monopoles and a massless cloud that increases from minimal size at the center to infinite size at the boundary of the figure . geodesics on this two - dimensional submanifold correspond to scattering solutions with vanishing angular momentum and internal charges . in contrast to the su(@xmath2 ) example we have studied in this paper , there are direct interactions between the two massive monopoles . consequently , there is nontrivial scattering even when the cloud is infinitely large . thus , each of the three geodesics that bound the moduli space in fig . 7 corresponds to a process in which the two massive monopoles start infinitely far apart , approach head - on , and then recede at right angles to their initial motion ; during the entire process the cloud remains infinitely large . numerical solutions of the geodesic equations were obtained by dancer and leese @xcite . examining these , we see behavior quite consistent with our interpretation of our su(@xmath2 ) solutions , in that the interaction between the cloud and the massive monopoles is greater if the massive monopoles are closer together at the time of the minimal cloud configuration . we see this in two ways . first , there are some geodesics that start in one leg of the figure , cross the line of trigonometric solutions and begin to move down another leg , and then , after crossing the line of hyperbolic solutions , cross a second line of trigonometric solutions and exit via third leg . these may be viewed as analogous to the su(@xmath2 ) scattering solutions in which the two massive monopoles reverse direction . as with the latter solutions , these `` three - leg trajectories '' are found only if the massive monopoles are sufficiently close in the minimal cloud configuration ; in other words , if the trajectory goes too far down the second leg , the massive monopoles can not reverse direction . the effects of cloud interactions on the `` two - leg trajectories '' that cross a line of hyperbolic solutions but only a single line of trigonometric solutions are more subtle . at large times these trajectories lie between the boundary and the line of hyperbolic solutions , with the cloud size increasing as the relative distance to the boundary decreases . comparing the distance of the geodesics from the boundary at points equally far down the initial and final legs thus gives a measure of the net interaction between the cloud and the massless monopoles , somewhat analogous to the su(@xmath2 ) energy transfer that we plotted in fig . examination of the dancer - leese geodesics shows that the greatest change between initial and final legs , and thus the greatest interaction , occurs if the geodesic crosses the line of hyperbolic solutions nearest the center of the figure ; i.e. , nearest the point of minimal massless monopole separation . before closing , we must address the validity of the moduli space approximation . there is a potential problem in this regard in any theory where the excitation of massless particle modes is a possibility . the fact that the asymptotic solution of eq . ( [ asymrbsolution ] ) has the cloud parameter @xmath28 increasing faster than the speed of light at large times is an indication that there actually is a breakdown of the moduli space approximation in the example we have studied . considerable insight in this issue can be obtained by studying @xcite an so(5 ) example with a spherically symmetric cloud . this symmetry makes it possible to compare the predictions of the moduli space approximation with a numerical solution of the full field equations , starting from a slowly varying initial configuration . it turns out that the moduli space approximation continues to be reliable until @xmath256 becomes of order unity . after that time , the edge of the cloud develops into a rather well - defined front expanding outward at the speed of light . behind this front , in the cloud interior , the time evolution of the fields continues to be well - described by the moduli space solution ( i.e. , as if @xmath257 ) . ahead of the front , the evolution of the fields is much slower , consistent with a slow linear growth of the cloud at a rate fixed by the initial value of @xmath256 . how does this affect the picture of monopole - cloud scattering that we have developed ? two cases can be distinguished . in one , a static solution with finite @xmath52 and @xmath28 is subject to an external disturbance that induces small , but nonzero , values for @xmath35 and @xmath256 . as long as @xmath256 remains small ( energy conservation ensures that @xmath258 will remain small ) the moduli space description of the interaction between the cloud and the massive monopoles should be reliable . eventually @xmath256 will become of order unity , but since @xmath259 when this happens , the moduli space approximation should continue to give a good description of the subsequent motion of the massive particles . thus the main modification to our picture is that the velocity of the cloud asymptotically approaches unity , rather than increasing without bound . in the other case , the initial low - energy configuration approximates a solution with large cloud that is contracting with the magnitude of @xmath256 being of order unity . we expect that the deviation from the moduli space solution will lead to the emission of long wavelength , low amplitude waves of the massless boson fields that propagate at the speed of light . this will leave behind a configuration whose evolution then proceeds much as in the previous case . in closing , we note that while our focus in this paper has been on the massless monopole clouds in classical multimonopole configurations , much of the underlying motivation was to gain insight into the states of the electrically - charged elementary particle sector that are dual to these . thus , it would be of great interest to study the bps and near - bps states containing massless and massive elementary particles with non - abelian charge with the aim of identifying the counterparts of the cloud parameter @xmath28 or of the asymptotic separation of the energy between the cloud and the massless monopoles . a natural set of parameters is the cloud size @xmath28 ; the massive monopole separation @xmath52 ; three spatial euler angles @xmath53 , @xmath54 , and @xmath262 ; three internal su(2 ) euler angles @xmath263 , @xmath210 , and @xmath264 ; and the relative u(1 ) phase @xmath265 . this is a total of nine but , as we discussed in sec . 2 , we only expect eight parameters . we will see later that this conflict is resolved by the identification of a spatial euler angle with one of the internal su(2 ) euler angles . the @xmath266 transform as doublets under both the spatial and the internal su(2 ) , with the upper and lower indices being spatial and internal , respectively . hence the @xmath266 for arbitrary values of the spatial euler angles must be of the form @xmath267 where @xmath268 and @xmath269 are functions of @xmath28 , @xmath52 , @xmath265 , and the internal euler angles . we choose a parameterization of @xmath270 such that @xmath271 where @xmath272 similarly , @xmath268 and @xmath269 can be written as linear combinations of @xmath273 with coefficients that depend only on @xmath52 , @xmath28 , and @xmath265 . by performing a gauge rotation , we can make @xmath268 proportional to @xmath274 . we then have @xmath275 the 2 relative phases among @xmath276 and @xmath277 can be absorbed in redefinitions of @xmath264 and @xmath262 . then , by a redefinition @xmath278 , @xmath279 and @xmath280 , we can make the @xmath281 all real and non - negative . equations ( [ rfromz ] ) and ( [ bfromz ] ) then give the constraints @xmath282 these are solved by @xmath283 , @xmath284 , @xmath285 . this gives us @xmath286 notice that @xmath264 and @xmath262 always come together in the combination @xmath287 . this is a manifestation of the axial symmetry of this configuration : for any spatial axial rotation that changes the parameter @xmath262 , one can do an opposite axial gauge rotation that changes @xmath264 , so that the metric is invariant . there are several interesting limits of this metric . if the relative distance between the two massive monopoles @xmath291 and the relative u(1 ) charge @xmath152 both vanish , it reduces to the metric for one massive and one massless monopole in so(5 ) @xcite : @xmath292 . \label{so5metric}\end{aligned}\ ] ] in the limit where we have minimal cloud @xmath178 and vanishing internal su(2 ) charges , we get the metric for two distinct massive su(3 ) monopoles @xcite . the angles @xmath262 and @xmath264 can be absorbed into a redefinition of @xmath265 , and we obtain @xmath293 + \frac{\kappa^2}{2 } \frac{r}{\kappa+\mu r } ( -d\chi+\cos\theta\ , d\phi)^2 \ , .\end{aligned}\ ] ] we can also consider the limit where @xmath47 is much greater than both @xmath294 and the initial energy of the cloud . to leading order , the massive monopole motion decouples from that of the cloud , and we may set @xmath295 . making the change of variables @xmath296 and @xmath297 , we get the eguchi - hanson @xcite metric . \right\ } \ , .\end{aligned}\ ] ] when @xmath299 , this reduces to eq . ( [ so5metric ] ) . in order to study the classical dynamics for the general case with nonzero angular momenta and internal charges , it is most convenient to first eliminate the redundant parameters . this can be done , e.g. , by simply choosing @xmath300 . we then write the angular part of the metric as @xmath301 where the @xmath302 ( @xmath303 ) are one - forms expressed in terms of the various rotational and gauge angles . these satisfy an algebra of the form @xmath304 where the @xmath305 are antisymmetric in @xmath306 and @xmath12 . we then define the charges to be @xmath307 . using eq . ( [ algebra ] ) , we can write the angular equations of motion for eq . ( [ angularpart ] ) as @xmath308 ordinarily , we could have chosen the @xmath118 to include the three @xmath309 and the three @xmath310 . however , after eliminating the redundant angular variable , this is not possible . instead , we choose @xmath311 with this choice , the @xmath305 are not constants , as they are in more familiar examples . the nonzero @xmath305 are given by @xmath312 @xmath313 can be obtained from eq . ( [ explicitmetric ] ) for the metric . it is a block diagonal matrix formed by three two - by - two blocks , with the first two blocks being identical . it is straightforward to verify that eq . ( [ jdoteq ] ) implies that the square of the angular momentum , eq . ( [ jsquare ] ) , and the square of the internal su(2 ) charge , eq . ( [ tsquare ] ) , are indeed constant . a. abouelsaood , phys . lett . * 125b * , 467 ( 1983 ) ; p. nelson , phys . * 50 * , 939 ( 1983 ) ; p. nelson and a. manohar , phys . lett . * 50 * , 943 ( 1983 ) ; a. balachandran , g. marmo , m. mukunda , j. nilsson , e. sudarshan , and f. zaccaria , phys . lett . * 50 * , 1553 ( 1983 ) ; p. nelson and s. coleman , nucl . phys . * b237 * , 1 ( 1984 ) . k. lee , e. weinberg , and p. yi , phys . lett . b * 376 * , 97 ( 1996 ) ; j.p . gauntlett and d.a . lowe , nucl . phys . * b472 * , 194 ( 1996 ) ; s.a . connell , _ the dynamics of the @xmath314 charge ( 1,1 ) magnetic monopoles , _ university of south australia preprint . |
motivated by the recent experiments on periodically modulated , two dimensional electron systems placed in large transversal magnetic fields , we investigate the interplay between the effects of disorder and periodic potentials in the integer quantum hall regime . in particular , we study the case where disorder is larger than the periodic modulation , but both are small enough that landau level mixing is negligible . in this limit , the self - consistent born approximation is inadequate . we carry extensive numerical calculations to understand the relevant physics in the lowest landau level , such as the spectrum and nature ( localized or extended ) of the wave functions . based on our results , we propose a qualitative explanation of the new features uncovered recently in transport measurements . | cond-mat0401007 | two - dimensional electron systems ( 2des ) placed in a uniform perpendicular magnetic field exhibit a rich variety of phenomena , such as the integer@xcite and fractional@xcite quantum hall effects . @xcite another well - studied problem is that of a 2des in a uniform perpendicular magnetic field subjected to a periodic potential . even before the discovery of the quantum hall effects , hofstadter@xcite showed that in this case , the electronic bands split into a remarkable fractal structure of subbands and gaps , the so - called hofstadter butterfly . two `` asymptotic '' regimes are usually considered : ( i ) if the magnitude of the periodic potential is very large compared to the cyclotron energy and the zeeman splitting , then one can use lattice models to describe the hopping of electrons between wannier - like states localized at the minima of the periodic potential , whereas ( ii ) if the magnitude of the periodic potential is small compared to the cyclotron energy , the periodic potential lifts the degeneracy of each landau level . in both cases , the resulting butterfly structure is a function only of the ratio between the flux @xmath0 of the magnetic field through the unit cell of the periodic lattice , and the elementary magnetic flux @xmath1 . remarkably , if @xmath2 of the first asymptotic case is equal to @xmath3 of the second case , their electronic structures are solutions of the same harper s equation.@xcite if the periodic potential is comparable to the cyclotron energy , landau level mixing must be taken into account ; although landau levels still split into subbands , the structure is no longer universal , but depends also on the ratio of the periodic potential amplitude and the cyclotron energy.@xcite experimentally , the case with a small periodic modulation can be realized more easily . this is because the periodic potential is usually imprinted at some distance from the 2des layer ; as a result , its magnitude in the 2des is considerably attenuated . the interesting cases to study experimentally also correspond to small values of @xmath2 ( of order unity ) , where the butterfly structure shows a small number of subbands separated by large gaps , and should therefore be easier to identify . periodic modulations have been created using lithography@xcite and holographic illumination . @xcite the lattice constants of the resulting square lattices are of order 100 nm . as a result , the condition @xmath4 ( for instance ) is satisfied for @xmath5 t. this is a very low value , in the shubnikov - de haas ( sdh ) regime , not the high-@xmath6 quantum regime . significant landau level mixing and complications from the fact that the fermi level is inside one of the higher landau levels for such small @xmath6-values make the identification of the hofstadter structure difficult . recently , a new method for lateral periodic modulation has been developed using a self - organized ordered phase of a diblock copolymer deposited on a gaas / algaas heterostructure.@xcite the polymer spheres create a 2d triangular lattice with a lattice constant of about 39 nm . the corresponding unit cell area is almost an order of magnitude smaller than those achieved in previous experiments , implying that the condition @xmath4 is now satisfied for very strong magnetic fields , @xmath7 t. at such high magnetic fields the system is in the strong quantum regime , and landau level mixing can be safely ignored . for the experimental 2des electron concentrations , the fermi level is in the spin - down lowest landau level.@xcite as a result , this experimental setup appears more promising for the successful observation of the butterfly . nevertheless , one must take into account the disorder which is present in the system ( without disorder , there is no integer quantum hall effect iqhe to begin with ) . if the disorder is very small compared to the periodic potential amplitude , one expects that the subbands of the hofstadter structure are `` smeared '' on a scale @xmath8 , where @xmath9 is the scattering time , and @xmath10 as disorder becomes vanishingly small . as a result , the larger gaps in the hofstadter structure should remain open at the positions predicted in the absence of disorder , and one expects a series of minima in the longitudinal conductivity as the fermi level traverses such gaps . the experiment indeed shows a very non - trivial modification of the longitudinal resistivity , with many peaks and valleys appearing in what is ( in the absence of the periodic modulation ) a smooth lorentz - like peak.@xcite however , the position of the minima in @xmath11 do not track the positions of the main gaps in the corresponding hofstadter butterfly structure . instead , the data suggests that in this experimental setup , disorder is not small , but rather large compared to the estimated amplitude of the periodic potential . this is not a consequence of poor samples , since these 2des have high mobilities . it is due to the fact that the periodic modulation is considerably attenuated in the 2des , leading to a small energy scale for the hofstadter butterfly spectrum as compared to @xmath8 . as a result , the hofstadter structure predicted in the absence of disorder is of little use for interpreting the experimental data . one might expect that in this case the periodic potential should have basically no effect on the disorder - broadened landau level . this is indeed true for the strongly localized states at the top and bottom of the landau level . however , states in the center of the landau level extend over many unit cells of the periodic potential , and , as we demonstrate in the following , are non - trivially modified by its presence . in this paper , we investigate numerically the behavior of a 2des subject to a perpendicular magnetic field , a periodic potential and a disorder potential , under conditions applicable to the experimental system . the effective electron mass in gaas is @xmath12 while the magnetic fields of interest are on the order of 10 t. under these conditions , the cyclotron energy @xmath13 , of the order of 200 k , is the largest energy scale in the problem . the zeeman energy @xmath14 for these fields is roughly 3 k , but electron interaction effects lead to a considerable enhancement of the spin splitting between the ( spin polarized ) landau levels , which has been measured to be 20 k.@xcite the amplitude of the periodic potential s largest fourier components is estimated to be of the order of 1 k , and the scattering rate from the known zero field mobility is estimated to be @xmath15 k. @xcite as a result of this ordering of energy scales , we neglect landau level inter - mixing and study non - perturbatively the combined effects of a periodic and a large smooth disorder potential on the electronic structure of the lowest landau level . previously , the effects of small disorder on a hofstadter butterfly have been perturbatively investigated using the self - consistent born approximation ( scba ) , @xcite and the combined effect of white - noise disorder and periodic modulation on hall resistance was studied following the scaling theory of iqhe.@xcite our results reveal details of the electronic structure not investigated previously . the two - lead geometry we consider is schematically shown in fig . [ fig1 ] : the finite 2des is assumed to have periodic boundary conditions in the @xmath16-direction ( along which the hall currents flow ) , and is connected to metallic leads at the @xmath17 and @xmath18 edges . in particular , in this paper we study the effects of the periodic potential on the extended states carrying longitudinal currents between the two leads , and identify a number of interesting properties , in qualitative agreement with simple arguments provided by a semi - classical picture . our main conclusion is that while the beautiful hofstadter structure is destroyed by large disorder , the system still exhibits very interesting and non - trivial physics . -direction , and is attached to metallic leads at the @xmath19 ends . ] the paper is organized as follows : in section [ sec2 ] we briefly review the computation of the hofstadter structure for a small - amplitude periodic potential . in section [ sec3 ] we describe the type of disorder potentials considered . section [ sec4 ] describes the numerical methods used to analyze the spectrum and the nature of the electronic states , with both semi - classical and fully quantum - mechanical formalisms . results are presented in section [ sec5 ] , while section [ sec6 ] contains discussions and a summary of our conclusions . to clarify our notation , we briefly review the problem of a free electron of charge @xmath20 moving in a 2d plane ( from now on , the @xmath21-plane , of dimension @xmath22 ) in a magnetic field @xmath23 perpendicular to the plane , as described by @xmath24 in the landau gauge @xmath25 , the eigenfunctions of the schrdinger equation @xmath26 are : @xmath27 with eigenenergies @xmath28 here @xmath29 is the magnetic length , @xmath30 is the cyclotron frequency , @xmath31 are the hermite polynomials and @xmath32 , respectively @xmath33 are the eigenspinors of @xmath34 : @xmath35 . the degeneracy of a landau level is given by the number of distinct @xmath36 values allowed . imposing cyclic boundary conditions in the @xmath16-direction , we find @xmath37 where @xmath38 is an integer . the allowed values for @xmath38 are found from the condition that the electron wave - functions , which are centered at positions @xmath39 [ see eq . ( [ 2.1 ] ) ] are within the boundary along the @xmath40-axis , i. e. @xmath41 . it follows that the degeneracy of each landau level is @xmath42 , with @xmath43 . consider now the addition of a periodic potential , with a lattice defined by two non - collinear vectors @xmath44 and @xmath45 , such that @xmath46 for any @xmath47 . the periodic potential has non - vanishing fourier components only at the reciprocal lattice vectors @xmath48 , where @xmath49 and @xmath50 are integers . thus : @xmath51 further , since @xmath52 is real , it follows that @xmath53 . in the absence of landau level mixing , the hofstadter spectrum for both square@xcite @xmath54,\ ] ] and triangular@xcite @xmath55 @xmath56\ ] ] periodic potentials , with nonzero fourier components only for _ the shortest reciprocal lattice vectors _ , have been studied extensively in the literature.@xcite the parameter defining the spectrum is the ratio between the flux @xmath57 of the magnetic field through a unit cell and the elementary flux @xmath58 . for @xmath59 , where @xmath60 and @xmath61 are mutually prime integers , the original landau level is split into @xmath61 sub - bands . we would like to emphasize a qualitative difference between the two types of potentials : the square potential in eq . ( [ 2.6 ] ) is particle - hole symmetric , since @xmath62 . as a result , the sign of its amplitude is irrelevant . on the other hand , the triangular potential does not have this symmetry . with the sign chosen in eq . ( [ 2.7 ] ) and @xmath63 , @xmath64 has deep local minima at the sites of the triangular lattice , whereas the maxima are relatively flat and located on a ( displaced ) honeycomb lattice . as a result , the sign of @xmath64 is highly relevant . the second fact that must be mentioned is that the choice made in eqs . ( [ 2.6 ] ) and ( [ 2.7 ] ) is rather simple , since it aligns the periodic potential with the edges of the sample in a very specific way . in general , however , one could consider the case where the periodic lattice is rotated by some finite angle with respect to the sample edges ; study of such cases will be discussed in future work . finally , it may seem that this choice of periodic potentials is very restrictive also because only the shortest lattice vectors have been kept in the fourier expansion . in fact , the methods we employ can be directly used for potentials with more fourier components , but their inclusion leads to no new physics . real samples always have disorder . the current consensus is that high - quality gaas / algaas samples exhibit a slowly varying , smooth disorder potential . in a semi - classical picture , the allowed electron trajectories in the presence of such disorder follow its equipotential lines.@xcite closed trajectories imply localized electron states , while extended trajectories connecting opposite edges of the sample are essential for current transport through the sample ( for more details , see sec . [ classical ] ) . in typical experimental setups,@xcite dopant si impurities with a concentration of @xmath65 @xmath66 are introduced in a thin layer of 6 nm in thickness , located 20 nm above the gaas / algaas interface . typically , up to 10% of the si atoms are ionized . a small fraction of the ionized electrons migrate to the gaas / algaas interface where they form the 2d electron gas . the electrostatic potential created by the ionized impurities left behind is the major source of disorder in the 2des layer . on the length - scale we are interested in , there are @xmath67 to @xmath68 such ionized si impurities per @xmath69 . the resulting disorder potential must be viewed as a collective effect of the density fluctuation of the ionized impurities@xcite rather than a simple summation of the coulomb potential of a few impurities . the electrostatic potential from si impurities is compensated and partially screened by other mobile negative charges in the system such as , for example , the surface screening effect by mirror charges considered by nixon and davies.@xcite an exact treatment of this problem is difficult , since one should consider the spatial correlation of the ionized impurities . @xcite one model used to describe such disorder consists of randomly placed gaussian scatterers.@xcite this model captures the main feature of a smooth disorder potential and supports classical trajectories on equipotential contours , but it has no natural energy / length scales associated with it . as a result , here we choose to also investigate a different model of the disorder , which incorporates the smooth character of the coulomb potential in real space . we generate a realization of the disorder potential in the following way : positive and negative charges , corresponding to a total concentration of @xmath70 are randomly distributed within a volume @xmath71\times[-l_y/2,l_y/2]\times[20\mbox{nm}+d,26\mbox{nm}+d]$ ] above the electron gas which is located in the @xmath72 plane . here , we choose @xmath73 nm as an extra spacer since the electronic wave - functions are centered about 3 - 5 nm below the gaas / algaas interface . since we are not simulating single impurities but density fluctuations , these charges are not required to be elementary charges . instead , we use a uniform distribution in the range @xmath74 $ ] for convenience ( a gaussian distribution would also be a valid choice ) , and sum up all coulomb potentials from these charges , using the static dielectric constant in gaas @xmath75.@xcite the resulting disorder potential has energy and length scales characteristic of the real samples . typical contours for such potentials are shown in sec . [ sec5 ] . in an infinite system , in the quantum hall regime , the existence of quantum hall steps implies the existence of critical energies at which the localization length diverges.@xcite this is the quantum analog of the two dimensional percolation problem in a smooth random landscape , for which there exists a single critical energy.@xcite in the case of potentials with electron - hole symmetry @xmath76 , the critical energy lies in the middle of the band ( @xmath77 ) , leading to percolating path at half filling . for a finite mesoscopic sample , however , not only does the percolating path ( critical energy ) @xmath78 deviate from this value , but in samples without a periodic boundary condition one need not have a percolating path traversing the system in the desired direction . this arises from the fluctuations near the edge of a mesoscopic system with free boundary conditions . we circumvent such a possibility by adding an extra smooth potential @xmath79 to the impurity - induced disorder potential @xmath80 , such that the total potential @xmath81 is zero on the opposite edges @xmath82 of the sample where the metallic leads are attached . the supplementary contribution @xmath79 can be thought of as simulating the effect of the leads on the disorder potential , since the metallic leads hold the potential on each edge constant by accumulating extra charges near the interface . therefore , physically we expect that the extra potential @xmath83 decays exponentially over the screening length @xmath84 inside the sample . this implies : @xmath85 @xmath86 where @xmath84 is taken to be @xmath87 nm in our calculation . . for both coulomb and gaussian model , @xmath88 is averaged over 116 disorder realizations . the relation between @xmath89 and @xmath52 and relevant parameters are discussed in the text . the standard deviation , @xmath90 for the coulomb model is @xmath91ev@xmath92 , and @xmath93ev@xmath92 for the gaussian model . ] in fig . [ fig1.5 ] , we plot the average of fourier transform of the magnitude of the random potential @xmath94 versus @xmath95 for the coulomb model and the gaussian model . the gaussian model is generated by adding 100 randomly placed gaussian scatterers on an area of @xmath96 , each contributing @xmath97 , where @xmath98 is uniformly distributed in @xmath99 $ ] mev , and @xmath100 is uniformly distributed in @xmath101 $ ] @xmath102 m . @xmath103 is related to @xmath52 by @xmath104 , where the summation is over all the wavevectors involved in the fast fourier transformation . the gaussian model has an arbitrary energy scale which is fixed by the maximum value of the distribution @xmath105 . here @xmath106 mev . as can be seen , @xmath89 of both models are decreasing functions of @xmath61 . the trends of decay are exponential at large @xmath61 . at small @xmath61 , the two models behave differently . despite the difference , both models lead to the same qualitative results , although , as expected , minor quantitative differences are present . this shows that the physics we uncover is independent of the particular type of slowly - varying disorder potential considered , and therefore should be relevant for the real samples . in this section we discuss the numerical methods we use , including derivations of some relevant formulas . as already stated , we focus on the case where the amplitudes of the periodic and disorder potentials are very small compared to the cyclotron energy and the zeeman splitting , and therefore inter - level mixing is ignored . the semi - classical approach is valid@xcite for the integer quantum hall effect in the presence of a slowly varying , smooth disorder potential and large magnetic fields ( such as we consider ) , so that the magnetic length @xmath107 which determines the spatial extent of the electron wave - functions is much smaller than the length scale of variation of the smooth disorder potential , @xmath108 . then , semi - classically the electron moves along the equipotential contours of the disorder potential @xmath52 , in the direction parallel to @xmath109 . since the kinetic energy is quenched in the lowest landau level , the total energy of the electron simply equals the value of the disorder potential on the equipotential line on which its trajectory is located . as a result , the density of states in the semi - classical approach is directly given by the probability distribution for the disorder potential , which can be calculated by randomly sampling the potential energy and plotting a histogram of the obtained values.@xcite in sec . [ sec5 ] we compare the results obtained within this semi - classical approach with fully quantum mechanical results . as expected , the agreement is good if only the disorder potential is present . however , if the periodic modulation is also included , the lattice constant @xmath110 provides a new length - scale which is comparable to the magnetic length @xmath107 , and the semi - classical picture breaks down . quantum mechanical calculations are absolutely necessary to quantitatively treat this case . as shown in sec . [ sec2 ] , for a finite sample of size @xmath111 at a given magnetic field @xmath6 , the degeneracy of the unperturbed landau level is @xmath112 . since the disorder varies very slowly , we need to consider systems with @xmath113 to properly account for its effects . as a result , the number of states in a landau level can be as large as @xmath67 in our calculations . storage of the hamiltonian as a dense matrix requires considerable amount of computer memory and its direct diagonalization is prohibitively time - consuming . sparse matrix diagonalization techniques could be employed , but they are less efficient when all eigenvectors are needed , and also have some stability issues . here we introduce the numerical methods we use to compute densities of states and infer the nature ( localized or extended ) as well as the spatial distribution of the wave - functions , while avoiding direct diagonalization . since inter - level mixing is ignored , the hilbert subspaces corresponding to different spin - polarized landau levels do not hybridize . each hilbert subspace @xmath114 has a basis described by eq . ( [ 2.1 ] ) , containing @xmath115 orthonormal vectors indexed by different @xmath36 values . in order to compute matrix elements of the total hamiltonian in such a basis , we use the following identity derived in ref . ( notice their different sign convention for @xmath36 . if @xmath116 , the overlap is zero ) : @xmath117 where @xmath118^{n - n ' } & & \\ \times e^{-{1 \over 2 } q}q^{{1 \over 2}|n'-n|}l_m^{(|n'-n|)}(q ) , & & \end{aligned}\ ] ] with @xmath119 , @xmath120 and @xmath121 the minimum and the maximum of @xmath122 and @xmath123 respectively , and @xmath124 the associated laguerre polynomial . when band - mixing is neglected @xmath125 and @xmath126 . for the first landau level , @xmath127 . ( [ 2.5 ] ) gives us the matrix elements for the square [ eq . ( [ 2.6 ] ) ] or triangular [ eq . ( [ 2.7 ] ) ] periodic potentials . in either case , there are fourier components corresponding to @xmath128 and @xmath129 . since only basis vectors for which the difference @xmath130 give non - vanishing matrix elements , we must choose the length @xmath131 of the sample to be a multiple integer of @xmath110 , the lattice constant . the matrix elements of the disorder potential are computed in a similar way . we use a grid of dimension @xmath132 to cover the sample and generate the values of the disorder potential on this grid . then , fast fourier transform ( fft)@xcite is used to find the long wavelength components of the disorder potential corresponding to the allowed values @xmath133 { 2\pi\over l_{x , y}}$ ] ( proper care is taken to define fourier components so that @xmath134 ) . the matrix elements of this discretized disorder potential are then computed using eq . ( [ 2.5 ] ) . in principle , finer grids ( increased values for @xmath135 and @xmath136 ) will improve accuracy . however , they also result in longer computation times , since they add extra matrix elements in the sparse matrix , corresponding to large wave - vectors . we have verified that a grid size of dimension @xmath137 is already large enough to accurately capture the landscape of a @xmath96 sample and the computed quantities have already converged , with larger grids leading to hardly noticeable changes . this procedure is also justified on a physical basis . first , the neglected large wave - vector components describe very short - range spatial features , which are probably not very accurately captured by our disorder models to begin with , and which are certainly not believed to influence the basic physics . secondly , this procedure insures that the actual disorder potential we use is periodic in the @xmath16-direction , since each fourier component retained has this property . this is consistent with our use of a basis of wave - functions which are periodic along @xmath16 . the matrix elements of the hamiltonian within a given landau level @xmath114 are then @xmath138 , where @xmath139 are given by eq . ( [ 2.2 ] ) and the matrix elements of both the periodic and the disorder part of the potential @xmath140 are computed as already discussed . this produces a sparse matrix , which is stored efficiently in a column compressed format . a quantity that can be computed without direct diagonalization is the filling factor @xmath141 as a function of fermi energy . the filling factor is defined as : @xmath142 where @xmath143 is the heaviside function and @xmath115 is the total number of states in the @xmath114 landau level . ( since we neglect landau - level mixing , we can define this quantity for individual levels . ) the filling factor tells us what fraction of the states in the given landau level are occupied at @xmath144 , for a given value of the fermi energy . this corresponds to the average filling factor measured in experiment and is also proportional to the integrated total ( as opposed to local ) density of states . the filling factor is straightforward to compute if the eigenenergies @xmath145 are known . however , we want to avoid the time - consuming task of numerical brute force diagonalization . the strategy we follow is a generalization to hermitian matrices of the method used in ref . . we restate the problem in the following way : assume we have a hermitian matrix of size @xmath146 ( no landau level mixing ) , given by the matrix elements of @xmath147 in the basis @xmath148 ( @xmath149 is the unit matrix ) . then , @xmath141 is proportional to the number of negative eigenvalues of the matrix @xmath121 . we now generate the quadratic form @xmath150 , and transform it into its standard form @xmath151 using the jacobian method described below . here , @xmath152 s are all real numbers , and the @xmath153 s are linear combinations of the @xmath154 s . this is a similarity transformation which retains the signature of the matrix . as a result , even though the numbers @xmath152 are not eigenvalues of @xmath121 , the number of negative eigenvalues equals the number of negative @xmath152 values . it follows that @xmath141 is obtained by simply counting the number of negative @xmath152 values for the given @xmath155 . the jacobian method is iterative in nature . first , all terms containing @xmath156 and @xmath157 are collected and the needed complementary terms are added to form the first total square @xmath158 , so that @xmath156 and @xmath157 are eliminated from the rest of the quadratic form @xmath159 . the procedure is then repeated for all @xmath160 and @xmath161 terms ( producing @xmath162 ) etc . , until all @xmath115 values @xmath152 are found . computationally , this can be done by scanning the lower or upper triangle of the hermitian matrix @xmath121 only once . the total number of operations is proportional to the number of nonzero elements of the matrix , meaning that for a dense matrix it scales with @xmath163 ( sparse matrices require much fewer operations ) . as a result , this procedure is much faster than brute force diagonalization which scales with @xmath164 ( for us , @xmath165 ) . the filling factor @xmath166 is a sum of step - like functions , with steps located at the eigenvalues . by scanning @xmath167 and identifying the position of these steps we can also find the true eigenvalues @xmath145 , with the desired accuracy . finally , the total density of states is given by @xmath168 . the advanced / retarded green s functions are the solutions of the operator equation @xmath169 where @xmath170 . ( in practice we use a set of small positive numbers , and use the dependence on @xmath171 to obtain results . ) if the exact eigenstates and eigenvalues of the total hamiltonian @xmath172 are known , @xmath173 ( no landau level mixing ) , it follows : @xmath174 the exact eigenstates can be expanded in terms of the basis states @xmath175 as @xmath176 since the states @xmath175 are localized near @xmath177 [ see eq . ( [ 2.1 ] ) ] , the coefficients @xmath178 describe the probability amplitude for an electron in the state @xmath179 . knowledge of these coefficients allows us to infer whether such states are extended or localized in the @xmath40-direction , i.e. whether they can carry currents between the leads . however , as already stated , we wish to avoid direct diagonalization . we can still infer whether the hamiltonian has extended or localized wave - functions near a given energy @xmath180 in the following way . we introduce the matrix elements : @xmath181 @xmath182 if landau level mixing is neglected , eq . ( [ 4.1 ] ) can be rewritten in the basis @xmath175 as : @xmath183\ ] ] @xmath184 we use the popular numerical library superlu,@xcite based on lu decomposition and gaussian reduction algorithm for sparse matrices , to solve these linear equations . consider now the matrix element @xmath185 corresponding to the smallest @xmath186 and the largest @xmath187 values . if all wave - functions with energies close to @xmath180 are localized in the @xmath40-direction , it follows that @xmath188 is a very small number , of the order @xmath189 , where @xmath190 is the localization length at the given energy . on the other hand , we expect to see a sharp peak in the value of @xmath188 if @xmath180 is in the vicinity of an extended state eigenvalue , since [ see eqs . ( [ 4.3],[4.4 ] ) ] both @xmath191 and @xmath192 are non - vanishing for an extended wave - function with significant weight near both the @xmath193 and the @xmath194 edges . moreover , the height of this peak scales like @xmath195 , so by varying @xmath171 we can easily locate the energies of the extended states . we can also use green s functions techniques to image the local density of states at a given energy @xmath167 . by definition ( and neglecting landau level mixing ) , the local density of states in the level @xmath114 is : @xmath196 @xmath197 where the second equality follows from eq . ( [ 4.2 ] ) . this function traces the contours of probability @xmath198 for electrons with the given energy @xmath167 . its direct computation , however , is difficult and very time - consuming . for the rest of this subsection , the discussion is restricted to the lowest landau level @xmath199 ( the value of @xmath200 is irrelevant ) . we know that in the lowest landau level , electronic wave - functions can not be localized in any direction over a length - scale shorter that the magnetic length @xmath107 . as a result , it suffices to compute a projected local density of states on a grid with @xmath201 ( or larger ) spacings . the projection is made on maximally localized wave - function , defined as follows . let @xmath202 be a point on the grid . we associate it with a vector : @xmath203 where we use the simplified notation @xmath204 ) ) and we take @xmath205 it is then straightforward to show that @xmath206 in other words , @xmath207 is an eigenstate of the first landau level strongly peaked at @xmath208 . ( the phase factor is due to the proper magnetic translation ) . we then define the projected density of states [ compare with eq . ( [ 5.1 ] ) ] : @xmath209 and use it to study the spatial distribution of the electron wave - functions at different energies . strictly speaking , the local density of states defined in eq . ( [ 5.1 ] ) can not be projected exactly on the lowest landau level , because the lowest landau level does not support a @xmath171-function ( @xmath210 , @xmath211 ) . however , the coherent states @xmath207 we select are the maximally spatially - localized wave functions in the lowest landau level , and have the added advantage that they can be easily stored as sparse vectors , because of their gaussian profiles [ see eq . ( [ 5.3 ] ) ] . moreover , in the limit @xmath212 ( @xmath213 ) where @xmath214 , the projected density of states @xmath215 . therefore , for the large @xmath6 values that we consider here , the projected density of states @xmath216 should provide a faithful copy of the local density of states . we compute the projected local density of states following the method of ref . . let @xmath217 be the vector with elements @xmath218 obtained from the representation of @xmath207 in the @xmath219 basis [ see eq . ( [ 5.2 ] ) ] , and let @xmath220 be the matrix of the hamiltonian @xmath221 in the @xmath219 basis . we generate the series of orthonormal vectors @xmath222 using : @xmath223 and for @xmath224 , @xmath225 the numbers @xmath226 and @xmath227 can be shown to be real . we do not have a `` terminator''@xcite to end this recursive series . instead , our procedure ends when the orthonormal set of vectors @xmath222 exhausts a subspace of the lowest landau level containing all states coupled through the disorder and/or periodic potential to the state @xmath207 ( i.e. , all states that contribute to the projected dos at this point ) . in the presence of disorder , this usually includes the entire lowest landau level . then , the projected density of states is given by eq . ( [ 5.5 ] ) , where the matrix element of the green s function is the continued fraction : @xmath228 @xmath229^{-1 } \right]^{-1 } \right]^{-1}\ ] ] because the hamiltonian is a sparse matrix , the generation of these orthonormal sets and computation of @xmath230 for all the grid points is a relatively fast procedure . moreover , this computation is ideally suited for parallelization , with different grid points assigned to different cpus . in this section we present numerical results obtained using these methods . we have analyzed over 20 different disorder realizations for samples of different sizes , and all exhibit the same qualitative physics . here , we show results for several typical samples . the lattice constant is always @xmath231 nm if periodic potential is present , as defined by the experimental system.@xcite profile of the disorder potential obtained from our coulomb model on a @xmath232m@xmath233 m sample , without ( upper panel ) and with ( lower panel ) the @xmath234 correction at the @xmath19 edges . the disorder potential varies between @xmath235 mev and 3 mev , on a spatial length - scale much larger than @xmath236 nm . the critical region containing extended states is in the vicinity of @xmath237 mev . the contours are shown for @xmath2380.0575 mev ( dashed ) , 0.17 mev ( thick solid ) and 0.31 mev ( thin solid ) . these energy values correspond to classical filling factors @xmath239=0.47 , 0.58 and 0.68 in the upper panel and @xmath239=0.45 , 0.56 , 0.66 in the lower panel . the difference is due to the supplementary smooth potential @xmath83.,title="fig : " ] profile of the disorder potential obtained from our coulomb model on a @xmath232m@xmath233 m sample , without ( upper panel ) and with ( lower panel ) the @xmath234 correction at the @xmath19 edges . the disorder potential varies between @xmath235 mev and 3 mev , on a spatial length - scale much larger than @xmath236 nm . the critical region containing extended states is in the vicinity of @xmath237 mev . the contours are shown for @xmath2380.0575 mev ( dashed ) , 0.17 mev ( thick solid ) and 0.31 mev ( thin solid ) . these energy values correspond to classical filling factors @xmath239=0.47 , 0.58 and 0.68 in the upper panel and @xmath239=0.45 , 0.56 , 0.66 in the lower panel . the difference is due to the supplementary smooth potential @xmath83.,title="fig : " ] , but different amplitudes of the triangular periodic potential ( a ) @xmath240=0 ; ( b ) @xmath240=0.05mev ; ( c ) @xmath240=0.5mev and ( d ) @xmath240=5 mev . as expected , agreement exists only in the limit @xmath241 . ] . we show only the center of the disorder - broadened lowest landau level , where the density of states is large . ] for the first sample , we consider @xmath242 ( @xmath243 t ) . the magnetic length is @xmath244 nm , and we choose a sample size @xmath245 m and @xmath246 m . with these choices , the landau level contains @xmath247 states . the disorder potential obtained with our scheme described in sec . [ sec3 ] is shown in fig . [ fig2 ] , both with and without the correction @xmath234 . an extended equipotential line appears , as expected , at @xmath248 . in figs . [ fig3 ] and [ fig4 ] we plot the filling factor @xmath249 and the corresponding total density of states @xmath250 as a function of @xmath167 ( computation details were given in sec . [ subs2 ] ) . these quantities are obtained in the semi - classical limit ( dashed line ) and with the full , quantum - mechanical treatment ( solid line ) . results are shown for 4 different cases : ( a ) only disorder potential and ( b , c , d ) disorder plus a triangular periodic potential with amplitudes @xmath251 , 0.5 and 5 mev , respectively . we only plot a relatively small energy interval where the dos is significant , and ignore the asymptotic regions with long tails of localized states . while the agreement between the semi - classical and quantum - mechanical treatment is excellent in the limit @xmath252 , the two methods give more and more different results as the periodic potential amplitude is increased . as already explained , this is a consequence of the fact that the magnetic length @xmath107 is comparable to the lattice constant @xmath110 , leading to a failure of the semi - classical treatment when this extra length - scale is introduced . in particular , in the case with the largest periodic potential [ panel ( d ) of figs . [ fig3 ] and [ fig4 ] ] we can clearly see the appearance of the 3 subbands expected for the hofstadter butterfly at @xmath242 , although the disorder leads to broadened and smooth peaks , and partially fills - in the gap between the lower two subbands . this picture [ panel ( d ) ] is quite similar to the density of states that ref . calculated using the self - consistent born approximation . this is expected since the scba approach is valid in the limit of strong periodic potential with weak disorder . however , the scba approach is not appropriate in the limit of moderate or strong disorder , where the higher order terms neglected in scba are no longer small . for disorder varying on a much longer length - scale than the periodic potential , one still expects that _ locally _ , on relative flat regions of disorder , the system exhibits the hofstadter - type spectrum . however , these spectra are shifted with respect to one another by the different local disorder values . if disorder variations are small , then the total spectrum shows somewhat shifted subbands with partially filled - in gaps , but overall the hofstadter structure is still recognizable . on the other hand , for moderate and large disorder , the detailed structure of the local density of states from various flat regions are hidden in the total density of states . all one sees are some broadened , weak peaks and gaps superimposed on a broad , continuous density of states . we now analyze the nature of the electronic states for these configurations . we start with the case which has only disorder . in fig . [ fig5 ] we plot @xmath253 as a function of the energy @xmath167 , for different values of @xmath171 ( computation details were given in sec . [ subs3 ] ) . as already discussed , extended states are indicated by large values of this quantity , as well as a strong ( roughly @xmath254 ) dependence on the value of the small parameter @xmath171 . figure [ fig5 ] reveals that as @xmath171 is reduced , resonant behavior appears in a narrow energy interval @xmath255 mev . panel ( a ) shows that results corresponding to @xmath256 ev and @xmath257 ev indeed differ by roughly 2 orders of magnitude , with @xmath257 ev showing sharper resonance peaks . the difference between results for @xmath257 ev and @xmath258 ev shown in panel ( b ) , is no longer so definite . the reason is simply that for such small @xmath171 , the denominator in the green s function expression is usually limited by @xmath259 and not by @xmath171 [ see eq . ( [ 4.4 ] ) ] , and the dependence on @xmath171 is minimal . only if @xmath167 is such that @xmath260 can we expect to see a @xmath171 dependence , and indeed this is observed at some energies . finally , in panel ( c ) we show the comparison with a larger energy interval . the value of the green s function decreases exponentially fast on both sides of the critical region , indicating strongly localized states . here , data for @xmath261 ev is a smooth curve , whose magnitude is much less than that of the other three values even for localized states . this is due to the fact that this @xmath171 is larger than typical level spacings . as a result , several levels contribute significantly to green s function at each @xmath167 value , and the destructive interference of the random phases of different eigenfunctions lead to the supplementary @xmath171-dependence . we conclude that the disorder potential has a critical energy regime of approximately @xmath262 mev width , covering less than 5% ( in energy ) and 20% ( in number of states ) of the disorder - broadened band with total width @xmath263 mev . the position of the critical energy interval is in agreement with the semi - classical results which suggest an extended state in the vicinity of @xmath237mev . by comparison with fig . [ fig3 ] , we can also see that this critical regime corresponds to a roughly half - filled band , in agreement with the experiment . the effect of an additional triangular periodic potential is shown in fig . [ fig6 ] , where we plot the same quantity shown in fig . [ fig5 ] for a fixed @xmath256 ev and different amplitudes @xmath264 , 0.05 , 0.5 and 5 mev , respectively . these results correspond to a different coulomb disorder potential ( not shown ) , as can be seen from the different location of its extended states . here we see how the narrow critical interval of extended states grows gradually as the amplitude of periodic potential is increased and finally exhibits the three well - separated extended subbands expected for @xmath265 in the limit of vanishing disorder . the three subbands can already be resolved for the moderate case @xmath266mev , although they are very wide and exhibit significant overlap . , as a function of energy . only the disorder potential of fig . [ fig2 ] is present . ( a ) comparison between @xmath267 and @xmath268 results ; ( b ) comparison between @xmath268 and @xmath269 results ; ( c ) comparison between results corresponding to @xmath270 and @xmath271 . ( the last three curves are indistinguishable to the eye on this scale . ) all @xmath171 values are in ev units . ] mev . as the amplitude @xmath240 of the periodic potential increases , the range of extended states increases dramatically . the left panel shows results for disorder - only and two relatively weak periodic potentials , while the right panel shows two larger periodic potentials where the three - subband structure expected for @xmath265 is clearly seen . ] qualitatively similar behavior is obtained if we use the gaussian scatterers model for disorder . a typical realization of this disorder is shown in fig . [ fig7bis ] . results for the green s function s values with such disorder are shown in fig . [ fig7 ] , for cases with pure disorder , and also cases with either a triangular or a square periodic potential . the magnetic field has been doubled , such that @xmath272 . similar to the case shown in fig . [ fig6 ] , the periodic potential leads to a widening of the critical regime . for large periodic potentials , the expected hofstadter - like three - subband structure emerges again . we conclude that coulomb and gaussian disorder models show qualitatively similar behavior . the three lines are equipotential contours close to the critical regime , with energies of -0.1 mev ( dashed ) , 0 mev ( thick solid ) and 0.1 mev ( thin solid ) . cyclic boundary condition are applied in the @xmath16 direction . ] . similar to results shown in fig . [ fig7 ] , we see that the periodic potentials widen the critical region . ] we now analyze the projected local density of states @xmath273 discussed in sec . [ subs4 ] , in order to understand the reason for this substantial widening of the critical region by even small periodic potentials . we consider a smaller sample , of size approximately @xmath274m@xmath275 m , and compute the projected density of states for 500 equally - spaced energy values , on a @xmath276 square grid and for a value @xmath257 ev . this @xmath171 value is comparable or smaller than the level spacing , so we expect to see sharp resonances from the contribution of individual eigenfunctions as we scan the energy spectrum . each computation generates a large amount of data ( roughly 24 m ) , corresponding to the 500 plots of the local density of states at the 500 values of @xmath167 . since we can not show all this data , we select a couple of representative cases and some statistical data to interpret the overall results . figures [ fig9 ] and [ fig10 ] show some of our typical results . the two figures are calculated for the same coulomb - disorder potential , for values of @xmath277 mev ( at the bottom of the band ) and @xmath278 mev ( close to , but below the band center ) respectively . each figure contains four panels , panel ( a ) shows the profile of the disorder potential as well as an equipotential line ( solid black ) corresponding to the value @xmath167 considered ; the other three panels show the projected density of states @xmath273 for ( b ) pure disorder ; ( c ) disorder plus triangular periodic potential with @xmath279 mev ; ( d ) disorder plus square periodic potential with @xmath279 mev . in fig . [ fig9 ] , this equipotential line ( which traces the semi - classical trajectory of electrons with the same energy @xmath167 ) for @xmath277 mev . panel ( a ) shows the profile of the disorder potential , and the equipotential contour ( black line ) corresponding to @xmath277 ev . the other three panels show @xmath280 for ( b ) disorder only ; ( c ) disorder plus triangular periodic potential with @xmath281 mev ; ( d ) disorder plus square periodic potential with @xmath282 mev . the width and length of the sample are both @xmath283 m , and @xmath242 . increased brightness corresponds to larger values . ] surrounds local minima of the disorder potential , suggesting localized electron states at such low energies . indeed , this is what panels ( b ) , ( c ) and ( d ) show . the projected density of states @xmath273 is large ( bright color ) at the positions where electrons of energy @xmath167 are found with large probabilities . for pure disorder , we observe only closed trajectories ( localized states ) , whose shape is in excellent agreement with the semi - classical trajectory , as expected . if a moderate periodic potential is added , the wave - functions spread over a larger area , and nearby contours sometimes merge together . instead of sharp lines , as seen in panel ( b ) , the contours now show clear evidence of interference effects of the wave - functions on the periodic potential decorating the electron reservoirs . some periodic modulations can also be observed in the background of panels ( c ) and ( d ) , especially for the square potential . these are _ not _ the direct oscillations of the periodic potentials , since the grid we use to compute these figures has a linear size equal to @xmath284 of the period @xmath231 nm of the periodic potential . capturing detailed behavior inside each unit cell would require a much smaller grid , which is not only time consuming , but also violates the requirement that the grid size be of order @xmath107 or larger . figure [ fig10 ] for an energy close to the band center , shows the same characteristics . for pure disorder , the electrons at this energy trace a sharp contour very similar , but for an energy @xmath278 mev close to the band center . ] to the corresponding equipotential line shown in panel ( a ) . electrons are still not delocalized , since this contour does not connect either pair of opposite edges . however , addition of the periodic potential now leads to extended states for both types of periodic potentials [ ( c ) and ( d ) ] at roughly @xmath278 mev , demonstrating the widening of the critical region with the addition of a periodic potential . physically , one can understand this spread of the wave function in the presence of the periodic potential using the semi - classical picture.@xcite if only a smooth disorder potential is present , the equipotential at any energy @xmath167 must be a smooth , continuous line . however , if a periodic potential with minima @xmath285 and maxima @xmath286 is superimposed over disorder , the new equipotential line now breaks into a series of small `` bubbles '' surrounding the disorder - only contour . this happens throughout the area defined by the equipotentials @xmath287 and @xmath288 of the disorder potential , since the addition of the periodic potential leads new regions in this area to have a total energy @xmath167 . quantum mechanically , we expect some tunneling inside this wider area and this is indeed what we observe in figs . [ fig9 ] and [ fig10 ] . this mechanism suggests _ enhanced delocalization _ on both sides of the critical region as localized wave functions spread out over larger areas , as well as a widening of the critical region itself , in agreement with our numerical results . this spreading of the wave functions in the presence of the periodic potential can also be characterized by counting , at a given energy @xmath167 , the number of grid points @xmath289 which have a value @xmath290 , where @xmath291 is some threshold value . for sufficiently large @xmath291 , this procedure counts grid points where electrons with energy @xmath167 are found with large probabilities , thus , in effect it characterizes the `` spatial extent '' of the wave functions . the results of such counting are shown in fig . [ fig11 ] for 500 energy values corresponding to the disorder potential analyzed in figs . [ fig9 ] and [ fig10 ] . there are a total of @xmath292 grid points on the sample . for the case of pure disorder ( black line ) we see that the largest values are found at energies just below 0 , where the extended states ( the critical region ) are found for this particular realization of disorder . because it is a smooth , sharp line , even the most extended trajectory has significant probabilities at only about 10% of the grid sites . for both higher and lower energies , this number decreases very fast , indicating wave functions localized more and more about maxima or minima of the disorder potential , as expected . addition of a small periodic potential increases this number substantially , clearly showing the supplementary spreading of the wave functions in the presence of the periodic potential . figure [ fig11 ] shows this effect for three types of periodic potential : triangular lattices with @xmath293 and @xmath294 ( upper panel ) , and square lattice in the lower panel . all three cases show significant enhancement , as compared to the pure disorder case . in addition , we see that while the square potential gives a fairly symmetrical enhancement , the triangular potential does not , with curves for @xmath295 not overlapping . this is a consequence of the asymmetric shape of the periodic potential , which has different values for its minima and maxima @xmath296 , as well as different arrangements for the points where minima / maxima appear ( triangular lattice vs. honeycomb lattice ) . [ fig11 ] clearly shows that @xmath297 favors increased delocalization below the critical energy regime , while @xmath298 favors increased delocalization above it . the reason for this different response to the two signs of the triangular potential can be nicely explained within the semi - classical framework . in fig . [ fig12 ] we show the equipotential lines corresponding to filling factors @xmath299 ( well below critical region ) and @xmath300 ( well above the critical region ) for a realization of coulomb disorder ( not shown ) plus a triangular potential with @xmath63 . areas with energy below the equipotential value are shaded . in this case we can clearly see that instead of the continuous , smooth trajectory expected for disorder - only cases , there are also extra `` bubbles '' regions connecting the areas between such contours . since the choice @xmath63 leads to deep minima at @xmath301 with triangular arrangement and relatively flat maxima at @xmath302 with honeycomb arrangement [ see eq . ( [ 2.7 ] ) ] , it follows that the triangular ( honeycomb ) `` bubbles '' region appear roughly in the area bounded by the equipotentials @xmath167 and @xmath288 ( respectively , @xmath287 and @xmath167 ) of the pure disorder potential . at low filling factors , the pure disorder @xmath167 equipotential is a collection of closed contours surrounding local minima [ see panel ( d ) of fig . [ fig9 ] for an illustration ] . it follows that for the choice @xmath303 , the more extended region with triangular `` bubbles '' will be found outside these `` islands '' and will lead to a spread of the wave function over considerably larger areas , as indeed seen in the upper panel of fig . [ fig12 ] . on the other hand , at large filling factors the disorder - only contours are `` islands '' surrounding the maxima of the disorder potential . in this case , contours between @xmath167 and @xmath288 are inside the @xmath167 contour , so the triangular `` bubbles '' region does not help to connect various `` islands '' as before . the honeycomb `` bubbles '' regions does this , but because @xmath304 the extension of the wave function between `` islands '' is significantly smaller in this case . in the quantum - mechanical case one expects interference ( due to tunneling ) effects between the small `` bubbles '' regions , and therefore a wave function which is extended over their entire area , as indeed we observe to be the case in figs . [ fig9 ] and [ fig10 ] . in other words , one expects that a triangular potential with @xmath63 will lead to considerable increase of the localization length , and respectively widening of the critical energy region , at filling factors below one - half , whereas @xmath294 will favor delocalization at filling factors above one - half , as seen in fig . [ fig11 ] . this asymmetry is therefore clearly a consequence of the asymmetry of the triangular potential , and is absent for a square potential with only lowest order fourier coefficients , which possesses electron - hole symmetry . this should have clear implications for the transport properties of the system . in this study we have investigated the effects of moderate - to - large smooth disorder on the hofstadter butterfly expected for 2des in a perpendicular magnetic field and a pure periodic modulation . the parameters of our study are chosen so as to be suitable for the interpretation of recent experiments on a two dimensional electron system in gaas / algaas heterostructure with periodic modulation provided by a diblock copolymer , @xcite the experiment shows that ( i ) the longitudinal resistance @xmath305 is still peaked approximately at half filling ; ( ii ) there are many reproducible oscillations in @xmath305 , indicating non - trivial electronic structures in the patterned sample ; ( iii ) the distribution of these oscillatory features is asymmetric , with most of them appearing on the high magnetic fields ( i.e. low filling factors @xmath306 ) side of the peak of @xmath305 ; and ( iv ) the temperature dependence of @xmath305 indicates that the asymmetric off - peak resistance is thermally excited , whereas the central @xmath305 peak ( close to half filling ) has metallic behavior . these observations can not be explained on the basis of the hofstadter structure.@xcite this is not surprising , since one expects that large disorder will modify this structure considerably . effects of small disorder on the hofstadter butterfly had been investigated previously using scba,@xcite but this basically perturbational approach is not appropriate for the case of moderate - to - large smooth disorder . instead , we identify and use a number of techniques which give the exact solution ( if electron - electron interactions , as well as inelastic scattering are neglected ) while avoiding brute force numerical diagonalizations . our results demonstrate that while the hofstadter butterfly is destroyed by large disorder , the effects of the periodic potential are non - trivial for states near the critical regime . firstly , they lead to a significant increase in localization lengths of the localized states at mesoscopic ( @xmath307 ) length scale and induce an effective widening of the critical regime near the critical regime . this is achieved through a spreading of the electron wave - function on the flat regions of the slowly varying disorder potential , where their behavior is dominated by the periodic modulation . this regime shows an interesting transition between the pure disorder and the pure periodic potential cases . in the case of pure disorder , the semi - classical approach tells us that at finite filling factors , some areas of the sample are fully occupied by electrons with the maximum possible density of @xmath308 ( these are the areas where the disorder potential has minima ) whereas other areas are fully devoid of electrons ( areas where the potential has maxima ) and the boundary between such regions is very sharp . on the other hand , for a pure periodic modulation all wave functions have translational invariance with the proper symmetry , and therefore electron densities are uniform over the entire sample ( up to small periodic modulations inside each unit cell ) . when both types of potential are present , with disorder being dominant , our results show _ three _ types of areas . there are regions which are fully occupied and regions which are completely devoid of electrons , as in the case of pure disorder . however , the periodic potential leads to a widening of the boundary between the two , where the wave functions interact with several oscillations of the periodic modulation and therefore have some partial local filling . as the critical regime containing wave functions percolating throughout the sample is approached , this spreading of the wave function becomes dominant in establishing the transport properties of the system . an equivalent way to say this is that the main effect of the periodic potential is to provide bridging between the fully - occupied electron `` puddles '' created by the disorder potential . since the connecting areas are relatively flat , the local wave functions respond to the local periodic potential , and therefore locally have a hofstadter - butterfly like structure . if the partial filling factor in such a region is inside the gap of the local hofstadter butterfly structure , one expects no transport through this local area . this should result in a dip in the longitudinal transport , since in such cases the periodic potential will not transport electrons from one `` puddle '' to another one . by contrast , if the local filling factor in such a region is inside a subband of a local hofstadter structure , this area will establish a link between different `` puddles '' and thus help enhance the transport through the sample . transport in this regime should show strong thermal activated behavior , in contrast to metallic transport in the critical regime where the wave functions connect opposite edges of the sample . as a result , one expects a series of local minima and maxima in the longitudinal resistivity on either side of the central peak induced by the extended states ( critical regime ) . furthermore , for an asymmetric triangular potential , this response should be strongly asymmetric , with the effect most visible on one side of the central peak . ( one must keep in mind that since tunneling leads to exponential dependencies , even small differences in the extent of the wave functions can have rather large effects on @xmath11 ) . such an asymmetry should also be present in longitudinal conductance of finite but low temperature , e. g. in the hopping regime which is sensitively dependent on the nature of the localized wavefunctions , as is indeed seen experimentally.@xcite to summarize , our qualitative explanation for the various experimental features are as follows : \(i ) the @xmath305 peak is roughly at the center of the band because the weak periodic potential can not establish a hofstadter - like structure over the whole band . instead , low and high @xmath239 states are strongly localized and do not transport longitudinal currents . \(ii ) new extended states induced by the periodic potential are responsible for the reproducible peaks and valleys appearing in @xmath305 . \(iii ) the periodic potential also leads to the expansion of localized wave functions , which contribute to the thermally activated conduction at lower filling factors . the detailed structure of the wave functions gives rise to the oscillations of the off - peak @xmath305 , similar to conductance fluctuations.@xcite finally , \(iv ) the asymmetry in @xmath305 is a manifestation of the asymmetry of the triangular potential , which has a stronger effect at low filling factors than at high filling factors for @xmath63 . we predict that this asymmetry should be absent for a symmetric square periodic potential . the weak point in our calculation is that we are unable to accurately model the potential in the real samples , because various screening effects have not been properly taken into account . also , we have no quantitative information about the magnitude of the periodic potential in the 2des layer , because of the additional strain@xcite contribution induced by the periodic decoration . as a result , we only claim qualitative agreement with the experiment , although our investigations show the same type of behavior for various types of disorder potentials and various ( small - to - moderate ) strengths of the periodic potential . the most direct check of this work would be an experimental demonstration that thermally activated conduction appears symmetrically on both sides of the @xmath305 peak for a periodic potential with square symmetry and primarily lowest fourier coefficients . limited computer resources restrict our calculations to samples no larger than @xmath309m@xmath310 m , while the sample used in experiment has a size of @xmath311m@xmath312 m . from a theoretical point of view , it is interesting to ask what is the thermodynamic limit . for pure disorder , it is believed that in this limit the typical size of wavefunction diverges at a single critical energy . since we can not pursue size - dependent analysis for samples larger than @xmath309m@xmath310 m , we do not know whether the small periodic potential will lead to a finite size critical regime in the thermodynamic limit , although this seems likely . from an experimental point of view , the interesting question is whether the hofstadter structure can be observed at all . our studies suggest that this may be possible for small mesoscopic samples , where the slowly - varying disorder has less effect . alternatively , one must find a way to boost the strength of the periodic modulations inside the 2des . we thank sorin melinte , mansour shayegan , paul m. chaikin and mingshaw w. wu for valuable discussions . we also thank prof . li kai s group in computer science department of princeton university for sharing their computer cluster with us . this research was supported by nsf grant dmr-213706 ( c.z . and r.n.b . ) and nserc ( m.b . ) . m.b . and r.n.b . also acknowledge the hospitality of the aspen center for physics , where parts of this work were carried out . s. melinte , m. berciu , c. zhou , e. tutuc , s. j. papadakis , c. harisson , e. p. de poortere , m. wu , p. m. chaikin , m. shayegan , r. n. bhatt and r. a. register , to appear in phys . ( cond - mat/0311400 ) . allowed trajectories are such that the total flux through the area enclosed by the trajectory is an integer number of elementary fluxes @xmath58 . this can be understood in the spirit of the bohr - sommerfeld quantization rule , since it ensures constructive interference of the wave function around the contour . except for the most localized states , found at the bottom and the top of the band , all other localized states are such that they enclose large numbers of magnetic fluxes . imposing the exact quantization condition ( which is numerically time consuming ) leads to a negligible change in the value of the allowed equipotential value with respect to a randomly chosen value . since such small changes do not influence the shape of the density of states , we ignore imposing this quantization condition in obtaining the semi - classical densities of states . |
we use a combination of a t hooft limit and numerical methods to find non perturbative solutions of exactly solvable string theories , showing that perturbative solutions in different asymptotic regimes are connected by smooth interpolating functions . our earlier perturbative work showed that a large class of minimal string theories arise as special limits of a painlev iv hierarchy of string equations that can be derived by a similarity reduction of the dispersive water wave hierarchy of differential equations . the hierarchy of string equations contains new perturbative solutions , some of which were conjectured to be the type iia and iib string theories coupled to superconformal minimal models of type . our present paper shows that these new theories have smooth non - perturbative extensions . we also find evidence for putative new string theories that were not apparent in the perturbative analysis . * non - perturbative string theory from water waves * * ramakrishnan iyer , clifford v. johnson , jeffrey s. pennington * _department of physics and astronomy _ _ university of southern california _ _ los angeles , ca 90089 - 0484 , u.s.a . _ _slac national accelerator laboratory _ _ stanford university _ _ stanford , ca 94309 , u.s.a . _ ramaiyer ,johnson1 , [ at ] usc.edu ;jpennin [ at ] stanford.edu | 1011.6354 | in the quest to better understand non perturbative phenomena in string theory , and to find clues as to the full nature of m theory@xcite , rich solvable examples are of considerable value . the type 0 minimal string theories ( formulated in refs . @xcite and refs . @xcite , and recognized as type 0 strings in ref . @xcite ) have highly tractable non perturbative physics , and despite being exactly solvable contain a rich set of phenomena such as holography and open closed dualities . like the original minimal strings@xcite , they have bosonic spacetime physics , although they have worldsheet fermions , and may be thought of as two dimensional supergravity coupled to @xmath6 superconformal minimal models , with a diagonal gso projection . the physics of the minimal strings can be succinctly formulated in terms of associated non linear ordinary differential equations , known as the string equations . they furnish asymptotic expansions for the free energy once the boundary conditions are fixed . these expansions take the form of a string world sheet expansion . typically ( for the cases of the type 0a and 0b systems coupled to the @xmath7 superconformal minimal models ) there are two perturbative regimes in which such an expansion can be developed . one is interpreted as purely closed string backgrounds with integer , @xmath8 , units of r r flux , while the other has both open and closed string sectors with @xmath8 d branes in the background d branes or flux units correspond to special soliton solutions , as shown in refs . @xcite . ] . remarkably , these models have a non perturbative completion ( first discovered in refs . @xcite in the context of the type 0a systems ) that connects these two asymptotic regimes , furnishing an example of a so called `` geometric transition '' between open and closed string backgrounds . while in the case of type 0a , complete ( numerical ) solutions of the exact string equations for any @xmath8 can be found , the connectedness of the type 0b case has been argued for on the basis of a large @xmath8 , t hooft like , limit where the interpolating solution can be found using algebraic techniques@xcite . this has more than just shades of the ads / cft correspondence@xcite , where in the prototype , the open string physics of @xmath8 d3branes can be rewritten in terms of that of closed strings in ads@xmath9 with @xmath8 units of r r flux . in the present context we have a precise analogue of this important example . there is of course a t hooft large @xmath8 limit there too , connecting supergravity to large @xmath8 yang mills . that we have here not just the analogue of the large @xmath8 limit but also knowledge of how to go beyond , may ultimately prove instructive . in ref . @xcite , we investigated a new system called the dispersive water wave ( dww ) hierarchy@xcite , which we argued should yield new string equations _ via _ a similarity reduction . these string equations turn out to be a painlev iv hierarchy of equations . we found that both the type 0a and 0b string theories were found to be naturally embedded within this framework . in addition to this new non perturbative connection , we found further that the two string theories are merely special points in a much larger tapestry of possibilities that also appear to be string theories . this is somewhat suggestive of an m theory , now for minimal strings , which is exactly solvable . it clearly deserves further study . among the other special points which suggested themselves ( using perturbation theory ) were some that we conjectured to be the type iia and iib string theories coupled to the @xmath0 superconformal minimal models . much of the analysis carried out to support this identification was using perturbative techniques . the subject of the present paper is to continue the study of this rich system into the non perturbative regime . we argue that the conjectured type ii theories have non perturbative completions by showing that the corresponding perturbative expansions match onto each other smoothly . this is accomplished using a combination of analytical and numerical techniques similar to those used for the type 0 theories coupled to the @xmath7 superconformal minimal models . the outline of the paper is as follows : in sections 2 and 3 , we review essential aspects of the well known type 0 theories and provide a summary of results about the dww system that will be needed later . we reproduce the analytic ( t hooft limit ) technique used for the type 0b string theory , as in ref . @xcite , in section 4 to keep the presentation self contained . we then reformulate it in a manner that allows us to extend the technique to our dww system . we apply this strategy to our system in section 5 for the first two even dww flows , which contain the type 0 and conjectured type ii theories of interest and find that the type ii theories possess smooth non perturbative solutions . in addition , we find evidence for the existence of new non perturbative solutions with novel asymptotics . in section 6 , we consider the first ( and simplest ) flow of the dww system . for this case , we are able to strengthen evidence for the existence of these new non perturbative solutions with exact numerical results . we end with a brief discussion in section 7 . an appendix details the expansions arising from the dww system for the first few flows along with some of their essential properties . we begin with a brief review of the type 0 string theories coupled to the @xmath7 superconformal minimal models of type @xmath10 . those familiar with these models can proceed directly to the next section . type 0a string theory coupled to the @xmath7 superconformal minimal models ( first derived and studied in refs.@xcite and identified as type 0a in ref . @xcite ) is described by the string equation @xmath11 where , for a particular model , @xmath12 is a real function of the real variable @xmath13 , a prime denotes @xmath14 , and @xmath15 and @xmath16 are real constants . the quantity @xmath17 is a function of @xmath12 and its @xmath13derivatives . in general @xmath12 additionally depends on couplings @xmath18 associated with flowing between the various models . then we have @xmath19 where the @xmath20 $ ] are polynomials in @xmath12 and its @xmath13derivatives , called the gelfand dikii polynomials@xcite . they are related by a recursion relation ( defining a recursion operator @xmath21 ) @xmath22 and fixed by the value of the constant @xmath23 and the requirement that the rest vanish for vanishing @xmath24 . the first four are : @xmath25 the @xmath26th model is chosen by setting all the @xmath27 to zero except @xmath28 and @xmath29 this number is chosen so that the coefficient of @xmath30 in @xmath17 is set to @xmath31 . as @xmath32 . if we had instead chosen @xmath33 , we would have chosen the coefficient of @xmath30 to be unity . ] the flows between various models are captured by the integrable kdv hierarchy@xcite . the function @xmath12 defines the partition function @xmath34 of the string theory @xmath35 @xmath36 where @xmath37 is the coefficient of the lowest dimension operator in the world sheet theory . the asymptotic expansions of the string equations for the first two cases are : @xmath38 @xmath39 @xmath40 @xmath41 the solutions for various @xmath26 for @xmath42 can be numerically and analytically shown to match onto the solution for @xmath43 , providing a unique@xcite non perturbative completion of the theory . ( see figure ( [ fig : plot ] ) for an example of a solution found using numerical methods . ) + in the @xmath44 regime , @xmath15 represents@xcite the number of background zz d branes @xcite in the model , while in the @xmath45 regime , @xmath15 represents the number of units of r r flux in the background@xcite . type 0b string theory coupled to the @xmath46 superconformal minimal models @xcite is described by the following string equations@xcite : @xmath47 where the @xmath48 and @xmath49 are polynomials of functions @xmath50 and @xmath51 ( and their derivatives ) , and @xmath16 and @xmath52 are real constants . the function @xmath53 defines the partition function of the theory @xmath35 @xmath54 where @xmath37 is the coefficient of the lowest dimension operator in the world sheet theory . the @xmath55th model is chosen by setting all @xmath56 to zero except @xmath57 and @xmath58 . these models have an interpretation as type 0b strings coupled to the @xmath59 superconformal minimal models only for even @xmath60 . the flows between various models are captured by the integrable zakharov shabat ( zs ) hierarchy@xcite . the asymptotic expansions of the string equations ( [ streqn0b ] ) for the first two even @xmath60 are : @xmath61 @xmath62\ , \quad ( x \rightarrow \infty ) \nonumber\\ { \widetilde w}(x ) & = & \frac{\nu q\sqrt{2}}{4|x|^{1/2 } } - \frac{\nu^2 q^2}{4 |x|^2 } + \frac{\nu^3}{|x|^{7/2}}\frac{5\sqrt{2}}{64 } q\left(1 + 4q^2\right)+\cdots \quad ( x \rightarrow -\infty)\end{aligned}\ ] ] @xmath63 @xmath64 in the @xmath45 regime , @xmath52 represents the number of background zz d branes in the model , while in the @xmath65 regime it counts the number of units of r r flux in the background@xcite . the structure of solutions with increasing @xmath55 is particularly rich @xcite ; the @xmath66 expansion for @xmath45 shown above breaks a @xmath67 symmetry due to the presence of r r fields in the dual string theory . unlike the 0a case , the solutions for @xmath68 have not been shown numerically to match onto those for @xmath69 so far . the highly non - linear nature of the string equations makes it difficult to numerically obtain smooth solutions connecting the two regimes for the 0b case . nevertheless , as argued in ref . @xcite and reviewed in section [ sec : type0expnmtch ] , these theories can be argued to be non - perturbatively complete in a particular ( t hooft ) limit . for the @xmath70 ( @xmath71 ) model @xmath72 super conformal minimal models is given by @xmath73 . for @xmath70 , the central charge of the @xmath74 superconformal minimal model is thus zero and we simply have the pure world sheet supergravity sector . ] , the full non perturbative solution is known since it can be mapped directly to the solution known for the @xmath71 type 0a case @xmath35 the morris map @xcite . the string equation for the 0a theory becomes the string equation for the 0b theory once one identifies @xmath15 with @xmath52 , but with the sign of @xmath75 reversed . analogues of the morris map for higher @xmath55 are not known . the dww system introduced in ref . @xcite leads to a painlev iv hierarchy of string equations : @xmath76 where @xmath77 and @xmath15 are real constants and @xmath16 plays the same role as for the type 0 theories ( note that here and in the rest of the paper , for any function @xmath78 , @xmath79 will denote @xmath80 ) . the functions @xmath81 and @xmath82 are generalizations of the two functions @xmath50 and @xmath83 used to describe the 0b theory . the polynomials @xmath84 $ ] and @xmath85 $ ] are defined by @xmath86\\ k_n[u , v]\end{array}\right ) \ \quad,\ ] ] where @xmath87 $ ] and @xmath88 $ ] are polynomials in @xmath81 , @xmath82 and their derivatives , similar to the polynomials @xmath89 and @xmath90 for the 0b theory . they satisfy the recursion relation @xmath91\\ k_{n+1}[u , v]\end{array}\right ) \ = r \left(\begin{array}{c } l_n[u , v]\\ k_n[u , v]\end{array}\right ) \ \quad .\end{aligned}\ ] ] where @xmath92 is the recursion operator of the dww hierarchy , given by @xmath93 the first few @xmath94 and @xmath95 are as follows : @xmath96 the @xmath55th model is chosen by setting all @xmath97 equal to zero except for @xmath98 and @xmath58 which is chosen to be a numerical factor to fix the normalization . the parameter @xmath58 can be replaced by @xmath99 in order to make direct contact with recent literature which discusses this system in a much different ( mathematical ) context@xcite . analytic expansion solutions to the string equations ( [ dwwstring1 ] ) and ( [ dwwstring2 ] ) for all @xmath55 can be organized into five main classes on the basis of the @xmath100 ( or @xmath101 ) behavior of @xmath81 and @xmath82 , as follows , @xmath102 they follow interesting patterns , while the expansions in class 5 are the analogues of the @xmath67 symmetry - breaking solutions of the 0b theories . ] with increasing @xmath55 and have been explored in detail in ref . representative solutions for the first few @xmath55 have been provided in the appendix for ease of reference . in addition to the type 0 theories reviewed earlier , these solutions encode new string theories , some of which were conjectured to be type ii string theories coupled to super conformal minimal models as reviewed below . the type 0 theories reviewed earlier can be recovered completely from this system of equations by applying appropriate constraints @xcite . setting @xmath103 and @xmath104 in ( [ dwwstring1 ] ) and ( [ dwwstring2 ] ) results in the type 0a and type 0b theories respectively . ] . consistency of the two equations then requires that @xmath105 for type 0a and @xmath106 for type 0b , while the free parameter counts the branes and fluxes in the respective theories . the function @xmath82 ( after appropriate redefinition ) encodes the free energies of the respective theories , according to equations ( [ 0afreeenergy ] ) and ( [ 0bfreeenergy ] ) . one can also recover solutions corresponding to the type 0 theories by setting @xmath107 or @xmath106 in appropriately chosen expansions from the five classes listed above . expansions in classes 1 and 2 ( labeled @xmath108 and @xmath109 ) reduce to those of the type 0a theory ( see ( [ 0aexpnsk=1 ] ) and ( [ 0aexpnsk=2 ] ) ) once we set @xmath105 . this helps fix the directions of the various expansions by requiring that @xmath82 be real , with the result that some expansions are real and others complex for a given @xmath55 . further details can be found in ref . @xcite . a full solution for @xmath82 is constructed by specifying its behavior in the two asymptotic regimes , as seen for the @xmath110 0a theory in figure ( [ fig : plot ] ) . this 0a solution is obtained from the dww expansions , @xmath108 and @xmath109 , by setting @xmath105 . using this scheme , the above expansions can be organized in the form of a square as shown in figure ( [ square2 ] ) . the four corners of the square correspond to four different string theories . the top corners represent the known type 0 strings coupled to the @xmath7 @xmath10 superminimal models . the bottom corners represent two new theories that were conjectured in ref . @xcite to be type iia and iib string theories coupled to the @xmath0 @xmath1 minimal models ( for even @xmath26 only ) . at each corner , one parameter from @xmath111 is frozen to the value indicated on the figure , while the other parameter counts the branes / fluxes in the appropriate regime . the horizontal sides of the square correspond to expansions in the @xmath112 direction , while the vertical sides represent expansions in the @xmath113 direction . for all @xmath55 , there are two special points on the vertical sides corresponding to @xmath114 , where the expansions @xmath115 and @xmath116 identically vanish . + it must be noted that the square fully exists only for even @xmath117 . for @xmath118 , the expansions labeled @xmath119 become complex as @xmath112 expansions and the expansions labeled @xmath120 do not appear for odd @xmath55 . the new theories were conjectured to be type ii strings coupled to super minimal matter by matching the genus zero contributions from the asymptotic expansions ( the terms appearing at order @xmath121 in the free energy ) with the continuum calculation of their one loop partition functions . in the absence of non perturbative constraints ( like @xmath122 for the 0a theory ) leading to the conjectured type ii theories , the corresponding values of @xmath123 and @xmath124 in figure [ square2 ] , were obtained by systematic analysis of the properties of the genus zero contributions @xcite . however , pairing perturbative expansions in this manner does not guarantee the existence of a full non - perturbative solution with the desired properties . a type of t hooft limit will allow us to argue for the existence of such non - perturbative solutions for the type ii theories . we demonstrate how this limit works for the two parameter dww system in section ( [ sec : dwwexpnmtch ] ) . before that , we review the t hooft limit for the known type 0 theories . the theories encoded by the string equations described so far can be non perturbatively completed by matching the perturbative solutions for @xmath125 on to those for @xmath126 . numerical solutions have been found for the string equations for the 0a models that smoothly interpolate between the two asymptotic regimes as discussed earlier . see , e.g. , figure ( [ fig : plot ] ) . the string equations for the 0b models , however , are more difficult to analyze numerically and attempts to find smooth numerical solutions similar to the 0a ones have been unsuccessful thus far @xmath127 0b theory , where the numerical @xmath71 type 0a solution can be converted into it @xmath35 the morris map , as mentioned in section [ sec : type0breview ] . ] this difficulty is inherited by the dww string equations presented in section [ sec : dwwreview ] . it was shown in ref . @xcite that the asymptotic expansions for the 0b theories ( for example , those in equations ( [ 0bexpnsm=2 ] ) and ( [ 0bexpnsm=4 ] ) ) match on to each other in a t hooft limit , with the brane / flux counting parameter ( labeled @xmath52 for 0b and @xmath15 for 0a ) taken to be large . this was then taken to suggest that the theories are non perturbatively complete even when the parameter is finite . we review this limit in some detail , before analyzing our system in the same limit in section [ sec : dwwexpnmtch ] . the string equation for the simplest 0b theory with @xmath70 is @xmath128 obtained by using the explicit forms of the polynomials @xmath21 and @xmath129 and eliminating @xmath83 between the two equations ( [ streqn0b ] ) . consider the limit @xmath130 , @xmath131 with @xmath132 fixed . this is our t hooft limit with @xmath133 being the t hooft coupling and @xmath52 as counting the number of units of r r flux , this is recognized as the usual t hooft coupling @xmath134 familiar from higher - dimensional string theories . taking @xmath131 amounts to taking the limit @xmath135 . ] . defining @xmath136 , the string equation ( [ streqn0bm=2v2 ] ) becomes @xmath137 in the large @xmath52 limit , the first term containing the derivative can be neglected resulting in an algebraic equation for @xmath138 , @xmath139 for generic @xmath140 this equation has 3 solutions , only _ one _ of which is real for all @xmath140 ( an easy check is to consider @xmath141 ) . this solution reads @xmath142 \quad.\ ] ] for @xmath143 , the arguments of the square roots are positive and @xmath144 is real . since there are opposite signs in front of the square roots in the above expression , the half - integer powers of @xmath140 cancel when one expands @xmath144 for @xmath145 . for @xmath146 the arguments of the square roots are negative and the second and third terms in the solution equation are complex , but @xmath144 is real since all the imaginary contributions cancel out . the expansions for @xmath144 for large negative and positive @xmath140 are @xmath147 these expansions reproduce the coefficients of the highest powers of @xmath52 in the asymptotic expansions to the string equation ( [ 0bexpnsm=2 ] ) after remembering that @xmath148 . one smooth function ( [ vsolnm=2 ] ) captures the limiting behavior of large @xmath52 and connects = \int dx \left[\frac{1}{2}r'^{2 } + v(r^2)\right]$ ] whose equation of motion is ( [ streqn0bm=2v2 ] ) . since this action is bounded below , it is clear that a solution to this equation will exist . the original matrix model integral dual to the theory is well defined and convergent , so one expects a finite and real answer for the free energy @xmath149 . it is then natural to expect that equation ( [ streqn0bm=2v2 ] ) will have a unique real and smooth solution . ] the two asymptotic regions smoothly . this limit has been interpreted in @xcite as a t hooft limit where only spherical topologies with boundaries ( d branes ) or fluxes survive . for @xmath150 , the 0b theory exhibits the gross witten phase transition at @xmath151 ( namely @xmath152 for @xmath153 and @xmath154 for @xmath69 ) . this transition can be smoothed out by either the genus expansion with @xmath155 or by the expansion in the t hooft parameter for large @xmath52 . a similar analysis shows that the asymptotic expansions for the @xmath66 theory ( [ 0bexpnsm=4 ] ) match smoothly onto each other . the analysis is somewhat complicated because @xmath83 can not be eliminated in favor of @xmath50 , unlike the @xmath70 case and one has to use the asymptotic expansions for both @xmath50 and @xmath83 . the expansions for @xmath83 are @xcite : @xmath156 in the @xmath130 limit with @xmath157 , @xmath158 and @xmath159 , the derivative terms in the string equation for this case are suppressed , resulting in @xmath160 the second equation , on solving for @xmath161 , gives @xmath162 from equation ( [ 0bexpnsm=4omega ] ) , it can be seen that @xmath163 is negative , so that only the solution @xmath164 can be chosen . such multiple solutions will be prominent in our more general analysis in the next section . ] for real @xmath161 . substituting @xmath165 into the first equation gives , after some rearrangement : @xmath166 defining @xmath167 reduces the above equation to a quadratic in @xmath168 , @xmath169 solving for @xmath168 as a function of @xmath170 @xmath171 will allow a solution for @xmath172 . the solution @xmath173 is always positive and non zero as a function of @xmath170 . to get the asymptotics we desire , we focus on @xmath174 , which is negative and zero only for @xmath175 . using @xmath176 it is easy to see that as @xmath177 , @xmath145 and and as @xmath178 , @xmath179 . as @xmath140 changes continuously , we will always lie on the @xmath180 branch since the two branches never cross . the expansion for @xmath180 as a function of @xmath140 can then be worked out @xcite to be @xmath181 the coefficients in these expansions agree with the leading powers for large @xmath52 in the expansions of @xmath83 in equation ( [ 0bexpnsm=4omega ] ) and can again be interpreted as a limit in which only spherical topologies with boundaries or fluxes survive .. this shows that an appropriately chosen solution of equation ( [ vheqnm=4 ] ) interpolates between the two asymptotic regimes in the limit of large @xmath52 . this t hooft limit removes the derivative terms from the string equations , resulting in algebraic equations that are simpler to analyze . it can then be sensibly thought of as _ algebraic _ limit . including the derivatives gives terms _ sub leading _ in powers of @xmath52 and presumably does not introduce any singularities that would destroy the smooth interpolation . leading powers of @xmath52 likely smooth out any sharp transitions ( as we saw for the gross witten phase transition above ) from large negative @xmath75 to large positive @xmath75 . it is interesting to work out the t hooft limit of the type 0a theories ( with string equations ( [ streqn0a ] ) ) . since , as already discussed , solutions of the full equations have been obtained numerically , it is instructive to compare the t hooft limit ( _ i.e. , _ algebraic ) results to the numerical results . ( note that while the last section s type 0b results were a review , the type 0a analysis is presented here for the first time . ) the string equation for this theory is , @xmath182 consider the limit @xmath183 , @xmath184 with @xmath185 fixed . this is the t hooft limit with @xmath186 being the t hooft coupling . defining @xmath187 , the above string equation becomes @xmath188 in the large @xmath15 limit , the derivative terms can be neglected to give the simple algebraic equation @xmath189 as for the type 0b case of the last section , the solutions to this cubic equation ) to obtain this equation directly from the corresponding 0b algebraic equation ( [ 0balgeqnm=2 ] ) . ] can be expanded as a taylor series for @xmath190 and seen to reproduce the coefficients of the highest powers of @xmath15 in the asymptotic expansions ( [ 0aexpnsk=1 ] ) . one smooth function essentially connects the two asymptotic regions , as in the 0b theory . figure ( [ 0ak1g1 ] ) shows a comparison between the solution to equation ( [ 0aalgeqnforgk=1 ] ) and the solution to the exact string equations ( [ streqn0ak=1 ] ) obtained by numerical methods , both with @xmath191 . the two solutions deviate slightly from each other in the interior , and asymptotically , they match . this can be seen in figure ( [ 0ak1g1diff ] ) , which plots the difference between the exact and algebraic solutions , which goes to zero for large @xmath192 . physics sub leading in powers of @xmath15 contribute to the exact solution , while only the highest powers of @xmath15 contribute to the algebraic solution . the inclusion of the derivative terms does not introduce any singularities . + the string equation for the @xmath110 0a theory has @xmath193 . in the large @xmath183 and @xmath184 limit , with @xmath194 held fixed , the derivative terms drop out . defining @xmath195 results in the algebraic equation @xmath196 the solutions to this equation reproduce the coefficients of the highest powers of @xmath15 in the asymptotic expansions ( [ 0aexpnsk=2 ] ) . figure ( [ 0ak2g1 ] ) shows a comparison between the solution to equation ( [ 0aalgeqnforgk=2 ] ) and the solution to the exact @xmath110 0a string equation obtained by numerical methods . the difference between the two solutions is plotted in figure [ 0ak2g1diff ] and can be seen to approach zero for large @xmath192 . the deviation for finite @xmath75 and asymptotic matching is evident , and can be explained by the additional contributions coming from sub leading powers of @xmath15 in the exact solution . + a modification of the above limit ( one that does not affect the physics ) will allow us to apply it to the painlev iv hierarchy of string equations . the need for such a modification will be explained in the next section . instead of rescaling variables in the 0b ( 0a ) theory to eliminate @xmath52 ( @xmath15 ) , one can remove the derivatives by first rescaling @xmath197 and then taking the limit @xmath198 in that order , without affecting the physics . note that the rescaled @xmath52 is large , so this is still a large @xmath52 limit . the limit @xmath199 amounts to taking @xmath135 , as is clear from the definition of @xmath200 ( @xmath201 for @xmath70 and @xmath202 for @xmath203 ) . this modified limit therefore extracts the physics same as the t hooft limit of the previous subsections . in this modified limit , the following equations are obtained for the type 0b theories for @xmath70 , @xmath204 and for @xmath66 , @xmath205 these are the same as equations ( [ 0balgeqnm=2 ] ) and ( [ vheqnm=4 ] ) with @xmath206 , @xmath207 and @xmath208 , but with the rescaled parameter @xmath52 explicitly present . for the 0a theory , after rescaling @xmath209 , the following equations are obtained for @xmath71 and @xmath110 , respectively , @xmath210 in fact , for general @xmath26 , the type 0a string equation ( [ streqn0a ] ) effectively reduces to @xmath211 where @xmath212 is obtained from @xmath213 after dropping all the derivatives of @xmath12 . the string equations of the dww hierarchy ( reviewed in section [ sec : dwwreview ] ) are quite general , encoding the string equations of type 0a and of type 0b as special cases . this generality , however , comes at a price : all but the simplest cases are too complicated to be solved using standard numerical methods . this complicates our task of performing a non perturbative analysis of the string theories proposed in ref . @xcite , which were based on strictly perturbative considerations . to facilitate this analysis , we examine the dww string equations in a ( modified ) t hooft limit . the result of this study is considerable evidence that the conjectured type ii theories possess smooth solutions connecting the perturbative expansions . this analysis also uncovers new non - perturbative solutions , absent from the earlier perturbative analysis , which we suspect may encode new unidentified string theories . in section [ sec : type0expnmtch ] we described how to take the t hooft limit of the type 0a and type 0b theories . implicit in these methods was the fact that the type 0 theories each depend on a single parameter , which counts branes or fluxes . the full dww equations , in contrast , have two free parameters , so it is not immediately clear how to implement the t hooft limit in this more general case . the perturbative analysis of ref . @xcite suggests that , in addition to the type 0 theories , the full dww string equations also describe type ii theories whose branes and fluxes are counted by a single parameter . this motivates our assumption that _ new theories are described by a one dimensional subspace of the parameters @xmath77 and @xmath15_. we further restrict our study to subspaces defined by _ linear _ combinations of @xmath77 and @xmath15 . this is the simplest possibility and although there could be interesting one - dimensional subspaces corresponding to non - trivial curves in @xmath214 parameter space , we will not consider them here . given this assumption and the scaling prescription described in section [ sec : modlim ] , we can utilize the t hooft limit to analyze the string equations ( [ dwwstring1 ] ) and ( [ dwwstring2 ] ) . for given @xmath215 and @xmath216 , let @xmath217 be the constraint defining the one - dimensional parameter subspace of a new theory under consideration . since the constraint is ( by assumption ) satisfied by the solutions of interest , we can impose the constraint on the dww equations themselves . we are left with a single free parameter which we can recale as in section [ sec : modlim ] to take the t hooft limit . in this way we obtain algebraic equations which are simple to analyze . unfortunately , in the search for new theories , we do not _ a priori _ know the values of @xmath215 and @xmath216 , so the above method is not directly applicable . nevertheless , we can proceed by adopting a slightly different perspective . note that the constraint @xmath218 implies that the free parameter can be taken to be @xmath77 or @xmath15 , or any linear combination of them ( as long as it is independent of the constraint ) . rescaling the free parameter by @xmath219 is therefore equivalent to rescaling @xmath220 and @xmath221 . this procedure is independent of the details of the original constraint , so it can be implemented without actually knowing the values of @xmath215 and @xmath216 . the result is a set of algebraic equations which depends on @xmath77 and @xmath15 . moreover , the solutions to these equations can be used to deduce information about the original constraint . to see this , note that since @xmath216 is a constant , after taking the limit the constraint takes the form @xmath222 . therefore , the original constraint is satisfied by any solution of the algebraic equations which satisfies @xmath222 . turning this around , if a particular solution to the algebraic equations only exists when @xmath222 , then this is a strong indication that it descends from the special solution to the full equations subject to @xmath223 . that is to say , it is an algebraic approximation to the full solutions of some new string theory . note that @xmath216 is not determined from this approach , but @xmath215 is . to obtain the algebraic equations , we begin with the dww string equations ( [ dwwstring1 ] ) and ( [ dwwstring2 ] ) , @xmath224 in the modified t hooft limit discussed above , @xmath225 the string equations simplify to @xmath226 where @xmath227 and @xmath228 are the polynomials in @xmath229 and @xmath230 obtained from @xmath231 and @xmath232 after dropping all the derivative terms . before analyzing these equations in full generality , let us discuss a few special cases to illustrate our methods . 1 . @xmath122 ( type 0a ) + the restriction to type 0a requires @xmath55 to be even , in which case @xmath233 . thus , setting @xmath122 in first equation of ( [ dwwalgeqns ] ) implies @xmath123 . this constraint can be written as @xmath234 with @xmath235 . as described above , this constraint lifts to @xmath236 at the level of the full equations . if we had nt known the parameter constraint of the 0a theory ( @xmath237 ) , the algebraic equations alone would indicate that @xmath238 , a constant . + alternatively one could start by looking for solutions to the equations ( [ dwwalgeqns ] ) with @xmath123 . a subset of smooth solutions , identified by their boundary behavior , will correspond to asymptotic expansions in classes 1 and 2 with @xmath77 fixed ( recall that these correspond to 0a ) . the boundary behavior of the smooth solutions is matched with the behavior of expansions in classes 1 and 2 , which are the asymptotic expansions of the type 0a theory . 2 . @xmath239 ( type 0b ) + setting @xmath239 forces @xmath106 and the first equation gives @xmath240 . for even @xmath55 , these reduce to the 0b algebraic equations after redefining the variables appropriately . the constraint @xmath106 can be rewritten as @xmath234 with @xmath241 and lifts to a constraint on the full equations which takes the form @xmath242 , a constant . the true parameter constraint for the full @xmath243 theory is @xmath244 ; the algebraic equations alone do not determine the precise value of @xmath216 . + as in the 0a case above , one could look for smooth solutions to the algebraic equations subject to @xmath106 . if they obey the correct boundary conditions ( expansion classes 1 and 3 ( or 5 ) ) one could identify these as 0b solutions . our strategy to demonstrate the existence of smooth non perturbative solutions to the type 0 and type ii theories and identify potentially new theories can be summarized as follows . take the limit ( [ modlimdww ] ) of the full dww equations to obtain a set of algebraic equations . search for solutions to these equations subject to one of three constraints , and for simplicity we exclude such cases from the present analysis . ] : @xmath245 by matching the asymptotic behavior of these solutions onto the various expansions , we identify the type 0 and type ii string theories . we also find new solutions with asymptotics different from those of the type 0 or type ii theories . we speculate that these solutions correspond to some new unknown string theories . as emphasized above , the precise parameter constraints of these new theories can not be fully determined from the algebraic equations . the string equations for @xmath70 reduce in the algebraic limit ( [ modlimdww ] ) to @xmath247 they produce a total of nine asymptotic expansions which fall into 4 classes ( see appendix [ appexpns ] ) . as the expansions within each class are related by various @xmath248 symmetries @xcite , we label each expansion only by a subscript specifying its class . the solutions corresponding to the three constraints in parameter space are described below . 1 . @xmath123 + a plot of solutions to the algebraic equations ( [ n=2algeqn ] ) with @xmath123 and @xmath249 is shown in figure ( [ n=2c0 ] ) . the plot on the right shows two smooth solutions that connect regions of large @xmath113 and @xmath112 . in our conventions , these are @xmath250 and @xmath251 solutions asymptotics @xmath252 and @xmath112 asymptotics @xmath253 by @xmath254 . ] . the plot on the left shows the solutions which do not join negative asymptotics to positive asymptotics , a feature we expect to find in any solutions describing new string theories . for this reason , these types of solutions will not be our primary interest . + the @xmath255 solutions correspond to the 0a theory ( with @xmath71 ) coupled to pure supergravity . it is unclear if the @xmath256 solutions correspond to a consistent new theory . based on the square of theories in figure ( [ square2 ] ) , it is tempting to conclude that this theory might be the type iia string theory coupled to the @xmath257 @xmath1 superconformal minimal model . however , as mentioned in the discussion below figure ( [ square2 ] ) , the @xmath119 expansion is complex as a @xmath112 expansion for @xmath258 . for the particular case of @xmath70 and @xmath259 , @xmath119 happens to be real , but this is an exception . we find it unlikely that the corresponding solution to the full equations exists and encodes a consistent theory , but we leave this question for future investigation . @xmath106 + figure ( [ n=2g0 ] ) shows the algebraic solutions for @xmath244 and @xmath260 . the right figure shows two smooth solutions , @xmath261 and @xmath262 , connecting the negative and positive regions . + the @xmath261 solutions are algebraic approximations to the full solutions of the 0b theory ( with @xmath70 ) coupled to pure supergravity . analogous to the previous case with @xmath259 , it is unclear if the @xmath263 solutions descend from a consistent new theory . it is tempting to conclude that these are type iib string theories coupled to the @xmath257 @xmath1 superminimal model , but this is likely incorrect because @xmath119 is complex as a @xmath112 expansion for @xmath118 . 3 . @xmath264 + we consider @xmath265 . the case of @xmath266 is completely analogous . figure ( [ n=2ceqg ] ) shows solutions for @xmath267 and @xmath249 . the solutions in the right figure are @xmath268 and @xmath269 and connect the negative region smoothly to the positive region . + the fixed value of @xmath270 can not be deduced from the algebraic solutions . the string equations in the t hooft limit , after setting @xmath271 , are too long to be reproduced here . the polynomials @xmath231 and @xmath232 reduce to , @xmath272 from which the corresponding algebraic string equations can be easily obtained using ( [ dwwalgeqns ] ) . there are a total of twenty five expansions @xcite , falling into five classes . solutions to the algebraic string equations with the three constraints are shown below . 1 . @xmath123 + figure ( [ n=4c0 ] ) shows solutions with @xmath259 and @xmath249 . there are three solutions that interpolate between the negative region and the positive region , @xmath273 , @xmath274 and @xmath275 . + the @xmath255 solutions represent type 0a coupled to the @xmath276 @xmath10 superminimal model . the @xmath275 solutions fit with our conjectured type iia string theory coupled to the @xmath277 @xmath1 superminimal model and are analogues of the @xmath67 symmetry breaking solutions of the 0b theory . + the @xmath278 solutions are new . on setting @xmath259 in the _ full _ expansions , the powers of @xmath15 in both correspond to a parameter counting branes . an underlying string theory with these solutions , if it exists , would have branes in both asymptotic regimes type expansions ( for @xmath107 ) in both asymptotic directions for @xmath75 . ] . further work is needed to conclude if such an underlying theory exists . 2 . @xmath106 + figure ( [ n=4g0 ] ) shows solutions with @xmath244 and @xmath279 . the solutions which interpolate between @xmath113 and @xmath112 are @xmath280 , @xmath274 and @xmath281 . the @xmath282 solutions correspond to the @xmath67 symmetry breaking solutions of type 0b string theory coupled to the @xmath276 superminimal model , while the @xmath283 solutions fit with our conjectured type iib string theory coupled to the @xmath277 @xmath1 superminimal models . + the @xmath284 solutions with @xmath106 are new , similar to those above with @xmath259 . further work is required to understand the existence and nature of such theories . @xmath264 + again we consider @xmath265 since the case of @xmath266 is completely analogous . figure ( [ n=4ceqg ] ) plots the solutions for @xmath279 and @xmath285 . the four smooth solutions which interpolate between @xmath113 and @xmath112 are two @xmath282 solutions , one @xmath286 solution and one @xmath284 solution . ( the two @xmath282 solutions can be obtained from one another by applying @xmath67 symmetries on the signs of the parameters ) . all of these solutions are new , and it is possible that they correspond to well - defined underlying string theories , but we leave any definitive claims to future work . the string equations for @xmath287 are , @xmath288 where we have used the relation @xmath289 . these equations are simple enough to allow for complete numerical solutions , which will help us classify new non - perturbatively complete solutions . we will also examine these equations in the limit ( [ modlimdww ] ) , under which they reduce to @xmath290 where we have used @xmath291 . these equations admit four asymptotic expansions , which we label as : @xmath292 the solutions with the three different parameter constraints are presented below . 1 . @xmath123 + a plot of solutions to the algebraic equations ( [ n=1algeqn ] ) with @xmath123 is shown in figure ( [ c0gm2 ] ) . there are two smooth solutions interpolating between @xmath112 and @xmath113 , labeled @xmath293 in our convention . + + + as outlined in the appendix ( [ braneflux ] ) , @xmath115 with @xmath123 can be interpreted as a flux expansion ( for odd @xmath55 ) in @xmath15 , while @xmath116 with @xmath123 is a brane expansion in @xmath15 . these interpretations are possible strictly for @xmath123 ; a finite value of @xmath77 results in powers of @xmath200 that do not allow @xmath115 to be interpreted as a flux expansion . so a string theory corresponding to this solution ( if it exists ) should require @xmath259 in the exact string equations . 2 . @xmath106 + the plots for this case are shown in figure ( [ cm2g0 ] ) . there are three smooth solutions , two of which are @xmath294 and @xmath295 solutions . the third solution is parallel to the @xmath75axis and exactly equals @xmath296 and @xmath297 . it is a 0b solution , the ` topological point ' of the 0b theory @xcite with @xmath298 . + the @xmath294 and @xmath283 solutions are new . for any value of @xmath15 , @xmath115 and @xmath119 have powers of @xmath77 and @xmath200 consistent with a parameter counting branes . this suggests that a string theory underlying such solutions ( if it exists ) should have branes in _ both _ asymptotic regimes . + + + interestingly , algebraic @xmath294 solutions do not exist for @xmath299 . nevertheless , we have demonstrated the existence of numerical solutions to the full equations in this and other cases . see figure ( [ cwithg0full ] ) . these solutions decreases . there are a number of qualitative features of this family of solutions that are akin to those seen in studies of the type 0a case in refs . @xcite , when examining the case of @xmath300 . ] were obtained using the @xmath301 algorithm in @xmath302 . one lesson learned is that the failure to find algebraic solutions in a t hooft limit is not a guarantee that smooth solutions to the full equations do not exist . + + 3 . @xmath264 + figure ( [ c2g2 ] ) shows algebraic solutions where @xmath267 and @xmath249 . these are @xmath303 and @xmath304 solutions in our notation . + the constraint @xmath305 on the algebraic equations lifts to @xmath306 on the full solutions . the constant @xmath216 can not be determined from algebraic solutions alone . in this case , however , a thorough numerical analysis of the full equations is possible and it shows that @xmath307 . examples of these numerical solutions are displayed in figure ( [ cplusg ] ) . + + + + the numerical solution strongly suggests that such new solutions are not just an artifact of the t hooft limit and should be taken seriously as new examples of non perturbatively complete solutions . such solutions were not apparent in the perturbative analysis of ref . @xcite and lead us to believe that the modified t hooft limit presented here is a good way of unearthing new non perturbative solutions . we have presented a modification of the t hooft limit , first used by the authors of ref . @xcite , to argue for the existence of new non perturbative solutions to string equations that are difficult to obtain numerically . we have analyzed the painlev iv hierarchy of string equations introduced in ref . @xcite in this limit , showing that examples of the conjectured type ii string theories coupled to the @xmath0 superconformal minimal models have well defined non perturbative solutions . as in the case of type 0a , higher @xmath26 solutions are likely to exist as a consequence of these@xcite , since the underlying integrable flow structure should evolve lower @xmath26 solutions into higher @xmath26 ones . although this limit results only in the highest powers of the brane / flux parameter surviving ( so that we have only spherical topologies with boundaries or fluxes ) , it is likely that smooth solutions exist for the full string equations too . the higher genus terms in the free energy are obtained by including the derivative corrections to the string equations and from the lesson of type 0a ( where we can compare to numerical solutions of the full equations see section [ sec : modlim ] ) it seems that they do not introduce any singular behaviour . this can presumably be checked through further numerical work for the 0b and conjectured type ii theories . we have uncovered a number of clear examples of new non perturbative solutions which also seem to be string like . it would be interesting to determine if these indeed correspond to new string theories . this could presumably be checked using perturbative techniques of the sort that we used in ref . @xcite to identify the type ii theories . we have demonstrated , in at least the @xmath287 case , that such new solutions are not simply an artifact of the t hooft limit , and that smooth solutions to the full equations can be obtained numerically obeying the same parameter constraints . it would be interesting to find new solutions of this type numerically for higher @xmath55 . this work was supported by the us department of energy . we list the asymptotic expansions for @xmath287 , @xmath70 and @xmath66 here . for each @xmath55 , there are multiple expansions within each class that can be related to each other by various @xmath67 symmetries . we list these explicitly for @xmath287 . the symmetries for @xmath70 and @xmath66 are similar to those for @xmath287 and can be found in ref . @xcite . the other expansions within each class can be obtained by the following symmetry operation : @xmath311 since @xmath115 and @xmath119 contain only even powers of @xmath15 , the are invariant under this map ; however , @xmath312 . thus there are two expansions in the @xmath116 class , and , together with @xmath115 and @xmath119 , these comprise four total @xmath287 expansions . in this case , there are four relevant classes of expansions @xmath313 here , we see that the string coupling is @xmath314 . there are @xmath315 distinct expansions related to each other by various discrete symmetries as for @xmath287 . the following five classes of expansions are obtained : @xmath316 here , the string coupling is @xmath317 . there are a total of _ twenty five _ expansions related by various @xmath67 symmetries . these have been explicitly listed in @xcite . recall the interpretation given to the parameter @xmath15 of the 0a theory ( and also to @xmath52 of the 0b theory ) . in the @xmath318 regime , @xmath15 represents the number of background zz d branes in the model , with a factor of @xmath15 for each boundary in the worldsheet expansion . since an orientable surface with odd ( even ) euler characteristic must contain an odd ( even ) number of boundaries , @xmath15 must be raised to an odd ( even ) power if @xmath200 is . in addition , the power of @xmath15 must be less than or equal to the power of @xmath200 . in the @xmath45 regime , @xmath15 represents the number of units of r r flux in the background , with @xmath319 appearing when there is an insertion of pure r so in this case both @xmath15 and @xmath200 should appear with even powers . in applying these observations to the dww expansions above , we immediately notice the remarkable fact that the various expansions have powers of the parameters which somehow allow for interpretations as counting branes or fluxes . this is by no means guaranteed , and indeed its occurrence was one of our main motivations for in depth study of the system . the presence of two parameters , however , leads to a few subtleties . for example , in some expansions an interpretation in terms of branes is only possible if one of the two parameters is set to zero . we summarize these below . * @xmath120 contains powers of @xmath15 consistent with a parameter counting branes . this remains true for any value of @xmath77 . * @xmath120 contains powers of @xmath77 consistent with a parameter counting fluxes . however , for arbitrary @xmath15 , @xmath200 appears with odd powers , inconsistent with our requirements for a description of fluxes , as mentioned above . this problem is avoided if we set @xmath244 since this forces the odd powers of @xmath200 to vanish . * for even @xmath55 , @xmath115 contains powers of @xmath15 and @xmath200 consistent with a parameter counting fluxes . this is true for all values of @xmath77 . for odd @xmath55 , the powers of @xmath15 are still consistent with the flux interpretation , but there are odd powers of @xmath200 which are inconsistent with fluxes . these odd powers can be removed by setting @xmath259 . * for even @xmath55 , @xmath115 contains powers of @xmath77 and @xmath200 consistent with a parameter counting fluxes . this is true for all values of @xmath15 . for odd @xmath55 , the powers of @xmath77 are consistent with a parameter counting branes . in this interpretation , there are no contributions from surfaces with only one boundary . * @xmath116 contains powers of @xmath15 consistent with a parameter counting branes . this is true only for @xmath259 . * @xmath116 contains powers of @xmath77 consistent with a parameter counting branes . this is true only for @xmath244 . * we notice that associating one boundary to each factor of @xmath77 _ and _ @xmath15 also produces a consistent worldsheet expansion . this encourages us to speculate these expansions might capture @xmath77 and @xmath15 simultaneously counting branes . * @xmath119 contains powers of @xmath77 consistent with a parameter counting branes . this remains true for any value of @xmath15 . * @xmath119 contains powers of @xmath15 consistent with a parameter counting fluxes . however , for arbitrary @xmath77 , @xmath200 appears with odd powers , inconsistent with our requirements for a description of fluxes , as mentioned above . this problem is avoided if we set @xmath259 since this forces the odd powers of @xmath200 to vanish . * @xmath320 contains powers of @xmath15 consistent with a parameter counting branes . this is true only for @xmath259 . * @xmath320 contains powers of @xmath77 consistent with a parameter counting branes . this is true only for @xmath244 . * as for @xmath116 , we notice that associating one boundary to each factor of @xmath77 _ and _ @xmath15 also produces a consistent worldsheet expansion . this encourages us to speculate these expansions might capture @xmath77 and @xmath15 simultaneously counting branes . * these expansions do not exist for @xmath321 . c. m. hull and p. k. townsend , `` unity of superstring dualities , '' http://dx.doi.org/10.1016/0550-3213(94)00559-w[_nucl . phys . _ * b438 * ( 1995 ) 109137 ] , http://arxiv.org/abs/hep-th/9410167[arxiv:hep-th/9410167 ] . p. k. townsend , `` the eleven - dimensional supermembrane revisited , '' http://dx.doi.org/10.1016/0370-2693(95)00397-4[_phys . lett . _ * b350 * ( 1995 ) 184187 ] , http://arxiv.org/abs/hep-th/9501068[arxiv:hep-th/9501068 ] . e. witten , `` string theory dynamics in various dimensions , '' http://dx.doi.org/10.1016/0550-3213(95)00158-o[_nucl . phys . _ * b443 * ( 1995 ) 85126 ] , http://arxiv.org/abs/hep-th/9503124[arxiv:hep-th/9503124 ] . t. r. morris , `` 2-d quantum gravity , multicritical matter and complex matrices , '' . fermilab - pub-90 - 136-t . s. dalley , c. v. johnson , and t. r. morris , `` multicritical complex matrix models and nonperturbative 2- d quantum gravity , '' http://dx.doi.org/10.1016/0550-3213(92)90217-y[_nucl . phys . _ * b368 * ( 1992 ) 625654 ] . s. dalley , c. v. johnson , and t. r. morris , `` nonperturbative two - dimensional quantum gravity , '' http://dx.doi.org/10.1016/0550-3213(92)90218-z[_nucl . _ * b368 * ( 1992 ) 655670 ] . c. v. johnson , t. r. morris , and a. watterstam , `` global kdv flows and stable 2-d quantum gravity , '' http://dx.doi.org/10.1016/0370-2693(92)90112-h[_phys . lett . _ * b291 * ( 1992 ) 1118 ] , http://arxiv.org/abs/hep-th/9205056[arxiv:hep-th/9205056 ] . s. dalley , c. v. johnson , t. r. morris , and a. watterstam , `` unitary matrix models and 2-d quantum gravity , '' http://dx.doi.org/10.1142/s0217732392002226[_mod lett . _ * a7 * ( 1992 ) 27532762 ] , http://arxiv.org/abs/hep-th/9206060[arxiv:hep-th/9206060 ] . c. crnkovic , m. r. douglas , and g. w. moore , `` physical solutions for unitary matrix models , '' http://dx.doi.org/10.1016/0550-3213(91)90413-r[_nucl . * b360 * ( 1991 ) 507523 ] . t. j. hollowood , l. miramontes , a. pasquinucci , and c. nappi , `` hermitian vs. anti - hermitian 1-matrix models and their hierarchies , '' http://dx.doi.org/10.1016/0550-3213(92)90457-m[_nucl . phys . _ * b373 * ( 1992 ) 247280 ] , http://arxiv.org/abs/hep-th/9109046[arxiv:hep-th/9109046 ] . i. r. klebanov , j. m. maldacena , and n. seiberg , `` unitary and complex matrix models as 1-d type 0 strings , '' http://dx.doi.org/10.1007/s00220-004-1183-7[_commun . * 252 * ( 2004 ) 275323 ] , http://arxiv.org/abs/hep-th/0309168[arxiv:hep-th/0309168 ] . d. j. gross and a. a. migdal , `` nonperturbative two - dimensional quantum gravity , '' http://dx.doi.org/10.1103/physrevlett.64.127[_phys . * 64 * ( 1990 ) 127 ] . e. brezin and v. a. kazakov , `` exactly solvable field theories of closed strings , '' http://dx.doi.org/10.1016/0370-2693(90)90818-q[_phys . * b236 * ( 1990 ) 144150 ] . m. r. douglas and s. h. shenker , `` strings in less than one - dimension , '' http://dx.doi.org/10.1016/0550-3213(90)90522-f[_nucl . phys . _ * b335 * ( 1990 ) 635 ] . j. e. carlisle , c. v. johnson , and j. s. pennington , `` baecklund transformations , d - branes , and fluxes in minimal type 0 strings , '' http://dx.doi.org/10.1088/1751-8113/40/41/013[_j . phys . _ * a40 * ( 2007 ) 1245112462 ] , http://arxiv.org/abs/hep-th/0501006[arxiv:hep-th/0501006 ] . j. e. carlisle , c. v. johnson , and j. s. pennington , `` d - branes and fluxes in supersymmetric quantum mechanics , '' http://arxiv.org/abs/hep-th/0511002[arxiv:hep-th/0511002 ] . j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' http://dx.doi.org/10.1023/a:1026654312961[_adv . * 2 * ( 1998 ) 231252 ] , http://arxiv.org/abs/hep-th/9711200[arxiv:hep-th/9711200 ] . s. s. gubser , i. r. klebanov , and a. m. polyakov , `` gauge theory correlators from non - critical string theory , '' http://dx.doi.org/10.1016/s0370-2693(98)00377-3[_phys . * b428 * ( 1998 ) 105114 ] , http://arxiv.org/abs/hep-th/9802109[arxiv:hep-th/9802109 ] . e. witten , `` anti - de sitter space and holography , '' _ adv . * 2 * ( 1998 ) 253291 , http://arxiv.org/abs/hep-th/9802150[arxiv:hep-th/9802150 ] . r. iyer , c. v. johnson , and j. s. pennington , `` string theory and water waves , '' http://arxiv.org/abs/1002.1120[arxiv:1002.1120 [ hep - th ] ] . d. kaup , `` finding eigenvalue problems for solving nonlinear evolution equations , '' http://dx.doi.org/10.1143/ptp.54.72[_prog . * 54 * ( 1975 ) 7278 ] . r. gordoa , n. joshi , and a. pickering , `` on a generalized @xmath322 dispersive water wave hierarchy , '' http://dx.doi.org/10.2977/prims/1145477227[_publ . inst . math . * 37 * ( 2001 ) no . 3 , 327347 ] . i. m. gelfand and l. a. dikii , `` asymptotic behavior of the resolvent of sturm - liouville equations and the algebra of the korteweg - de vries equations , '' _ russ . math . surveys _ * 30 * ( 1975 ) 77113 . m. r. douglas , `` strings in less than one dimension and the generalized kdv hierarchies , '' http://dx.doi.org/10.1016/0370-2693(90)91716-o[_phys . lett . _ * b238 * ( 1990 ) 176 ] . t. banks , m. r. douglas , n. seiberg , and s. h. shenker , `` microscopic and macroscopic loops in non - perturbative two- dimensional gravity , '' http://dx.doi.org/10.1016/0370-2693(90)91736-u[_phys . _ * b238 * ( 1990 ) 279 ] . a. b. zamolodchikov and a. b. zamolodchikov , `` liouville field theory on a pseudosphere , '' http://arxiv.org/abs/hep-th/0101152[arxiv:hep-th/0101152 ] . v. e. zakharov and a. b. shabat , `` integration of nonlinear equations of mathematical physics by the method of inverse scattering . ii , '' _ funct . * 13 * ( 1979 ) 166174 . t. r. morris , `` multicritical matter from complex matrices , '' _ class . grav . _ * 9 * ( 1992 ) 18731881 . c. v. johnson , `` string theory without branes , '' http://arxiv.org/abs/hep-th/0610223[arxiv:hep-th/0610223 ] |
an articulation point in a network is a node whose removal disconnects the network . those nodes play key roles in ensuring connectivity of many real - world networks , from infrastructure networks to protein interaction networks and terrorist communication networks . despite their fundamental importance , a general framework of studying articulation points in complex networks is lacking . here we develop analytical tools to study key issues pertinent to articulation points , e.g. the expected number of them and the network vulnerability against their removal , in an arbitrary complex network . we find that a greedy articulation point removal process provides us a novel perspective on the organizational principles of complex networks . moreover , this process is associated with two fundamentally different types of percolation transitions with a rich phase diagram . our results shed light on the design of more resilient infrastructure networks and the effective destruction of terrorist communication networks . | 1609.00094 | a fundamental challenge in studying complex networked systems is to reveal the interplay between network topology and functionality@xcite . here we tackle this challenge by investigating a classical notion in graph theory , i.e. articulation points . a node in a network is an articulation point ( ap ) if its removal disconnects the network or increases the number of connected components in the network@xcite . those ap nodes play important roles in ensuring the connectivity of many real - world networks . for example , in infrastructure networks such as air traffic networks or power grids , aps , if disrupted or attacked , pose serious risks to the infrastructure @xcite . in wireless sensor networks , failures of aps will block data transmission from one network component to others @xcite . in the yeast protein - protein interaction network , lethal mutations are enriched in the group of highly connected proteins that are aps @xcite . identification of aps also helps us better solve other challenging problems , e.g. the calculation of determinants of large matrices @xcite ; and the vertex cover problem on large graphs , which is a notorious np - complete problem @xcite . representing natural vulnerabilities of a network , aps are potential targets of attack if one aims for immediate damage to a network . for example , in the terrorist communication network of the 9/11 attacks on u.s . ( fig.1a ) , cutting off those ap members ( shown in red ) could have caused a great deal of disruption in the terrorists ability to communicate @xcite . note that the removal of an ap in a network may lead to the emergence of new aps in the remainder of the network . a brute - force strategy of network destruction is to _ iteratively _ remove the most destructive ap that will cause the most nodes disconnected from the giant connected component ( gcc ) of the current network . given a limited `` budget '' ( i.e. the number of nodes to be removed ) , this ap - targeted attack ( apta ) strategy is very efficient in reducing the gcc , compared with other existing strategies @xcite . indeed , we find that for a small fraction of removed nodes apta leads to the quickest reduction of gcc for a wide range of real - world networks from technological to infrastructure , biological , communication , and social networks ( see appendix [ sec.si-i-a ] ) . depending on the initial network structure , apta would either completely decompose the network or result in a residual gcc that occupies a finite fraction of the network . this residual gcc is a biconnected component ( or bicomponent ) , in which any two nodes are connected by at least two independent paths and hence no ap exists @xcite . for accuracy , we will call it residual giant bicomponent ( rgb ) hereafter . this rgb naturally represents a _ core _ that maintains the _ structural integrity _ of the network . identification and removal of aps also provide us a novel perspective on the organizational principles of complex networks . for example , in a terrorist communication network , each ap member can be considered as a _ messenger _ of a particular subnetwork , because any information exchange between that subnetwork and the rest of the network passes through the ap . all the aps and their associated subnetworks in the original network constitute the _ first layer _ of the terrorist network . after removing all the aps in the original network , the first layer is peeled off , new aps emerge and the _ second layer _ of the network is exposed . we can repeat this greedy aps removal ( gapr ) process until there is no ap left in the network . figure 1 illustrates this novel network decomposition process in the 9/11 terrorist communication network , which has 62 terrorists . we find that this network consists of 3 layers and an rgb of 26 nodes . ( note that the rgb associated with the gapr process is similar to but not necessarily the same as that of the apta process , see appendix [ sec.si-i-b ] ) interestingly , among those 26 rgb nodes , 16 of them are hijackers in the 9/11 terrorist attack , which in total has 19 hijackers . in a sense , this rgb serves as a _ core _ maintaining the _ functionality _ of this covert network , which has a particular goal hijacking . note that some of the hijackers in the rgb are not hubs ( i.e. highly connected nodes ) , but only have two or three neighbors in the network . hence they can not be easily identified through traditional network decompositions , e.g. the classical @xmath0-core decomposition @xcite . despite the importance of aps in ensuring the connectivity of many real - world networks , we still lack a deep understanding on the roles and properties of aps in many complex networks . we do nt have a theory to quantify if a given network has overrepresented or underrepresented aps comparing to its randomized counterparts . moreover , the implication of the rgb serving as a core that maintains the structural integrity and/or functionality of a network has never been explored before . in this article we offer an analytical framework to study those fundamental issues pertinent to aps in complex networks . we first study the fraction of aps ( @xmath1 ) and the relative size of the rgb ( @xmath2 ) in a wide range of real - world networks . the aps of a network can be identified all together by a linear - time algorithm based on depth - first search @xcite . as for rgb , we find that the two rgbs associated with the apta and gapr processes significantly overlap for many real - world networks ( appendix [ sec.si-i-b ] ) . hereafter we focus on the rgb obtained from the gapr process , which is deterministic and hence analytically solvable . we find that many real networks have a non - ignorable fraction of aps and a rather small rgb ( fig.2a ) . one may expect that infrastructure networks should have a relatively small fraction of aps and a large rgb , and hence are very robust against ap removal . ironically , this is not the case . the power grids in two regions of u.s . have almost the largest fraction ( @xmath3 ) of aps among all the real networks analyzed in this work . and they have almost no rgb . the road networks of three states in u.s . have almost 20% of aps , and a small rgb ( @xmath4 ) . in contrast , among all the 28 food webs we analyzed , 22 of them have no aps ( and hence @xmath5 ) , indicating that those networks are biconnected and the extinction of one species will not disconnect the whole network . more interestingly , we find that most of the real networks analyzed here have either a very small rgb or a rather big one ( fig . 2a light red and green regions ) . later we will show that this phenomenon is related to a novel discontinuous phase transition associated with the gapr process . to identify the topological characteristics that determine these two quantities ( @xmath6 and @xmath7 ) , we compare @xmath6 ( or @xmath7 ) of a given real network with that of its randomized counterpart . to this aim , we randomize each real network using a complete randomization procedure that turns the network into an erds - rnyi ( er ) type of random network with the number of nodes @xmath8 and links @xmath9 unchanged @xcite . we find that most of the completely randomized networks possess very different @xmath6 ( or @xmath7 ) , comparing to their corresponding real networks ( fig . this indicates that complete randomization eliminates the topological characteristics that determine @xmath1 and @xmath2 . by contrast , when we apply a degree - preserving randomization , which rewires the links among nodes while keeping the degree @xmath0 of each node unchanged , this procedure does not alter @xmath1 and @xmath2 significantly ( fig . 2d , e ) . in other words , the characteristics of a network in terms of @xmath1 and @xmath2 is largely encoded in its degree distribution @xmath10 . most of the real - world networks display slightly smaller @xmath6 and bigger @xmath7 than their degree - preserving randomized counterparts . we attribute these differences to higher - order structure correlations , such as clustering @xcite and degree assortativity @xcite , which are eliminated in the degree - preserving randomization . the results of real - world networks prompt us to analytically calculate @xmath1 and @xmath2 for networks with any prescribed degree distribution . this is achieved by analyzing the discrete - time dynamics of the deterministic gapr process , which generates a series of snapshots for the remainder network with a clear temporal order @xmath11 . here , @xmath12 is the total number of gapr steps , which is also the number of layers peeled off during the gapr process . we denote the fraction of aps and the relative size of the gcc in the original network as @xmath13 and @xmath14 , respectively . removal of the original aps leads to a new fraction of aps @xmath15 and a smaller gcc of relative size @xmath16 . we repeat this process and denote the fraction of aps and the relative size of the gcc of the network snapshot at time step @xmath17 as @xmath18 and @xmath19 , respectively . at the end of the gapr process , we have @xmath20 and @xmath21 . based on the configuration model of uncorrelated random networks @xcite , we can analytically calculate @xmath18 and @xmath19 for networks with arbitrary degree distributions at any time step @xmath17 . this enables us to further compute @xmath12 and @xmath7 ( see appendix [ sec.si-ii ] for details ) . we first investigate the fraction of aps in the original network , i.e. @xmath22 . we calculate @xmath6 in two canonical model networks : ( 1 ) er random networks with poisson degree distributions @xmath23 , where @xmath24 is the mean degree ; and ( 2 ) scale - free ( sf ) networks with power - law degree distributions @xmath25 , where @xmath26 is often called the degree exponent ( fig . the fraction of aps is trivially zero in the two limits @xmath27 and @xmath28 , and reaches its maximum at a particular mean degree @xmath29 . for er networks , we find that @xmath30 , which is larger than @xmath31 , the critical point of ordinary percolation where the gcc emerges . the phenomenon that @xmath6 displays a unimodal behavior and the fact that @xmath32 can be explained as follows . the process of increasing the mean degree @xmath24 of an er network can be considered as the process of randomly adding links into the network . when the mean degree @xmath24 is very small ( nearly zero ) , there are only isolated nodes and dimers ( i.e. components consisting of two nodes connected by one link ) , and thus @xmath33 . with @xmath24 gradually increasing but still smaller than @xmath34 , the network is full of finite connected components ( fccs ) , most of which are trees ( fig . 3c ) . hence , in the range of @xmath35 most of the nodes ( except isolated nodes and leaf nodes ) are aps , and adding more links to the network will decrease the number of isolated nodes and dimers and hence increase the number of aps . when @xmath36 , the gcc develops and occupies a finite fraction of nodes in the network ( highlighted in light blue in fig . 3d - f ) . in this case , we can classify the links to be added to the network into two types : ( i ) links inside the gcc ( yellow dashed lines ) ; and ( ii ) links that connect the gcc with an fcc or connect two fccs ( green dashed lines ) . the probability that an added link is type - i ( or type - ii ) as a function of the mean degree @xmath24 is shown in fig . 3a ( light blue region ) . adding type - i links to the network will never induce new aps , and may even convert the existing aps ( see fig.3d , e , nodes in black boxes ) back to normal nodes . by contrast , adding type - ii links will never decrease the number of aps and could convert normal nodes ( see fig.3c - e , nodes in orange boxes ) to aps . the contributions of these two types of links to @xmath1 compete with each other . at the initial stage of this range of @xmath24 ( @xmath37 ) , since the gcc is still small , most of the added links are type - ii ( fig . 3a , green dashed line ) , and thus @xmath1 continues to increase ( fig . 3a , red line ) . at certain point @xmath29 , where the peak of @xmath1 locates , the contribution of type - i links to @xmath6 overwhelms that of type - ii links , hence @xmath1 begins to decrease . when the mean degree @xmath24 is large enough , the network itself becomes a bicomponent without any ap . the phenomena that @xmath38 is even more prominent for sf networks , where @xmath39 and @xmath40 . this is because , for sf networks , even though the gcc emerges at lower @xmath34 , its relative size is rather small at the initial stage of its emergence , which results in larger @xmath41 ( fig . 3b ) . we now systematically study the behaviors of @xmath19 and @xmath2 as functions of the mean degree @xmath24 for infinitely large er networks . to emphasize the @xmath24-dependence , hereafter we denote @xmath19 and @xmath2 of er networks as @xmath42 and @xmath43 , respectively . as shown in fig . 4a ( thin grey lines ) , after any _ finite _ steps of gapr , the gcc always emerges in a continuous manner , suggesting a _ continuous _ phase transition . hereafter we will call it gcc percolation transition . for @xmath17 steps of gapr , the gcc percolation transition displays a critical phenomenon : @xmath44 for @xmath45 , where @xmath46 is the critical mean degree , i.e. the percolation threshold , and @xmath47 is the critical exponent . we find that as @xmath17 increases , @xmath46 becomes larger and larger , but eventually converges to @xmath48 . note that , for any finite @xmath17 , the critical exponent @xmath47 associated with the gcc percolation transition is the same : @xmath49 ( see fig . 4c , thin grey lines ) . by contrast , if we allow for _ infinite _ steps of gapr ( i.e. we stop the process only if there is no ap left ) , the size of the resulting gcc ( which is nothing but the rgb ) , denoted as @xmath50 , displays a remarkable _ discontinuous _ phase transition : @xmath50 abruptly jumps from zero ( when @xmath51 ) to a finite value at @xmath52 , and then increases with increasing @xmath24 ( see fig.4a thick black line ) . hereafter we will call it rgb percolation transition . if we denote the jump size as @xmath53 , we find that @xmath54 with critical exponent @xmath55 when @xmath56 ( fig . 4c , thick black line ) , suggesting that the rgb percolation transition is actually a _ hybrid _ phase transition @xcite . in other words , @xmath50 has a jump at the critical point @xmath57 as a first - order phase transition but also has a critical singularity as a second - order phase transition . interestingly , the gcc and rgb percolation transitions have completely different critical exponents associated with their critical singularities ( fig . we also calculate the total number of gapr steps @xmath58 needed to remove all aps of an infinitely large er network of mean degree @xmath24 ( fig . we find that @xmath58 is finite for @xmath51 ; diverges when @xmath59 ; and is infinite for any @xmath60 . the divergence of @xmath58 displays a scaling behavior @xmath61 with critical exponent @xmath62 when @xmath59 ( fig . 4d , red line ) . the nature of the discontinuous rgb percolation transition and the behavior of @xmath58 can be revealed by analyzing the dynamics of the gapr process . in particular , we can calculate @xmath63 , @xmath64 , as well as a key quantity in the gapr process , i.e. the average number of newly induced aps per single ap removal : @xmath65 for @xmath66 and at different mean degrees @xmath24 ( fig . 4e - g ) . for @xmath67 ( supercritical region ) , the fraction of aps exponentially decays as @xmath68 after an initial transient time ( fig . 4f , green lines ) , where @xmath69 is the characteristic time scale . in this region , with increasing @xmath17 , @xmath70 quickly reaches an equilibrium value @xmath71 , which is smaller than 1 ( fig . 4 g , green lines ) . consequently , @xmath63 converges to a finite value for @xmath72 ( fig . 4e green lines ) , resulting in a finite @xmath2 . since @xmath58 is infinite in this region , we can use @xmath69 to characterize the relaxation behavior of gapr process . we find that @xmath69 increases as @xmath24 decreases ( fig . 4b , green line ) , and diverges as @xmath73 when @xmath74 ( fig.4d , green line ) . note that as @xmath24 decreases and approaches @xmath57 from above , the equilibrium value @xmath75 gradually approaches 1 ( fig.4 g ) . when @xmath76 ( i.e. right above the criticality ) , the fraction of aps decays in a power - law manner for large @xmath17 , i.e. @xmath77 with @xmath78 ( fig . 4f and inset , black line ) , rendering @xmath79 ( fig.4 g , black thick line ) . consequently , @xmath63 converges to a finite value in the @xmath80 limit ( fig . 4e , black thick lines ) , leading to a finite @xmath2 . the fact @xmath79 suggests that in average every removed ap will induce one new ap at the next time step , and hence the gapr process will continue forever . this explains why @xmath81 diverges . note that the equilibrium value @xmath79 can only be reached when @xmath64 displays a power - law decay as @xmath80 . this is because , as long as @xmath64 is finite at any finite @xmath17 , the gapr will gradually dilute the network , rendering a larger and larger value of @xmath70 as @xmath17 grows . when @xmath82 ( i.e. right below the criticality ) , as well as in the entire subcritical region ( @xmath51 ) , after an initial decay , @xmath64 begins to exponentially grow with increasing @xmath17 ( fig . 4f , red lines ) . consequently , @xmath70 is initially less than 1 , but then becomes drastically larger than 1 ( fig . g , red lines ) , which causes @xmath63 quickly decays to zero ( fig . 4e , red lines ) , and hence @xmath58 is finite and the rgb dose not exist . the sudden collapse of the rgb upon an infinitesimal decrease in @xmath24 suggests the discontinuous nature of the rgb percolation transition in er networks . note that at time step @xmath12 , the network will break into pieces and there is no ap left . hence in the last few gapr steps the growth of @xmath64 will slow down and eventually decrease ( fig . 4f , g , tails in the red lines ) . for finite - size networks sampled from a network ensemble with a prescribed degree distribution , the value of @xmath2 at criticality @xmath52 is subject to large sample - to - sample fluctuations , being either zero or a large finite value ( fig . 5a and inset ) , which is another evidence of discontinuous phase transition @xcite . this discontinuous phase transition also partially explains the fact that real - world networks have either a very small or a rather big rgb ( fig . 2a ) . the nature of the rgb percolation transition in sf networks is qualitatively the same as that in er networks . the transition from the non - rgb phase to the rgb phase is discontinuous ( fig . the critical point @xmath83 increases with decreasing @xmath26 . also , the jump of the rgb relative size at criticality increases as @xmath26 approaches 2 ( fig . the @xmath84 phase diagram of sf networks is shown in fig . the whole diagram consists of three phases . for @xmath85 ( grey region ) , there exists no gcc in the network , and hence no rgb . for @xmath86 ( light blue region ) , even though the gcc may survive after certain finite steps of gapr , the rgb still does not exist . since in both regimes there is no rgb , we call them non - rgb - i phase and non - rgb - ii phase , respectively . the transition between these two phases is the ordinary gcc percolation transition , which is continuous ( thick dashed line ) . note that , in non - rgb - ii phase , the phase transition associated with the emergence the gcc after any finite @xmath17 steps of gapr is still continuous ( thin dot - dashed lines ) . for @xmath87 ( light yellow region ) , the network suddenly has an rgb . this regime is referred to as the rgb phase . as we mentioned above , the transition between the non - rgb - ii phase and the rgb phase is discontinuous ( thick solid line ) . we have performed extensive numerical simulations to confirm our analytical results ( see appendix [ sec.si-iii ] for details ) . network science has been tremendously developed in the last decade and significantly advanced our understanding of many complex systems @xcite . in particular , structural transitions in complex networks have been extensively studied and found to affect many network properties @xcite . however , some basic graph theoretical concepts are still not fully understood in the context of complex networks , hindering the further development of network science . articulation point is one of them . in this article , we propose an analytical framework to systematically investigate ap - related issues in complex networks . many interesting phenomena of aps are discovered and explained for the first time . in particular , the novel network decomposition method based on the gapr process and the concept of the rgb offer us a brand - new perspective on the organizational principles of complex networks . the presented results shed light on the design of more resilient infrastructure networks and more effective destructions of malicious networks such as the criminal or terrorist communications networks @xcite . the framework presented here also raises a number of questions , answers to which could further deepen our understanding of complex networks . for example , although our analytical work focuses on uncorrelated networks , all the quantities considered here can be efficiently calculated for large real - world networks , providing a platform to systematically address the role of higher - order structure correlations on the fraction of ap nodes and the relative size of the rgb @xcite . moreover , the functionalities of the rgb nodes in different real - world networks can be further studied and related to the nature of those networks . taken together , our results indicate that many aspects of articulation points can be systematically investigated , opening new avenues to deepening our understanding of complex networked systems . ) . ] ) versus relative size of the rgb ( @xmath88 ) in a wide range of real networks . ( * b*-*c * ) @xmath89 and @xmath90 calculated from the fully randomized counterparts of the real networks , compared with the values ( @xmath91 and @xmath88 ) calculated from the real networks . ( * d*-*e * ) @xmath92 and @xmath93 calculated from the degree - preserving randomized counterparts of the real networks , compared with the values calculated from the real networks . all data points and error bars ( s.e.m . ) in * b*-*e * are determined from 100 realizations of the randomized networks . the dashed lines ( @xmath94 ) are guide for eyes . for detailed description of these real networks and references , see appendix [ sec.si-iv ] . ] * fractions of aps as functions of mean degree in two canonical model networks . * ( * a * ) erds - rnyi random networks @xcite ; ( * b * ) scale - free networks with different degree exponents @xmath26 . in ( * a * ) , the fraction of aps is shown in red line . the probabilities of adding type - i ( yellow dashed line ) and type - ii links ( green dashed line ) are also shown . in ( * b * ) , we use the static model @xcite to construct sf networks with asymptotically power - law degree distribution @xmath25 . simulation results ( symbols ) are generated with network size @xmath95 . error bars ( s.e.m . over 128 realizations ) are smaller than the symbols . lines are our theoretical predictions . ( * c - f * ) illustrations of aps ( red nodes ) , type - i links ( yellow dashed lines ) , and type - ii links ( green dashed lines ) in er random networks at different mean degrees . note that adding a single type - ii link at most convert two normal nodes to aps ( orange boxes ) , while adding a single type - i link could convert much more aps back to normal nodes ( black boxes ) . this explains why the peak of @xmath1 emerges even though the probability of adding type - ii links is still larger than that of adding type - i links . the largest connected component is highlighted in light blue in * d*-*f*. steps of gapr , @xmath42 , as functions of the mean degree @xmath24 . note that @xmath96 corresponds to the ordinary percolation ( thin purple line ) ; @xmath42 with finite @xmath17 ( only @xmath97 are shown here ) corresponds to the gcc percolation ( thin grey lines ) ; and @xmath98 corresponds to the rgb percolation ( thick black line ) . ( * b * ) total number of the gapr steps @xmath58 for @xmath51 ( red line ) and the characteristic number of the gapr steps @xmath69 for @xmath60 ( green line ) as functions of the mean degree @xmath24 . ( * c * ) the critical scaling behavior of @xmath42 and @xmath43 for the gcc and rgb percolation transitions , respectively . ( * d * ) the divergence of @xmath58 and @xmath69 associated with the rgb percolation transition . temporal behavior of @xmath42 ( * e * ) , @xmath99 ( * f * ) , and @xmath70 ( * g * ) at critical ( thick black lines ) , subcritical ( red lines , @xmath100 , respectively ) , and supercritical ( green lines , @xmath101 , respectively ) regions of the rgb percolation transition . at criticality , @xmath102 decays in a power - law manner for large @xmath17 ( inset of * f * ) . ] in the erds - rnyi network ( red ) , and scale - free networks with different degree exponents , @xmath1034.0 ( green ) , 3.0 ( blue ) , and 2.5 ( yellow ) , constructed from the static model @xcite . lines are our theoretical predictions . simulation results ( symbols ) are calculated from networks of size @xmath95 . error bars ( s.e.m . over 128 realizations ) are generally smaller than the symbols , except at criticality . the deviation of simulation results from our theoretical prediction for @xmath104 owes to degree correlations present in the constructed networks , which become prominent as @xmath105 @xcite . inset displays the distribution of @xmath2 generated from 51,200 er networks of size @xmath95 at criticality . the bimodal distribution of @xmath2 indicates that it undergoes a discontinuous jump from nearly zero to a large finite value at the critical point . ( * b * ) phase diagram associated with the gapr process in sf networks . the rgb percolation transition , the gcc percolation transitions ( only @xmath97 are presented here ) , and the ordinary percolation transition are shown in thick solid line , thin dot - dashed lines , and thick dashed line , respectively . in the limit of large @xmath26 , the phase boundaries for er networks are recovered ( highlighted by arrows : @xmath106 ( thick solid arrow ) ; @xmath107 and @xmath108 ( thin dot - dashed arrows ) ; @xmath109 ( thick dashed arrow ) ) . ] a common measure of the structural integrity of a network is the size of its giant connected component ( gcc ) . articulation points ( aps ) are natural targets if we aim for immediate damage to the gcc . to quickly destruct the gcc , we can target the most destructive ap whose removal will cause the most nodes disconnected from the gcc . this leads to a brute - force strategy of network destruction : step-1 : : identify all the aps in the current network , which can be achieved by a linear - time algorithm based on depth - first search @xcite . step-2 : : for each ap , calculate its destructivity " , i.e. how many nodes will be disconnected from the gcc after its removal . step-3 : : rank all the aps based on their destructivities , and remove the most destructive one together with the links attaching to it . ( in case the most destructive aps are not unique , randomly choose one of them to remove . ) step-4 : : repeat steps 1 - 3 until the network is totally destroyed or left with a residual giant biconnected component ( rgb ) , in which no aps exist . hereafter we call this iterative process ap - targeted attack ( apta ) . to demonstrate the efficiency of apta in reducing the size of the gcc , we compare it with two other network destruction strategies based on the following node centrality measures : * * degree*. in this strategy , we iteratively remove the node with the highest degree in the current network @xcite . ( in case the highest - degree nodes are not unique , we randomly choose one of them to remove . ) degrees of the remaining nodes will be recalculated after each node removal . * * collective influence*. the collective influence ( ci ) of node @xmath110 is defined as the product of its reduced degree @xmath111 and the total reduced degree of all nodes at distance @xmath112 from it , i.e. @xmath113 , where @xmath114 is the frontier of a ball of radius @xmath112 centered at node @xmath110 ( i.e. the set of the @xmath112-th nearest neighbors of node @xmath110 ) @xcite . in each step , we compute the ci of every node , and remove the node with the largest ci . ( in case the highest - ci nodes are not unique , we randomly choose one of them to remove . ) it has been shown that , in locally tree - like large networks , the ci - based attack gives the optimal threshold , i.e. the minimal fraction of removed nodes that totally destroys the gcc @xcite . note that , for @xmath115 , this method is reduced to the degree - based network destruction . we compute the relative size of the gcc , denoted as @xmath116 , as a function of the fraction of removed nodes @xmath117 by using the three above - mentioned network destruction strategies in two classical model networks ( fig . [ fig.1]a ) and five different types of real - world networks ( including technological , infrastructure , biological , communication , and online social networks ) ( fig . [ fig.1]b - f ) . interestingly , we find that , given a limited budget " ( i.e. a small fraction of nodes to be removed ) , apta is the most efficient strategy in reducing gcc for both model and real networks ( fig . [ fig.1 ] ) . it turns out that even though ap removal maximizes only the _ local _ damage to the network in every step , and does not aim for a systemic or global breakdown of the network , for a certain range of @xmath117 this is highly efficient in damaging the entire system . moreover , in some of the networks , such as the road network ( fig . [ fig.1]b-2 ) , the power grid ( fig . [ fig.1]b-3 ) , and the world wide web ( www ) ( fig . [ fig.1]c-1 ) , apta is optimal even in the whole range of @xmath117 . the reason that for those networks apta leads to smaller critical @xmath117 than the state - of - the - art ci - based method is because these networks are either spatially embedded or / and rich in loops , in which the ci - based method is not guaranteed to be efficient . note that , even for networks with rather big rgbs ( fig . 1a-2 , a-4 , f-3 ) , atpa is still very efficient till the gcc becomes biconnected ( or becomes an rgb ) , where the process of apta terminates . comparison of different network destruction strategies in reducing the size of the giant connected component for both model ( * a * ) and real networks ( * b - f * ) . for model networks , the data points and error bars ( defined as s.e.m . ) are determined from 32 independent network instances of size @xmath118 . the real - world networks used in this figure are : openflights ( * b-1 * ) , roadnet - tx ( * b-2 * ) , pg - weststate ( * b-3 * ) , nd.edu ( * c-1 * ) , oregon2 - 010331 ( * c-2 * ) , p2p - gnutella31 ( * c-3 * ) , _ caenorhabditis elegans _ ( * d-1 * ) , _ homo sapiens _ ( * d-2 * ) , cellphone ( * e-1 * ) , email - enron ( * e-2 * ) , ucionline ( * e-3 * ) , epinions ( * f-1 * ) , slashdot-1 ( * f-2 * ) , twitter ( * f-3 * ) , wikivote ( * f-4 * ) , youtube ( * f-5 * ) ( see section iv for details).,scaledwidth=80.0% ] the rgb can also be obtained by the process of greedy aps removal ( gapr ) . performing gapr on a given network consists of the following steps : step-1 : : identify all the aps in the current network by the depth - first search algorithm @xcite . step-2 : : remove all the aps and the links attaching to them . step-3 : : repeat steps 1 - 2 until no aps exist in the network . for a general network of finite size , gapr and apta do not necessarily yield exactly the same rgb . in fact , the rgb obtained from the former is unique ( because the gapr process is deterministic ) , while the rgb obtained from the latter could be history - dependent ( there may be multiple most destructive aps in each step ) . for a given network , we denote the sets of nodes in the two rgbs associated with the two processes as @xmath119 and @xmath120 , respectively . the overlap between the two rgbs can be measured by the classical jaccard index : @xmath121 in fig . [ fig.2 ] , we show the relative sizes of the two rgbs , as well as the overlap between them . we find that the two rgbs significantly overlap for both model networks and real - world networks . the relative sizes of the two rgbs associated with the gapr ( pink bars ) and the apta ( blue bars ) processes , and the overlap between the two rgbs ( green squares ) in model networks and real - world networks . for each type of model network , the results ( with error bars defined as s.e.m . ) are obtained from 32 independent network instances with @xmath118 . for each real - world network , the results ( with error bars defined as s.e.m . ) are averaged over 32 independent realizations of apta processes . ] in our theoretical framework , we study the gapr process in large uncorrelated random networks with arbitrary degree distribution @xmath10 and finite mean degree @xmath122 . in the thermodynamic limit , these networks are locally tree - like and do not contain loops of finite length . this tree ansatz directly leads to three important network properties @xcite : 1 . all of the finite connected components ( fccs ) are trees . loops ( of infinite length ) only exist in the gcc ; 2 . there can be only one gcc in a network ; 3 . for any node in a network , its neighbors become disconnected or independent from each other if the node and its attaching links are removed . we emphasize that , even though our theoretical framework is based on the tree ansatz , its predictions are fairly precise for finite and loopy networks , provided that the density of loops is not excessively large ( see section iii for numerical simulations ) . the deterministic gapr process is naturally associated with a temporal order . at @xmath123 , we have the original network . at @xmath124 , i.e. the last time step of gapr , no ap exists in the network , and the gapr process terminates . at each time step @xmath17 , we denote the fraction of aps and the relative size of the gcc as @xmath125 and @xmath126 , respectively , where @xmath24 is the mean degree of the original network . the rgb is nothing but the gcc at the last time step , whose relative size is denoted as @xmath127 . at each time step @xmath17 during the gapr process in a network @xmath128 , we classify the remaining nodes into the following three categories or states : 1 . @xmath129-nodes : nodes that belong to fccs ; 2 . @xmath130-nodes : nodes that are aps ; and belong to the gcc ; 3 . @xmath131-nodes : nodes that are not aps ; and belong to the gcc . note that the notations @xmath132 and @xmath133 here have totally different meanings from the critical exponents mentioned in the main text . according to the tree ansatz , the state of a randomly chosen node @xmath110 can be determined by the states of its neighbors in @xmath134 , i.e. the induced subgraph of @xmath128 with node @xmath110 and all its links removed . therefore , at each time step @xmath17 , we need to know the probability that , following a randomly chosen link to one of its end nodes , this node belongs to any of the above categories after this link is removed . these probabilities are denoted as @xmath129 , @xmath130 , and @xmath131 , respectively . to be precise , hereafter when we consider the state of a neighbor of a given node @xmath110 , we mean the state of the neighbor in the induced subgraph @xmath134 . also , when we mention the state of an end node of a chosen link @xmath112 , we mean the state of the node in @xmath135 , i.e. the induced subgraph of @xmath128 with link @xmath112 removed . according to the definitions of @xmath130 and @xmath131 , the probability that an end node of a randomly chosen link belongs to the gcc at time step @xmath17 is just @xmath136 . the gapr process can be fully characterized by the three sets of probabilities @xmath137 , @xmath138 , and @xmath139 . note that every node must belong to one of the three categories , which means the three sets of probabilities are not independent from each other . specifically , at time step @xmath17 , the probability @xmath131 can be derived by the other two sets of probabilities through the following normalization condition : @xmath140 figure [ fig.3 ] shows the diagrammatic representations of these probabilities ; the probability that an end node of a randomly chosen link belongs to the gcc at time step @xmath17 ; and the normalization equation [ eq.2 ] . ( a ) diagrammatic notations for the probabilities that , following a randomly chosen link to one of its end nodes , we arrive at @xmath129-node , @xmath130-node , and @xmath131-node , respectively , at time step @xmath17 . ( b ) the probability that an end node of a randomly chosen link belongs to the gcc at time step @xmath17 is given by @xmath136 . ( c ) diagrammatic representation of eq . [ eq.2 ] for the normalization condition of the three sets of probabilities @xmath137 , @xmath138 , and @xmath139 . ] thanks to eq . [ eq.2 ] , hereafter all the equations will be written in terms of @xmath129 and @xmath130 only . we can calculate @xmath137 and @xmath138 in an iterative way . at first , we consider the initial time step @xmath123 . the self - consistent equations for @xmath141 and @xmath142 are given by [ eq.3 ] @xmath143 @xmath144,\ ] ] where @xmath145 is the degree distribution of the nodes that we arrive at by following a randomly chosen link ( a.k.a . the excess degree distribution ) @xcite . we derive the above equations based on the following observations : * @xmath141-node : its neighbors can only be @xmath141-nodes ; * @xmath142-node : since it is an ap node , at least one of its neighbors is an @xmath141-node . moreover , since it belongs to the gcc , at lease one of its neighbors is not an @xmath141-node . note that , due to the tree ansatz , removal of an ap in the gcc can only cause a finite number of nodes disconnected from the gcc , and there exists no such an ap that its removal breaks down the gcc into two infinitely large components . the diagrammatic representations of eqs . [ eq.3a ] and [ eq.3b ] are shown in fig . [ fig.4 ] . diagrammatic representations of eqs . [ eq.3a ] and [ eq.3b ] . ] for the @xmath17-th gapr time step ( @xmath146 ) , we can compute @xmath129 and @xmath130 as follows : [ eq.5 ] @xmath147 @xmath148 the derivations of eqs . [ eq.5a ] and [ eq.5b ] are based on the following observations : * @xmath129-node : first of all , its neighbors can only be @xmath129-nodes or @xmath149-nodes with @xmath150 ( because if one of its neighbors is @xmath151-node with @xmath150 , this node will be an ap before time step @xmath17 and hence would have already been removed ; if one of its neighbor is @xmath130-node , this node will belong to the gcc at time step @xmath17 ) . second , its neighbors can not be simultaneously all @xmath149-nodes with @xmath152 . otherwise this node will be a leaf node before time step @xmath153 , and will become an isolated node before the @xmath17-th time step . in this case , we can not reach this node through a randomly chosen link at time step @xmath17 . * @xmath130-node : first of all , its neighbors can not be @xmath151-nodes with @xmath150 , otherwise this node would have been removed before time step @xmath17 . second , since it is an ap node , at least one of its neighbors is an @xmath129-node . finally , since this node belongs to the gcc , at least one of its neighbors is neither @xmath129-node nor @xmath149-node with @xmath150 . the diagrammatic representations of eqs . [ eq.5a ] and [ eq.5b ] are shown in fig . [ fig.5 ] . diagrammatic representations of eqs . [ eq.5a ] and [ eq.5b ] . ] combining eqs . [ eq.3 ] and [ eq.5 ] , we obtain a complete set of discrete - time dynamic equations for the gapr process : [ eq.6 ] @xmath154 @xmath155 @xmath156 where @xmath157 is the generating function of the excess degree distribution @xmath158 . by solving the above self - consistent equations , we obtain @xmath137 and @xmath138 , which govern the whole process of gapr . with these two sets of probabilities , we can compute any quantities of interest , such as the total number of gapr steps , the fraction of aps , the relative size of the gcc and the rgb , and so on . it should be noted from eq . [ eq.6 ] that @xmath129 and @xmath130 depend not only on the previous time step @xmath153 , but also on the entire history of gapr , from the 0-th to the @xmath159-th time steps . hence to calculate @xmath12 , the total number of gapr steps , we have to solve the equations from the initial step to the @xmath12-th step when the gapr process stops . since the gapr process terminates when there is no aps left in the network , @xmath12 can be determined by requiring @xmath160 @xmath161 note that , according to eq . [ eq.6b ] , @xmath162 is also zero , otherwise @xmath163 will have a non - zero solution . consider the fraction of aps at time step @xmath17 during the gapr process , @xmath64 , which also represents the probability that a randomly chosen node is an ap at @xmath17-th step . we can calculate @xmath99 as the sum of the following two parts : 1 . @xmath164 : fraction of aps that belong to the gcc ; 2 . @xmath165 : fraction of aps that belong to fccs . with similar theoretical considerations as deriving equations for @xmath129 and @xmath130 , we have @xmath166 @xmath167 these two equations can be understood as follows : * ap in fccs , @xmath165 : first of all , since the node belongs to an fcc , its neighbor can only be @xmath129-nodes or @xmath149-nodes with @xmath150 . second , since the node is an ap , at least two of its neighbors are @xmath129-nodes . otherwise , it would be a leaf node or an isolated node at time step @xmath17 . * ap in the gcc , @xmath164 : first of all , its neighbors can not be @xmath151-nodes with @xmath150 , otherwise this node would have already been removed before time step @xmath17 . second , since it is an ap node , at least one of its neighbors is an @xmath129-node . finally , since this node belongs to the gcc , at least one of its neighbors is neither @xmath129-node nor @xmath149-node with @xmath150 . the diagrammatic representations of eqs . [ eq.9a ] and [ eq.9b ] are shown in fig . [ fig.6 ] . diagrammatic representations of eqs . [ eq.9a ] and [ eq.9b ] . ] combing eqs . [ eq.9a ] and [ eq.9b ] , we obtain the fraction of aps at time step @xmath17 as @xmath168 where @xmath169 is the generating function of the degree distribution @xmath10 . the fraction of aps in the original network , @xmath170 , can be obtained by substituting @xmath123 into eq . [ eq.10 ] the above equation , yielding @xmath171 another quantity of interest is the relative size of the gcc , @xmath42 , after performing a given @xmath17 steps of gapr in a network . we can compute @xmath63 as follows : @xmath172 where the kronecker delta function @xmath173 if @xmath174 , and 0 otherwise . this equation can be understood as follows : * for any node in the gcc at time step @xmath17 , at least one of its neighbors belongs to the gcc at current time step . moreover , none of its neighbors is @xmath151-nodes with @xmath150 , otherwise this node would have already been removed as an ap before time step @xmath17 . * the first term on the right hand side of eq . [ eq.13 ] accounts for the nodes with at least two neighbors belonging to the gcc . * the second term accounts for the nodes with only one neighbor in the gcc . it should be noted that , for each of these nodes , at least one of its neighbors is @xmath129-nodes or @xmath175-nodes . otherwise , this node would be a leaf node before time step @xmath17 , and thus will become an isolated node at time step @xmath17 . * we only consider nodes with degree larger than 2 in the first two terms . this is because any leaf nodes will become isolated ones right after the first step of gapr . if we include the ordinary percolation as the 0-th time step in our gapr process , leaf nodes can also belong to the gcc at the 0-th step , which is the third term . the diagrammatic representation of eq . [ eq.13 ] is shown in fig . [ fig.7 ] . after some algebra , we can rewrite @xmath176 in terms of generating functions : @xmath177 note that the relative size of the gcc for ordinary percolation can be recovered by substituting @xmath123 into the above equation , yielding @xmath178 diagrammatic representation of eq . [ eq.13 ] . ] if we keep performing gapr in a network until there is no ap left , depending on the initial structure of the network , it could be totally destructed or left with an rgb . the relative size of the rgb can be calculated as @xmath179 which is based on the following observations : * for any node in the rgb , its neighbors can not be @xmath129-nodes at any @xmath17 . otherwise , this node is removable as an ap . * since the rgb is biconnected , each of its nodes must be connected to at least two other rgb nodes . the probability that an end node of a randomly chosen link belongs to the rgb is @xmath180 . the diagrammatic representation of eq . [ eq.16 ] is shown in fig . [ fig.8 ] . after some algebra , eq . [ eq.16 ] can be rewritten in terms of generating functions as : @xmath181 note that the rgb is nothing but the gcc at the last time step @xmath12 , we can also obtain the above equation by substituting @xmath124 and also @xmath182 , @xmath183 , @xmath184 into the eq . [ eq.14 ] of @xmath42 . diagrammatic representation of eq . [ eq.16 ] . ] the developed theoretical framework works for complex networks with any prescribed degree distribution . in this work , we consider two canonical model networks : erds - rnyi ( er ) networks and scale - free ( sf ) networks , which have specific degree distributions . for erds - rnyi random network , the degree distribution @xmath10 in the thermodynamic limit follows the poisson distribution , i.e. @xmath185 where @xmath186 is the mean degree . the generating functions are given by @xmath187 substituting the above equation into eqs . [ eq.6a ] , [ eq.6b ] , and [ eq.6c ] , we obtain the dynamic equations of the gapr process on erds - rnyi network : @xmath188\ ] ] @xmath189 - \exp \left [ c \left ( \sum_{\tau=0}^{t-2}\beta_\tau-1 \right ) \right]\ ] ] @xmath190+\exp \left [ c\left(\sum_{\tau=0}^{t-1}\beta_\tau-1\right)\right].\ ] ] similarly , the fraction of aps is given by @xmath191.\ ] ] hence the fraction of aps in the original network at @xmath123 is given by @xmath192 the relative size of the gcc at the @xmath17-th time step is given by @xmath193 \nonumber \\ & & - \left ( 1- \delta_{\tau,0 } \right ) c\left(1-\sum_{\tau=0}^{t}\alpha_{\tau } - \sum_{\tau=0}^{t-1}\beta_{\tau } \right ) \exp \left [ c\left(\sum_{\tau=0}^{t-2}\beta_{\tau}-1 \right ) \right].\end{aligned}\ ] ] the relative size of the rgb is @xmath194 -c \left(1-\sum_{\tau=0}^{t}\alpha_{\tau}-\sum_{\tau=0}^{t}\beta_{\tau}\right ) \exp \left [ c\left(\sum_{\tau=0}^{t}\beta_{\tau } -1 \right)\right].\ ] ] we use the static model to generate asymptotically sf networks with tunable network size @xmath8 , mean degree @xmath24 , and degree distribution exponent @xmath195 @xcite . this model can be described as follows : step-1 : : we start with @xmath8 isolated nodes , which are labeled from 1 to @xmath8 . each node is assigned a probability @xmath196 , where @xmath197 . step-2 : : we independently pick up two nodes according to their assigned probabilities , and add a link between these two nodes if they have not been connected before . self - links and double - links are forbidden . step-3 : : repeat step-2 until @xmath198 links have been added into the network . in the thermodynamic limit , the degree distribution of the static model can be analytically derived @xcite : @xmath199^k}{a k!}\int_1^{\infty } dx \exp[-c(1-a)x ] x^{k-1 - 1/a}.\ ] ] the generating function of the degree distribution @xmath10 is given by @xmath200^k}{a k!}\int_1^{\infty } dy \exp[-c(1-a)y ] y^{k-1 - 1/a } x^k \nonumber\\ = & & \frac{1}{a } \int_1^{\infty } dy \exp[-c(1-a)y ] y^{-1 - 1/a } \sum_{k=0}^{\infty } \frac{[c(1-a)xy]^k}{k ! } \nonumber \\ = & & \frac{1}{a } \int_1^{\infty } dy \exp[-c(1-a)(1-x)y ] y^{-1 - 1/a},\end{aligned}\ ] ] where @xmath201^k}{k!}$ ] is used in the last step . similarly , the generation function of the excess degree distribution @xmath158 is derived as @xmath202 y^{-1/a}.\ ] ] by defining @xmath203 @xmath204 and @xmath205 can be rewritten as @xmath206\ ] ] @xmath207\ ] ] where the fact @xmath208 has been used . substituting the above equations into eqs . [ eq.6a ] , [ eq.6b ] , and [ eq.6c ] , we obtain the dynamic equations of the gapr process for the static model : @xmath209\ ] ] @xmath210 - e_{\frac{1}{a } } \left [ c \left ( 1-a \right ) \left(1- \sum_{\tau=0}^{t-2}\beta_\tau \right ) \right ] \right\}\ ] ] @xmath211 - e_{\frac{1}{a } } \left [ c \left ( 1-a \right ) \sum_{\tau=0}^{t}\alpha_\tau \right ] \nonumber \\ & & - e_{\frac{1}{a } } \left [ c \left ( 1-a \right ) \left ( 1-\alpha_t - \sum_{\tau=0}^{t-1}\beta_\tau \right ) \right ] - e_{\frac{1}{a } } \left [ c \left ( 1-a \right ) \left ( 1-\sum_{\tau=0}^{t-1}\beta_\tau \right ) \right ] \bigg\}.\end{aligned}\ ] ] similarly , the fraction of aps is given by @xmath212 -e_{1+\frac{1}{a } } \left [ c \left ( 1-a \right ) \sum_{\tau=0}^{t}\alpha_\tau \right ] \right\ } \nonumber \\ & & - \frac{c\alpha_t(1-a)}{a } e_{\frac{1}{a } } \left [ c \left ( 1-a \right ) \left ( 1- \sum_{\tau=0}^{t-1}\beta_{\tau } \right ) \right].\end{aligned}\ ] ] hence the fraction of aps in the original network at @xmath123 is @xmath213- \frac{c\alpha_0(1-a)}{a } e_{\frac{1}{a } } [ c(1-a)].\ ] ] the relative size of the gcc at the @xmath17-th time step is given by @xmath214 - e_{1+\frac{1}{a } } \left [ c \left ( 1-a \right ) \left ( 1- \alpha_{t } - \sum_{\tau=0}^{t-1}\beta_{\tau } \right ) \right ] \right\ } \nonumber \\ & & -\left ( 1- \delta_{\tau,0 } \right ) \frac{c(1-a)}{a } \left(1-\sum_{\tau=0}^{t}\alpha_{\tau } - \sum_{\tau=0}^{t-1}\beta_{\tau } \right ) e_{\frac{1}{a } } \left [ c \left ( 1-a \right ) \left(1- \sum_{\tau=0}^{t-2}\beta_{\tau } \right ) \right].\end{aligned}\ ] ] the relative size of the rgb is @xmath215 - e_{1+\frac{1}{a } } \left [ c \left ( 1-a \right ) \left ( 1- \sum_{\tau=0}^{t } \beta_{\tau } \right ) \right ] \right\ } \nonumber \\ & & - \frac{c(1-a)}{a } \left(1-\sum_{\tau=0}^{t}\alpha_{\tau}-\sum_{\tau=0}^{t}\beta_{\tau } \right ) e_{\frac{1}{a } } \left [ c \left ( 1-a \right ) \left(1- \sum_{\tau=0}^{t}\beta_{\tau } \right ) \right].\end{aligned}\ ] ] figure [ fig.9 ] shows the simulation results of the fraction of aps , @xmath125 , as a function of mean degree @xmath24 at different time steps for er networks and sf networks ( constructed from the er model and the static model ) with different degree exponents @xmath26 . for comparison , analytical results for infinitely large networks are also shown ( in lines ) . the comparison between simulation results and analytical predictions of the relative size of the gcc , @xmath126 , is shown in fig . [ fig.10 ] . note that , in figs . [ fig.9 ] and [ fig.10 ] , the deviation of the simulation results from the theoretical prediction for sf networks with exponent @xmath104 owes to degree correlations in the constructed scale - free networks , which become prominent as @xmath105 @xcite . in fig . [ fig.11 ] , we show the simulation results of the distribution of the relative size of the rgb , @xmath2 , at criticality for both er network and sf networks with different degree exponents . the bimodal distribution is another evidence that @xmath2 undergoes a discontinuous jump from zero to a large finite value at the critical point @xcite . the fraction of aps , @xmath125 , as a function of mean degree @xmath24 at different time steps @xmath17 in the gapr process , for er networks ( a ) and sf networks with different degree exponents @xmath216 ( b - d ) . symbols are numerical results calculated from the gapr procedure on finite discrete networks constructed with er model ( a ) and the static model ( b - d ) with @xmath95 . the data points and error bars ( defined as s.e.m . ) are determined from 128 independent network realizations . lines are analytical results for infinitely large system ( @xmath217 ) . ] the relative size of the gcc , @xmath126 , as a function of mean degree @xmath24 at different time steps @xmath17 in the gapr process , for er networks ( a ) and sf networks with different degree exponents @xmath216 ( b - d ) . symbols are numerical results calculated by performing a given steps of gapr procedure on finite discrete networks constructed with er model ( a ) and the static model ( b - d ) with @xmath95 . the data points and error bars ( defined as s.e.m . ) are determined from 128 independent network realizations . thin lines are analytical results for @xmath126 with finite @xmath17 , and thick lines are analytical results for the relative size of the rgb , which is @xmath218 . ] the distribution of the relative size of the rgb , @xmath2 , at criticality for both er network ( a ) and sf networks with different exponents @xmath216 ( b - d ) . each distribution is generated by performing gapr process on 12,800 different finite discrete networks of size @xmath95 . ] all the real - world networks analyzed in this work are listed and briefly described in the following tables . for each network , we show its type , name and reference ; number of nodes ( @xmath8 ) and edges ( @xmath9 ) ; and a brief description ( unless it is clear from the network type ) . 78ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop _ _ ( , ) @noop ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _ _ ( , ) pp . @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _ _ ( , ) pp . @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) http://toreopsahl.com/2011/08/12/why-anchorage-is-not-that-important-binary-ties-and-sample-selection/ [ `` , '' ] ( ) @noop `` , '' ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop _ _ ( , ) @noop _ _ ( , ) @noop `` , '' @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _ _ ( , ) pp . @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop _ _ ( , ) * acknowledgement . * we thank shlomo havlin , mehran kardar , wei chen , endre cska , abhijeet sonawane , yandong xiao , chuliang song for valuable discussions . this work was partially supported by the john templeton foundation ( award no . 51977 ) and national natural science foundation of china ( grants nos . 11505095 and 11374159 ) . * author contributions . * y .- y.l . conceived and designed the project . l.t . performed all the analytical calculations and extensive numerical simulations , as well as empirical data analysis . all authors analyzed the results . l.t . and y .- y.l . wrote the manuscript . d .- n.s . edited the manuscript . |
we present the first images of the j=54 and j=1615 lines of the dense gas tracer , cyanoacetylene , hcn , in an external galaxy . the central 200 pc of the nearby star - forming spiral galaxy , ic 342 , was mapped using the vla and the plateau de bure interferometer . hcn(54 ) line emission is found across the nuclear mini - spiral , but is very weak towards the starburst site , the location of the strongest mid - ir and radio emission . the j=1615 and 109 lines are also faint near the large hii region complex , but are brighter relative to the 54 line , consistent with higher excitation . the brightest hcn emission is located in the northern arm of the nuclear minispiral , 100 pc away from the radio / ir source to the southwest of the nucleus . this location appears less affected by ultraviolet radiation , and may represent a more embedded , earlier stage of star formation . excitation temperatures are consistent with those determined from co ; the gas is dense , and cool , t k. so as to not violate limits on the total h mass determined from co , at least two dense components are required to model ic 342 s giant molecular clouds . these observations suggest that hcn(54 ) is an excellent probe of the dense , quiescent gas in galaxies . the high excitation combined with faint emission towards the dense molecular gas at the starburst indicates that it currently lacks large masses of very dense gas . we propose a scenario where the starburst is being caught in the act of dispersing or destroying its dense gas in the presence of the large hii region . this explains the high star formation efficiency seen in the dense component . the little remaining dense gas appears to be in pressure equilibrium with the starburst hii region . | 1105.1792 | little is known about the properties of the dense gas component on gmc scales in external galaxies yet it is dense gas ( @xmath5 @xmath6 ) that most directly correlates with global star formation rates ( e.g. , * ? ? ? * ; * ? ? ? * ) . to understand the evolution and regulation of current star formation in galaxies , dense gas properties need to be accurately characterized . in a recent imaging survey of millimeter - wave molecular lines in the nearby spiral nucleus of ic 342 , a surprising degree of morphological variation is observed in its dense gas @xcite , which was not anticipated from co or hcn ( e.g. , * ? ? ? * ; * ? ? ? these and other observations ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) confirm that co and hcn fail to fully constrain excitation and chemical properties of the dense clouds . various probes of the dense component exist , each with various strengths and weaknesses ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? hcn(10 ) is the most commonly used dense gas tracer , primarily because it is bright , however , it is optically thick in most starburst environments ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , has a widely spaced rotational ladder with its j@xmath73 transitions in the submm and can be sensitive to chemical effects ( pdr effects and ir pumping ; e.g. * ? ? ? * ; * ? ? ? * ) . here we present a study of the physical conditions of the very dense gas component of a nearby starburst nucleus using hc@xmath0n . with its large electric dipole moment ( @xmath8 = 3.72 debye versus 3.0 debye for hcn ) , low opacity and closely spaced rotational ladder accessible to powerful centimeter and millimeter - wave interferometers , hc@xmath0n is well - suited to high resolution imaging of the structure and excitation of the densest component of the molecular gas ( eg . * ; * ? ? ? * ; * ? ? ? hc@xmath9n(54 ) has a critical density of @xmath10 @xmath6 and an upper level energy of e / k = 6.6 k ; the 1615 line has a critical density of @xmath11 @xmath6 and an upper level energy of e / k = 59 k. we have observed the j = 54 and j = 1615 lines of hc@xmath0n at @xmath122 resolution in ic 342 with the very large array ( vla ) and the plateau de bure interferometer ( pdb ) , respectively . ic 342 is one of the closest ( d@xmath133 mpc , or 2 = 29 pc ; eg . * ? ? ? * ; * ? ? ? * ) , large spirals with active nuclear star formation @xcite . it is nearly face - on , with bright molecular line emission both in dense clouds and a diffuse medium @xcite . the wealth of data and proximity of ic 342 permit the connection of cloud properties with star formation at sub - gmc spatial scales . the data published here represent the first high resolution maps of hc@xmath9n(54 ) and hc@xmath9n(1615 ) in an external galaxy . excitation is a key component in the interpretation of molecular line intensities . so these two maps are compared with the previously published lower resolution hc@xmath0n(109 ) map made with the owens valley millimeter array @xcite , to constrain the physical conditions of the densest component of the ism in the center of ic342 , and to correlate properties of the dense gas with star formation , diffuse gas , and chemistry . aperture synthesis observations of the hc@xmath0n j=54 rotational line ( 45.490316 ghz ) towards ic 342 were made with the d configuration of the vla on 2005 november 25 ( vla i d : am839 ) . the synthesized beam is 195@xmath1415 ( fwhm ) ; position angle ( pa)=-11.2 . fifteen 1.5625 mhz channels were used , for a velocity resolution of 10.3 km s@xmath15 , centered at @xmath16 km s@xmath15 . the phase center is @xmath17(j2000 ) = 03@xmath1846@xmath19483 ; @xmath20(j2000 ) = 6805470 . amplitude tracking and pointing was done by observing the quasar 0228 + 673 every 45 minutes . phases were tracked by fast switching between the source and the quasar 0304 + 655 with 130s/50s cycles . absolute flux calibration was done using 3c48 and 3c147 and is good to @xmath13510% . calibration and analysis was performed with the nrao astronomical image processing software ( aips ) package . the naturally weighted and cleaned datacube has an rms of 0.70 mjy beam@xmath15 . correction for the primary beam ( @xmath131 at 45 ghz ) has not been applied . the shortest baselines in the dataset are @xmath215 k@xmath22 , corresponding to scales of @xmath1340 ; structures larger than this are not well - sampled . the hc@xmath0n(16 - 15 ) line emission in the center of ic 342 was observed with 5 antennas using the new 2 mm receivers of the iram plateau de bure interferometer ( pdbi ) on 2007 december 28 and 31 in c configuration with baselines ranging from 24 to 176 m . the phase center of the observations was set to @xmath17(j2000 ) = 03@xmath2346@xmath1948105 ; @xmath20(j2000 ) = 68054784 . nrao150 and 0212 + 735 served as phase calibrators and were observed every 20 min . flux calibrators were 3c84 and 3c454.3 . the calibration was done in the gildas package following standard procedures . the hc@xmath0n(16 - 15 ) line at 145.560951ghz was observed assuming a systemic velocity of @xmath24 km s@xmath15 and a spectral resolution of 2.5mhz ( 5.15 km s@xmath15 ) . the average 2 mm continuum was obtained by averaging line - free channels blue and redward of the hc@xmath0n and h@xmath4co lines and subtracted from the @xmath25 datacube to obtain a continuum - free datacube . the final naturally weighted and cleaned datacube with 10kms@xmath15 wide channels has a clean beam of [email protected] , pa=46@xmath27 and an rms of 1.7 mjy / beam . single - dish observations of hc@xmath0n(1615 ) find a peak brightness temperature of 6.7 mk in a 16.9@xmath28 beam @xcite . convolving our map to this beam and sampling it at the same location yields a peak brightness of 7.1 mk , which agrees within the uncertainties of both datasets . therefore no flux is resolved out of the interferometer maps , as expected for these high density tracers . continuum subtracted integrated intensity line maps of hc@xmath0n(54 ) , hc@xmath0n(109 ) and hc@xmath0n(1615 ) in ic 342 are shown in figure [ inti ] . figure [ inticont ] shows the 54 transition overlaid on the 7 mm continuum image generated from offline channels . locations of giant molecular cloud ( gmc ) cores @xcite and the optical clusters ( e.g. * ? ? ? * ) , compared with hc@xmath0n(54 ) and hcn(10 ) @xcite , are also shown in figure [ inticont ] . figure [ spec ] shows the hc@xmath9n(54 ) and hc@xmath0n(16 - 15 ) spectra ( in flux units ) taken over the same 2aperture centered on each cloud . hc@xmath0n(54 ) emission picks out most clouds seen in other dense gas tracers ( e.g hcn(10 ) . the only labeled cloud not clearly detected in hc@xmath9n(54 ) is gmc b , the cloud associated with the nuclear star - forming region ( @xmath29 ; * ? ? ? * ; * ? ? ? * ) . positions of the gmcs measured in hc@xmath0n are consistent with those fitted in c@xmath1o(2 - 1 ) to within a beam @xcite . an additional gmc is detected in hc@xmath0n(54 ) just south of gmc c3 , labeled c4 . unlike c@xmath1o(2 - 1 ) which peaks at gmc c2 , hc@xmath0n(54 ) emission peaks farther north , towards gmc c1 , at a distance of 105 pc from the nucleus suggesting changes in excitation across gmc c. gmcs a and d are resolved into two components , but we do not discuss them as separate entities given the lower signal - to - noise , other than to state that there is a clear difference in velocity centroid between the 54 and 1615 transitions ( e.g. figure [ spec ] ) indicating there is likely an excitation gradient across gmc a. no emission @xmath30 is detected towards the weak co(2 - 1 ) feature associated with the central nuclear star cluster @xcite . the j=1615 line intensity is comparable to that of j=54 line in all clouds . while dominated by gmc c , hc@xmath0n(16 - 15 ) is detected ( or tentatively detected ) towards all clouds , except gmc d. fluxes for gmcs d and d are quite uncertain since they are just inside the half power point of the pdbi primary beam . that gmc c is much brighter than the other clouds in both transitions indicates large quantities of dense gas are present in this cloud . hc@xmath0n(1615 ) favors c2 over c1 . gmc b , while still faint in an absolute sense , is significantly brighter in hc@xmath0n(1615 ) relative to 54 . this is not unexpected given the higher excitation requirements of the 1615 line and that both b and c2 have 7 mm continuum sources associated with massive star formation . comparisons of the hc@xmath9n ( 54 ) , ( 109 ) and ( 1615 ) maps provide a chance to establish the excitation of the densest molecular cloud gas component . line intensities from the j=54 and j=16 - 15 lines were measured over 2 apertures centered on each of the gmcs . table [ tinti ] records gaussian fits to each spectrum along with integrated line intensities . peak antenna temperature ratios are calculated for hc@xmath0n(1615)/hc@xmath0n(54 ) , hereafter denoted r@xmath32 . r@xmath32 ranges from less than 0.06 up to @xmath130.5 , with gmc b having the highest value ( table [ trat ] ) . the 54 and 1615 transitions generally bracket the peak of the level populations , so we achieve good constraints on gas excitation . lte excitation temperatures , t@xmath33 , implied by r@xmath32 range from @xmath1210 k to @xmath718 k ( where we have neglected t@xmath34 in this determination ) . excitation is lowest towards gmcs d , d and e. these excitation temperatures t@xmath33 are similar to those found from the presumably much less dense gas traced in c@xmath1o ( table 2 ; * ? ? ? * ) . only the gmc c clouds have significantly higher t@xmath33 in hc@xmath9n towards c1 - c3 t@xmath33(hc@xmath0n ) are about a factor of two greater than t@xmath33(c@xmath1o ) . the resolution of the hc@xmath0n(109 ) data is significantly lower than it is for the 54 and 1615 lines . to compare j=54 and j=10 - 9 line intensities , the 54 data were convolved to the resolution of the ( 109 ) data ( @xmath35 * ? ? ? * ) , then integrated intensity ratios , hereafter r@xmath36 , were sampled at the locations of r@xmath32 . though at lower resolution than r@xmath32 , we make the approximation that r@xmath36 does not change on these sub - gmc scales . while leading to larger uncertainties , this provides a way to include all three transitions in modeling dense gas excitation at very high resolution . @xmath37 range from 0.25 to 1.4 ( table [ tinti ] ) . gmc b has the highest value of 1.4 . the remainder of the gmcs have ratios of @xmath38 . excitation temperatures implied by these ratios range from t@xmath39 616 k , consistent with those derived from @xmath40 separately . t@xmath33 derived from r@xmath36 are similar to those derived from c@xmath1o . the only exception here is gmc a , the cloud where pdrs ( photon - dominated regions ) dominate @xcite . towards gmc a t@xmath33(c@xmath1o ) and t@xmath33(r@xmath32 ) are 12 - 13 k , twice that of t@xmath33(r@xmath36 ) . before modeling the densities implied by the line ratios , we test whether ir pumping can be responsible for the observed excitation . ir pumping is important if , @xmath41 where @xmath42 is the einstein @xmath43 of the corresponding @xmath44 vibrational transitions at @xmath45 m , @xmath46(45 @xmath8 m ) is the 45 @xmath8 m intensity as seen by the hc@xmath0n molecules and @xmath47 is the einstein @xmath48 of the rotational state ( e.g. * ? ? ? it is extremely difficult to estimate the applicable 45 @xmath8 m ir intensity , but it is expected to be most intense towards the starburst gmc ( b ) . a detailed assessment of ir pumping must await high resolution mir maps , but we constrain @xmath46(45 @xmath49 ) in several ways . first , we take the 45 @xmath8 m flux from @xcite and scale it by the fraction of total 20 @xmath8 m flux that comes from within 2.1@xmath28 of the starburst as found by @xcite and then average over that aperture . for the average @xmath46(45 @xmath8 m ) calculated this way , @xmath50(45 @xmath8 m ) is @xmath51 ( @xmath52 ) times too low to pump the 1615 ( 54 ) transition . alternatively if we ( very conservatively ) take the total mir luminosity from the central 30@xmath28 and assume it comes from a blackbody of the observed color temperature ( e.g. @xmath1350 k ; * ? ? * ) then @xmath46(45 @xmath8 m ) is still at least an order of magnitude too low to meet the inequality in eq . [ pump ] for both transition . ir pumping rates only become comparable to @xmath47 for the 54 transition if the total ir luminosity originates from a @xmath132.5 pc source with a source temperature of @xmath53100 k. we conclude that ir pumping is not important for the 1615 transition of hc@xmath0n in any reasonable geometry of the ir field . for the 54 transition to be sensitive to ir pumping , the ir source must be warm , opaque and extremely compact . therefore ir pumping is neglected for all clouds in the following discussion . the values of the excitation temperature constrain the density and kinetic temperatures , @xmath54 and @xmath55 , and the physical conditions of the clouds driving the excitation . a series of large velocity gradient ( lvg ) radiative transfer models were run to predict the observed intensities and line ratios for a given @xmath54 , t@xmath56 and filling factor , @xmath57 , of the dense component ( e.g. , * ? ? ? single component lvg models are instructive , particularly when lines are optically thin , as is the case for hc@xmath0n . the lvg model used is that of @xcite , adapted to hc@xmath0n , with levels up to j=20 included . collision coefficients are from @xcite . a range of densities , @xmath54 = 10@xmath5810@xmath59 , and kinetic temperatures , t@xmath56 = 0100 k , was explored . column densities based on lte excitation ( table [ trat ] ) , are calculated at 2@xmath28 resolution from : @xmath60 using molecular data of @xcite , hc@xmath0n(54 ) intensities , and t@xmath33(hc@xmath0n ) from table [ trat ] . hc@xmath0n abundances are found to be , @xmath61 , with the highest values towards c1 and d. while uncertain these abundances agree with those found in @xcite and are typical of galactic center hc@xmath0n abundances ( e.g. * ? ? ? * ; * ? ? ? * ) and good enough for constraining @xmath62 . the ratio of cloud linewidth to core size is @xmath63 for the gmcs . therefore , we adopt a standard model abundance per velocity gradient of @xmath62 = 10@xmath64 km s@xmath15 , but run models for values of @xmath65 - @xmath66 . antenna temperatures are sensitive to the unknown filling factor . in these extragalactic observations , the beam corresponds to scales large compared to cloud structure , and hence filling factors are not directly known . to first order , the line ratios , @xmath37 and r@xmath32 are independent of filling factor if we assume that hc@xmath0n(54 ) , ( 109 ) and ( 1615 ) originate in the same gas . so model ratios are compared to the observed data to constrain parameter space . for the parameter space implied by the line ratios model brightness temperatures are determined . a comparison of the model brightness temperature to the observed brightness temperature sets the required areal filling factors for that solution . figure [ lvg ] displays the results of the lvg modeling . acceptable @xmath67 ( t@xmath56,@xmath54 ) parameter spaces are shown for each line ratio ( r@xmath36 , thick solid gray line ; r@xmath32 , thick solid black line ) . also shown in figure [ lvg ] are the acceptable solutions obtained from the c@xmath1o(21)/c@xmath1o(10 ) line ratio ( thin gray lines ; * ? ? ? figure [ levels ] and table [ tlvg1 ] display the model flux versus upper j state of the line for the t@xmath68 = 10 k , 30 k and 50 k solutions ( also 70 k solutions for gmc b ) . hc@xmath0n(54 ) has a critical density of @xmath69 @xmath6 and an upper energy state of 6.55 k. hc@xmath0n(1615 ) has a critical density , @xmath70 @xmath6 and an upper energy state of 59 k. hc@xmath0n(109 ) values are intermediate . therefore @xmath37 and r@xmath32 are sensitive probes of @xmath54 between the range of @xmath71 and @xmath72 @xmath6 . observed ratios constrain @xmath54 to @xmath730.1 dex at a given t@xmath56 and to @xmath730.3 dex for all modeled t@xmath74 k. for t@xmath75 40 k , ratios are largely insensitive to t@xmath56 ( curves are horizontal ) . the lvg models do not constrain density when t@xmath56 is low . thus to narrow the range of possible solutions requires two or more line ratios , or an external constraint on one of the axes . kinetic temperatures of the dense component at 2@xmath28 scales in ic 342 remain largely unconstrained to date . one method of constraining the kinetic temperature is the peak t@xmath76 at this resolution of the optically thick co(2 - 1 ) line . using this method , t@xmath77 3545 k towards gmcs a , b and c , and t@xmath77 1520 k for gmcs d , d and e @xcite . these values are not highly discordant from arcminute resolution t@xmath78 k measurements from nh@xmath0 @xcite or the far - infrared dust temperature of 42 - 55 k @xcite . however , the co(2 - 1 ) traces gas with densities two to three orders of magnitude lower than these hc@xmath0n transitions , so it is not clear that this is the relevant t@xmath56 for the dense cores of the gmcs . valid solutions are found for the gmcs , with the exception of gmc d , which is the most uncertain due to its distance from the center of the field . ( no fit is attempted for gmc d because both hc@xmath0n(54 ) and hc@xmath0n(1615 ) are upper limits or tentative detections . ) the solutions show modest cloud excitations given the strong nuclear star formation . good agreement is observed between solutions found using r@xmath32 and r@xmath36 when kinetic temperatures are low ( t@xmath79 k ) . given uncertainties in line intensities and @xmath80 , higher temperatures can not be ruled out . while t@xmath56 itself is less well constrained the combination @xmath54t@xmath56 is well determined . since line brightnesses are the measured quantity , as gas temperatures are raised , densities or filling factors must decrease to compensate . the nature of the solutions are such that the densities decrease more than the filling factors ( table [ lvg ] ) . gmcs a , c1 and c3 are best fit with @xmath81 @xmath6 and t@xmath77 20 k. for t@xmath82 k derived densities drop to @xmath83 @xmath6 , while they drop to @xmath84 @xmath6 for t@xmath56 = 100 k. changes in @xmath62 do not strongly influence the derived solutions . towards these clouds filling factors are @xmath850.03 - 0.1 . for comparison , a 1 pc@xmath58 cloud would have @xmath86 0.0012 for the aperture size and ic 342 s distance . therefore these gmcs appear to have a few dozen dense clumps similar to the larger clumps found in galactic gmcs ( e.g. * ? ? ? * ; * ? ? ? gmc c2 appears to have similar densities but slightly elevated temperatures , t@xmath77 30k . in gmcs d and d densities are at least a factor of four lower than the other clouds . little can be said about the kinetic temperatures of these gmcs . gmc d is the only location with statistically discrepant solutions from r@xmath36 and r@xmath32 . here the ( 1615 ) line is somewhat brighter than would be expected from the r@xmath36 . the starburst gmc , b , is markedly different from the others . the combination of a low absolute hc@xmath0n brightness temperature and high ratios , @xmath37 and r@xmath32 , require higher densities of @xmath87 @xmath6 and a smaller ( @xmath880.02 ) filling factor . both line ratios seem to indicate a kinetic temperature of more than t@xmath77 40 k , and perhaps significantly higher , averaged over our 30 pc beam . the dense gas in gmc b , which is closest to the active star - forming region and ir source , is more compact , denser , and more highly excited than the other clouds . for our own galactic center it has long been argued that molecular clouds must be unusually dense , at least @xmath89 to withstand tidal forces ( eg . * ? ? ? * ) . if one assumes that gmc c , the most well defined cloud , is a point mass in a spherical mass potential of the galaxy , its roche limit would be at a radius of 25 pc . the radius of the combined c1-c2-c3 complex is @xmath725 pc . in ic 342 , the proximity to the nucleus where the strong noncircular velocity field @xcite together with feedback from the starburst @xcite suggests that these clumps have densities large enough to maintain their identity but likely will not remain gravitationally bound to each other . for all gmcs ( including gmc b ) , t@xmath33(hc@xmath0n ) @xmath21 t@xmath33(c@xmath1o ) . if a model in which all the h@xmath4 exists in one uniform component , then the similarity of t@xmath33 from both hc@xmath0n and c@xmath1o implies that the densities of the molecular clouds are high enough ( @xmath90 @xmath6 ) to thermalize both c@xmath1o and hc@xmath0n across the nucleus . ( lvg solutions for c@xmath1o j=21 and 10 lines from @xcite are shown with the hc@xmath9n solutions in figure [ lvg ] . ) in this monolithic model the low observed t@xmath33 of 1020 k demand that the thermalized clouds must be quite cool . these temperatures are similar to those of dark clouds in the disk of our galaxy , which is somewhat surprising given the elevated star formation rate in the nucleus of ic 342 . however this model of a monolithic , dense , cold ism seen in both hc@xmath0n and c@xmath1o runs into problems with total mass constraints . if the ism is uniformly this dense and cold then the total dense gas cloud mass implied by the required densities and filling factors become very large , greater than permitted based on the optically thin c@xmath1o line emission . in table [ tlvg1 ] the total mass of dense gas , @xmath91 , is approximated from the lvg solutions assuming @xmath92 , with r defined as @xmath93 . table [ trat ] lists the total molecular gas mass , @xmath94 , estimated from c@xmath1o(2 - 1 ) data ( see * ? ? ? * ) assuming @xmath95 . one can see that @xmath91 is typically larger than @xmath94 for t@xmath96 30 k and diverges rapidly as t@xmath56 drops towards the lvg favored values . in short if gmcs have uniformly such high densities and cold temperatures then the clouds would contain too much h@xmath4 mass for what is observed in c@xmath1o . to match the line ratios while not violating mass constraints requires a multi - component dense ism . a possible two component model of the dense gas is discussed in @xmath97 [ cloudsxco ] . a consideration of the conversion factor between co intensity and h@xmath4 column density , @xmath98 , in the nuclear region of ic 342 suggests that the model of a single , high density and relatively cool ism is not consistent with observations of co isotopologues . moreover these clouds are unlikely to resemble galactic disk gmcs in their internal structure and dynamics . the well known galactic relation between co intensity and h@xmath4 column density , @xmath99 can be explained if gmcs consist of optically thick ( in co ) turbulent clumps in virial equilibrium ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? for the clumps to emit in hc@xmath0n densities must be large enough so that t@xmath76 can legitimately be approximated by t@xmath56 . if we adopt this model , then @xmath98 is approximated by : @xmath100~\rm cm^{-2}\ , ( k~km\,s^{-1})^{-1 } \label{eqxco}\ ] ] where @xmath101 is the volume filling factor and @xmath102 is the density of the _ clumps _ @xcite . from the brightness of the co isotopic lines , @xmath103 on 2 scales @xcite . the conversion factor @xmath98 within the central few hundred pc of ic 342 has been determined to be @xmath104 @xcite . if the monolithic model presented in the previous section is correct and clouds are virialized clumps in equilibrium , then @xmath105 must be @xmath1060.75 to match the observed conversion factor . table [ tlvg1 ] records @xmath105 for each of the lvg solutions across the nucleus of ic 342 . for all solutions , @xmath105 is greater than the required value , especially at low t@xmath56 . for the single component lvg favored t@xmath56 of @xmath1310 - 20 k towards the more quiescent clouds , @xmath107575 , implying @xmath108 should be several times larger than @xmath109 . however @xmath98 in the nucleus of ic 342 is 34 times _ lower _ than the galactic conversion factor , based on both optically thin isotopologues of co as well as dust emission @xcite . lower conversion factors appear to be the norm for the nuclei of gas - rich star forming galaxies , including our own ( eg . * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in the nuclear region of ic 342 dense gas kinetic temperatures do not appear to be higher than the galactic disk value by an amount large enough to offset the observed increase in density over typical galactic disk - like clouds ; t@xmath110100 k would be required . recently @xcite completed a more generalized treatment of the conversion factor including radiative transfer and concludes that the true exponential dependences of @xmath54 and t@xmath56 are weaker than 0.5 and -1 , respectively . however in all cases modelled , the dependences remain positive for @xmath54 and negative for t@xmath56 . in summary , it is not possible to reconcile bright hc@xmath0n emission from uniformly cool , dense gas ( low t@xmath56 , high @xmath54 and high @xmath111 ) with the known total amount of h@xmath4 present if virialized constant density clumps are adopted . the lack of an apparent connection between @xmath105 of the dense component and the observed conversion factor indicates that the clouds in the nucleus of ic 342 can not be treated as simple virialized collections of uniform density clumps . one might expect that a large fraction of the c@xmath1o emission could originate from a moderate density component that is distinct from the denser , hc@xmath0n - emitting gas . ( note that @xmath113co traces a distinct component more diffuse than that traced both by c@xmath1o @xcite and hc@xmath0n . ) as a first order extension to the basic virialized clumps model , we imagine two spatially well - mixed sets of clumps ; one low density c@xmath1o emitting , @xmath114 , and one high density , @xmath115 , that emits both c@xmath1o and hc@xmath0n . assuming the fraction , by number , of clumps with high density is @xmath116 , then if both clumps have the same t@xmath56 the virialized clumps @xmath98 relation becomes : @xmath117~\rm cm^{-2}\ , ( k~km\,s^{-1})^{-1 } , \label{eqxco2}\ ] ] where @xmath118 . the ratio of dense gas to total h@xmath4 becomes , m@xmath119/m@xmath120 = @xmath121 . for a given t@xmath56 , @xmath114 is chosen to match the c@xmath1o lvg solution @xcite , while @xmath115 is chosen to match the hc@xmath0n lvg solutions , and @xmath122 . table [ tlvg2 ] displays the adopted c@xmath1o lvg solutions , along with the new @xmath123 and m@xmath119/m@xmath124 for the two component model , assuming , for simplicity , that t@xmath56 is the same for both components . @xmath123 and hence @xmath98 are decreased from the simple one component model by factors of at least four . this simple extension results in closer agreement between the observed and calculated values of x@xmath112 ( especially when including the somewhat super - virial line widths ) , while maintaining most of the mass in the dense ( low filling factor ) component . it is almost certainly the case that this extension is an oversimplification . in reality , we expect the clouds to exhibit a continuum of densities . however the lvg modeling demonstrates that at least these three components are required to match the multi - transition observations of @xmath113co , @xmath125co , c@xmath1o and hc@xmath0n . this is consistent with conclusions from recent single - dish modeling of higher j transitions of hc@xmath0n @xcite , but we find multiple components are required to match intensities not only between gmcs but within individual gmcs . hcn(10 ) emission is the workhorse for relating quantities of dense gas to star formation ( e.g. * ? ? ? it is interesting to compare conclusions about the dense component from hc@xmath0n with those of the more commonly used hcn(10 ) . hcn(10 ) has been mapped at similar spatial resolution by @xcite . figure [ inticont ] compares hc@xmath0n(54 ) to hcn(10 ) . while hc@xmath0n generally traces the same dense gmcs seen in hcn their relative brightnesses are rather different . in hcn(10 ) gmc a , b and c are all within 10 % of the same brightness and gmcs d and e are 100 % and 50 % weaker , respectively . whereas in hc@xmath0n(54 ) and ( 1615 ) , c dominates , b is nearly absent and a is not significantly different from d and e. comparisons with hc@xmath0n clearly demonstrate that there is larger variations in dense gas properties than hcn indicates . the dominate difference is enhanced hcn towards the starburst and gmc a relative to gmc c. unlike the hc@xmath0n transitions , hcn(10 ) is optically thick and has slightly larger ( 25 - 50 % ) filling factors @xcite . as kinetic temperatures increase optically thick transitions brighten more rapidly than ( lower excitation ) optically thin transitions . hence it is expected that the hcn emission should favor somewhat warmer dense gas . likely this effect results in the much brighter relative hcn(10 ) intensities towards the starburst . the relative enhancement towards gmc a is less clear . however , this cloud is known to be dominated by pdrs @xcite , strongly influenced by mechanical feedback from the nuclear cluster @xcite and has a complicated hc@xmath0n temperature and velocity structure ( sections [ morp ] and [ ratio ] ) . this cloud must have a complex density and temperature structure , potentially with a warmer intermediate density medium ( table [ trat ] ) . moreover , hcn abundances can be elevated in pdrs . we conclude that while hcn(10 ) does a good job locating the dense gas but it does a poorer job tracing small scale variation in the properties of the dense gas when a mix of strong star formation and quiescent gas are present . it is expected that such effects should become more important as specific star formation rates increase and spatial scales decrease . the brightest hc@xmath0n emission and the brightest star - forming regions do not coincide . most of the current star formation , traced by bright infrared and thermal radio continuum emission @xcite , is situated about 50 pc to the southwest of the dynamical center , in the vicinity of gmc b. the brightest hc@xmath0n emission by far is on the northeast side of the nucleus , in the northern molecular arm centered at gmc c2 . this region has free - free emission amounting to only a third of the brightness of the strongest radio source , associated with gmc b. the faintness , in absolute terms , of hc@xmath0n(1615 ) towards gmc b is unexpected and important reflecting a much lower areal filling of warm , very dense gas here . either the number of cloud clumps or their size is small relative to the other gmcs . on the other hand , the excitation of hc@xmath0n does reflect the presence of young forming stars . hc@xmath0n(1615 ) and hc@xmath0n(109 ) are relatively brighter towards the current starburst . this suggests that the absence of a correlation between hc@xmath0n(54 ) intensity and star formation is partly due to depopulation of the lower energy transitions . gmcs that are faint in the hc@xmath9n j=1615 transition and not associated with strong star formation show up well at the lower j transitions of hc@xmath0n , as would be expected for gas of lower excitation . clearly there is a large amount of dense gas currently not actively forming stars , that shows up in the low excitation transitions of hc@xmath0n . this is consistent with the fact that both hnc(10 ) and n@xmath4h@xmath126(10 ) , generally considered dense quiescent gas tracers , are found to be very bright towards ic 342 s nucleus @xcite . hc@xmath0n(54 ) appears to be an excellent extragalactic probe of the dense , quiescent molecular gas component not yet involved in the current starburst . to quantitatively compare star formation with dense gas , a star formation rate is derived from the 45 ghz continuum flux ( spectral indices measured between 2 cm and 7 mm demonstrate that the vast majority of flux toward gmcs b and c is thermal bremsstrahlung ) . the star formation rate is then compared with @xmath91 ( table [ tlvg2 ] ) to estimate a dense gas star formation efficiency , sfe@xmath127 . dense gas depletion timescales , @xmath128= 1/sfe@xmath119 , are also computed . in table [ tlvg2 ] the ratio of the observed lyman continuum ionization rate ( e.g. * ? ? ? * ; * ? ? ? * corrected for distance ) to m@xmath119 derived from the lvg analysis is reported . towards the non - starburst gmcs @xmath129 , while towards gmc b this ratio is 10 - 30 times larger . if one adopts the conversion from n@xmath130 to star formation rate of @xmath131 ( e.g. * ? ? ? * ) , then sfe@xmath119 for the starburst is @xmath132 , or dense gas depletion times of @xmath133 2 - 3 myr ! this highly enhanced sfe@xmath119 is a direct consequence of the faint hc@xmath0n emission here . even towards the non - starburst clouds sfe@xmath119 are @xmath134 yr@xmath15 . these efficiencies are sufficiently short that they imply dense gas consumption timescales that are non - negligible fractions of the expected gmc lifetimes . the dense sfe is rather high across the nucleus , but the extreme value towards gmc b is remarkable . @xcite argued that intense star formation is suppressed along the spiral arms being triggered when the inflowing molecular gas collides with the inner ring molecular gas . therefore it is reasonable that in a relative sense sfe@xmath119 is lower away from the central ring . however this leaves unexplained why gmc b s sfe@xmath119 is so much larger than gmc c , though its positions at the arm / inner ring intersection is the same . we suggest this is a sign of the evolution of star formation across the nucleus that is impacted by radiative and mechanical feed - back _ from within _ the molecular cloud . a possible cause of the different sfe@xmath119 between gmc c and b is that we are observing the clouds at high enough spatial resolution to begin to identify the changing internal structure of the clouds in the presence of the starburst . over the lifetime of a cloud sfes vary . in the earliest stages of a star formation episode sfes will appear low because elevated star formation rates have yet to convert the bulk of the molecular material to stars . towards the final stages of a gmcs evolution instantaneous sfes appear to increase dramatically as the cloud clumps are consumed , destroyed or dispersed . so observed instantaneous sfes are expected to vary widely throughout the lifetime of an individual gmc and relative to lifetime averaged sfes typically considered in extragalactic studies . if starburst b is a few myrs more evolved than the other gmcs , especially the dynamical similar c , then we may be witnessing the consumption , dispersal or destruction phase of the remaining dense clumps in the presence of the expanding hii region . the magnitude of the dense gas consumption times for b are indeed shorter than the lifetime of the gmc . several lines of evidence suggest that the ( weaker ) star formation towards c2 may be at a somewhat earlier phase . these include less extended hii regions ( e.g. * ? ? ? * ) , bright hot core - like species ch@xmath0oh @xcite , nh@xmath0(6,6 ) @xcite and co(65 ) ( e.g. * ? ? ? * ) , and more mm dust continuum emission ( e.g. * ? ? ? * ) . in this context it is interesting to compare the thermal pressure of the starburst hii region with that of the dense clumps along the same line of sight . assuming a strmgren sphere of r@xmath133 pc , n@xmath135 s@xmath15 and t@xmath136 = 8000 k , parameters determined for the main starburst hii @xcite , the thermal pressure of such an hii region would be @xmath137t@xmath136 = @xmath138 @xmath6 k. this value is equal given the uncertainties to @xmath139t@xmath140 @xmath6 k for the dense component of gmc b , hinting at pressure balance between the dense clumps and the hii region . in addition , the filling factor of the hii region is larger than implied by the hc@xmath0n lvg analysis . so it is possible that hc@xmath0n emission towards gmc b comes from dense clumps embedded within the hii region . the above analysis suggests the following physical picture for the faint hc@xmath0n emission towards the starburst . the starburst towards b is more evolved . the hii region at gmc b has had time to expand , destroying or dispersing the dense gas , which is now in the form of smaller clumps and/or more diffuse gas . clumps that remain there must have high pressure to survive . near the younger star forming cloud gmc c ( particularly toward gmc c2 ) , the hii regions may just be developing , and have not had time to disperse the clouds . the dense clumps here would be more abundant and still present a hot core - like chemistry . the clouds further from the central are on average less dense and at the current epoch remain largely quiescent , except possibly d , where the second high excitation component could be associated with the presence of shocks @xcite . we have imaged the hc@xmath9n j=54 line in the nucleus of ic 342 with the vla and the hc@xmath0n(1615 ) line with the pdbi at @xmath141 resolution . these are the first maps of these transitions in an external galaxy . we have detected emission extended along the nuclear mini - spiral " in ( 54 ) and more concentrated emission in ( 16 - 15 ) , with relative abundance of @xmath142 . _ hc@xmath0n emission is not tightly correlated with star formation strength . _ dense gas _ excitation _ however , follows star formation more closely . gmc b , which is weak in all the hc@xmath0n lines , is relatively stronger in the higher j lines . lvg modeling indicates that the hc@xmath9n - emitting gas has densities of 10@xmath143 . in ic 342 , physical conditions of the densest component are fairly constant away from the immediate environment of the starburst , though beyond the central ring densities begin to fall . comparison with the c@xmath1o observations of @xcite reveal excitation temperatures similar to c@xmath1o values indicating either that the molecular gas is dense and cool ( t@xmath144 k ) or that there are multiple gas components where the densities and kinetic temperatures of each component conspire to give similar overall excitation . the strong over - prediction of the amount of gas mass present if densities are large and temperatures cool , favors a multi - component ism with at least two components beside the diffuse one seen in @xmath113co(21 ) . the actual ism is likely a continuum of cloud densities with different densities dominating different tracers . hc@xmath0n also differs in morphology from hcn(10 ) , with hcn(10 ) being much brighter towards the starburst . this is further evidence that there are multiple dense gas components . of particular interest , the starburst site ( gmc b ) exhibits the largest difference in intensity between hc@xmath0n ( both transitions ) and hcn(10 ) . the faintness of the hc@xmath0n here suggests that the brightness of hcn(10 ) is not due solely to large quantities of dense gas . a comparison between the gmcs with the largest star formation rates and similar dynamical environments , b and c , hint at an explanation . while gmc b shows higher excitation , the low brightness of this cloud indicates that it is composed of a relatively small amount of warm , dense clumps . the smaller amount of dense gas at the site of the strongest young star formation indicate high star formation efficiencies in the dense gas . towards gmc c , hc@xmath0n , ch@xmath0oh and nh@xmath0 are more intense and the fraction of millimeter continuum from dust is higher . this indicates that gmc c is in an early , less evolved ( hot core - like ) state . the extreme dense gas star formation efficiency observed towards gmc b reflects the fact the main burst is in a more evolved state . the dense clumps towards the starburst are being dispersed or destroyed in the presence of the hii region . the little dense gas remaining appears to be in pressure equilibrium with the hii region . the larger opacity of hcn(10 ) relative to hc@xmath0n elevates its brightness temperature in this warm gas and lowers its critical density permitting it to remain excited in the somewhat lower density component . we conclude that evla observations of hc@xmath0n(54 ) can be a powerful probe of dense , quiescent molecular gas in galaxies , and when combined with high resolution imaging of the higher j transitions of hc@xmath0n with current and upcoming mm interferometers ( like alma ) provide tight constraints on dense molecular gas properties in stronger or more widespread starbursts , where changes like those localized to gmc b are expected to permeate much of the ism . we thank chris carilli and miller goss for assistance with the vla observations . we also thank philippe salome and nemesio rodriguez for their help with the pdbi observations and data reduction . we thank the referee for several suggestions that improved the presentation . d.s.m . acknowledges support from the national radio astronomy observatory which is operated by associated universities , inc . , under cooperative agreement with the national science foundation and nsf grant ast-1009620 . this work is also supported by nsf grants ast-0307950 and ast-0506469 to j.l.t . aalto , s. , polatidis , a. g. , httemeister , s. , & curran , s. j. 2002 , , 381 , 783 aladro , r. , martn - pintado , j. , martn , s. , mauersberger , r. , & bayet , e. 2011 , , 525 , a89 becklin , e. e. , gatley , i. , mathews , k. , neugebauer , g. , sellgren , k. , werner , m. k. , & wynn - williams , c. g.1980 , , 236 , 441 brandl , b. r. , et al . 2006 , , 653 , 1129 bussmann , r. s. , et al . 2008 , , 681 , l73 costagliola , f. & aalto , s. , 2010 , , 515 , 71 dahmen , g. , huttemeister , s. , wilson , t. l. , & mauersberger , r. 1998 , , 331 , 959 de vicente , p. , martn - pintado , j. , neri , r. , & colom , p. 2000 , , 361 , 1058 downes , d. , radford , s. j. e. , giulloteau , s. , guelin , m. , greve , a. , & morris , d. 1992 , , 262 , 424 fuente , a. , martin - pintado , j. , cernicharo , j. , & bachiller , r. 1993 , , 276 , 473 gao , y. & solomon , p. m. 2004 , , 606 , 271 garca - burillo , s. , martn - pintado , j. , fuente , a. & neri , r. 2000 , , 355 , 499 garca - burillo , s. , martn - pintado , j. , fuente , a. & neri , r. 2001 , , 563 , l27 garca - burillo , s. , martn - pintado , j. , fuente , a. , usero , a. & neri , r. 2002 , , 575 , l55 graci - carpio , j. , garca - burillo , s. , planesas , p. , fuente , a. , & usero , a. 2008 , , 479 , 703 green , s. & chapman , s. 1978 , , 37 , 169 harris , a. i. , hills , r. e. , stutzki , j. , graf , u. u. , russell , a. g. & genzel , r. , 1991 , , 382 , l75 harrison , a. , henkel , c. , & russell , a. 1999 , , 303 , 157 ho , p. t. p. , martin , r. n. , & ruf , k. 1982 , , 113 , 155 hunter , s. d. et al . 1997 , , 481 , 205 ishizuki , s. , kawabe , r. , ishiguro , m. , okumura , s. k. , morita , k .- i . , chikada , y. , & kasuga , t. 1990 , , 344 , 224 kennicutt , r. c. , jr . 1998 , , 36 , 189 knudsen , k. k. , walter , f. , weiss , a. , bolatto , a. , riechers , d. a. , & menten , k. 2007 , , 666 , 156 krips , m. , neri , r. , garca - burillo , s. , martn , s. , combes , f. , graci - carpio , j. , & eckart , a. 2008 , , 677 , 262 karachentsev , i. d. 2005 , , 129 , 178 lafferty , w. j. & lovas , f. j. 1978 , j. phys . chem . ref . data , 7 , 441 larson , r. b. 1981 , , 194 , 809 lo , k. y. et al . 1984 , , 282 , l59 maloney , p. , & black , j. h. 1988 , , 325 , 389 mauersberger , r. , henkel , c. & sage , l. j. 1990 , , 236 , 63 meier , d. s. & turner , j. l. 2001 , , 551 , 687 meier , d. s. , & turner , j. l. 2004 , , 127 , 2069 meier , d. s. & turner , j. l. 2005 , , 618 , 259 meier , d. s. , turner , j. l. & hurt , r. l. 2000 , , 531 , 200 meier , d. s. , turner , j. l. , & hurt , r. l. 2008 , , 675 , 281 montero - castao , m. , herrnstein , r. m. , & ho , p. t. p. 2006 , , 646 , 919 morris , m. , turner , b. e. , palmer , p. & zuckerman , b. 1976 , , 205 , 82 myers , p. c. , & benson , p. j. 1983 , , 266 , 309 narayanan , d. , cox , t. j. , shirley , y. , dav , r. , hernquist , l. , & walker , c. k. 2008 , , 684 , 996 papadopoulos , p. p. 2007 , , 656 , 792 rickard , l. j , & harvey , p. m. 1984 , , 89 , 1520 saha , a. , claver , j. & hoessel , j. g. 2002 , , 124 , 839 sakamoto , s. 1996 , , 462 , 215 schinnerer , e. , bker , t. , & meier , d. s. 2003 , , 591 , l115 schinnerer , e. , bker , t. , meier , d. s. , & calzetti , d. 2008 , , 684 , l21 schulz , a. , gsten , r. , kster , b & krause , d. 2001 , , 371 , 25 scoville , n. z. , & sanders , d. b. 1987 , in interstellar processes , ed . d. j. hollenbach and h. a. thronson , jr . , ( dordrecht : reidel ) , 21 scoville , n. z. , yun , m. s. , & bryant , p. m. 1997 , , 484 , 702 smith , p. a. , brand , p. w. j. l. , mountain , c. m. , puxley , p. j. , & nakai , n. 1991 , , 252 , 6 solomon , p. m. , rivolo , a. r. , barrett , j. , & yahil , a. 1987 , , 319 , 730 stark , a. a. , bally , j. , wilson , r. w. , & pound , m. w. 1989 , in the center of the galaxy , ed . m. morris , ( dordrecht : kluwer ) , 129 strong , a. w. et al . , 1988 , , 207 , 1 tsai , c .- w . , turner , j. l. , beck , s. c. , crosthwaite , l. p. , ho , p. t. p. & meier , d. s. 2006 , , in press turner , j. l. , & ho , p. t. p. 1983 , , 268 , l79 turner , j. l. , hurt , r. l. 1992 , , 384 , 72 turner , j. l. , hurt , r. l. , & hudson , d. y. 1993 , , 413 , l19 usero , a. , garca - burillo , s. , fuente , a. , martn - pintado , j. & rodrguez - fernndez , n. j. 2004 , , 419 , 897 usero , a. , garca - burillo , s. , martn - pintado , j. , fuente , a. , & neri , r. 2006 , , 448 , 457 vanden bout , p. a. , loren , r. b. , snell , r. l. & wootten , a. 1983 , , 271 , 161 wall , w. f. 2007 , , 379 , 674 wu , j. , evans , n. j. , ii , gao , y. , solomon , p. m. , shirley , y. l. , & vanden bout , p. a. 2005 , , 635 , l173 zinchenko , i. , pirogov , l. , & toriseva , m. 1998 , , 133 , 337 lcccccccc a & [email protected] & [email protected] & 22@xmath1452 & 29@xmath1455 & [email protected] & [email protected] & 16@xmath1453 & 44@xmath14510 + b & @xmath125.3 & @xmath121.0 & & & [email protected] & [email protected] & 19@xmath1453 & 28@xmath14510 + c1 & [email protected] & [email protected] & 48@xmath1451 & 39@xmath1453 & [email protected] & [email protected] & 47@xmath1451 & 43@xmath1454 + c2 & [email protected] & [email protected] & 50@xmath1451 & 35@xmath1454 & [email protected] & [email protected] & 45@xmath1451 & 47@xmath1454 + c3 & [email protected] & [email protected] & 49@xmath1451 & 31@xmath1455 & [email protected] & [email protected] & 45@xmath1451 & 47@xmath1454 + d & [email protected] & [email protected] & 51@xmath1456 & 58@xmath14520 & @xmath121.0 & @xmath122.5 & & + d & [email protected] & [email protected] & 52@xmath1452 & 40@xmath1457 & @xmath1461.0 & [email protected] & 57@xmath1458 & 44@xmath14526 + e & [email protected] & 4.4@xmath1451 & 12@xmath1452 & 28@xmath1457 & @xmath121.0 & @xmath122.5 & & + [ tinti ] lcccccccc a & [email protected] & [email protected] & [email protected] & [email protected] & 13@xmath1454 & 4(22 ) & 1(-9 ) & 4.4(5 ) + b & 2.5@xmath1451 & 37@xmath14519 & @xmath530.52 & @xmath5318 & 19@xmath1458 & 4(22 ) & @xmath78(-10 ) & 7.6(5 ) + c1 & [email protected] & [email protected] & [email protected] & [email protected] & 8@xmath1453 & 7(22 ) & 3(-9 ) & 6.8(5 ) + c2 & [email protected] & [email protected] & [email protected] & [email protected] & 8@xmath1453 & 1(23 ) & 2(-9 ) & 1.0(6 ) + c3 & [email protected] & [email protected] & [email protected] & [email protected] & 8@xmath1453 & 5(22 ) & 1(-9 ) & 4.9(5 ) + d & @xmath1460.15 & @xmath135 & @xmath120.17 & @xmath1213 & 6@xmath1454 & 2(22 ) & 3(-9 ) & 6.4(5 ) + d & @xmath130.24 & @xmath136 & [email protected] & [email protected] & @xmath136 & 6(22 ) & 1(-9 ) & 5.8(5 ) + e & [email protected] & [email protected] & @xmath120.057 & @xmath1210 & 7@xmath1453 & 6(22 ) & 1(-9 ) & 1.1(6 ) + [ trat ] lccccccc a & 10 & 5.15 & 0.091 & 38 & 4.7(6 ) & 11 & 6.16 + & 30 & 4.63 & 0.051 & 6.9 & 6.0(5 ) & 1.4 & 6.11 + & 50 & 4.39 & 0.067 & 3.1 & 5.2(5 ) & 1.2 & 6.09 + b & 10 & 5.75 & 0.015 & 75 & 1.3(6 ) & 1.7 & 6.75 + & 30 & 5.22 & 4.7(-3 ) & 14 & 6.5(4 ) & 0.084 & 6.70 + & 50 & 4.87 & 6.0(-3 ) & 5.5 & 4.2(4 ) & 0.055 & 6.57 + & 70 & 4.81 & 5.4(-3 ) & 3.6 & 3.1(4 ) & 0.041 & 6.66 + c1 & 10 & 5.23 & 0.13 & 41 & 9.7(6 ) & 14 & 6.23 + & 30 & 4.69 & 0.066 & 7.4 & 1.0(6 ) & 1.5 & 6.17 + & 50 & 4.50 & 0.075 & 3.6 & 8.0(5 ) & 1.2 & 6.19 + c2 & 10 & 5.36 & 0.12 & 48 & 1.3(7 ) & 13 & 6.36 + & 30 & 4.80 & 0.056 & 8.4 & 1.0(6 ) & 1.0 & 6.28 + & 50 & 4.64 & 0.057 & 4.2 & 7.3(5 ) & 0.73 & 6.34 + c3 & 10 & 5.44 & 0.076 & 53 & 7.1(6 ) & 14 & 6.44 + & 30 & 4.86 & 0.033 & 9.0 & 5.3(5 ) & 1.1 & 6.34 + & 50 & 4.70 & 0.034 & 4.5 & 3.8(5 ) & 0.0.76 & 6.40 + d & 10 & @xmath124.49 & @xmath70.040 & & & & @xmath125.49 + & 30 & @xmath124.10 & @xmath70.040 & & & & @xmath125.58 + & 50 & @xmath123.88 & @xmath70.060 & & & & @xmath125.58 + d & 10 & & & & & & + & 30 & & & & & & + & 50 & & & & & & + e & 10 & 4.90 & 0.12 & 28 & 4.0(6 ) & 3.6 & 5.90 + & 30 & 4.36 & 0.096 & 5.1 & 8.3(5 ) & 0.75 & 5.84 + & 50 & & & & & & + [ tlvg1 ] lccccccc a & 30 & 2.6 & 0.015 & 3.01 & 1.1 & 0.62 & 4.0(45 ) + & 50 & 2.4 & 0.017 & 2.82 & 0.55 & 0.63 & 4.0(45 ) + b & 30 & 2.8 & 3.2(-4 ) & 2.83 & 0.87 & 0.078 & 5.2(46 ) + & 50 & 2.5 & 4.6(-4 ) & 2.54 & 0.37 & 0.097 & 4.2(46 ) + & 70 & 2.4 & 4.0(-4 ) & 2.44 & 0.24 & 0.092 & 4.4(46 ) + c1 & 30 & 2.4 & 0.017 & 3.03 & 1.1 & 0.77 & 1.9(45 ) + & 50 & 2.1 & 0.021 & 2.90 & 0.56 & 0.84 & 1.8(45 ) + c2 & 30 & 2.4 & 0.021 & 3.03 & 1.1 & 0.77 & 2.1(45 ) + & 50 & 2.1 & 0.013 & 2.87 & 0.54 & 0.83 & 1.9(45 ) + c3 & 30 & 2.3 & 5.0(-3 ) & 2.80 & 0.84 & 0.69 & 4.1(45 ) + & 50 & 2.1 & 6.3(-3 ) & 2.64 & 0.42 & 0.72 & 4.0(45 ) + d & 30 & @xmath122.5 & & & & & + & 50 & @xmath122.3 & & & & & + d & 30 & & & & & & + & 50 & & & & & & + e & 30 & 2.3 & 0.030 & 2.94 & 0.99 & 0.78 & 1.6(45 ) + & 50 & 2.0 & & & & & + [ tlvg2 ] |
using the refurbished 305 m arecibo gregorian telescope , we detected 43 low surface brightness ( lsb ) galaxies from the catalog of oneil , bothun , & cornell ( 1997a ) . the detected galaxies range from 22.0 mag arcsec (0 ) 25.0 mag arcsec , with colors ranging from the blue through the first detection of a very red lsb galaxies ( bv =.7 to 1.7 ) . the m/l of these galaxies ranges from 0.1 / 50 / , showing this sample to range from very gas poor to possibly the most gas rich galaxies ever detected . one of the more intriguing results of this survey is that the galaxies with the highest m/l correspond to some of the reddest ( optically ) galaxies in the survey , raising the question of why star formation has not continued in these galaxies . since the average h i column density in these systems is above the threshold for massive star formation , the lack of such may indicate that these galaxies form some kind of `` optical core '' which trace a much more extended distribution of neutral hydrogen . alternatively , a model in which no stars more massive than two solar masses form in these systems can explain the presence of both blue and red gas rich lsb galaxies . moreover , under this model the baryonic mass fraction ( ) of lsb galaxies is the same as for galaxies of higher surface brightness , thus perhaps escaping the dilemma proposed by mcgaugh & de blok ( 1997a , 1997b ) with respect to lsb and high surface brightness galaxies defining the same tully - fisher relation . a subset of the detected lsb galaxies have rotational velocities 200 km s and yet are at least an order of magnitude below l in total luminosity . as such , they represent extreme departures from the standard tully - fisher relation . in fact , our sample does not appear to have any significant correlation between the velocity widths and absolute magnitudes , with only 40% of the galaxies falling within the 1 low surface brightness galaxy tully - fisher relation . overall , the discovery of very red , sub - l but very gas rich lsb galaxies in the nearby universe has increased , once again , the overall parameter space occupied by disk galaxies . = -0.25 in = -0.25 in = -0.25 in = 6.5 in = 8.5 in 0.2 in | astro-ph9909129 | our knowledge of the local galaxy population , and by inference the total galaxy population , has been biased towards the bright , high surface brightness ( hsb ) galaxies found in our current galaxy catalogs . low surface brightness ( lsb ) galaxies , those systems with central surface brightness lower than the natural sky brightness can not , however , be ignored , as they are both a significant contribution to the total mass found in galaxies and are critical to our understanding of the distribution of galaxy types . without adequate representation of the lsb class of galaxies we can not fully uncover the range of evolutionary paths available for galaxies to follow . lsb galaxies are important in a number of other contexts as well : * while the overall space density of lsb disks remains unknown , significant numbers of galaxies with @xmath2(0 ) @xmath13 mag arcsec@xmath0 have been detected in ccd surveys , suggesting that the trend remains fairly constant ( oneil _ et.al . _ 1997b ) or even rises towards fainter @xmath14 ( schwartzenberg _ et.al . _ 1995 , dalcanton _ et.al . _ 1997 ) . as a result , there is a strong possibility that lsb galaxies are the major baryonic repository in the universe ( i.e. impey & bothun 1997 ; oneil & bothun 1999 ) . * increasing the space density of galaxies at z = 0 , in principle , poses a feasible solution to the dilemma of the apparent excess of faint blue galaxies at intermediate redshifts . while there is much to be worked out in the details of such a solution ( see ferguson & mcgaugh 1995 ) , the discovery of faint , red , z=0 lsb spirals by oneil _ et.al . _ ( 1997b ) is promising in this regard . * measurements of the lsb rotation curves by pickering _ et.al . _ ( 1997 ) and mcgaugh & de blok ( 1997a ) have yielded the result that the fundamental shape / mass density of the dark matter halo appears to be different in lsb galaxies compared to hsb galaxies of the same circular velocity . additionally , these rotation curves have shown that the baryonic mass fractions ( @xmath8 ) in lsb galaxies is , on average , a factor of 3 lower than hsb galaxies of the same circular velocity . these observations suggest that lsb and hsb disks are fundamentally distinct with the distinction being physically defined by differences in their respective halos . although a coherent picture of the underlying mechanism of lsb formation and evolution has not yet been found , previous studies have shown that most of the initial , and somewhat naive , concepts of lsb galaxies are at odds with the observations . for example , the idea that lsb galaxies are `` faded '' versions of hsb galaxies due to the dimming of an aging stellar population is shown by the optical colors , metallicity , and gas content of lsb galaxies to be completely wrong ( bell , _ _ 1998 ; mcgaugh 1994 ; schombert , _ et.al . _ 1990 ; van der hulst , _ et.al . _ 1993 ) . in addition , the measured low h i surface densities ( van der hulst _ et.al . _ 1993 ) in lsb galaxies raise the important question : why are they any stars in these systems at all ( i.e. oneil _ et.al . _ 1998 ; mcgaugh & de blok 1997b ; oneil , verheijen & mcgaugh 1999 ) ? a recent survey by oneil , _ et.al . _ ( 1997a , b ) has extended the known range of properties of lsb galaxies , including the discovery of an intriguing new class of fairly red lsb disks whose overall color is , in fact , quite consistent with a conventional fading scenario . as this survey contains lsb galaxies with a wide range of properties , i.e. colors from the very blue through the very red , scales lengths ranging from 1 - 4 kpc , central surface brightnesses from near the freeman value ( @xmath2(0 ) = 22.0 mag arcsec@xmath0 ) through @xmath2(0 ) = 25.0 mag arcsec@xmath0 , further study of this catalog presents an excellent opportunity for understanding the global properties of lsb systems . in particular , the complete properties of the lsb galaxies discovered by oneil _ et.al . _ ( 1997a , b ) could not be ascertained as no redshift measurements existed . to further elucidate the nature of this new lsb sample , we undertook to detect the entire catalog of oneil , _ et.al . _ ( 1997a , b ) with the refurbished arecibo gregorian telescope . the layout of this paper is as follows : in section 2 the observations , data reduction , and analysis is discussed . section 3 describes the data analysis and presents the basic observational parameters which can be derived from the radio and optical observations . section 4 briefly describes some new inferences on the star formation history of lsb spirals that can now be made with the discovery of red , but very gas rich , lsb galaxies . finally , section 5 discusses how the types of lsb galaxies discussed here do not readily conform to the standard tully - fisher relation as defined by samples of cluster spirals . concluding remarks are contained in section 6 . using the refurbished arecibo gregorian telescope we attempted detection of the complete list of lsb galaxies in the catalog of oneil , bothun , & cornell ( 1997a ) ( obc from now on ) available to the arecibo sky . data was taken using the l - wide and l - narrow receivers , both with 2 polarizations , 1024 channel sub - correlators , and 9 channel sampling . each of the four sub - correlators has a 25 mhz bandpass , resulting in 5.2 km / s resolution ( at 1420 mhz ) . the l - narrow receiver ranged from 1370 mhz 1432.5 mhz and was centered at 1401.25 mhz , providing an overlap of two correlators across most of the spectrum and thereby effectively doubling the integration time . the l - wide receiver was centered at 1392.5 mhz with a range from 1350 mhz 1435 mhz , resulting in only a 5 mhz overlap on each side . in both cases the overlap more than eliminated any problems which otherwise may have arisen due to poor performance in the outer 50 channels of each correlator . during the observations the l - narrow receiver was set to linear polarizations ( a or b ) while the l - wide was set to circular polarization ( a @xmath15 _ i_b ) . the data were taken between 24 june 1998 and 27 april 1999 . a minimum of one 5 minute on / off pair was taken of each galaxy , followed by a 10 second on / off calibration pair . when possible , if a galaxy remained undetected after the first 5 minute pair and that galaxy was either particularly interesting ( very low @xmath2(0 ) or high b@xmath3v ) , and/or looked like a potential detection , at least one ( and on occasion many ) more 5 minute on / off pair was taken of it . table [ tab : obs ] lists the total number of observations done on each galaxy , as follows : column 1 lists the galaxy name ; columns 2 and 3 give the right ascension and declination ( in b1950 coordinates ) of the galaxies as determined from the digital sky survey ( dss ) ; columns 4 and 5 list the number of 5 minute pair taken with each receiver . if the pairs were of a different length of time ( i.e. 2 minutes ) , that time is given in parenthesis ; column 6 lists whether or not a galaxy was detected . all data was analyzed using the analyz software package ( deich 1990 ) . the two polarizations , as well as any overlapping channels were initially combined to determine whether or not a detection had occurred . once an initial detection was made , each polarization and ( when the detection lay within more than on sub - correlator ) each sub - correlator was analyzed to insure the galaxy could be retroactively detected within each system , providing a minimum of 2 4 semi - independent detections and minimizing the chance of false positives ( i.e. rfi noise ) . after a detection was confirmed , the two polarizations ( 3-channel boxcar and hanning smoothed ) of each observation of that galaxy were combined , and the resultant data displayed against correlator channel . to avoid possible problems resulting from a gain loss in the outer 50 channels of each sub - correlator , any detection which lay within 100 channels of a sub - correlator edge was not used , as there was always enough overlap in the sub - correlators for at least one ` clean ' galaxy image . otherwise each sub - correlator detection was analyzed individually and the results were averaged . in the cases that more than one observation of a galaxy existed , and the detection in question had a low s / n , the multiple observations were averaged before the above process was applied . if more than one observation of the galaxy under study was available yet the s / n was high enough within only one on / off pair for reliable analysis , each observation was analyzed separately with the averaged results being recorded in table [ tab : data ] . corrections had to be applied to the data to accommodate variations in the gain and system temperature of the telescope with zenith angle . the correction was determined through tracking a number of previously studied continuum sources the entire time they were visible to the arecibo sky ( za @xmath16 19.6@xmath17 ) . the obtained fractional temperatures ( @xmath18 ) were fit against the known values , as published in the nrao vla sky survey ( condon , _ 1998 ) . eight sources were observed with the l - narrow receiver and seven with the l - wide receiver between 01 june through 14 june , 1998 . fourth - order best fit curves were found for this data and the gain and temperature variance , with zenith angle , was obtained for each receiver . to insure any observed difference in g / t between the l - wide and l - narrow receivers was systematic , five of the continuum sources observed with the l - wide receiver were subsequently observed with the l - narrow , and the ratio of fractional temperatures versus zenith angle were determined . the result is a linear plot which shows slight systematic variation with zenith angle , l - wide having the lowest sensitivity at low zenith angles . the average of the plot for all five continuum sources is @xmath19 = 0.840 @xmath15 0.002 ( 3.0@xmath17 @xmath16 za @xmath16 18@xmath17 ) . as a final check , we interspersed observations of a subset of galaxies from the catalog of lewis , helou , & salpeter ( 1985 ) with the lsb galaxy survey described in this paper . on average our l - narrow results differed from those of lewis , helou , & salpeter by 2.9@xmath15 0.7% , those with the range in differences being @xmath32% to 10% of their value . the l - wide was comparable , with an average offset of 7.1@xmath15 1.5% and a range of @xmath312% 18% . see lewis , _ et.al . _ ( 1999 ) for further details . baselines were fit to the data using a modified version of the galpak base module , which fits a polynomial to all the data within a specified region . for this data we let the polynomial vary in order from 15 for the region lying within 100150 channels on either side of the galaxy edges . the most sensible best fit line ( typically of order 1 ) was then used as the baseline subtraction . again , that region within 100 channels of a sub - correlator edge was avoided for this analysis . once the baselines were subtracted , the velocities were corrected to the heliocentric coordinate system and the velocity , velocity widths , and the flux of the h i profiles were determined . in each case , the values were found four ways at 20% of the mean profile flux , at 50% of the mean profile flux , at 20% of the peak flux , and at 50% of the peak flux . ( as a fraction of the detected galaxies do not show a two - horned profile , no attempt was made to incorporate horns into the data analysis . ) during the flux analysis , galaxy edges were estimated by eye . although this introduces some measurement error , the error lies well below the 5 10 km / s width error which exists due to the somewhat low s / n in the 21-cm data . error analysis was done as follows as no difference existed between the fluxes determined using the four methods listed above , the determined flux errors are completely derived from the analysis of hsb galaxies in the catalog of lewis , helou , & salpeter ( 1985 ) . thus the l - narrow fluxes are given errors of @xmath15 5% while the fluxes determined from l - wide data have 10% errors . it is important to note that , as determined from the lewis , helou , & salpeter ( 1985 ) galaxies , it is possible that the flux errors are systematic , and result in our presenting too low a value for m@xmath4 . comparing our central velocity values with those of lewis , helou , & salpeter ( 1985 ) shows virtually no error ( the highest error being 1% ) , showing there is no significant contribution from systematic errors . the low s / n and lack of symmetry for the galaxies in this survey , though , did result in a typical scatter between the four different velocity measurements of @xmath15 2 km s@xmath0 , on average . if the error was larger , it is noted in table [ tab : data ] . the low s / n and asymmetric profiles of the lsb galaxies in this survey also resulted in considerable scatter in the velocity width measurements . comparing our results with lewis , helou , & salpeter ( 1985 ) suggests our data may have systematically wider profiles by 1 2 % . additionally , there was a typical scatter between the two width determinations ( peak and mean ) for each fraction ( 20% and 50% ) of 3% , resulting in a total 5% error for the widths ( again unless otherwise noted in table [ tab : data ] ) . as discussed in bothun & mould ( 1987 ) , errors in 21-cm line width measurements are likely the dominant source of observational scatter in the tully - fisher relation . this problem is made worse if most of the h i profiles do not have very steep sides . a quick perusal of figure 1 immediately suggests that line width errors are larger than `` normal '' for this sample . the information obtained from the observations described herein are listed in table [ tab : data ] and are arranged as follows . all 21-cm profiles are shown in figure [ fig : hiprof ] . * column 1 : * the galaxy name as it appears in obc . * column 2 : * the morphological type of the galaxy , as determined from the optical images of obc . * column 3 : * the heliocentric velocity of the galaxy in km s@xmath10 , determined as described above . the listed velocities have errors of @xmath15 5 km s@xmath10 , unless otherwise noted . * column 4 : * the apparent ( johnson ) b magnitude for the galaxies , as given in obc . the error , also as given in obc , is @xmath15 0.1 . * column 5 : * absolute b magnitude , determined using @xmath20 ( weinberg 1972 ) . for this , and all other derivations in this paper , h@xmath21=75 km s@xmath10 mpc@xmath10 , c=2.99793 x 10@xmath22 km s@xmath10 , and q@xmath21=0.1 . no k - correction was applied is it would be less than 0.1 mag . * columns 6 & 7 : * the average uncorrected velocity widths at 20% and 50% of the peak and mean flux . errors are @xmath15 10 km s@xmath10 , unless otherwise noted . * column 8 : * the total flux , after correction for gain and temperature variances with zenith angle . no correction for partial resolution was made although from the asymmetric appearance of some of the profiles ( e.g. p3 - 3 , p5 - 4 ) it is possible there was some degree of poor pointing . * column 9 : * the total h i mass , in 10@xmath23@xmath6 , as found from @xmath24 where the distance is in mpc and the flux ( @xmath25 ) is in jy km s@xmath10 . * column 10 : * the h i mass to luminosity ratio in units of @xmath6 /@xmath7 , where the luminosity was determined from the magnitude listed in column 4 . * column 11 : * the h i mass to luminosity ratio in units of @xmath6 /@xmath7 . in this case the luminosity was determined through the total integrated blue magnitude ( m(@xmath26 ) ) given in obc and converted to absolute magnitude using equation [ eqn : absm ] . * column 12 : * the b@xmath3v color as given in obc . * column 13 ; * the galaxy s optical inclination , determined from the obc data using the iraf ellipse parameter . the inclination listed is simply @xmath27 the inclination error is @xmath15 5@xmath17 unless otherwise noted . table [ tab : nondet ] lists all of the galaxies for which the attempts at detection failed during this survey . column 1 of the table gives the galaxy name , while columns 2 and 3 provide the 1@xmath12 detection limits for both the l - narrow and the l - wide set - ups , found from the r.m.s . error in the baselines . it is likely that were a galaxy to to lie at least 1.5@xmath12 above the baseline it would have been detected . it is important to note , however , that this simply limits the maximum h i density of the undetected galaxies _ if they lay within the 0 11,500 km s@xmath10 survey boundary . _ as many galaxies were found near the 11,500 km s@xmath10 survey boundary , it is quite possible that a planned search at higher redshift will readily detect many of the ` missing ' galaxies . the distribution of the galaxies in the survey follows the large - scale hsb galaxy distribution . figure [ fig : rav ] shows a cone diagram for all the galaxies in pegasus and cancer fields of obc with known velocities ( obtained either through this survey or through ned , the nasa extragalactic database ) . as can be readily seen , all the galaxies lie within the hsb defined galaxy clusters , and none of the detected galaxies fill in the cluster voids . the one exception is u1 - 4 ( not shown ) which appears to be a genuine isolated system . that none of the detected galaxies fill in the cluster voids is in agreement with both the result of schombert _ et.al . _ ( 1995 ) for lsb galaxies off the second palomar sky survey and the uppsala general catalog search by bothun , _ _ ( 1986 ) . the reader should also note that these galaxy clusters do not match the extreme regions of galaxy density , such as the coma or virgo clusters . we make no statement on existence of lsb galaxies in cluster cores , already shown to be rare by bothun _ et.al . _ ( 1993 ) . a follow - up study into the clustering statistics of all these galaxies ( oneil & brandenburg 1999 ) should clarify this issue . for now it suffices to note that no trend was found between a galaxy s velocity or cluster identity and its color , surface brightness , rotational velocity , inclination , and absolute magnitude . it is interesting to note that with the inclusion of our lsb galaxies , galaxy groups now found to contain the full range of galaxy types , from dwarfs to intrinsically luminous lsb galaxies , and from the very blue , star - forming galaxies to those galaxies with apparently old stellar populations . the galaxies for the sample span a range of morphological types from late - type spiral to dwarf irregulars . the hubble types are given in table [ tab : data ] , following the prescription of schombert , pildis , & eder ( 1997 ) and sandage & binggeli ( 1984 ) . this system is based mostly on the presence , and size of a bulge component . objects with a distinct bulge are classed sb to sc depending on bulge dominance . objects with a disk - like shape and some central concentration are classed as sm . objects with no shape or concentration are classed as i m . most of the objects with double - horned h i profiles , indicative of rotation , displayed disk - like morphological appearance . eight of the 43 objects ( 19% ) are classed as i m or dwarf - like . all of these dwarf galaxies have luminosities and sizes on the low end of the sample scale , justifying their classification as dwarf in terms of size and magnitude . several of the disk galaxies in the sample have magnitudes and scale lengths comparable to the class of dwarf spirals ( schombert _ et.al . _ 1995 ) , but many also have rotational masses similar to the most massive disk galaxies known ( see section 5 ) . this immediately suggests detection of objects that strongly violate the tully - fisher relation . figures [ fig : bcomp ] and [ fig : morph ] compare m@xmath5 with the total neutral hydrogen mass , color , central surface brightness , and scale length for all the galaxies detected with our survey . in figure [ fig : bcomp ] these results are compared with other samples ( de blok , van der hulst , & bothun 1995 ; de blok , van der hulst , & mcgaugh 1996 ; schombert , _ _ 1995 ; matthews , & gallagher 1997 ; bothun , sullivan , & schommer 1982 ; becker , _ et.al . _ 1988 ) . the most relevant comparison might be with bothun ( 1982 ) who reported the results for normal spirals in the pegasus and cancer regions . with respect to figure 3a , our sample clearly defines the same locus of points with the obvious exception of 3 outliers c1 - 2 , c5 - 5 , and n9 - 2 ( having extremely high m@xmath4/l@xmath5 ratios ) . however , most of the points sit above the ridge line fit to these samples , indicating that , on average , our sample galaxies have higher gas to star ratios . the situation summarized in figure 3b however is quite different . here we can see that our sample has clearly extended the color - magnitude relation for spiral galaxies with gas in that we have detected a number of red galaxies with luminosities below that of l@xmath11 . previous reports of very red gas - rich disk galaxies ( e.g. schommer & bothun 1983 , van der hulst _ et.al . _ 1987 ) have all been relatively luminous spirals . to our knowledge , this survey is the first to have discovered red , gas - rich objects of relatively low intrinsic luminosity . overall , however , there clearly is no color - magnitude relation defined for the late type samples shown here . these objects clearly define a very diverse galaxy population . in figure [ fig : morph]_a _ we plot the absolute b magnitude versus @xmath2(0 ) for the galaxies in this survey . by selection , there are no galaxies with @xmath2(0 ) brighter than 22.0 mag arcsec@xmath0 . the void in the top right corner is most likely artificial , created by the inability of obc to identify small , compact galaxies ( the ` star - like galaxies ' of arp 1965 ) . had high enough resolution images been taken by obc , that corner might disappear . indeed , that corner is often inhabited by blue compact dwarfs which have become the source of intense recent study ( e.g. marlowe _ et.al . _ 1999 ) . the more interesting void in figure [ fig : morph]_a _ is at the lower left corner . this is the region where very lsb yet intrinsically very luminous , and physically quite large , galaxies would lie ( the malin i galaxies ) . this lack of intrinsically luminous , very lsb galaxies can also be seen in figure [ fig : morph]_b _ , which shows a void in the very lsb , large scale length regime . this result either shows that the intrinsic space density of these physically large lsb disks is low ( but see sprayberry _ _ 1995a ) or that we have not sampled the correct regions to find them ; that is , they tend to avoid the extended cluster environment ( see hoffman , silk , & wyse 1992 ) are are only found in isolated , very low density environments . the use of the color - gas content plane as a diagnostic for studying the evolution / star formation history of galaxies was pioneered by tinsley ( e.g. tinsley 1972 ) . bothun ( 1982 ) present a codified version of this as it applies to cluster spirals . one of the main results of that analysis was that , for a sufficiently large sample of spirals , there really was no correlation between disk color and the gas - to - star ratio . figures [ fig : red ] _ a _ and _ b _ show the h i mass and m@xmath4/l@xmath5 versus b @xmath3 v color for the galaxies in our survey where we confirm the absence of any correlation . note the presence of some galaxies in our sample which have extreme values of m@xmath4/l@xmath5 . more to the point , our samples clearly shows that there is no definitive trend toward lower m@xmath4/l@xmath5 or m@xmath4 with redder colors . in fact , if a trend does exist , it is toward _ higher _ h i mass and m@xmath4/l with increasing b@xmath3v , raising an interesting question if a galaxy has been capable of forming and evolving at least one generation of stars to reach b@xmath3v @xmath281.0 , and the galaxy still contains a significant fraction of neutral hydrogen , why has it been unable to continue this process ? that is , why hasnt such a disk exhausted its gas and assumed its endpoint evolution in this diagram at the lower right where the bulk of s0 galaxies reside ? clearly , we have detected what could best be described as a `` dormant '' population of disk galaxies . one solution to the above questions may be found by appealing to kennicutt s ( 1989 ) argument that a critical density of gas must be present for star formation to occur . thus it could be argued that star formation in these red , high m@xmath4/l@xmath5 galaxies occurred only in those regions dense enough to support it , and that the bulk of the gas is located in the less dense regions which have been unable to commence self - gravitation . in this case , these galaxies would represent `` optical '' cores inside a much more extended gas distribution . certainly individual examples of this have been found ( e.g. ngc 2915 meurer _ et.al . _ 1996 ) , but these seem to be quite rare . indeed , a systematic search for such kinds of objects from a seed sample of blue compact dwarfs failed to detect any examples of large h i mass objects with small optical sizes ( i.e. salzer , _ et.al . _ 1993 ) . previous studies of red , h i rich disk galaxies ( e.g. van der hulst _ et.al . _ 1987 ) have shown them to be large scale length objects with generally strong color gradients . their overall properties imply that the bulk of the star formation has occurred within the inner two scale lengths and the gas at larger radii is mostly unprocessed . the red h i rich galaxies in this survey , however , are not of large scale length as shown in figure [ fig : mlbalp ] . in fact , for the extreme cases , it can be seen that the galaxies with m@xmath4/l@xmath5 @xmath28 9 @xmath6 /@xmath7 are confined to the @xmath26 @xmath16 2 kpc range . possibly this is an indication that these galaxies are indeed `` optical '' cores of a much larger h i extent . however , if their h i size is similar to their optical size then we have discovered a class of relatively compact galaxies which have an average column density of approximately 10@xmath29 @xmath6 /100 kpc@xmath30 = 10@xmath31 atoms per cm@xmath0 ( e.g. above the threshold ) , but which are not forming stars , or at least are not forming massive stars . it is this latter possibility that makes the discovery of these objects relevant to the apparent contradiction discussed by mcgaugh & de blok ( 1997a ) regarding the low @xmath8 value that results for lsb galaxies if they are fit with a standard dark matter halo . this low @xmath8 value is follows directly from assuming a low m / l for the stellar population . typically a value of ( m / l)@xmath5 of 12 is assumed as this is consistent with the relatively blue colors of the mcgaugh & de blok disks . but what if this is wrong ? what if we are dealing with an anomalous stellar population because of an anomalous imf ? suppose , for the sake of argument , that we force the @xmath8 to be the same for both lsb and hsb galaxies . one motivation for doing this is to better understand why it is that lsb disk galaxies apparently define the same tully - fisher ( tf ) relation as hsb disk galaxies ( e.g. sprayberry _ et.al . _ 1995b ; zwaan _ et.al . _ 1997 ) . this issue has been addressed by dalcanton _ et.al . _ ( 1997 ) , mo , mao , & white ( 1998 ) and steinmetz & navarro _ et.al . _ ( 1999 ) who can recover this `` universal '' tf relation by assuming that galactic disks are sub - maximal and it is the halo properties that therefore dominate the mass profile . support for this general picture comes from a recent study by courteau & rix ( 1999 ) who show that the tf relation and its scatter are best understood if the disks of hsb spirals are indeed sub - maximal . on the other hand , beauvais & bothun ( 2000 ) present fairly compelling evidence that disks are likely maximal ( see also sackett 1997 ; giraud 1998 ) . mo , mao , & white ( 1998 ) explore the efficiency of the conversion of baryonic mass into stars . variations in this efficiency can obviously lead to different loci of points in the tf relation . in the extreme case of no conversion into stars , obviously a galaxy has no luminosity to even place it on the tf relation ! mo , mao , & white ( 1998 ) find that surface mass density is a key parameter in determining this efficiency ( see also bothun 1990 ) and predict that low surface mass density systems have low efficiencies leading to the existence of relatively gas - rich but dim galaxies . such galaxies , are in fact , the kind that we have discovered here . however , the luminosity evolution of galaxies depends on factors other than the efficiency of gas to star conversion . there is the critical issue of the mass function of stars that are formed during the conversion process . while discussing whether or not there is a universal imf is well beyond the scope of this paper , we amplify the issue that was initially raised in mcgaugh and de blok ( 1998 ) at face value , the data seem to argue that two potentials of much different baryonic mass fraction , but similar circular velocity , produce the same amount of light ( hence making a universal tf relation ) . while this issue can be resolved using modified newtonian dynamics ( mond e.g. milgrom 1983 ) , that solution seems no more extreme than a solution which seeks to change the stellar population so that the baryonic mass fractions in lsb galaxies in fact , is the same as in hsb galaxies . for @xmath8 to be the same , the assumed ( m / l)@xmath5 for blue , lsb disks would have to be 10 , instead of 12 . while there may be some dynamical constraints of having ( m / l)@xmath5 this high ( quillen & pickering 1997 ) , values this large can not be explicitly ruled out for the objects in the de blok and mcgaugh ( 1998 ) sample . they are only indirectly ruled out by the standard ( m / l@xmath5)/@xmath33 arguments which flow from an assumed imf . therefore , one is obligated to produce a stellar population which basically has ( m / l)@xmath5 @xmath34 10 ( the value for a typical elliptical galaxy ) and @xmath35 @xmath34 0.3 . this is a tough problem and its solution requires an extreme stellar population , one that is presently deficient in stars on the giant branch , so that the high ( m / l)@xmath5 is provided by main sequence stars of mass less than 0.3 @xmath6 and the blue colors are supplied by a small amount of a and f stars . this population of a and f stars must remain sufficiently small so that their subsequent giant branch evolution does not significantly alter ( m / l)@xmath5 . note that a galaxy with such a stellar population will not remain dim forever , only for the first few billion years of its life . after 10 gyr , the heavily populated lower main sequence will begin to evolve onto the giant branch and the luminosity will increase . this extreme model , in fact , is observationally testable if near ir colors could be obtained . dwarf dominated integrated light will be significantly bluer than giant dominated integrated light in @xmath36 . we also note that it is very difficult to produce any stellar population with giant branch dominated light that has @xmath33 @xmath9 1.0 ( see tinsley 1978 ; bothun 1982 ) and ages that do nt exceed 12 gyrs . the reddest objects in our sample approach @xmath33 = 1.2 . since there is unlikely to be significant internal extinction in these systems ( see below ) , these very red colors imply either 1 ) a rather old galaxy ( unlikely ) or 2 ) the presence of very cool giants usually only seen in metal - rich populations ( again unlikely ) , or 3 ) a large contribution to the light from main sequence stars of spectral type k5 m0 . in the galaxies star forming ( on ) state it is blue . in the galaxies non - starforming ( off ) state it fades and reddens . in table [ tab : botmod ] , we present three models ( discussed in detail below ) which show the on and off states that can be achieved under our dwarf dominated integrated light scenario . while this scenario is clearly non - standard it does produce both red and blue gas - rich lsb galaxies with high values of @xmath37 in both cases . as such , this scenario provides an alternative to apparent significant discrepancy between the different @xmath8 values inferred for lsb and hsb disks . however , to be fair , in reaching this conclusion , mcgaugh & de blok ( 1997a , b ) actually compare the @xmath8 of lsb disks with clusters of galaxies , which typically have @xmath8 in the range 0.1 - 0.3 ( e.g. the baryon catastrophe see white _ et.al . _ 1993 ) . by contrast , lsb disks are factor of 3 - 5 lower . direct comparison to hsb disks , however , is more problematical as most rotation curves do nt go out far enough to accurately sample the halo . thus , @xmath8 depends upon the scale of measurement in the hsb sample . adding to the confusion are the results of zaritsky _ et.al . _ ( 1997 ) whose work on satellite galaxies and the total extent of the halos around hsb galaxies leads to @xmath8 0.05 , similar to those of lsb disks but significantly less than what is seen in clusters . so while its clear that @xmath8 for lsb disks is significantly lower than that measured for clusters of galaxies , it remains ambiguous whether or not @xmath8 is also significantly lower compared to individual hsb disks . this lack of clarity on the @xmath8 issue , however , does not deter us from pursuing our alternative stellar population model . indeed , in a rigorous but now largely forgotten paper , larson ( 1986 ) makes a set of compelling arguments about the need for a bi - modal imf to fully reconcile the chemical evolution of galaxies with their observed colors and mass - to - light ratios . one of those modes involves a peak at m @xmath34 0.25 @xmath6 and that is the mode we have adopted as the dominant mode of star formation in our models . to account for the presence of very blue lsb galaxies and the observation ( e.g. mcgaugh 1994 ) that some lsb galaxies have a small number ( e.g. 16 ) of h ii regions we have added a few o , a and f stars to this mode . however , as pointed out in de blok ( 1997 ) as well as oneil _ et.al . _ ( 1998 ) , there is ample evidence that lsb disks are generally deficient in the formation of massive stars . the models of oneil _ et.al . _ ( 1998 ) assert that this deficiency is a statistical result of a low surface density of gas and the difficulty of gathering sufficient gas mass over a relatively small spatial scale to produce the kind of massive star formation that is observed in hsb disks . under this set of assumptions we ran a large series of models and show the results for 3 specific cases in table [ tab : botmod ] . in each case , the on state is achieved by allowing @xmath34 2% of the available gas mass to to participate in star formation . values for @xmath33 and @xmath37 for the stellar population are listed for the on and off states . the extreme nature of these models is readily apparent as the giant branch contribution to the integrated light at b is never larger than 10% . in models 1 and 2 we truncate star formation at 2 @xmath6 so that only a and f stars are produced . the main difference between the two models is that model 2 peaks at a lower main sequence mass to drive the colors as red as possible . still , none of our models can reach the more extreme @xmath33 values in this sample ( e.g. @xmath33 @xmath9 1.2 ) . in model 3 , we add o - stars to model 2 in order to accommodate the existence of the observed h ii regions . this addition does not appreciable lower the @xmath37 for the stellar population but does push the @xmath33 color blueward by @xmath34 0.1 magnitude . since only a relatively small amount of gas is involved in the off to on transition , then our models suggest that both blue and red h i rich lsb galaxies can co - exist for billions of years , given their amounts of available h i. note as well that in all three cases , the integrated light in the on state is dominated by the small number of o , f and a stars . as the light from these stars fades , the luminosity of the galaxy will decrease by a factor of 3 - 4 and the galaxy will return to its red off state . in fact , if in its on state the galaxy falls on the standard tf relation , it will surely be well below this position as it transitions to the off state . in the next section , we will show that these red h i rich lsbs are significantly under - luminous with respect to their circular velocity . we have pursued this alternative stellar population model strictly as means of determining the plausibility of producing a stellar population which can be blue ( @xmath33 @xmath34 0.4 ) but which can have a significantly higher @xmath37 ratio so as to raise the value of @xmath8 for lsb disks to the level observed for clusters of galaxies . the models presented in table 4 achieve this and we leave it up to the reader to assess their plausibility in comparison to other alternatives ( e.g. a non - universal value for @xmath8 , mond , etc ) . clearly , further resolution of this intriguing matter requires actual measurements of the h i distributions in these red h i rich objects . either the gas is well below the critical density thus effectively shutting off star formation , or star formation is proceeding without the accompanying massive star formation . while that would be very strange , we argue that the position of these galaxies in the color - gas content plane is already very strange so something quite different is likely going on . the fact that we can produce blue lsb galaxies in a model in which only 2% of the gas supply needs to participate in a star formation event is consistent with what appears to be occurring in ugc 12695 , perhaps the bluest lsb detected to date . ugc 12695 , in fact , has a total h i mass remarkably near its total dynamical mass ( @xmath38@xmath6 , m@xmath39@xmath6 ) , yet recent hst wide field planetary camera-2 and vla observations show that the majority of star formation in ugc 12695 is occurring in isolated regions . thus , although ugc 12695 has a high total h i mass , the surface density of the gas is often below the kennicutt star formation threshold ( oneil , verheijen , & mcgaugh 1999 ; oneil , _ _ 1998 ) . if the light from this small amount of star formation fades , its quite likely that ugc 12695 would be a red , lsb , h i rich object . in 1977 tully & fisher published an empirical correlation between galaxies h i velocity width and their absolute magnitude ( or diameter ) . since that time the tully - fisher ( tf ) relation has been used in over 100 different investigations that attempt to determine the relative distances between disk galaxies and/or clusters of galaxies . this multitude of different investigations clearly establishes that the tf relation exists in a variety of different samples . the recognition that the potentials of disk galaxies are dark matter dominated makes the existence of the tf relation very difficult to understand from first principles because , apparently , it requires that the amount of luminosity trace the dark matter in a way that scales exactly with @xmath40 . this fine tuning , in essence , requires that the value of @xmath8 in disk galaxy potentials be nearly constant . the required constancy of @xmath8 should be a troubling point . the existence of the tf relation implies that just enough baryonic gas creeps into these dark potentials and then has the correct star formation history to produce the right amount of light for a given circular velocity . worse still is the apparent physical dilemma or paradox that results from the apparent ability for samples of lsb and hsb spirals to define the same tf relation . simple virial arguments coupled with the constancy of @xmath37 fail to reproduce this observation if lsb and hsb disk galaxies are dominated by dark matter halos of the same form ( see bothun & mcgaugh 1999 ) . the only way out of this dilemma is to postulate that fine tuning exists between central surface brightness and @xmath37 in such a way so as to preserve a universal tf relation . in this section , however , we will document how our , relatively extreme sample of lsb galaxies , in fact does not conform well to the standard tf relation . while hints of this have been seen in other samples ( e.g. matthews , van driel , & gallagher 1998 ) those hints are mostly based on behavior at the low line width end . our sample contains galaxies with substantial circular velocity that nevertheless strongly deviate from the tf relation . to compare our results with those of previous galaxy surveys , we modeled our magnitude and line width corrections after those of tully & fouque ( 1985 ) and zwaan , _ et.al . _ ( 1995 ) . thus we corrected the magnitudes for galactic and internal extinction to face - on orientation , using @xmath41 with @xmath42=0.55 and _ f_=0.25 . the line widths were corrected for random motion effects and inclination , using @xmath43 with @xmath44 and w@xmath45 , the turbulent motion parameter , being set to 14 km s@xmath10 . figure [ fig : tf](a ) shows w@xmath46 versus b@xmath47 with the 1@xmath12 and 2@xmath12 ranges of zwaan , _ et.al . _ overlaid on the plot . the zwaan , _ et.al . _ data , based off broeils ( 1992 ) findings , is @xmath48 only 77% of the data falls within the 2@xmath12 lines of zwaan , _ et.al . _ , and a mere 40% fall within the 1@xmath12 lines . clearly this sample does not produce a well - defined tf relation for lsb galaxies in the way that the sample of zwaan _ et.al . _ does . before delving into the significance of figure [ fig : tf](a ) , a further look into potential errors is required . the first question which needs to be addressed is that of potential biases introduced by above corrections , as all were done as if this sample was comprised of hsb galaxies . the application of these `` corrections '' however , to lsb galaxies may be somewhat dubious . to begin with , there is a fair bit of evidence that suggests lsb galaxies have very low amounts of internal extinction and may , in fact , be extinction free . for instance , sprayberry _ et.al . _ ( 1995b ) did not apply any extinction corrections to the magnitude and formed a tf sample with a slope very near the virial slope of @xmath310 . the extinction correction assumes the galaxies are at least moderately optically thick . in figure 8 we plot the b@xmath3v vs b@xmath3i colors of the galaxies in obc . a nice locus of points is found whose position matches simple expectations of stellar population models . where the red galaxies driven so by excessive amounts of dust we would observe significant shifts away from this relation . the tightness of the two color diagram thus provides another indication that lsb galaxies are relatively dust - free ( figure [ fig : bvbi ] ) . additionally , it should be noted that the six most deviant points in figure [ fig : tf](a ) ( p1 - 3 , p9 - 4 , c4 - 1 , c5 - 5 , c6 - 1 , c8 - 3 ) range in color from b@xmath3v = 0.5 1.0 and in inclination from 45@xmath17 through 73@xmath17 . it is therefore likely that , if anything , the extinction corrections applied to the data are larger than necessary , and the true b magnitudes are fainter than shown in figure [ fig : tf](a ) . this , then , would shift the data points downward , and farther from the tully - fisher relation . perhaps a better gauge of the departure of our lsb sample from the standard tf relation is provided in figure [ fig : tfi ] which shows the i - band tully - fisher relation for the data in this paper , with the data and @xmath49 lines from pierce & tully ( 1988 ) overdrawn . from this it can be seen that the scatter is as deviant in the i band as in the b for our data . it should be noted that the two data points ( u1 - 4 , and i1 - 1 ) which lie above pierce & tully s tf relation in figure [ fig : tfi ] all belong to galaxies which are not a part of any known cluster and hence we can not assign some mean `` cluster '' distance to these galaxies as we can for the bulk of the objects in our sample ( see figure 2 ) . the next correction was done to compensate for the inclination of the galaxies when trying to determine their rotational motion . this correction is fairly straightforward provided the true inclination of a galaxy is known . the inclination of the galaxies in this survey were determined by obc assuming their true ( face - on ) shape is a perfect circle . lsb galaxies tend to be more amorphous in appearance than their hsb counterparts , however , making optical determinations of their true inclination difficult . by their very nature , lsb galaxies are difficult to fit isophotes to and there is a fair amount of variation in position angle and ellipticity between adjacent isophotes . while some of this noise averages out , there is likely an increase scatter in the tf relation for this sample just as a result of `` inclination '' noise . it is unclear , however , if there is any systematic effect present in applying the inclination correction to the measured line width . it should be noted that the @xmath15 5% error in inclination is accounted for in the error bars of figure [ fig : tf ] . the final correction undertaken was to compensate for the random motion of the gas contained within each galaxy . this is the most uncertain of all the corrections , as the parameter which describes the importance of turbulent motion , w@xmath45 , is unknown for lsb galaxies . for hsb galaxies , w@xmath45 appears to be of order 10 - 15 km / s , averaged over the entire disk ( see data in beauvais & bothun 1999 ) . for lsb galaxies one might expect w@xmath45 to be larger due to the lower restoring force perpendicular to the disk ( due to low surface mass density ) . on the other hand , w@xmath45 also is a measure of the kinematic feedback of star formation to the gas and lsb galaxies have quite low current star formation rates . thus it is possible that w@xmath45 is either significantly greater than ( or smaller than ) that used in the above equations , which would shift the points in figure [ fig : tf ] to the left ( or right ) . since the application of our corrections to lsb galaxies may have limited validity , it seems worth just plotting the data without the corrections . this is done in figure [ fig : tf](b ) which shows the same galaxies without the corrections for inclination , internal extinction , and random motions ( only the galactic extinction correction remains ) . the situation with respect to the uncorrected data , in comparison with the lines of zwaan , _ _ ( 1995 ) , is very similar to that in the corrected data , strongly indicating that the observed ( large ) deviations from the tf relation are real , and not the result of large systematic errors associated with applying the corrections . the most obvious question to raise in regards to our results is why our survey did , and previous surveys did not , find galaxies deviating so widely from the tully - fisher relation . one possibility is simply the sensitivity of our survey and the size of this lsb sample . the combination of the large number of data points and the increased sensitivity of the arecibo gregorian system make for the ready detection of low s / n galaxies ( i.e. p6 - 1 and p1 - 3 ) . the second part of the answer to the above question is that the discovery of a group of galaxies which do not follow the tully - fisher relation is not entirely unanticipated . for instance , it should be clear that when a galaxy is a big bag of baryonic gas and has not produced any stars ( light ) yet , that it can not lie on the tf relation . so , one naively expects that galaxies with high fractional gas contents should systematically be under - luminous for their line width and hence fall - off the tf relation . however , no sample to date really shows this , and attempts to look for a correlation of the residuals with gas - content ( e.g. mcgaugh & de blok 1997b ) have not produced anything convincing . what is unique about our data set is the appearance of galaxies with relatively large line widths that are deviating from the tf relation in this expected sense ( e.g. figure 9 ) . in general those are the points with @xmath50 @xmath9 200 km sec@xmath10 . at the lower line width end , a number of studies have showed that the dispersion around the tf relation starts to increase and some galaxies are located well below the nominal relation ( e.g. matthews , van driel , & gallagher 1998 ) . in general , our sample has h i profile morphologies consistent with rotating disks ( e.g. we see 2 horns ) . some low line width galaxies show gaussian or triangle profile morphologies . inclusion of such galaxies invariably increases the scatter at the faint end because it is unclear if there is even a kinematic disk present that is rotating at some angle with respect to the line of site . so one should nt pay to much attention to the behavior in the tf plane for those galaxies with rotational line widths @xmath51 80 km sec@xmath10 . in figure [ fig : resid ] we plot the residuals from figure [ fig : tf]_a _ as a function of m@xmath4/l@xmath5 . a definitive trend toward the more h i rich objects lying the farthest off the tully - fisher relation is evident . this reflects a substantial reservoir of unprocessed gas and hence these potentials are under - luminous for their rotational velocity . while this is completely obvious , this is really the first sample where this effect is directly seen . this , of course , does nt mean that there is nt a tf relation , it just means that the tf relation as defined by the luminosity versus circular velocity plane is not appropriate . if we replace luminosity by baryonic mass then there will be a large correction that is driven by the gas mass fraction . discussion of the existence of a `` baryonic '' tully - fisher relation , however , is beyond the scope of this paper . however , de blok _ et.al . _ ( 1999 ) have undertaken a systematic investigation of the baryonic tf relation to conclude a ) that such a relation exists and b ) it has relatively low scatter . under the idea of constant @xmath8 per potential , a good baryonic tf relation must exist . galaxies with high gas mass fraction must therefore naturally depart from the l - v relation in the sense that is finally observed in our sample . the previous lack of detection of galaxies lying outside the 2 @xmath12 range of the h i tully - fisher relation in figure [ fig : tf ] thus appears to be due to selection effects similar to those which created the ` freeman law ' and delayed the detection of lsb galaxies . also akin to the original optical non - detection of lsb galaxies , these selection effects have been discussed in the literature , most notably by briggs ( 1998 ) and disney & banks ( 1998 ) . the h i tully - fisher relation , then , like the freeman law , is a relation demonstrating our limited view of the universe , as only galaxies with similar gas mass fractions have been surveyed to date . since gas turns into stars , the evolution in surface brightness of any disk galaxy is equivalent to evolution in decreasing gas mass fraction . the difficulty of finding high gas mass fraction galaxies is therefore equivalent to the general difficulty of detecting galaxies of extremely low surface brightness . the fact that such galaxies are now shown to exist in the nearby universe is once again a testimony to the slow evolutionary rates of some galaxies ( e.g. bothun _ et.al . _ 1990 ) . were we to have done an all sky survey at 21-cm before any optical surveys its likely that a ) freeman s law would have never existed and b ) the tully - fisher relation would have been harder to find . as more observations are done , and as the sensitivity of telescopes increase , we predict that most all possible regions of the h i velocity width versus magnitude plot should become populated . indeed , one can imagine populating the upper left region simply as a result of a small starburst in one of these dormant galaxies . the sample of salzer _ et.al . _ ( 1999 ) may contain such objects , but of course , such objects are * not * lsb galaxies . using the refurbished arecibo gregorian telescope we detected 43 galaxies with 22.0 mag arcsec@xmath0 @xmath1 @xmath2(0 ) @xmath1 25.0 mag arcsec@xmath0 . the detected galaxies range in color from the very blue through the the first 21-cm detection of very red lsb galaxies , while their h i mass - to - luminosity ratios range from reasonably gas poor ( m@xmath4/l@xmath5 = 0.1 @xmath6 /@xmath7 ) to possibly the most gas rich galaxies ever detected ( m@xmath4/l@xmath5 = 46 @xmath6 /@xmath7 ) . analysis of the structural properties of these galaxies show a diverse population , ranging from dwarfs to intrinsically luminous systems , though no large , malin i - type galaxies were identified . we found no correlation between the galaxies color and m@xmath4/l@xmath5 , confirming earlier results for hsb galaxies . we have also discovered a contingent of red lsb galaxies with m@xmath4/l@xmath5 @xmath52 9 @xmath6 /@xmath7 . these galaxies are generally small and if the h i distribution is similar to the optical distribution then such red galaxies are in a regime where their is no massive star formation despite having an average column density above the critical value . either star formation is occurring without any massive star formation ( e.g. the model in table [ tab : botmod ] ) , or the h i distribution is very extended and we have merely detected the `` optical core '' of this extended gaseous distribution . finally , perhaps the most interesting discovery of this survey is the presence of large line width galaxies which are substantially under - luminous for their circular velocity and hence represent a significant departure from the tully - fisher relation . moreover , our sample shows a clear correlation between residuals from the tf relation and gas content in the expected sense . such a correlation exists primarily because most galaxies that form the tf relation have a very similar gas mass fraction and are in a similar evolutionary state in their history of gas - to - star conversion . strong deviations away from this relation , such as those exhibited by our sample , indicate these lsb galaxies are not in that same state of evolution . as much of their baryonic matter is still in the form of gas , there should be no expectation that these galaxies would fall in the same place on a luminosity vs circular velocity diagram for galaxies with lower gas content , unless the percentage of dark matter in these systems was very high . this in turn implies an unusual conspiracy - namely , the baryonic mass fraction of galaxies varies with surface brightness in such a way so as to preserve the tully - fisher relation . while this conspiracy can be resolved using mond ( e.g. mcgaugh & de blok 1998 ) , it is also possible to achieve a universal baryonic mass fraction using the kind of stellar population model here , which does adequately account for the simultaneous existence of both gas - rich blue and red lsb galaxies . under that model , we expect very red gas - rich lsb galaxies to strongly deviate from the tully - fisher relation in the sense observed here . the next step in unraveling all of this requires detailed investigations of the actual gas distributions in these red , gas - rich lsb galaxies . many thanks to mike davis for all his help and patience in determining the gain / temperature zenith angle corrections for the l - wide receiver . this research made use of ned , the nasa extragalactic database . work on lsb galaxies at the university of oregon is supported by nsf grant ast 9617011 . arp h. , apj 1965 , 142 , 402 + beauvais , c. , & bothun , g. 2000 , in preparation + beauvais , c. , & bothun , g. 1999 , preprint + becker , r. , mebold , u. , reif , k. , van woerden , h. 1988 a&a 203 , 21 + bell , e. , _ et.al . _ 1998 mnras 302l , 55 + bothun , g. , & mcgaugh , s. 1999 , in preparation + bothun , g. , _ et.al . _ 1993 aj 106 , 530 + bothun , g. , schombert , j. , impey , c. , schneider , s. 1990 apj 360 , 427 + bothun , g. 1990 _ asp conference series 10 : evolution of the universe of galaxies _ + bothun , g. & mould . j. 1987 apj 313 , 629 + bothun , g. , beers , t. c. , mould , j. r. , huchra , j. p. 1986 apj 308 , 510 + bothun , g.d . , sullivan , w. t. , iii , schommer , r. a. 1982 aj 87 , 725 + bothun , g.d . 1982 apjs 50 , 39 + briggs , f. 1998 in _ the low surface brightness universe _ eds . j. davies , m. disney , s. phillipps + broeils , a.h . 1992 , ph.d . thesis , univ . groningen + condon , _ _ 1998 aj 115 , 1693 + courteau , s. & rix , h - w . 1999 apj 513 , 561 + dalcanton , j. , _ et.al . _ aj 114 , 635 + de blok , w.j.g , _ et.al . _ 1999 , in preparation + de blok , w.j.g 1997 ph.d . dissertation university of groningen : groningen + de blok , w.j.g . , van der hulst , j.m , mcgaugh , s. 1996 baas 189 , 8402 + de blok , w.j.g . , van der hulst j.m . , & bothun , g. 1995 mnras 274 , 235 + deich , w. 1990 _ analyz user s guide _ arecibo observatory , arecibo + disney , m. & banks , 1998 in _ the low surface brightness universe _ eds . j. davies , m. disney , s. phillipps + ferguson , h. & mcgaugh , s. 1995 apj 440 , 470 + giraud , e. 1998 aj 116 , 2177 + hoffman , y. , silk , j. , wyse , rosemary f. g. 1992 apj 388l , 13 + impey , c. & bothun , g. 1997 ara&a 35 , 267 + kennicutt , r. , jr . 1989 apj 344 , 685 + larson , r.b . 1986 mnras 218 409 + lewis , b.m . , oneil , k. , eder , j. , & davis , m. _ the spectral line manual _ , in _ the arecibo user s handbook _ , c. salter , ed . 1999 + lewis , b.m . , helou , g. , & salpeter , e.e . 1985 apjs 59 , 161 + marlowe _ et.al . _ 1999 , preprint + matthews , l. , van driel , w. , & gallagher , j. 1998 aj 116 , 2196 + matthews , l. , & gallagher , john s. 1997 aj 114 , 1899 + meurer , g. , _ et.al . _ 1996 aj 111 , 1551 + mcgaugh , s. & de blok , w.j.g . 1998 apj 508 , 132 + mcgaugh , s. & de blok , w.j.g . 1997a apj 481 689 milgrom , m. 1983 apj 270 , 365 + mo , h. j. , mao , shude , & white , simon d. m 1998 mnras 295,319 + oneil , k. , & brandenburg , h. 1999 , in preparation + oneil , k. verheijen , m. & mcgaugh , s. 1999 preprint + oneil , k. & bothun , g. 1999 preprint + oneil , k. _ et.al . _ aj 1998 116 , 570 + oneil , k. bothun , g. & cornell , m , 1997a , aj 113 , 1212 ( obc ) + oneil , k. _ et.al . _ 1997b aj 114 , 2448 + pickering , t. _ et.al . _ 1997 aj 114 , 1858 + pierce , m. & tully , r. 1988 apj 330 , 579 + quillen , a. c. & pickering , t. e 1997 aj 113 , 2075 + sackett , p. 1997 apj 483 , 103 + salzer , j. j. , _ et.al . _ 1999 , in preparation + salzer , j. j. , _ et.al . _ 1993 a&as 182 , 3109 + sandage , a. , binggeli , b. 1984 aj 89 , 919 + schombert , j. , pildis , r. , eder , j. 1997 apjs 111 , 233 + schombert , j. , _ et.al . _ 1995 aj 110 , 2067 + schombert , j. , g bothun , g. , impey , c , & mundy , l 1990 aj 100 1523 schommer , r & bothun , g. 1983 aj 88 , 577 + schwartzenberg , j.m . , _ _ 1995 mnras 275 , 121 + sprayberry , d. , impey , c. d. , bothun , g. d. , irwin , m.j . 1995a aj 109 , 558 + sprayberry , d. , bernstein , g. m. , impey , c. d. , bothun , g. d. 1995b apj 438 , 72 + steinmetz , m. & navarro , j. 1999 apj 513 , 555 + tinsley , b. 1972 a&a .20 , 83 + tully , r.b . , & fouque , p. 1985 apjs 58 , 67 + tully , r.b . , & fisher , j.r . 1977 a&a 54 661 + van der hulst , j , _ et.al . _ 1993 aj 106 548 van der hulst , j.m . , skillman , e. d. ; kennicutt , r. c. , bothun , g. d. 1987 a&a 177 , 63 + weinberg , s. 1972 _ gravitation and cosmology _ john wiley & sons : new york + white , s. _ et.al . _ 1993 nature 366 , 429 + zaritsky , d. , _ et.al . _ 1997 apj 478 53 + zwaan , m. , briggs , frank h. , sprayberry , david , sorar , ertu 1997 apj 490 , 173 + zwaan , m. , van der hulst , j. , de blok , w. , & mcgaugh , s. 1995 mnras 273 , l35 + |
in quantum mechanics , stringnet condensed states a family of prototypical states exhibiting non - trivial topological order can be classified via their long - range entanglement properties , in particular topological corrections to the prevalent area law of the entanglement entropy . here we consider classical analogs of such stringnet models whose partition function is given by an equal - weight superposition of classical stringnet configurations . our analysis of the shannon and renyi entropies for a bipartition of a given system reveals that the prevalent volume law for these classical entropies is augmented by subleading topological corrections that are intimately linked to the anyonic theories underlying the construction of the classical models . we determine the universal values of these topological corrections for a number of underlying anyonic theories including su(2) , su(n) , and su(n) theories . | 1309.3793 | the concept of topological order long - range order beyond the conventional paradigm of symmetry broken order has profoundly broadened our view of strongly correlated systems over the last decades @xcite . archetypal realizations of actual materials exhibiting such unconventional order include the ( fractional ) quantum hall liquids @xcite discovered over 30 years ago , sr@xmath2ruo@xmath3 as a possible realization of a @xmath4 superconductor @xcite or in the more recent past the discovery of topological band insulators @xcite . further insight into the rich physics of topologically ordered systems has been achieved by the analytical understanding of exactly solvable spin models including the toric code @xcite or kitaev s honeycomb model @xcite . despite all this progress it has remained a challenging task to unambiguously identify a topologically ordered state in a given system both in experiments or theoretical approaches . on the analytical side , concepts from quantum information theory have proven particularly helpful in generating a powerful measure of topological order in model systems . a key insight is that the topological order present in a quantum many - body system is reflected in its entanglement properties . in its most pronounced incarnation the long - range order present in topologically ordered systems leads to long - range entangled states @xcite , a scenario that applies to most of the examples of topologically ordered states above . the notable exception are the topological band insulators which exhibit short - range entanglement @xcite . here we will focus on the former class of long - range entangled states . a key concept to measure entanglement in a quantum many - body system is the entanglement entropy calculated by dividing the system into two parts @xmath5 and @xmath6 as illustrated in fig . [ fig : partition ] , obtaining the reduced density matrix of one partition by tracing out the other , e.g. @xmath7 , and then collapsing the information in the reduced density matrix into the so - called entanglement entropy @xmath8 \,.\ ] ] despite the relatively broad - brush character of the entanglement entropy studying its dependence on the geometry of the bipartition has been shown to allow for a rather general classification of ground states of interacting quantum many - body systems . for systems restricted to two spatial dimensions the entanglement entropy exhibits a so - called ` area law ' or more specifically a ` boundary law ' for all gapped ground states ; it grows with the length @xmath9 of the boundary between the partitions @xmath5 and @xmath6 @xmath10 where any additional terms indicated by the dots are subleading terms of order @xmath11 , i.e. vanish in the limit @xmath12 . one of the more striking features of this boundary law is the occurrence of a subleading constant contribution that for a smooth boundary , i.e. one without any sharp corners , is independent of the size or the geometry of the bipartition and thus indicates long - range entanglement . this constant @xmath13 is often called the _ topological entropy _ @xcite , since it strictly vanishes for disordered or conventionally ordered states while for a topologically ordered state it always remains finite . in fact , the topological entropy has to lock into a universal value , which is tightly connected to the effective topological field theory describing the non - local topological order in the system at hand . it was shown @xcite that @xmath14 where @xmath15 is the so - called total quantum dimension , an important characteristic of the effective topological field theory . an alternative perspective to understand the origin of a finite constant @xmath13 for a topologically ordered state emerges from considering the entropic contribution arising solely from the boundary . in contrast to a conventionally ordered or entirely disordered system , the states on the boundary of a topologically ordered system are found to be subtly constrained which then leads to a non - trivial entropic contribution of this boundary . we will discuss details of both perspectives in the remainder of the manuscript . despite the obvious fact that classical systems do not exhibit any entanglement , the notion of the entanglement entropy has a straight - forward companion for classical statistical systems the shannon entropy , which we can analogously define for a bipartition of the system as illustrated in fig . [ fig : partition ] . it is then given as @xmath16 where @xmath17 denotes the set of possible classical configurations in @xmath5 and @xmath18 is the statistical probability of a given configuration @xmath19 . the analogy to the entanglement entropy goes even one step further , as the shannon entropy also exhibits a rather characteristic scaling with the size and the geometry of the bipartition . in its most general form it follows a ` volume law ' of the form @xmath20 where @xmath21 is the volume of partition @xmath5 and @xmath9 is again the length of the boundary of the partition . the dots indicate subleading terms that vanish in the limit of @xmath22 . the occurrence of a non - vanishing constant term @xmath23 in this volume law for the classical shannon entropy is again tightly connected to the occurrence of long - range information in the classical system . a prototypical family of such classical systems exhibiting long - range information are classical variants of so - called stringnet states , which we will introduce and discuss in much detail in the following . the main result of this manuscript is that for this broad variety of classical systems we establish a universal equation relating the constant term @xmath23 to a characteristic feature of the topological field theory underlying these stringnet states . the methods , which allow us to derive this relation , are based on recent work by fendley and simon @xcite . we should note that a related question was discussed previously by castelnovo and chamon in @xcite , where the authors studied classical variants of a class of quantum double models , including kitaev s toric code @xcite . the authors found that the values of the classical and topological entropy are identical for this class of models . this statement is , in general , no longer true for the classical variants of the stringnet states . instead , we find that @xmath24 where @xmath25 is the number of abelian particles in the anyonic theory underlying the stringnet states . our results shed new light on the question on what kind of information in the quantum model can be retained in the corresponding classical variant . our discussion of the above results is structured as follows in the remainder of the manuscript : we provide a general introduction to the classical shannon and renyi entropies in section [ sec : shannon entropy ] . in section [ sec : classical_stringnet_models ] we introduce the stringnet models , first in the context of quantum double models based on the so - called su(2)@xmath0 anyon theories , and then describe their classical analogs . in section [ sec : crossing_symmetry ] we outline the methods we use to analytically compute the renyi entropies for these classical stringnet analogs with the results for the renyi entropies and the topological entropy of eq . presented in section [ sec : renyi entropy ] . the manuscript finishes with an outlook on more general stringnets , a summary of our main results , and several appendices providing the details of our calculation . in quantum mechanics , one of the most important measures of entanglement in a system of multiple quantum mechanical degrees of freedom is the entanglement entropy . we will briefly recap this notion in the following and then turn to the classical analogs of these concepts . let us consider a quantum system with ground state @xmath26 and a bipartition along a smooth cut ( without any sharp corners ) that divides the system into two parts called @xmath5 and @xmath6 . as already outlined in the introduction we can characterize the entanglement of the two partitions by calculating the reduced density matrix of one of the two partitions , say partition @xmath5 , by tracing out the other one to obtain @xmath27 we can now readily diagonalize this reduced density matrix to obtain its eigenvalues @xmath28 , which are the probabilities to find system @xmath5 in the quantum state corresponding to the respective eigenvector . in terms of these eigenvalues , the entanglement entropy can thus be readily calculated as @xmath29 note that for a quantum system the so - calculated entanglement entropies @xmath30 and @xmath31 are equal , i.e. @xmath32 . to identify a classical analog to the entanglement entropy we note that the definition of the entanglement entropy in the form of eq . has a straightforward interpretation also for classical systems . in particular , we can replace the quantum mechanical probabilities @xmath28 to find the system in a certain state @xmath33 by their classical counterparts to find subsystem @xmath5 in a particular classical configuration @xmath34 . the classical analog of the entanglement entropy is then simply given by the well - known shannon entropy @xmath35 where the sum now runs over the set of all possible configurations @xmath17 of subsystem @xmath5 . similar to the quantum mechanical case , the probability of a given classical configuration @xmath34 is obtained by summing over all possible complementary configurations in partition @xmath6 , i.e. @xmath36 where @xmath37 is the energy of the configuration in the full system , @xmath38 is the boltzmann factor , and @xmath39 the temperature . note that in general the classical shannon entropy is not equivalent for the two partitions @xmath5 and @xmath6 , i.e. @xmath40 . instead of directly calculating the shannon entropy it is often more convenient to compute one of the renyi entropies @xmath41 \,,\end{aligned}\ ] ] where the index @xmath42 typically is an integer @xmath43 . the renyi entropies are bounded by each other by @xmath44 for @xmath45 , and recover the shannon entropy in the limit @xmath46 . they can be computed by considering @xmath42 copies ( replicas ) of the system @xcite . in doing so , the different replicas of part @xmath5 are required to have identical configurations , while the configurations in the replicas of part @xmath6 are independent of one another , see e.g. ref . @xcite for details of a numerical implementation . for the ground state of two - dimensional gapped quantum systems , the entanglement entropy obeys an area law , i.e. it grows as the length @xmath9 of the boundary of system @xmath5 instead of its volume @xmath47 the coefficient @xmath48 is non - universal , as it depends on the microscopic details of the model and can be changed continuously . more interesting is the presence of the subleading constant term @xmath13 , which is often called topological entropy and is robust against changing the microscopic details of the system . for a smooth cut it does neither depend on any length scale of the system nor the geometry of the bipartition and as such must be rooted in long - range entanglement . the seminal work of refs . @xcite showed that the topological entropy is universal and can be directly calculated from the total quantum dimension @xmath15 of the effective topological field theory . in particular , it is given by @xmath49 . as the topological entropy is subleading , it is often cumbersome to determine it to sufficient accuracy by doing a scaling analysis , though recent numerical investigations have been quite successful in doing so @xcite . in addition , there may be additional @xmath50 contributions to @xmath30 arising from sharp corners in the cut . thus , it is beneficial to use a setup , which was introduced by levin and wen @xcite , to extract the topological entropy from a set of different bipartitions of the system as illustrated in fig . [ fig : levinwencuts ] . by considering the four different bipartitions of fig . [ fig : levinwencuts ] one can cancel out the leading boundary term in eq . as well as all @xmath50 potential corner contributions to directly obtain the topological entropy from the linear combination @xmath51 which then leads to the final result @xmath52 in contrast to the quantum case , the leading term of the classical shannon entropy @xmath30 scales with the volume @xmath53 of the subsystem @xmath5 in the bipartition , i.e. it follows a volume law of the general form @xmath54 where @xmath9 is again the length of the boundary of the partition and the dots indicate subleading terms of order @xmath55 ) that vanish in the limit of @xmath22 . note again that a direct consequence of this volume law is that the shannon entropy is in general not symmetric with regard to the two partitions , i.e. @xmath40 . similar to the quantum case , we can also in the classical system identify a constant contribution @xmath23 , which reduces the entropy . this constant turns out to be independent of the geometry and size of the system , which is why we will call it ` classical topological entropy ' . as we will discuss in the subsequent sections we can uniquely determine the universal values for this classical topological entropy for a broad variety of classical topologically ordered systems . ) and the classical loopgas , when computing @xmath30 ( @xmath56 ) respectively @xmath31 ( @xmath57 ) . ] at this point , we only want to point out that the form of the scaling behavior of the classical shannon ( and renyi ) entropies still allows to use the setup suggested by levin and wen to directly evaluate the @xmath50 topological correction . one subtle difference to the quantum mechanical calculation is that in the classical case the topological correction @xmath23 is sensitive to the number of disconnected regions in partition @xmath6 . as an example , we show the constant contributions to the entropy for a quantum system and its related classical system for the four levin - wen partitions in fig . [ fig : levinwenpartitions ] . the first line indicates the topological entanglement entropy for the quantum mechanical system , the example being the so - called toric code model @xcite , which has total quantum dimension @xmath58 . note that in the quantum mechanical case the topological correction for a given bipartition is related to the number of boundaries between parts @xmath5 and @xmath6 ; the bipartitions involving @xmath59 and @xmath60 contribute twice the value than those bipartitions involving @xmath61 and @xmath62 . the classical analog of the toric code is the so - called loopgas at @xmath63 , whose classical topological entropy was already studied in ref . @xcite and found to be @xmath64 . as the shannon entropy is not symmetric in @xmath5 and @xmath6 , we show both the topological correction when computing @xmath30 ( in the second line ) as well as @xmath31 ( in the third line ) . note that the classical value is independent of the number of boundaries but instead measures the number of disconnected regions in the complementary part . for instance , the topological @xmath50 contribution in @xmath30 is non - vanishing only if part @xmath6 is disconnected . finally note that the similar setup introduced by kitaev and preskill @xcite for extracting the topological entropy is not suitable for the classical systems considered here , because all bi - partitions have connected parts @xmath5 and @xmath6 . as an example family of model systems exhibiting non - trivial topological order , we will consider so - called stringnet states and in particular their classical analogs . their quantum variants , so - called stringnet condensed states or simply stringnets , have been introduced by levin and wen @xcite in a rather general mathematical construction using so - called quantum doubles @xcite . for completeness , we will briefly recap this levin - wen construction in the following by first introducing so - called su(2)@xmath0 anyon theories and then outlining the quantum double construction before shifting gear to discuss the construction of their classical analogs . the most elementary building block for both the quantum and classical versions of stringnets are so - called su(2)@xmath0 anyon theories @xcite . these theories describe anyonic degrees of freedom with both abelian and non - abelian exchange statistics ( in two spatial dimensions ) . for a given level @xmath65 such an anyonic theory contains @xmath66 individual degrees of freedom , which for our purposes here we label by integers @xmath67 one can think of every degree of freedom as a representation of the quantum group su(2)@xmath0 or in analogy to the spin representation of su(2 ) , which corresponds to the limit @xmath68 , as generalized spins with even / odd integer labels corresponding to integer / half - integer spins . we can combine two representations of su(2)@xmath0 into one joint representation similar to combining two spin quantum numbers into one joint spin quantum number for conventional su(2 ) spins . this process , which for the anyon theories is often called fusion , has to obey very similar rules as those for combining two conventional su(2 ) spins . in particular , they have to obey the so - called fusion rules which also incorporate the cut - off @xmath65 in a consistent way @xmath69}l\ , , \end{aligned}\ ] ] where @xmath70 increases in steps of two . eq . can be written more compactly by introducing fusion coefficients @xmath71 , which are defined via @xmath72 note that for the su(2)@xmath0 anyon theories at hand these fusion coefficients are always either 0 or 1 . the simplest example is probably the anyon theory su(2)@xmath1 with anyonic degrees of freedom @xmath73 and @xmath74 , for which the above fusion rules become @xmath75 a slightly less trivial example is the anyon theory su(2)@xmath2 with anyonic degrees of freedom @xmath76 and @xmath77 , for which the fusion rules read @xmath78 where the fusion with the identity @xmath73 has been omitted , as it is trivial . in addition we also mention the fusion rules of su(2)@xmath79 with anyonic degrees of freedom @xmath80 and @xmath81 , for which the fusion rules read @xmath82 and finally the fibonacci theory , which is the even - integer subset of this su(2)@xmath79 anyonic theory @xmath83 one striking distinction between the fusion rules for su(2)@xmath1 in and for the remaining ones in , , and is the occurrence of representations , which fused with itself , generate more than one fusion outcome , e.g. representation @xmath74 in su(2)@xmath2 . such representations are called _ non - abelian _ as opposed to _ abelian _ representations such as the identity @xmath73 that always generate a unique fusion outcome @xcite . in order to understand this concept better , let us consider a set comprising multiple such non - abelian representations . their combined fusion will no longer be described by a single state but necessarily needs to be described by a set of states . in more technical terms , this manifold of states for a set of non - abelian degrees of freedom asymptotically grows exponentially with the total number of degrees of freedom . the base of this exponential growth is called the _ quantum dimension _ @xmath84 of the representation @xmath33 . non - abelian representations have quantum dimensions that are strictly larger than one , i.e. @xmath85 . in contrast , any representation @xmath86 that is abelian ( exhibiting only single fusion outcomes ) has quantum dimension @xmath87 . for a given anyon theory the _ total quantum dimension _ @xmath15 is then defined as a sum over all the quantum dimensions of the individual representations of the theory @xmath88 for all values of @xmath65 , the su(2)@xmath0 anyon theories always contain two abelian representations , labeled by @xmath73 and @xmath65 , with quantum dimensions @xmath89 . the remaining representations are all non - abelian and thus have quantum dimension @xmath90 . anyon theory , b ) the su(2)@xmath2 anyon theory , and c ) the fibonacci theory . ] with the su(2)@xmath0 anyon theories as elementary building blocks at our hand , we can now proceed to briefly recap the quantum double construction of levin and wen @xcite and introduce the quantum mechanical version of stringnet condensed states or simply stringnets . the quantum double construction of levin and wen creates a lattice model from an anyonic theory , such as one of the su(2)@xmath0 anyon theories introduced in the previous section . the elementary constituents of the lattice model are edges that carry an anyonic degree of freedom , which is captured by the respective anyon theory , i.e. it corresponds to one of the labels @xmath91 . the bonds form a lattice via _ trivalent _ vertices . an example of a particularly regular lattice construction would be the honeycomb lattice , which we will use in all visualizations in the following . at each vertex the fusion rules of the anyonic theory need to be fulfilled , thus constraining the possible labelings of edges around a given vertex . examples of allowed vertices for the su(2)@xmath1 , su(2)@xmath2 and fibonacci theory are given ( up to rotations ) in the various panels of fig . [ fig : allowed_vertices ] , respectively . . ] forming an entire graph out of a set of these allowed vertices and corresponding edges , one can then create a multitude of lattice configurations . by visual inspection of these configurations it becomes evident how the vertex constraints translate into constraints of the overall lattice configurations . for the case of the su(2)@xmath1 anyonic theory , which contains only abelian anyons , the resulting configurations will consist of closed loops only , as illustrated in fig . [ fig : loopgas ] . introducing a non - abelian anyon into the anyon theory , as it is the case for , e.g. the fibonacci theory , the resulting lattice configurations also include branches resulting in netlike configurations as illustrated in fig . [ fig : stringnet ] . for the abelian su(2)@xmath1 anyon theory . ] the set of allowed lattice configurations forms a basis for the part of the hilbert space , where the groundstate resides in . before moving on to the construction of the classical analog of these quantum states , we want to mention in passing a few additional constraints present in the quantum model ( but not the classical model to be introduced in the next section ) . most notable for the remainder of this manuscript is that a subset of allowed lattice configurations has vanishing weight for the groundstate , in particular all those configurations where the netlike configuration includes so - called tadpoles illustrated in fig . [ fig : stringnettadpole ] . the actual weight of an allowed lattice configuration will be finely tuned by a number of parameters depending on the specifics of the underlying anyon theory ( such as , e.g. , a @xmath92-isotopy parameter for the inclusion of closed loops ) . further , the levin - wen construction goes beyond the construction of quantum ground states , but also allows a description of excited states , e.g. states where the vertex constraint is not fulfilled at individual vertices or which include one of the aforementioned tadpoles . finally we note that in more mathematical terms , the quantum double construction is routed in the so - called drinfeld double @xcite , which takes a topological quantum field theory ( tqft ) @xmath93 and its time - reversal conjugate @xmath94 to construct a doubled tqft that is time - reversal invariant @xcite . for more details on this construction the reader is referred to the original math literature @xcite or the more physical introduction in ref . note that the total quantum dimension of the quantum double model is the square of the total quantum dimension of the anyonic theory , i.e. @xmath95 , resulting in a topological entropy @xmath96 . we now turn to the classical analogs of the quantum mechanical stringnet states introduced as groundstates of the quantum double construction of levin and wen in the previous section . to do so , we first note that the quantum ground states of levin and wen can generally be written as a superposition of stringnet configurations @xmath97 , which fulfill the vertex constraints at every single vertex . the wavefunction can then be expressed as @xmath98 where the coefficients @xmath99 are real , but not necessarily all positive or non - zero @xcite . we can now construct a _ classical _ stringnet by defining a partition function that corresponds to the equal - weight superposition of allowed stringnet configurations @xmath97 of the quantum double construction , i.e. @xmath100 = \sum_s 1 \ , , \label{eq : partition_function}\end{aligned}\ ] ] where @xmath101 indicates the number of vertices in the lattice . note that whether a given stringnet configuration @xmath97 is allowed or not , again depends solely on the _ local _ constraints implemented around every vertex . in fact , one can explicitly count the number of allowed configurations contributing to the partition function . going through a sequence of combinatorial steps ( detailed in appendix [ sec : stringnet ] ) , one finds that @xmath102=\sum_{j=0}^k \left(\frac{\mathcal{d}}{d_j}\right)^v \approx 2\mathcal{d}^v \,,\end{aligned}\ ] ] where the @xmath84 indicate the quantum dimension of the anyons in the underlying anyon theory su(2)@xmath0 and @xmath15 is the total quantum dimension of the theory . in the large - volume limit , the explicit sum can be approximated as a power of the total quantum dimension , which connects the original meaning of the quantum dimension in the anyonic theory with its analogous role in the purely classical model . before we turn to a general discussion of the classical stringnets defined in this section , we will briefly layout the more technical details of the methods and in particular the role of a so - called crossing symmetry for the analytical calculations of the renyi entropies for these classical stringnets . in the following we derive several formulas , which are important to the computation of the shannon or renyi entropies of the classical stringnet models . our analytical approach is heavily based on the concept of a so - called crossing symmetry , which was recently introduced by fendley and simon as a method to compute the exact partition function of a special class of classical lattice model @xcite . for later convenience , we will keep the discussion in this section very general , such that it is applicable to _ any _ crossing symmetric model . as the following discussion of crossing symmetric models is very brief , the interested reader is referred to ref . @xcite for a more complete treatment . in the following we consider models that are defined on a trivalent graph with @xmath103 possible states labeled by @xmath104 living on the edges of the graph . the edges are in general oriented , which is indicated by an arrow in the visualizations . we assign each degree of freedom a conjugate via a permutation @xmath105 of @xmath106 $ ] of maximally order two , i.e. @xmath107 . the conjugate of @xmath86 is denoted by @xmath108 and conjugation is implemented by reversing the arrow on the edge . the physical significance of the conjugation will become clear below . we define local boltzmann weights for the vertices that depend on the labels ( and orientations ) of the three edges around the vertex . when defining the weights @xmath109 we use the reference orientation that all arrows are pointing inwards , as is visualized in fig . [ fig : weight_directed ] . the vertex weights can alternatively be written as a ` weight matrix ' @xmath110 defined as @xmath111 which will become convenient later on . note that the second matrix index is conjugated compared to the definition of the boltzmann weights , see also fig . [ fig : weight_directed ] . we consider only models that are isotropic , which implies that the value of the boltzmann weight is the same for cyclic permutations of the indices , i.e. @xmath112 . the partition function @xmath113 $ ] of a graph with @xmath101 vertices is then defined as the sum over all edge labelings of the product of all weights @xmath114&= \sum_{\mbox{\small edge labels } } \prod_{\mbox{\small vertices } \nu } w(i_\nu , j_\nu , l_\nu)\,.\end{aligned}\ ] ] the model is called crossing symmetric if the boltzmann weights defined on the vertices fulfill the crossing symmetry relation @xmath115 which is visualized in fig . [ fig : crossing symmetry ] . crossing symmetry implies that the weight matrices commute . in ref . @xcite fendley and simon showed that crossing symmetry allows for calculating the partition function exactly . using the crossing symmetry relation successively , one can equate the partition function of different graphs . this in turn allows a transformation to a graph , where the partition function can be calculated straightforwardly . an example of such a transformation is illustrated in fig . [ fig : crossing symmetry z ] . the main insight is that the graph can always be deformed into a chain of ` bubbles ' , where a bubble can be interpreted as a matrix that in the following is denoted by @xmath39 and defined through the weight matrices @xmath116 by definition , the @xmath39-matrix commutes with all the weight matrices . the graphical interpretation of the bubble as a matrix is visualized in fig . [ fig : t - matrix ] . . , scaledwidth=30.0% ] as a result of the transformation of the graph , the evaluation of the partition function is reduced to diagonalizing an @xmath117 matrix @xmath114&= \mbox{tr}\left(t^{v/2}\right)\ , . \label{eq : partition function}\end{aligned}\ ] ] the crossing symmetry relation seems at first glance rather restrictive . however , there are a number of interesting classical models that obey the relation . in particular , it was shown in ref . @xcite that the classical stringnet models discussed in section [ sec : classical_stringnets ] are crossing symmetric . to be more precise , the degrees of freedom living on the edges are the representations of the su(2)@xmath0 anyon theory . admissible stringnet configurations have the property that the fusion rules are fulfilled at each vertex . this constraint corresponds to a local boltzmann weight of @xmath118 where the ordering of the labels is not important , i.e. @xmath119 the conjugate @xmath108 of the degree of freedom @xmath86 is defined as the unique element , such that the fusion of @xmath86 and its conjugate contains the identity element @xmath73 , i.e. @xmath120 . from the fusion rules of the su(2)@xmath0 anyon theory , it can readily be seen that the identity element only occurs when @xmath121 . hence , all degrees of freedom are their own conjugate in this case and we can consider the edges as un - oriented . however , this is no longer true for more complicated anyon theories , such as e.g. su(n)@xmath0 . another class of crossing symmetric models are the ones based on finite groups . the degrees of freedom living on the edges are now taken to be elements of a finite group @xmath122 with the identity element denoted by @xmath123 . in this manuscript , we are most interested in models , where the group operation @xmath124 is fulfilled at each vertex , i.e. @xmath125 the conjugate element @xmath108 of @xmath86 is defined as the unique inverse , such that @xmath126 . when @xmath122 is non - abelian the ordering of the indices in @xmath109 is important , i.e. in general @xmath127 . as a result , we need to introduce an orientation for each vertex ( in addition to the orientation of the edges ) . here , we use the convention that the ordering of the indices corresponds to an anti - clockwise orientation of the vertex . we should comment on that eq . is , in fact , much more restrictive than is needed for crossing symmetry , as was shown in ref . @xcite . however , the models defined by are the most interesting from the perspective of calculating the renyi entropies , as it turns out that they have the maximal possible value of the topological entropy , namely @xmath128 , where @xmath129 is the number of elements in @xmath122 . having already derived an explicit expression for the partition function in eq . , let us now continue with discussing the relevant steps to compute the renyi entropies . when dividing the system into two partitions @xmath5 and @xmath6 , we divide the set of vertices spatially into vertices in @xmath5 and @xmath6 respectively , thus cutting the graph spatially on the edges . the total number of these ` boundary edges ' is denoted by @xmath9 in the following . as the two subsystems @xmath5 and @xmath6 are only coupled via the boundary links , it is useful to keep the boundary configuration , denoted by @xmath130 , explicit . a visualization of a bipartition of the system with boundary configuration @xmath131 can be found in fig . [ fig : boundary ] . in order to compute the renyi entropies , we need an expression for the probability @xmath132 of a given configuration @xmath133 , where the set @xmath134 incorporates all edge labels in the volume ( excluding the boundary ) of one of the partitions , say @xmath5 . in terms of the local boltzmann weights @xmath109 , the probability is given by @xmath135 where @xmath136 indicates a sum over all edge labels of links in @xmath6 ( again excluding the boundary ) . inserting the probability into the definition of the renyi entropy with index @xmath42 yields @xmath137 \,.\end{gathered}\ ] ] in order to simplify this expression , we define a ` boundary weight ' @xmath138 ( and similarly for @xmath5 ) by @xmath139 where @xmath140 denotes the volume of @xmath6 and is defined such that @xmath141 is the total number of vertices in @xmath6 . this definition ensures that the contributions of the volume and the boundary to the renyi entropy are nicely separated in the final result . @xmath142 is nothing but the total weight of classical configurations in @xmath6 , given a particular boundary configuration @xmath131 . it can be computed exactly by using crossing symmetry to transform the graph of subsystem @xmath6 into one , where the internal summation over all edge labelings in @xmath6 is expressed as a matrix power of the @xmath39 matrix : @xmath143_{\alpha_1,\beta}\mbox{tr}\left(\phi^{(\beta)}t^{v_b/2}\right)\ , . \end{aligned}\ ] ] an example of such a transformation is illustrated in fig . [ fig : crossing symmetry boundary ] . as the weight matrices @xmath144 commute with each other as well as with the @xmath39-matrix , we can simultaneously diagonalize them and eq . can be readily expressed in terms of the eigenvalues of the matrices . note that we must define a reference configuration on the boundary edges , in the case where edges have an orientation . in the following , we will use the convention that the arrows point towards the vertices in the respective volume , i.e. @xmath6 in case at hand . consequently , we need to take the conjugate boundary configuration @xmath145 for the boundary weight of the other partition , which is @xmath5 in this case . if the replicated system is crossing symmetric , i.e. if the weights @xmath146 obey the crossing symmetry relation , we can also compute @xmath147 exactly , using the weight matrices and the @xmath39-matrix that are obtained from the weights @xmath146 instead . as a result , we can find a compact expression for the renyi entropy of subsystem @xmath5 with index @xmath42 @xmath148\,,\end{aligned}\ ] ] for all classical models , where the model as well as the replicated model i.e. the model , where the local boltzmann weights are given by @xmath146 are crossing symmetric . the requirement that both @xmath109 as well as @xmath146 obey the crossing symmetry relation seems at first glance rather restrictive . however , we note that for the classical stringnet models based on the su(2)@xmath0 anyon theories the weights @xmath109 are all either @xmath73 or @xmath74 . hence , the replicated model is trivially crossing symmetric . the same applies to a few other stringnet models , e.g. those based on the su(n)@xmath0 theories with @xmath149 . it is also valid for the classical models based on finite groups that were discussed in the previous section . in this section , we give a brief outline on how to derive the shannon and renyi entropies of classical stringnet models and discuss their properties . we focus on the model based on the su(2)@xmath0 anyon theory , for which we provide the basic steps . a detailed derivation can be found in appendix [ sec : stringnet ] . let us first discuss a bipartition , where part @xmath5 and @xmath6 are both connected , for instance @xmath61 and @xmath62 in fig . [ fig : levinwenpartitions ] . we note that for the classical stringnet based on the su(2)@xmath0 anyon theory , all representations are self - conjugate and the local boltzmann weight of a vertex is either @xmath73 or @xmath74 . consequently , the expression for the renyi entropy simplifies to the following @xmath150\ , .\end{aligned}\ ] ] the partition function @xmath113 $ ] , see eq . , can be approximated by @xmath151\approx 2\mathcal{d}^v\end{aligned}\ ] ] in the limit of large volume @xmath101 . in the following , we assume that both the volume of partition @xmath5 and @xmath6 are large , such that the approximation above is valid . the boundary weights @xmath152 can then be expressed in terms of the quantum dimensions of the representations in the su(2)@xmath0 anyon theory @xmath153 where @xmath154 is the number representations of type @xmath33 in the boundary configuration . deriving eq . makes use of the fact that the eigenvalues of the weight matrices are related to the quantum dimensions , which we derive in appendix [ sec : stringnet ] . crossing symmetry implies that the ordering in the boundary configuration is irrelevant , only the number of each type of representation enters the expression for the weight . the summation over all boundary configurations can be performed by noting that the summands are nothing but multinomial coefficients of @xmath155 with @xmath156 and the renyi entropy becomes @xmath157\nonumber\\ & \approx v_a\log \mathcal{d } + \ell \log\left[\frac { \mathcal{d}^{2n}}{\mathcal{d}^{+}_{n+1 } } \right]^{1/(n-1)}\ , , \end{aligned}\ ] ] where the second line is valid in the limit of a long boundary length @xmath9 . note that there is no constant contribution to the renyi entropy , when part @xmath5 and @xmath6 are both connected . the constraint of the boundary there are only @xmath158 instead of @xmath159 boundary configurations that contribute to the entropy is exactly compensated by the constraint from the volume , i.e. the factor 2 in the boundary weights @xmath160 in eq . as well as the partition function in eq . . note that this additional factor came about because there are two abelian representations in the su(2)@xmath0 model , which resulted in that the highest eigenvalue of @xmath39-matrix was doubly degenerate . let us now assume that part @xmath5 is connected , but there are two disconnected regions in @xmath6 denoted by @xmath161 and @xmath162 , e.g. bipartition @xmath163 in fig . [ fig : levinwenpartitions ] . the boundary configuration for @xmath164 is given by @xmath165 with length @xmath166 . the expression for the weights @xmath160 in eq . is unchanged , except that the weight for part @xmath5 has a boundary that is the combination of @xmath167 and @xmath168 . that the boundaries combine and can be written as one boundary of length @xmath169 is ensured by crossing symmetry . the calculation of the renyi entropy proceeds along the same lines as the one above , but in this case the boundary and volume constraints do not compensate each other . instead the constraints arising from the volume are only partly compensated by the boundary , which yields an additional @xmath170 to the previous result : @xmath171-\log 2 \\ \approx v_a\log \mathcal{d } + \ell \log\left[\frac { \mathcal{d}^{2n}}{\mathcal{d}^{+}_{n+1 } } \right]^{1/(n-1)}-\log 2\ , , \end{gathered}\ ] ] where the second line is valid when @xmath172 and @xmath173 are both large . if we had computed @xmath174 instead , we had recovered expression with @xmath5 and @xmath175 interchanged . the calculation above is straightforward to generalize to arbitrary number of regions in @xmath5 and @xmath6 . we find that in the limit , where the individual volumes and boundary lengths are all large , the renyi entropy is given by : @xmath176^{1/(n-1 ) } -(n_b-1)\log 2\ , , \end{aligned}\ ] ] where @xmath177 is the number of disconnected regions in part @xmath6 . analytic continuation to @xmath178 , gives the following result for the shannon entropy : @xmath179\\ -(n_b-1)\log 2\ , .\end{gathered}\ ] ] let us emphasize again that the total quantum dimension @xmath15 of the anyon theory appears as coefficient in the volume term , but not in the topological entropy . the latter is instead given by the logarithm of the number of abelian representations , which is two for _ all _ su(2)@xmath0 anyon theories . in appendix [ sec : effective_loop_model ] we present an alternative description of the su(2)@xmath0 models , which gives an intuitive understanding of this , at first glance surprising , finding that the topological entropy does not depend the level @xmath65 . it turns out that all classical stringnet models based on the su(2)@xmath0 anyon theories can be mapped to generalized loop models , for which an topological entropy of @xmath180 is expected , independent on the details of the loop model . so far , we have focused our discussion on classical stringnets rooted in the su(2)@xmath0 anyon theories . however , the quantum double construction of levin and wen and as such also the construction of their classical analogs can in principle be applied to a much broader class of anyonic theories ( or in more general words topological quantum field theories ) . for the su(2)@xmath0 anyon theories we have found that the topological entropy is @xmath181 , independent of the level @xmath65 of the theory . in fact , this result can readily be generalized to a broader class of anyon theories to @xmath182 where @xmath25 is the number of abelian representations in the underlying anyon theory . it might thus be interesting to go to anyonic theories that contain more than the strictly two abelian representations in the su(2)@xmath0 theories . one prominent example is the family of su(n)@xmath0 anyon theories , which we will touch on in the next section . another example would be models based on finite groups @xmath122 , which we will briefly discuss in the subsequent section . we first consider stringnet models based on the su(n)@xmath0 anyon theories , which are related the the more conventional su(n ) algebra , in that only a finite number of representations is kept similar to the su(2)@xmath0 deformation of su(2 ) discussed before . the theory contains @xmath183 representations , @xmath103 of which are abelian . the fusion rules can again be constructed similar to the fusion rules of su(2)@xmath0 in a consistent way that incorporates the cutoff @xmath65 of the deformation , though no closed expression are known for @xmath184 . constructing the classical stringnet model based on these su(n)@xmath0 anyon theories we can again define a set of admissible lattice configurations , where the fusion rules are fulfilled at every single vertex . a ( weighted ) sum over these admissible lattice configurations will then define the partition function of the classical model . similar to our discussion of the su(2)@xmath0 stringnet models we define the local boltzmann weight of a vertex via the fusion rules as @xmath185 which not only enforces the fusion rules at the vertex , but also assigns weights to the various admissible vertex configurations . in particular , it should be noted that for @xmath186 not all vertices enter with the same weight , but depending on the multiplicity of certain fusion channels @xmath187 the weights can actually vary in contrast to the previously discussed case of su(2)@xmath0 anyon theories . the so - defined classical su(n)@xmath0 stringnet model obeys the crossing symmetry relation @xcite . however , as a consequence of the unequal boltzmann weights for @xmath186 the replicated system used for calculation of the renyi entropies no longer obeys the crossing symmetry . as such , our results can be directly generalized only to the case of su(n)@xmath1 and su(n)@xmath2 anyon theories , while the su(n)@xmath0 theories with @xmath186 require additional work beyond the scope of the current manuscript . let us first get some intuition by discussing the su(n)@xmath1 anyon theory , which turns out to be rather accessible . it has @xmath103 representations , which can be labeled consecutively by integers @xmath188 . all representation are abelian with fusion rules @xmath189 where we note that these fusion rules are equivalent to the group operation of the cyclic group @xmath190 . the fusion matrices are then given by @xmath191 and the @xmath39-matrix is diagonal with @xmath192 from the definition of the weight one can deduce that only configurations , where the boundary representations fuse to the identity , contribute to the renyi entropies @xmath193 inserting the weights into and noting that there are @xmath194 allowed boundary configurations for each individual boundary , one obtains the renyi entropy @xmath195 where @xmath177 is the number of disconnected regions in @xmath6 . we again see that the renyi entropy follows a volume law augmented by a boundary term and a topological correction of the form @xmath196 this result is precisely in line with the statement that only the number of abelian anyons in the underlying anyonic theory contributes to the topological correction . the calculation for su(n)@xmath2 is substantially more technical , but yields precisely the same value for the topological entropy @xmath197 the interested reader is referred to appendix [ sec : stringnet_su(n ) ] for some of the details of this calculation . in addition to anyon models described so far , the quantum double construction , the formulation of its classical analog as well as the methods described to calculate the classical entropies in section [ sec : renyi entropy ] can also be used to study classical stringnet models based on finite groups , both abelian and non - abelian . such a group is denoted by @xmath122 in the following and its elements by @xmath198 with @xmath123 being the identity element . the fusion rules are replaced by the usual group operation @xmath124 , such that @xmath199 implies a fusion coefficient @xmath200 where @xmath108 is the ( unique ) inverse of @xmath86 defined by @xmath201 . the boltzmann weights are thus defined as @xmath202 we want to emphasize , the outcome of ` fusing ' two particles is always unique given an ordering . thus , all elements are abelian according to our previous definition counting the number of fusion outcomes , even though the underlying group may be non - abelian . in particular , the individual quantum dimensions are all @xmath203 and the total quantum dimension @xmath204 , where @xmath129 is the total number of elements in the group . the calculation of the renyi entropy is very similar to the one outlined for su(n)@xmath1 . in particular , eq . still holds when replacing @xmath103 by @xmath129 @xmath205 finally , we should emphasize that our results for classical models based on finite groups are fully consistent with the results previously reported by castelnovo and chamon in ref . see also ref . @xcite for related work on kagome spin - ice . to summarize our results , the manuscript at hand provides a detailed introduction of the construction of classical stringnet models from quantum double models based on a given anyon theory . for the family of su(2)@xmath0 anyon theories , we have carefully analyzed the topological entropies arising as subleading contribution in the renyi entropies . in particular , we have derived the precise form of the volume law governing the renyi entropies of order @xmath42 for a subsystem @xmath5 @xmath206^{\frac{1}{(n-1)}}-(n_b-1)\log 2 \ , , \end{aligned}\ ] ] where @xmath177 is the number of disconnected regions in part @xmath6 , @xmath15 is the total quantum dimension of the anyon theory and @xmath207 . analytic continuation to @xmath178 yields the shannon entropy @xmath208\\ -(n_b-1)\log 2\ , .\end{gathered}\ ] ] these results for the su(2)@xmath0 anyon theories can readily be generalized to su(n)@xmath0 theories with @xmath149 as well as models based on finite groups to give a topological entropy of @xmath209 where @xmath25 is the number of abelian representations in the underlying anyon theory . eq . is the main result of our manuscript . let us finally reemphasize that the classical and quantum variants of topological entropy are encoding substantially different aspects of the system , even though they are defined in a very similar manner . while the quantum topological entropy arises from a constraint on the boundary , the classical entropy rather originates from a constraint on the volume . in particular , in the quantum system the topological @xmath50 contribution is proportional to the number of individual boundaries . for the classical counterpart , the shannon ( or renyi ) entropy of subsystem @xmath5 has a topological contribution , which is proportional to the number of disconnected regions in subsystem @xmath6 , or rather @xmath210 . even though the actual values of the classical and quantum topological entropy can turn out to be the same , in particular when considering classical models based on finite groups @xcite , they are in general sensitive to different features in the topological field theory . while the quantum version is sensitive to all representations in the quantum double model of the underlying anyon theory , the classical one is sensitive only to the abelian ones in the anyon theory resulting in vastly different estimates for su(2)@xmath0 anyon theories with @xmath211 . unfortunately , the proof leading to does not directly generalize to su(n)@xmath0 for arbitrary @xmath65 , as the summation over configurations in part @xmath5 and the boundary can not be performed analytically . the form of eq . suggests that the topological contribution is again given by @xmath212 . however , within our current approach we can not rigorously proof that there are no other contributions to the topological entropy arising , e.g. , from the summation over boundary configurations . for these cases numerical simulations might be useful to shed more light on this question . an important issue that was not discussed in this manuscript is the one of stability . the main feature of the topological entropy of quantum systems is that it is robust against any kind of local perturbations . it was already noted in ref . @xcite that the classical topological entropy is not robust against softening the vertex constraint . let us for instance consider the classical loop model . as soon as there is a finite ( even if infinitessimal ) probability of open loops , the topological entropy vanishes for large enough system sizes . another important perturbation in the classical stringnet model is introducing a string tension . some guidance on this issue arises from the classical loop model . the latter is dual to the 2d ising model with the loop tension in the loop model corresponding to finite temperature in the ising model . as such we know that the ( topologically non - trivial ) loopgas phase persists up to a finite , critical loop tension , corresponding precisely to the critical temperature of the dual ising model . as such it is reasonable to expect that the topological entropy remains constant up to the critical value of the loop tension . the topological entropy of the classical system can then be used to characterize an entire phase in full analogy to its quantum counterpart . + * acknowledgements * m.h . thanks eddy ardonne , steve simon , and joost slingerland for interesting and stimulating discussions . s.t . thanks r.g . melko for an inspiring discussion that has led to the idea for the calculations in the manuscript at hand . we also thank t. quella for a critical reading of the manuscript . we acknowledge partial support from sfb tr 12 of the dfg . in this appendix we derive the analytic expression of the renyi entropies for su(2)@xmath0 stringnets in the limit of large volume . we focus on a bipartition , where part @xmath5 and @xmath6 are both connected . generalizing to disconnected regions in @xmath6 and/or @xmath5 is straightforward . the weight matrices and the @xmath39 matrix are simultaneously diagonalized by the modular @xmath213-matrix , which is known explicitly for su(2)@xmath0 , see e.g. ref . @xcite : @xmath214 with @xmath215 . the explicit form of the @xmath213-matrix is in fact not important for the proof . however , we will use that the @xmath213-matrix elements are directly related to the quantum dimensions @xmath84 of the representations of the su(2)@xmath0 anyon theory : @xmath216 where @xmath15 is the total quantum dimension . consequently , the eigenvalues of the weight matrices and the @xmath39-matrix @xmath217_{i , j}&=\delta_{i , j}\lambda_{j}^{(\alpha ) } \nonumber\\ \left[s^\dagger t s\right ] _ { i , j}&=\delta_{i , j } t_{j } \end{aligned}\ ] ] are also related to the quantum dimensions : @xmath218 in particular , @xmath219 which can readily be derived from . for su(2)@xmath0 there are exactly two abelian representations , labeled by @xmath73 and @xmath65 , with @xmath220 . all other represenations are non - abelian and have , therefore , larger quantum dimensions . thus , the highest eigenvalue of the @xmath39-matrix has value @xmath221 and is two - fold degenerate . as a consequence , we find that the partition function grows asymptotically as a power of the total quantum dimension @xmath15 @xmath222=\sum_{j=0}^k \left(\frac{\mathcal{d}}{d_j}\right)^v\approx 2\mathcal{d}^v \end{aligned}\ ] ] in the limit of large @xmath101 . let us now proceed by deriving an explicit formula for the boundary weight . in the following , we assume that the volumes of both subsystems are large . the boundary weight of a boundary configuration @xmath223 and volume @xmath101 is by definition @xmath224_{0,\beta } \mbox{tr}\left[\phi^{(\beta ) } t^{v/2}\right]\nonumber\\ & = \sum_{\beta , s , j=0}^k s_{0j } \lambda^{(\alpha_1 ) } _ j \ldots \lambda^{(\alpha_\ell)}_j s_{\beta j } \lambda^{(\beta)}_s\left(\frac { \mathcal{d } } { d_s}\right)^v \ , .\end{aligned}\ ] ] using the explicit expression of @xmath225 in terms of the @xmath213-matrix elements , performing the summation over @xmath226 and noting that @xmath227 yields @xmath228 in the large volume limit only the terms with @xmath229 survive , the others are exponentially suppressed . by noting that eq . relates the @xmath73th and @xmath65th eigenvalue of the weight matrices to the quantum dimension , we arrive at the final result for the boundary weight @xmath230 where @xmath231 and @xmath232 is the multiplicity of the representation labeled by @xmath33 on the boundary . the ordering of the representations is unimportant due to crossing symmetry . using the explicit form of the boundary weights , the renyi entropy can be computed straightforwardly . @xmath233 \nonumber\\ & = \frac{1}{1-n}\log\left[\sum_{n_{0}+\ldots+n_{k}=\ell}\frac{\ell!}{n_{0}!\ldots n_{k}!}\right . \nonumber\\ & \left . \times \prod_{j=0}^{k}\left(d_j\right)^{n_{j}(n+1 ) } ( 1+(-1)^{\sigma})^{n+1}\right]\nonumber\\ & + \frac{1}{1-n}\log\left[\mathcal{d}^{v_a+nv_b - n(v_a+v_b+2\ell)}2^{-n}\right ] \,,\end{aligned}\ ] ] where the binomial factors enumerate the number of configurations for given @xmath234 . using @xmath235 and @xmath236 we find that @xmath237\ , , \end{aligned}\ ] ] where the last two lines are multinomial expansions of @xmath238 with @xmath239 we can simplify eq . by noting that the contribution from @xmath240 vanishes exponentially in the limit of large @xmath9 and obtain @xmath241^{1/(n-1)}\ , .\end{aligned}\ ] ] in the limit @xmath242 , we find the following limiting behavior for the shannon entropy : @xmath243\ , .\end{aligned}\ ] ] let us now briefly comment on the su(n)@xmath0 models and give the final result for the renyi entropies of the stringnet models based on the su(n)@xmath2 anyonic theories . the calculation proceeds along the same lines as in the previous appendix . thus , we only comment on the few details that differ from the previous calculation . the su(n)@xmath0 anyonic theories have @xmath103 abelian representations . a similar calculation as the one done in eqs . and shows that the @xmath39-matrix has an @xmath103-fold degenerate , highest eigenvalue @xmath221 . in the su(2)@xmath0 models , we saw that the eigenvalues of the weight matrices have special properties , namely @xmath244 , where 0 and @xmath65 were labeling the abelian representations . a similar property is valid for su(n)@xmath0 . in order to see this let us first introduce some necessary notation . in the su(n)@xmath0 anyon theories the representations can be labeled by @xmath103 component vectors @xmath245 @xcite , such that the sum of all the vector entries is @xmath65 . the abelian representations are labeled by the vectors , where one component is @xmath65 and all the others are 0 . for instance , we can label all the abelian representations consecutively by vectors @xmath246 with @xmath247 and @xmath248 , where @xmath249 denotes the identity representation . it can be shown by using the outer automorphism of su(n)@xmath0 that some of the eigenvalues of the weight matrices are related by phases @xcite : @xmath250 \lambda_{\boldsymbol\mu_0}^{(\boldsymbol\alpha)}\ , . \end{aligned}\ ] ] note that in contrast to the index @xmath251 , which labels an abelian representation , @xmath223 may label any representation , in particular it may label one of the non - abelian ones . the phases are @xmath103th roots of unity in analogy to what was found for the su(2)@xmath0 anyon theory . in analogy to the previous calculation , we can identify @xmath252 with the quantum dimension of the representation labeled by @xmath253 , i.e. @xmath254 . the boundary weights are computed in same way as shown in appendix [ sec : stringnet ] . restricting the discussion to systems where the volumes of each of the subsystem is large and using to relate the eigenvalues of the weight matrices to the quantum dimensions , we find the following form of the boundary weights @xmath255 \prod_{i=1}^{\ell}{d_{\boldsymbol \alpha_i } } \ , \mathcal{d}^{v}\nonumber\\ & = \left \ { \begin{array}{cc } n \prod_{i=1}^{\ell}{d_{\boldsymbol\alpha_i } } \ , \mathcal{d}^{v } & \mbox { for } \sum_{j=1}^\ell \sum_{s=0}^{n-1 } \frac{s \alpha_{js}}{n } \in \mathbb{n } \\ 0 & \mbox { otherwise } \ , . \end{array } \right . \end{aligned}\ ] ] we note that the form of the boundary weight is very similar to what was found in appendix [ sec : stringnet ] . the remaining calculation proceeds analogously to the one in appendix [ sec : stringnet ] and the scaling form of the renyi entropy in the limit of large volumes and boundaries looks very similar to the result of the stringnet based on the su(2)@xmath0 anyon theory @xmath256^{1/(n-1)}\!\!-(n_b-1)\log n\ , , \end{aligned}\ ] ] except that the total quantum dimension is now the one of the su(n)@xmath2 anyon theory and the topological contribution is proportional to @xmath257 , as there are @xmath103 abelian representations in su(n)@xmath2 . the number of disconnected regions in subsystem @xmath6 is again denoted by @xmath177 . @xmath258 is defined as in eq . . stringnets using a generalized loop model . the vertices are in one - to - one correspondence to the ones in panel b ) of fig . [ fig : allowed_vertices ] . note that a vertex with two lines on each outgoing leg is absent in the theory . ] in this appendix , we want to present an alternative description on the classical stringnets based on the su(2)@xmath0 anyon theories that is useful for understanding why the classical topological entropy is independent on the level @xmath65 . the main idea is to map the su(2)@xmath0 stringnets to models of non - crossing loops , where at most @xmath65 loops are allowed on each edge . the correspondence between the classical stringnet and this generalized loop model is illustrated in fig . [ fig : su(2)_2loop ] for the case of @xmath259 the upper panel shows the allowed vertices for su(2)@xmath2 and the lower panel the corresponding vertices in the loop model . in order to reproduce the allowed vertices for su(2)@xmath0 one needs to put additional constraints on the non - crossing loop model . for instance , the loop configuration with two strings on each edge is not allowed . the topological entropy is by definition a quantity that is not sensitive to local details . in contrast , it indicates a global conservation law , which on the classical level is enforced by local ( hard ) constraints . for loop models , this constraint is the absence of open strings . thus , as long as the partition function contains loops on all length scales with a finite weight , the topological entropy should be given by @xmath180 independent of which particular loop model is studied . thus , it seems reasonable that the topological entropy of all su(2)@xmath0 models is given by @xmath260 . for a more mathematical introduction the reader is referred to the early chapters of z. wang , _ topological quantum computation _ , conference board of the mathematical sciences * 112 * , american mathematical society ( 2010 ) . note the subtle difference to the loopgas configurations of the toric code model , which are equal - weight superpositions of closed loop configurations , similar to those occurring for the su(2)@xmath1 anyon theory , where all coefficients @xmath99 are strictly positive and of identical magnitude . |
we examine the decoherence properties of a quantum open system as modeled by a quantum optical system in the markov regime . we look for decoherence in both the environment induced superselection ( eis ) and consistent histories ( ch ) frameworks . we propose a general measure of the coherence of the reduced density matrix and find that eis decoherence occurs in a number of bases for this model . the degree of `` diagonality '' achieved increases with bath temperature . we evaluate the decoherence functional of consistent histories for coarse grained phase space two - time projected histories . using the measures proposed by dowker and halliwell we find that the consistency of the histories improves with increasing bath temperature , time and final grain size and decreases with initial grain size . the peaking increases with increasing grain size and decreases with increasing bath temperature . adopting the above proposed measure of `` coherence '' to the decoherence functional gives similar results . the results agree in general with expectations while the anomalous dependence of the consistency on the initial grain size is discussed . | gr-qc9306004 | the emergence of classical dynamics from the underlying quantum behaviour still continues to be one of the foremost questions in physics today . two ingredients essential to this transition are ( i ) the establishing of suitable `` classical '' correlations between appropriate coordinates and their conjugate momenta ( ie . classical equations of motion ) and ( ii ) the suppression or destruction of quantum correlations . the need for the former is self - evident while the latter is necessary to explain the general absence of macroscopic quantum superpositions . a paradigm which primarily addresses ( ii ) and which attempts to describe this transition is _ decoherence_. in this school of thought , decoherence is the general term which describes a dynamical process whereby the quantum correlations in the system in question are diminished or destroyed . these processes can be split into those which deviate from ordinary quantum mechanics , eg . grw theory and quantum mechanical spontaneous localisation qmsl @xcite , and those processes which stay within the confines of quantum mechanics . we will only consider the latter in this paper . the two primary paradigms which attempt to describe the transition from quantum to classical using pure quantum mechanics are environmental induced superselection ( eis ) and consistent histories ( ch ) @xcite . although much has been written concerning these two theories , calculations comparing the two in non - trivial models are lacking . even in relatively simple models the computations quickly become lengthy and difficult to interpret . in this paper we examine the dynamics of a simple quantum optical system from the viewpoints of both eis and ch . upon comparison we find that they generally agree , that is , if eis occurs in a particular basis then histories in that basis become more consistent . in section ii we derive the general solution to the quantum optical master equation in the markov regime for a system comprising of an harmonic oscillator linearly coupled to an infinite bath of harmonic oscillators with the usual quantum optical approximations . we solve for the evolution of the reduced density matrix of the system in the position , momentum , number and coherent state bases . in section iii we discuss eis in these bases using a general measure of `` diagonality '' which quantifies the peaking of the density matrix about the diagonal . this measure is sensitive to the coherence of the system state . we solve for the general behaviour of this measure in the position basis for an arbitrary initial system state and examine the particular case of a mixture of two coherent states . we find that eis occurs to some extent in all of the above bases with @xmath0 becomming more `` diagonal '' with time and increasing bath temperature . in section iv we calculate the decoherence functional of ch for two - time projected coarse - grained histories on phase - space . as in all previous calculations , we use quasi - projectors , that is , operators which are complete but are not exclusive , to define the coarse - grained histories on phase - space . we then compute the degree of consistency and peaking of the decoherence functional using measures advanced by dowker and halliwell @xcite . we find that the consistency and peaking behave as expected except for an anomalous _ decrease _ in consistency with increased initial grain size . this behaviour is also found in all previous calculations . we also adopt the above proposed measure for diagonality in the eis framework to quantify the `` coherence '' of the decoherence functional . we find a similar dependence of this measure on the model parameters to that displayed by the the dowker - halliwell measure . in the conclusion we discuss the results of the eis and ch calculations and the anomalous dependence of the consistency in the ch picture on the initial graining . we also discuss the legitimacy of using the `` coherence '' measure in the ch picture . we argue that in the case of a set of many alternatives , this `` coherence '' measure gives a good indication that the probability sum rule is met to the given degree . before examining the dynamics of the quantum optical master equation we first make a few remarks concerning both the eis and ch scenarios . environmentally induced superselection was primarily developed by zurek in the early 1980 s @xcite in relation to quantum non - demolition measurements of gravitational wave antennas . since then , the theory has steadily evolved to encompass more complicated and realistic models . essentially , the system in question is coupled to a large external bath via a particular interaction with hamiltonian @xmath1 . if the interaction is such that the reduced density matrix of the system becomes approximately diagonal in a particular basis in a time much less than the relaxation time then we say eis has occurred . the process is aimed primarily at removing the quantum coherence from the system . the system s reduced density matrix in this particular basis is approximately diagonal and is given a purely ignorance interpretation . that is , the system is now in one of the states on the diagonal but we do not know which one . a number of calculations have shown that in some cases eis is very successful and can damp away the quantum superpositions very quickly indeed @xcite . other models have shown however , that the spectral properties of environment play a significant role in determining the decoherence time @xcite . in this paper we will work with the standard ohmic environment . it is also the case that the initial conceptions of eis have changed . the concept of a pointer basis @xmath2 , the particular basis in which the reduced density matrix becomes diagonal , strictly holds only in those models for which the interaction hamiltonian commutes with the number operator of the pointer basis @xmath3 . for more realistic cases and for the models treated in this paper , no exact pointer basis exist . work on a suitable generalization of the concept of a pointer basis , that is , a system basis which is least effected by the interaction with the bath , is currently being pursued @xcite . also , there does not appear to be a general concrete description concerning the `` diagonality '' of a reduced density matrix in a particular basis . some measures have been advanced for gaussian states . for particular non - gaussian states a fringe visibility measure based on the appearance of the wigner function has been proposed @xcite . this however , appears to quantify the amount of interference in the state rather than the coherence . eis is the process whereby the reduced state of the system deviates more and more from a description as a single ray in hilbert space . such a process reduces the interferences present in the state . however , pure states , with unitary evolution , can display little interference . such processes do not display eis . with this in mind we propose and investigate a measure of the `` diagonality '' of the state in a number of bases in a quantum optical model . we find that the coherence of the reduced density matrix of the system decays and gives clear evidence of eis in a number of bases . however , the degree of diagonality depends on the basis chosen . it also depends on the model parameters in an expected manner , becoming more diagonal with increasing bath temperature . we have not examined the attainment of ( i ) above in the eis framework . this is because there appears to be some dispute concerning the quantification of classical correlations in the state . previously , some have examined the wigner function of the state for peaks @xcite . however , the wigner function is not a true probability distribution and does not correspond to the results of any conceivable measurement on the system . the @xmath4 function of quantum optics _ does _ correspond to the result of a particular measurement scheme @xcite . more generally , there exist true probability distribution functions corresponding to more general measurements @xcite . to establish the existence and to quantify `` classical '' correlations one should examine such distributions . work towards this goal is in progress @xcite . finally , the ignorance interpretation of the resulting diagonal reduced density matrix seems part and parcel of eis . this feature of the framework has surprisingly not attracted much attention in the literature . if eis is to describe a truly ontological reality then zurek s arguments may not suffice @xcite . it is clearly capable of describing an empirical reality @xcite but are it s aspirations higher than just such a description ? in order not to confuse the reader with various different meanings of the term decoherence , we will refrain from the more usual nomenclature of `` decohering histories '' and use the more descriptive terminology `` consistent histories '' . consistent histories was initially researched by griffiths and omns @xcite and later on independently rediscovered and greatly expanded on by gell - mann and hartle @xcite . this paradigm is more ambitious than that of eis as it attempts to describe the emergence of the complete quasi - classical world of familiar experience from the underlying quantum world . central to this framework is the notion of coarse - grained histories , each with an assigned probability . a primary goal is the identification of complete sets of coarse - grained histories which exactly satisfy the probability sum rules . although much has been written concerning this theory , there are only a few calculations on realistic models @xcite . quantum optics is an avenue where both quantum and classical behaviour are accessible to the experimentalist . we will in particular look at coarse - grained histories on phase - space itself . previous calculations have concentrated on position projected histories with momenta information gained through time - of - flight . in this paper we look for consistent histories in _ both _ position and momentum . we also note that the interpretation of the diagonal decoherence functional is typically a relative state interpretation and thus the interpretational problems concerning which history is realized are moot ( however see @xcite ) . we begin by looking at the optical master equation for a system linearly coupled via the usual quantum optical coupling to a bath of harmonic oscillators @xcite . for completeness and to relate this to earlier work let us first consider a system with a hamiltonian @xmath5 in the quantum optical regime we can write the master equation for the reduced density matrix of the system as @xcite @xmath6+\lambda\rho={\cal l}\rho\;\ ; , \label{eq1.1}\ ] ] @xmath7 + [ a\rho , a^{\dagger } ] \right\ } + \frac{\gamma\bar{n}}{2}\left\{[a^{\dagger } , \rho a ] + [ a^{\dagger}\rho , a ] \right\}\;\;.\label{eq1.2}\ ] ] this may be converted into a partial differential equation for a quasi - distribution function ( qdf ) , @xmath8 ( or @xmath9 for short ) where @xcite @xmath10 @xmath11\;\;,\label{eq3}\ ] ] @xmath12 where we have the normalization @xmath13 . the family of distribution functions @xmath14 encompasses the more familiar distributions . for @xmath15 , @xmath16 is the p distribution of quantum optics . for @xmath17 , @xmath18 is the familiar wigner distribution while for @xmath19 , @xmath20 is the q or husimi distribution . the different distributions correspond to different operator orderings . from ( [ eq1.1 ] ) we can derive a partial differential equation for both @xmath9 and @xmath21 from first principles using equations ( [ eq2 ] , [ eq3 ] , [ eq4 ] ) . however , using the general rules derived in vogel and risken @xcite we can easily obtain @xmath22\partial_{\alpha^ { * } } -[f_{1}^{*}+2\alpha^{*}f_{2}^{*}+\alpha\omega]\partial_{\alpha } -s[f_{2}\partial^{2}_{\alpha\alpha^{*}}-f_{2}^ { * } \partial^{2}_{\alpha\alpha^{*}}]\right\}w(\alpha , s , t)\\ & & + \frac{\gamma}{2}\left\{\partial_{\alpha}\alpha+ \partial_{\alpha^{*}}\alpha^ { * } + [ 2\bar{n}+1-s]\partial^{2}_{\alpha\alpha^{*}}\right\}w(\alpha , s , t ) \;\;,\label{eq6}\end{aligned}\ ] ] @xmath23 -[\xi f_{1}+\xi^{*}f_{1}^ { * } ] -s[\xi^{2}f_{2}-\xi^{*\,2}f_{2}^ { * } ] + 2[f_{2}\partial_{\xi^{*}}-f_{2}^{*}\partial_{\xi}]\right\}\chi(\xi , s , t)\\ & & + \frac{\gamma}{2}\left\{(s-1 - 2\bar{n})\xi\xi^{*}-\xi^{*}\partial_{\xi^ { * } } -\xi\partial_{\xi}\right\}\chi(\xi , s , t)\;\;,\label{eq7}\end{aligned}\ ] ] where we have dropped the @xmath24 dependence of @xmath25 and @xmath26 . we see that the pde for @xmath27 is first order . the general solution for this model was treated in @xcite . we will instead specialize to the case of a single harmonic oscillator with constant frequency linearly coupled to the bath oscillators . going to the interaction picture we get @xmath28\partial^{2}_{\alpha\alpha^ { * } } \right\}w(\alpha , s , t)\;\;,\label{eq8}\ ] ] @xmath29 using the method of characteristics we will solve for the dynamics of this model for an initial state @xmath30 where @xmath31 . thus @xmath32 is a general superposition of coherent states at @xmath33 . ( coherent states are eigenstates of the lowering operator of the oscillator algebra i.e. @xmath34 ) . since the coherent states are overcomplete any initial state may be represented thus . let us solve ( [ eq9 ] ) for @xmath35 when @xmath36 . from ( [ eq3 ] ) and using @xmath37 we get @xmath38 using this we can calculate the form of the characteristic function @xmath27 at @xmath33 to be @xmath39\;\;.\label{eq10}\ ] ] for @xmath19 we can rewrite ( [ eq9 ] ) as @xmath40 where @xmath41 and @xmath42 are now treated as independent variables . looking at the characteristics of ( [ eq12 ] ) we see that @xmath43 integrating this we obtain three constants , @xmath44@xmath45 solving for @xmath46 and @xmath27 in terms of @xmath47 and inserting these into ( [ eq10 ] ) at @xmath33 gives a relationship between the @xmath47 which is valid for @xmath48 . re - inserting the @xmath24 dependence back into the @xmath47 and going back to the heisenberg picture yields @xmath49\;\ ; , \label{eq14}\ ] ] where @xmath50 , @xmath51 and @xmath52 . inverting ( [ chi_def ] ) to obtain the @xmath4 distribution we get @xmath53\;\;,\label{eq15}\end{aligned}\ ] ] where we have used ( [ appendix1 ] ) from appendix a. to obtain the off - diagonal elements @xmath54 we use the identity @xmath55 and inserting ( [ eq14 ] ) we get @xmath56\langle\tilde{\alpha}| \tilde{\beta}\rangle \;\;.\label{eq18}\end{aligned}\ ] ] using equation ( [ eq18 ] ) we can now examine the general case where @xmath57 and can evaluate @xmath58 to be @xmath59\ , \;\;,\label{eq20}\ ] ] where @xmath60 is the markov super - operator ( [ eq1.1 ] ) . with this the dynamics is completely solved . however , to compare the decay of off - diagonal elements of the reduced density matrix in other bases we will also compute @xmath32 in the position @xmath61 , momentum @xmath62 and number basis . to compute the off - diagonal elements in the position basis we again use equation ( [ eq17 ] ) , giving @xmath63 where @xmath64 is given in ( [ eq14 ] ) . from appendix b we have @xmath65\ , \delta(\xi_{x}+\sqrt{\frac{\omega}{2\hbar}}(x - y))\sqrt{\frac{\omega}{2\hbar } } \;\;.\label{xy1}\ ] ] for @xmath66 we can perform the gaussian integrations in ( [ xy ] ) and use the result ( [ xy1 ] ) to finally get @xmath67\;\ ; , \label{eq21}\end{aligned}\ ] ] where @xmath68 and @xmath69 . using similar methods we can calculate the elements in the momentum basis to be @xmath70\;\ ; , \label{eq22}\end{aligned}\ ] ] where @xmath71 and @xmath72 . to calculate the elements of @xmath32 in the number basis is slightly tedious . noting that @xmath73 and the identity @xmath74 we find , with some work , that @xmath75 where @xmath76 we note that even in the case where @xmath77 equations ( [ eq24 ] ) and ( [ eq24a ] ) do not reduce to a simple tractable forms . in what follows we will not make much use of equations ( [ eq24 ] ) and ( [ eq24a ] ) except to note that they become diagonal as @xmath78 . finally we note that the results for @xmath79 found here agree with those found in @xcite . we now wish to examine the behaviour of the off - diagonal elements of @xmath32 in a number of different bases and to determine whether eis occurs . there have been a number of proposals for the measure of the degree of diagonality of the density matrix @xcite . these have mostly been defined however for a gaussian @xmath32 . here we expand these definitions to a measure @xmath80 of the `` diagonality '' of a general density matrix in a particular basis @xmath81 ( @xmath81 here denotes a general basis i.e. x , p , n or coherent state ) . from recent work @xcite there appears to exist some competition between the attainment of a stable predictable pointer basis and the attainment of eis type decoherence . in this paradigm one must be careful not to confuse the loss of coherence with the loss of interference . in this paper we interpret eis as a process whereby the _ coherence _ of the state is lost . from zurek s earlier work the signature of decoherence was the vanishing of the _ coefficients _ of the off - diagonal elements of the reduced density matrix of the system in a given basis . the decay of the interference in a particular basis was due to the decay of these coefficients . however , the loss of interference can also be brought about by sending the overlap of the off - diagonal states in that basis to zero . for example , in young s slits , when one sets the slit separation to be very large in comparison to the screen distance , the magnitude of the interference terms decays due to the vanishing of the overlap of these off - diagonal elements in the position basis . ( there may , of course , be significant overlap in other bases , eg . the momentum basis . ) in this case the loss of interference is _ not _ due to the decay of the coefficients associated with these off - diagonal elements . the state remains pure , no coherences have been destroyed yet the quasi - probability distribution function _ tends _ towards ( but is never exactly ) that displayed by a true probability distribution . recently decoherence in the eis paradigm has been defined as the absence of interference fringes in the wigner function in the special case of a superposition of two gaussians @xcite . these interference fringes decay in magnitude for large separations of the two gaussians . they never disappear except in the case of infinite separation . the fringes also vanish dynamically with time if one introduces a coupling with an infinite bath . in the latter case coherence has been lost and the off - diagonal coefficients decay while in the former case the interference terms vanish because the overlap of the off - diagonal elements in the chosen basis goes to zero . thus , while the loss of coherence necessitates the loss of interference the loss of interference need not result in a loss of coherence . in this paper we follow the spirit of earlier work and define environment induced superselection as the dynamical process whereby the reduced state of the system deviates from a single ray in hilbert space and moves towards a description as an improper mixture ( for more information on improper mixtures see @xcite ) . the most obvious measure of the deviation of a given state @xmath32 from being a ray in hilbert space is the statistical distance between the state @xmath32 and the closest mixed state @xcite @xmath82 where @xmath83 is the bures metric on the state space of all pure and impure states @xmath32 @xcite . equation ( [ eq25 ] ) is very difficult to evaluate in general . instead we will generalize a measure of coherence for gaussian @xmath32 first introduced by morikawa @xcite and used later in @xcite . for @xmath32 of the form @xmath84\;\;,\ ] ] morikawa defined a measure of coherence to be @xmath85 when @xmath86 the magnitude of the elements of the density matrix in the position basis are peaked preferentially along the diagonal @xmath87 . for @xmath32 pure , @xmath88 . the natural generalization of this measure to non - gaussian @xmath32 is to compute the ratio of the off - diagonal variance to the on - diagonal variance of the squared magnitude of @xmath32 in a specified basis @xmath89 , that is @xmath90 for @xmath32 gaussian and @xmath91 , @xmath92 from ( [ eq26 ] ) . for pure states , @xmath93 and @xmath94 . this makes sense as no pure state undergoing unitary evolution can lose it s coherence . the offset in the denominator gives the variance of @xmath95 about the mean @xmath96 . distributions of @xmath97 which are concentrated along @xmath98 result in @xmath99 . as @xmath100 becomes more concentrated along @xmath101 , @xmath102 decreases to zero . from ( [ eq27 ] ) the value of @xmath102 depends on the basis chosen . this we expect as some basis will display a greater peaking about their diagonal than others . the advantage of the dimensionless measure ( [ eq27 ] ) is that one can compare @xmath102 and @xmath103 for various bases @xmath104 . one can also consider bases which are discrete . in this case the integrals in ( [ eq27 ] ) become discrete sums . heuristically the measure @xmath102 tells us how damped the off - diagonal coherences are in the basis @xmath104 . for pure states , no matter what the configuration , these `` off - diagonal '' coherences never vanish . for every contribution to the `` on - diagonal '' elements of @xmath32 there are equal contributions to the `` off - diagonal '' elements . a classic example of this is the superposition of two localized coherent states @xmath105 . for a pictorial description of @xmath97 for this state see zurek @xcite . in this example the heights of the four peaks in zurek s plots do not change with increasing @xmath106 . no coherence is lost , however the interference diminishes . a quantification of the amount of interference present in a state is generally given by the magnitude of a typical off - diagonal element . for the special case of two superposed gaussians a measure of the fringe visibility of the ripples of the wigner function has been proposed @xcite . another quantification of the degree of `` off - diagonality '' would be @xmath107 along the off - diagonal . because this measure has dimensions of @xmath108 , comparisons of @xmath109 between bases are made complicated through the different length scales associated with each basis . for a comparison between bases one should use the dimensionless measure ( [ eq27 ] ) . we also note that while the discrete version of ( [ eq35 ] ) vanishes for @xmath110diagonal ( eg . spin 1/2 particle ) , for @xmath111 the continuous version of ( [ eq35 ] ) evaluated in the coherent state basis increases like @xmath106 . although ( [ eq27 ] ) and ( [ eq35 ] ) seem to be natural candidates for a measure of `` diagonality '' in the eis sense we see that the prototypical state ( [ state ] ) is not diagonal nor even approximately diagonal in the coherent state basis . it may be said that since the ripples in the wigner function decrease in size as @xmath112 one should treat this state in this limit as `` interferenceless '' or `` classical '' , this i believe is incorrect . the wigner function for this state always possesses negative regions for all @xmath113 . the interference represented by these would be manifested in the expectation values of observables which have support in these regions . the effects of eis are to cause these negative regions to disappear in a finite time . however , as pointed out by bell , positivity of the wigner function is a necessary but _ not _ sufficient condition that the wigner distribution correctly represents a true probability distribution @xcite . this connection will be made clearer in @xcite . finally we can use the relation @xmath114 to rewrite ( [ eq35 ] ) as @xmath115 \right\}\;\;,\label{eq30a}\end{aligned}\ ] ] where we have used @xmath116 in what follows we will use @xmath102 , ( [ eq27 ] ) as a measure of diagonality . taking @xmath77 we can compute @xmath117 recalling that @xmath50 , ( [ eq14 ] ) , this result would indicate that we obtain a greater degree of diagonalisation in the @xmath61 and @xmath62 basis than in the coherent state basis @xmath81 . at @xmath33 , since the system begins in a pure state , @xmath118 for all bases . thus environmental induced superselection has occurred in all three bases . from ( [ eq24a ] ) , @xmath119 and thus @xmath32 also diagonalises in the number basis , however the exact form of @xmath120 is quite complicated and will not be given here . for the case of a completely general initial state we only quote the result in the position basis . for @xmath121 one finds @xmath122 } { \sum_{ijkl}\,n_{ij}n_{kl}\langle\alpha_i|\alpha_j\rangle \langle\alpha_k|\alpha_l\rangle e^{a_+^2\sigma_+^2+a_-^2\sigma_-^2+b } \left[1 + 2a_+^2\sigma_+^2 + 8\langle x\rangle(\frac{\langle x\rangle}{\sigma_+^2}-a_+)\right]}\right\}\;\;,\label{19a}\end{aligned}\ ] ] where @xmath123\;\;,\\ a_+&=&\frac{\sqrt{\sigma_+\sigma_-}}{2\sigma_+^2 } \left[\alpha_i^*(t)+\alpha_k^*(t)+\alpha_j(t)+\alpha_l(t)\right]\;\;,\end{aligned}\ ] ] and @xmath124 \left(\frac{\sigma_-}{2\sigma_+}\right)\;\;.\ ] ] a similar result holds for @xmath125 where we replace @xmath126 with @xmath127 \;\ ; , \nonumber\\ a_- & = & \frac{-i\sqrt{\sigma_+\sigma_+}}{2\sigma_+^2 } \left[\alpha_i^*(t)+\alpha_k^*(t)-\alpha_j(t)-\alpha_l(t)\right]\;\;.\end{aligned}\ ] ] to obtain a clearer understanding of @xmath128 let us compute @xmath129 for the particular state @xmath130 where @xmath131 . denoting @xmath132 we find that @xmath133 where @xmath134 , ( from above ) and @xmath135\;\ ; , \label{omegax}\ ] ] where @xmath136 are defined in ( [ xy1 ] ) and evaluated at @xmath33 . since @xmath137 we see that the coherence of a mixture of two glauber states is less than the coherence of just one . the of the mixture @xmath138 decreases inversely as @xmath139 . thus packets which are separated in position by amounts significantly greater than their half widths possess lower @xmath61-coherence . also @xmath140 depends very weakly on the momentum @xmath141 except for small @xmath142 where the overcompleteness property of the glauber states cause interference - like behaviour in the @xmath61 basis . a plot of ( [ omegax ] ) is given in fig 1 . the dynamical evolution of @xmath143 is given again by equation ( [ omegax ] ) with @xmath144 given by ( [ xy1 ] ) while @xmath145\;\;,\\ \bar{p}(t)&=&e^{-\gamma t/2}\left[\bar{p}(0)\cos\omega t-\omega \bar{x}(0)\sin\omega t\right ] \;\;.\\\end{aligned}\ ] ] essentially , the evolution begins at @xmath33 with the two glauber states located symmetrically about the origin at positions @xmath146 . since the glauber states are overcomplete there is some overlap between the two . as time progresses , the state , represented loosely as two superposed gaussians in the wigner distribution , spirals in towards the origin while the half widths of each gaussian spreads with time for nonzero @xmath147 . the system motion complicates the interpretation and instead , one can return to the interaction picture . in the interaction picture , for a fixed choice of @xmath148 where @xmath149 , @xmath143 decreases below the value @xmath150 due to the large separation in position between the two gaussians . however , as @xmath78 , @xmath151 and thus the coherence of @xmath152 tends to the same value as the coherence of @xmath153 in this limit . this is not surprising as the final states are the same . as the bath temperature is raised the time for @xmath143 to reach a given value decreases since the gaussian s half widths become larger and thus @xmath150 decreases faster . also , we note that when the @xmath142 separation is large , the state may possess a _ lower _ coherence than in the equilibrium state at @xmath154 . this phenomena is also seen is the relaxation behaviour of squeezed states in a thermal bath . the coherence behaviour of extended quantum states and the relation of the @xmath128 measure of coherence to more information theoretic measures will be treated in another paper @xcite . # myfig1.ps#5.5#myfig1.ps ( 5,7 ) ( 0,0)(5,7 ) ( 0,7 ) ( 0,0 ) ( -2,-10 ) to conclude and summarize : we have described the dynamics of the reduced density matrix of a harmonic oscillator linearly coupled to an infinite bath of oscillators in the quantum optical regime . we have expressed the results in the position , momentum , number and coherent state basis . after considering a number of proposals for a measure of the `` diagonality '' or degree of coherence of the state @xmath32 we have advanced @xmath155 ( [ eq27 ] ) as a natural measure . we calculated this quantity in the above mentioned bases for an initial glauber state and found that coherence was lost in each basis . the coherence loss rate for this initial state was found to be identical in the @xmath61 and @xmath62 basis while the loss rate in the coherent state basis was smaller . expressions for the coherence of a general initial state were found in both the @xmath61 and @xmath62 bases . finally , the position coherence for an initial mixture of two coherent states was calculated and was found to be more `` diagonal '' than that of a single coherent state , especially of the position separation between the two coherent states is large . although the measure @xmath155 may seem to be a `` natural '' generalization of the early definitions @xcite it would be interesting to understand the information theoretic aspects of such measures . of more interest would be the link between the `` diagonality '' of @xmath32 and the representation of @xmath32 by a true probability distribution . this work is in progress @xcite . consistent histories is an alternate paradigm developed by hartle and gell - mann @xcite which attempts to describe the emergence of the `` quasiclassical '' world from the quantum world . the primary ingredients of this paradigm are coarse grained histories and the decoherence functional . essentially , a coarse grained history is a partial description of the temporal evolution of the total system+environment where one has `` coarse grained '' away information either by ( i ) ignoring some of the coordinates for all time ( ie . the environment ) ( ii ) only describing the configuration of the priveliged coordinates at certain times ( iii ) projecting the privileged system coordinates at specified time on to quantities of interest eg . binning , averaging etc . the resulting coarse grained histories may be described in some cases ( those treated here ) by history independent temporally ordered strings of projectors @xmath156 , where the @xmath157 are a complete and exclusive set of projectors on to a set of alternatives labeled by @xmath158 and where @xmath157 is the ith projector in the set @xmath159 . in the past , calculations have concentrated on sequences on projectors in position using the calderia - legget model @xcite . in this section we expand these calculations to projectors on phase - space and consider coarse grained phase - space histories of the privileged system harmonic oscillator in the quantum optical regime . in the consistent histories paradigm a probability is assigned to a coarse grained history via the rule @xmath160\;\;,\label{cheq1}\ ] ] where @xmath161 is the temporally ordered string of projectors corresponding to the coarse grained history @xmath162 ie . @xmath163 and where @xmath164 is the propagator for the full system - environment . we now assume that these projectors factor out the environment and can thus be written in the form @xmath165 . in the markov regime the dynamics of system s reduced density operator is purely local in time and @xmath166 satisfies partial differential equation which is first order in time and contains only system quantities . in this regime we can take the trace in ( [ cheq1 ] ) over then environment and arrive at @xmath167\right ] \;\;,\ ] ] where @xmath168 acts only on the system s hilbert space , @xmath169 is the reduced density operator for the system and we have again used ( [ eq1.1 ] ) to evolve the system s reduced density operator from @xmath170 to @xmath171 . a similar method is used in @xcite . thus in the markov regime the effects of the coupling to the environment are taken care of through the superoperator @xmath172 and the calculation ( [ cheq1 ] ) reduces to a calculation on the system only . at the heart of this formalism is the requirement that these probabilities satisfy the sum rule @xmath173 where @xmath174 is a finer graining of @xmath162 . for the sum rule ( [ cheq3 ] ) to hold with the graining @xmath174 we must have @xmath175=0\qquad\forall\ \tilde{h}^\prime,\;\tilde{h}\in h\;\;,\tilde{h}^\prime\neq\tilde{h}\;\ ; , \label{cheq4}\ ] ] where @xmath176={\rm tr}[p_{\tilde{h}}\rho p_{\tilde{h}^\prime}^\dagger ] \;\;,\label{decoherence_functional}\ ] ] is known as the decoherence functional @xcite . condition ( [ decoherence_functional ] ) is often termed _ weak consistency_. a stronger , although sometimes argued more physical , condition is @xmath176=0\qquad\forall\ ; \tilde{h}^\prime,\;\tilde{h}\in h\;\;,\tilde{h}^\prime\neq\tilde{h}\;\;,\ ] ] and is called _ medium consistency_. a major goal in this approach is the identification of all those sets of histories which are exactly consistent and from this , an identification of those histories which yield a `` quasiclassical '' world . we do not wish to give an in - depth review of this formalism and refer the reader the excellent articles of gell - mann and hartle @xcite . it has been recognized that sets of histories which are _ exactly _ consistent may be characterized by very exotic and un - physical system variables @xcite . to examine the consistency properties of histories parametrised by ( perhaps ) more interesting and physical variables ( ie . position and momentum ) a measure of the degree of consistency of a set of histories @xmath177 has been advanced by dowker and halliwell @xcite @xmath178 } { \sqrt{{\rm d}[\tilde{h},\tilde{h}]{\rm d } [ \tilde{h}^\prime,\tilde{h}^\prime]}}\;\;.\label{epsilon}\ ] ] as @xmath177 becomes more consistent @xmath179 decreases . dowker and halliwell have also argued that in general @xmath179 decreases as the graining size is increased . only for very artificial grainings , where the phases of the constituent histories in the coarser graining add constructively , does @xmath179 remain unchanged or actually increase . this point will be important later . as mentioned above , all previous calculations have looked at histories on position space . however we normally consider classical dynamics as operating within phase - space . it would be of great interest to see if phase - space histories of the system become more consistent with the addition of an interaction with an external bath . in the previous section on environmental induced superselection we saw that the quantum optical model gave eis in both the @xmath61 and @xmath62 bases . here we wish to see whether phase - space histories for the privileged harmonic oscillator displays consistency and if so , to what degree ? how does the degree of consistency depend on the model parameters ie . @xmath180 , @xmath147 graining etc ? in the following we will consider two - time projected histories on the phase - space of the privileged oscillator . we will again couple the system to an external bath and use the solutions of the master equation ( [ eq1.1 ] ) found in section ii to evolve @xmath169 between projections . we will calculate the decoherence functional and the degree @xmath179 of consistency of these histories . this @xmath179 will be a function of the equilibrium number of photons per mode in the bath ( @xmath181 ) , the time @xmath24 , the initial state of the system @xmath182 and the phase - space projectors chosen ( ie . the graining ) . we will choose gaussian type `` projectors '' on phase - space and will find that @xmath179 decreases with time @xmath24 , increasing bath temperature @xmath183 and final grain size @xmath184 but anomalously decreases with initial grain size @xmath185 . this anomalous behaviour can also be seen in other calculations of decoherence functionals for similar models @xcite . we conclude that the introduction of the bath causes the phase - space histories to become more consistent and less peaked about the classical path . finally we note that consistent histories calculations tend to give rather lengthy and hard to interpret results . we have found that working with this model has resulted in a more transparent description . we wish to calculate the decoherence functional for two - time projected histories on phase - space , that is @xmath186 & = & { \rm d}[\alpha_3,\alpha_2,\alpha_1,\sigma_2,\sigma_1,t]\nonumber\\ & = & { \rm tr}\left[p(\alpha_3,\sigma_2)\,e^{\int_{t_1}^{t_2}\,{\cal l}dt } \left[p(\alpha_2,\sigma_1)\rho(t_1)p^\dagger(\alpha_1,\sigma_1)\right]\right ] \;\;,\label{cheq6}\end{aligned}\ ] ] where the coarse grained history is defined by a projector @xmath187 on the phase - space of the system oscillator at @xmath188 of `` width '' @xmath189 . the trace in ( [ cheq6 ] ) is over the system s hilbert space . we will assume , as in @xcite , that @xmath190 is a _ quasi - projector_. in particular we will take @xmath190 to be gaussian with @xmath191 where @xmath192 is a glauber coherent state . these projectors are complete ie . @xmath193 , but are not exclusive @xmath194 using such gaussian `` quasi - projectors '' makes the calculation analytically tractable . in fact , since true projectors on phase - space do not exist @xcite the best one can do is to use such complete but nonexclusive `` quasi - projectors '' . indeed , these types of `` quasi - projections '' play a significant role in the more rigorous `` effects and operations '' theory of quantum measurement @xcite . to create other quasi - projectors which are more localized in phase - space one could use instead the marginally overcomplete set of coherent states @xmath195 where one places an ordinary glauber coherent state @xmath196 at each point of a lattice in phase space @xcite . we shall calculate ( [ cheq6 ] ) for an arbitrary @xmath197 by decomposing @xmath197 into coherent states and using ( [ eq20 ] ) to evolve to @xmath198 . we will then consider a @xmath197 resulting from the propagation from a time @xmath199 of @xmath200 which we shall choose to be @xmath201 . we will obtain the decoherence functional @xmath202 and from this calculate the degree of consistency @xmath179 and peaking @xmath203 about the classical trajectory . we will also calculate @xmath204 , a measure similar to ( [ eq27 ] ) , which measures the coherence in @xmath202 when considered as a density matrix . we find that @xmath179 and @xmath204 behave similarly with respect to their dependence on the parameters in the model . to begin , we take @xmath205 where we have normalized according to @xmath206 . using ( [ appendix1 ] ) , @xmath207 and the definition of a coherent state @xmath208 we can compute @xmath209 where @xmath210 and @xmath211\right\}\;\;. \label{cheq12}\end{aligned}\ ] ] using ( [ eq20 ] ) we can evolve ( [ cheq10 ] ) from @xmath212 to @xmath213 and denoting this by @xmath214 we get @xmath215 where @xmath216 where @xmath217 and @xmath218 . the final decoherence functional is thus @xmath219 and is thus a gaussian smearing of the q function of @xmath214 . note that although q functions are strictly positive for true density matrices , when @xmath220 , @xmath214 is _ not _ a true density matrix and thus @xmath221 $ ] may be complex . to calculate ( [ cheq15 ] ) is quite tedious but one finally arrives at @xmath222 where @xmath223 and @xmath224 . for simplicity we will now turn to the interaction picture and consider our projectors to be stationary in this picture ie . @xmath225 . with this , all @xmath26 dependence drops out of ( [ cheq16 ] ) . with a bit of rearranging ( [ cheq16 ] ) can be written in the more convenient form @xmath226where @xmath227 and @xmath228 we are now ready to choose @xmath201 . from ( [ eq1.1 ] ) we have @xmath229 where @xmath230 . since the action of the decoherence functional is linear we can use ( [ cheq17 ] ) to obtain the action on @xmath231 and can then perform the integrations in ( [ cheq20 ] ) to compute the complete decoherence functional . we finally get @xmath232 where the definitions of the various quantities appearing in ( [ cheq21 ] ) are @xmath233\;,$\\ $ s_2=\sigma_2 ^ 2+\kappa_2 \;,$ & \\ $ \kappa_2=1+\bar{n}(1-e^{-\gamma \delta t_2 } ) \;,$ & $ 2\upsilon{\bbox{\bbb a}}=z\left\ { \left [ 1-\frac{1}{s_2}e^{-\gamma\delta t_2}(1+\phi)\right]\delta_1^ * + \frac{2}{s_2}[s_1+\phi]e^{-\gamma\delta t_2/2}\delta_2^*\right\}\;,$\\ $ \kappa_1=1+\bar{n}(1-e^{-\gamma \delta t_1})\;.$ & \\ \end{tabular}\label{cheq22}\ ] ] thus @xmath234 and @xmath235 are the deviations of the projector positions at times @xmath212 and @xmath198 from the classical path of the damped coherent state @xmath236 . the diagonal elements of @xmath202 , ie . @xmath237 , are peaked along this classical path ( @xmath238 ) . we also note that because of the nonexclusivity of these quasi - projectors the diagonal elements of the decoherence functional do not sum exactly to one . for a decoherence functional of the form ( [ cheq21 ] ) it easy to show that the degree @xmath179 of consistency defined in ( [ epsilon ] ) is @xmath239 thus the maximum departure from perfect consistency between a pair of histories occurs for ( i ) those histories for which @xmath240 is such that @xmath241 ( ii ) for pairs of histories @xmath242 where the quasi - projectors @xmath243 and @xmath244 are separated by their full width at half maximum ( fwhm ) ie . @xmath245 . from ( i ) , ( ii ) and ( [ cheq23 ] ) we get @xmath246\right\}\nonumber\\ & \equiv & \exp\{-\bar{\epsilon}(\sigma_1,\sigma_2,\gamma,\delta t_2,\delta t_1,\bar{n})\}\;\;.\label{cheq24}\end{aligned}\ ] ] for @xmath247 we achieve good consistency . to examine the degree of peaking about the classical path we follow dowker and halliwell @xcite and look at the product of the determinant of the coefficients of @xmath248 with the fwhm of the initial and final projectors , that is @xmath249 . we get @xmath250\;\;.\label{cheq25}\ ] ] for @xmath251 we get significant peaking about the classical path . equations ( [ cheq21 ] , [ cheq22 ] , [ cheq24 ] , [ cheq25 ] ) are the main results of this section . let us examine the short and long time behaviour of @xmath252 and @xmath203 . for @xmath253 we have @xmath254\;\;. \label{cheq26}\ ] ] thus as the final graining size @xmath184 increases the degree of consistency increases . this is in agreement with the general arguments of dowker and halliwell . however , as the initial graining size @xmath255 is increased , the degree of consistency drops . this anomalous behaviour is also seen in other calculations @xcite . in the long time limit , with @xmath256 and letting @xmath257 we have @xmath258}\;\;,\label{cheq27}\ ] ] which again decreases with increasing @xmath255 . however , the dependence on bath temperature , through @xmath147 , is now apparent and thus the consistency increases with increasing @xmath183 . in the short time limit the peaking is @xmath259 while for long times @xmath260 ^ 2 } \left(\frac{1+\sigma_2 ^ 2}{1+\bar{n}+\sigma_2 ^ 2}\right)\;\;.\label{cheq29}\ ] ] the short time and long time limits of @xmath203 display the expected behaviour with the peaking getting better with increasing graining size and decreasing bath temperature and time . plots of the long time behaviors of @xmath252 and @xmath203 are shown in fig 2 . ( a)(b ) + # myfig2.ps#5.4#myfig2.ps ( 5,7 ) ( 0,0)(5,7 ) ( 0,7 ) ( 0,0 ) ( -2,-8 ) # myfig3.ps#5.4#myfig3.ps ( 5,7 ) ( 0,0)(5,7 ) ( 0,7 ) ( 0,0 ) ( 1,-8 ) we can also examine the degree of consistency by considering the decoherence functional as a density matrix in the space of the coarse grained histories @xmath261 . with this viewpoint we can adopt the definition ( [ eq27 ] ) to give a measure of the concentration of @xmath262 $ ] about the diagonal @xmath263 . for a decoherence functional of the form @xmath264\;\;,\ ] ] we can calculate @xmath265{\rm d}[\delta_2,\delta_1,-z]|z|^2 } { \int\,d^2\delta_1d^2z\ , { \rm d}[\delta_2,\delta_1,z]{\rm d } [ \delta_2,\delta_1,-z]|\delta_1-\langle\delta_1\rangle|^2}\;\;.\label{cheq30}\]]we can motivate this measure by noting that for each value of @xmath235 , @xmath266 $ ] is very like a density matrix with the on - diagonals labeled by @xmath234 and the off - diagonals labeled by @xmath267 . computing ( [ cheq30 ] ) for a given @xmath235 gives a measure of the degree of coherence between all the possible pairs of histories which have @xmath268 as their final quasi - projector . the maximum of ( [ cheq30 ] ) over @xmath235 gives the largest violation of consistency for the complete set of histories labeled by both @xmath235 and @xmath234 . before computing ( [ cheq30 ] ) we first obtain @xmath269 for a fixed @xmath270 to be @xmath271\delta_1\nonumber\\ & = & \frac{nc\delta_2}{a^2}e^ { -(ab - c^2)|\delta_2|^2/a}\;\;,\end{aligned}\ ] ] which gives @xmath272 when @xmath273 as expected . after some calculation one obtains the maximum violation of consistency to simply be @xmath274}{\left [ 1 + 2s_1-\frac{1}{s_2}e^{-\gamma \delta t_2}+\upsilon\left ( 4-\frac{1}{s_2}[3 - 2s_1]e^{-\gamma \delta t_2}\right)\right ] } \;\;.\label{cheq31}\end{aligned}\ ] ] in the short and long time limits we get @xmath275 from ( [ cheq33 ] ) we see that even when @xmath276 the coherence decays . this is due to the vacuum fluctuations . it is interesting to note that all dependence on @xmath277 drops out of these expressions in the long time limit . plots of @xmath204 for @xmath278 ( [ cheq31 ] ) , are shown in fig 3 . very similar dependence on the model parameters is shown between @xmath204 and @xmath252 , in particular the anomalous @xmath255 behaviour . ( a)(b ) + # myfig4.ps#5.4#myfig4.ps ( 5,7 ) ( 0,0)(5,7 ) ( 0,7 ) ( 0,0 ) ( -2,-8 ) # myfig5.ps#5.4#myfig5.ps ( 5,7 ) ( 0,0)(5,7 ) ( 0,7 ) ( 0,0 ) ( 1,-8 ) to conclude and summarize : we have examined the consistent histories framework within a quantum optical model using two - time projected phase - space histories . in the markovian regime the bath can be traced away and it s effect completely absorbed into the non - unitary evolution super - operator @xmath172 . this is used to evolve the reduced density matrix of the system between projections . since no true phase - space projectors exist we use `` quasi - projectors '' which are gaussian convolutions of coherent states . although these quasi - projectors are complete , they are not totally exclusive and therefore do not exactly meet the specifications of this paradigm . they , however , do provide a tractable model and have been used successfully in other calculations . moreover , such quasi - projectors arise naturally in more rigorous theories of quantum measurement . we first calculated the `` action '' of the decoherence functional for these coarse grained histories on the state @xmath231 at @xmath212 . with this information , the action of the decoherence functional on any state at @xmath212 can be evaluated through a decomposition of @xmath197 into coherent states . this was done for the particular case of @xmath201 where @xmath279 . the resulting decoherence functional was exponential in form and peaked on the classical path of the damped oscillator . using the measures of dowker and halliwell the degree of consistency and the degree of peaking about the classical path were calculated and the short and long time limits found . the behaviour of these quantities were mostly as expected : the consistency gets better with time , increasing bath temperature and final grain size while the peaking is better with increased final and initial grain size and small bath temperature . the consistency _ decreased _ with increased initial grain size . we also calculated @xmath204 , a measure similar to the one proposed in the eis section of this paper , which quantifies the amount of coherence in the decoherence functional when the decoherence functional is considered as a density matrix on the space of coarse grained histories . this measure displayed similar behaviour to the measure of consistency as advanced by dowker and halliwell . upon comparison we see that both eis and consistency among phase - space histories improves as the bath temperature increases . altering the coupling strength @xmath180 , changes only the time derivatives of the degree of eis or consistency achieved and not their actual values . a time for `` decoherence '' may be set equal to the time needed for @xmath155 to reach some predetermined value ie . @xmath280 , say . the anomalous dependence of @xmath252 on @xmath255 occurs in all the two - time projected history calculations to date . for this to occur in one model for a special type of graining might be understandable within the arguments given by dowker and halliwell concerning the constructive interference of the constituent histories in a very special coarse - graining . since the behaviour is seen in three different calculations with two different models @xcite one suspects that there is some feature common to all that is upsetting the expected behaviour of increased consistency with increased @xmath255 . an approximation common to all of these calculations is the use of `` quasi - projectors '' . it may be that the non - exclusivity of these projectors is somehow overcounting the contributions of `` off - diagonal '' histories in the decoherence functional . it is thus necessary to resort to true projectors . since there do not exist true projectors on phase - space one must consider true projectors on to other spaces . an obvious choice is either the position , momentum or number basis . choosing the latter is the most computationally efficient and was treated in @xcite . the results were very surprising and highlighted the differences between the two paradigms , eis and ch . essentially the goal of eis is the damping of the coherences present in the system s quantum state . the goal of ch , however , is the attainment of histories whose probabilities satisfy the sum rule . for this to occur the quantum _ interferences _ between the individual histories must vanish . the damping of quantum coherence and vanishing of quantum interferences need not be synonymous . the decay of coherence necessitates the vanishing of interference but not visa versa . this is clearly shown in the double slit examples of gell - mann and hartle . when the slit separation is much greater than the slit / screen separation the resulting coarse - grained histories are very consistent . however , if the separate beams are then later on recombined somehow , they will display interference . one then says that the histories which include this recombination are no longer consistent . thus , histories which may begin very consistent , may , later on , become very inconsistent and visa versa . this is so , even though the evolution can be strictly unitary ! from the viewpoint of eis , the coherence of the state is not lost under unitary evolution and thus eis does not occur . however , if one now inserts the young s slits in a bath of electrons @xcite , the resulting histories are even more consistent and furthermore they can not usually be made less consistent through recombination etc . this is so because the reduced state of the light beams has lost some of it s coherence to the bath of electrons . thus , both ch and eis has occurred in the reduced state of the light beams . since the measure @xmath155 was introduced in section iii as a measure of coherence , its value as a measure of consistency ( which involves the vanishing of interferences ) is slightly suspect . however , we would argue that the vanishing of interference between _ all _ pairs of possible histories in a set of very many alternatives necessitates the vanishing of interference between neighboring histories . for this to happen , as in the young slit case for small slit separation , one must have a loss of coherence ( usually by introducing a bath ) . finally , it would be of greater interest to have a more information theoretic formulation of the concept of `` diagonalisation '' and consistency . the introduction of a bath causes the reduced quantum state to lose it s non - local correlations preferentially in certain bases . to link this concept with that of a pointer basis - that basis least effected through the bath interaction - would be of interest . also to see how the decohered state ( either via eis or ch ) satisfies the bell inequalities would give , perhaps , a more physical measure of what `` diagonality '' really means . this work is in progress @xcite . also , the quantification of classical correlations via information theory in the ch approach is also important @xcite . in this paper we have shown that in this relatively simple model of a quantum open system eis and ch seem to behave similarly , becoming better with increasing bath temperature and time . the dependence of the consistency on graining also behaves almost as expected . this example shows how the two paradigms are similar . other calculations also demonstrate the differences between the two @xcite . clearly , if both paradigms are to describe an ontology it is essential to discover situations where they differ and to experimentally determine which is correct . the author thanks f. dowker , s. habib , b. l. hu r. laflamme , a. matacz , j. mccarthy , g. milburn , j. p. paz , m. simpson , w. unruh and w. zurek for interesting discussions . we also thank the university of british colombia , los alamos national laboratories and the university of maryland for their kind hospitality during the course of this research . in this appendix we compile a number of integral results useful for calculations involving coherent states . we refer the reader to @xcite for a very complete review of coherent states and their properties . here we list a number of results . @xmath281 @xmath282\,e^{bc / a}\;\;,\\ \int\frac{d^2\alpha}{\pi}\,|\alpha|^4\,e^{-a|\alpha|^2+b\alpha^*+c\alpha}&= & \frac{1}{a^5}\left[2a^2 + 4abc+b^2c^2\right]\,e^{bc / a}\;\;.\\\end{aligned}\ ] ] @xmath283+d_1\alpha_1 + e_1\alpha_1^{*\,2}+f_1\alpha_1 ^ 2\right.\nonumber\\ & & \left .- a_2|\alpha_2|^2+c_2\alpha_2^*+d_2\alpha_2 + e_2\alpha_2^{*\,2}+f_2\alpha_2 ^ 2\right\}= \frac{1}{\sqrt{\eta}}\,\exp(\lambda/\eta)\;\;,\end{aligned}\ ] ] where @xmath284 @xmath285 in this appendix we calculate the quantity @xmath286 . this is used in ( [ xy1 ] ) to obtain the resolution of the density matrix in the position basis . from the group properties of coherent states and the baker - campbell - hausdorff expansion we can get @xmath287|y\rangle \label{in1}\\ & = & e^{|\xi|^2/2+(\xi^{*\,2}-\xi^2)/4 } \exp\left[x\sqrt{\frac{\omega}{2\hbar}}(\xi^*-\xi)\right ] \langle x|\exp \left[i\frac{\hat{p}}{\hbar}\sqrt{\frac{\hbar}{2\omega}}(\xi^*+\xi)\right ] @xmath288 , @xmath289 in terms of @xmath290 and @xmath291 and again used bch identities to dis - entangle @xmath292/2}$ ] . now since @xmath293 the inner product on the right hand side of ( [ in2 ] ) becomes the delta function @xmath294 \delta(\xi_x+\sqrt{\frac{\omega}{2\hbar}}(x - y ) ) \sqrt{\frac{\omega}{2\hbar}}\;\;,\ ] ] where @xmath295 . collecting the terms we finally get , @xmath296\delta(\xi_x+\sqrt{\frac{\omega}{2\hbar}}(x - y ) ) \sqrt{\frac{\omega}{2\hbar}}\;\;.\ ] ] j. b. hartle , preprint ucsbth92 - 91 _ `` spacetime quantum mechanics and the quantum mechanics of spacetime '' _ , preprint ucsbth92 - 91 , to appear in proceedings of the 1992 les houches summer school , _ gravitation et quantifications _ , gr - qc/9304006 ( unpublished ) . m. gell - mann and j. b. hartle , in _ proceedings of the 25th international conference on high energy physics _ , south east asia theoretical physics association and physics society of japan , edited by k. k. phua and y. yamaguchi ( world scientific , singapore , 1990 ) . j. b. hartle , in _ quantum cosmology and baby universes _ , jerusalem winter school for theoretical physics , edited by s. coleman , j. b. hartle , t. piran , and s. weinberg ( world scientific , singapore , 1991 ) , pp . |
for any finite simple graph , the discrete dirac operator and the laplace - beltrami operator on the exterior algebra bundle are finite matrices , where , with denoting the cardinality of the set of complete subgraphs of . we prove the mckean - singer formula which holds for any complex time , where is the euler characteristic of . the super trace of the heat kernel interpolates so the euler - poincar formula for with the hodge theorem in the real limit . more generally , for any continuous complex valued function satisfying , one has the formula . this includes for example the schrdinger evolutions on the graph . after stating some immediate general facts about the spectrum which includes a perturbation result estimating the dirac spectral difference of graphs , we mention as a combinatorial consequence that the spectrum encodes the number of closed paths in the simplex space of a graph . we give a couple of worked out examples and see that mckean - singer allows to find explicit pairs of nonisometric graphs which have isospectral dirac operators . | 1301.1408 | some classical results in differential topology or riemannian geometry have analogue statements for finite simple graphs . examples are gauss - bonnet @xcite , poincar - hopf @xcite , riemann - roch @xcite , brouwer - lefschetz @xcite or lusternik - schnirelman @xcite . while the main ideas of the discrete results are the same as in the continuum , there is less complexity in the discrete . + we demonstrate here the process of discretizing manifolds using graphs for the * mckean - singer formula * @xcite @xmath19 where @xmath20 is the laplacian on the differential forms and @xmath21 is the euler characteristic of the graph @xmath9 . the formula becomes in graph theory an elementary result about eigenvalues of concrete finite matrices . the content of this article is therefore teachable in a linear algebra course . chunks of functional analysis like elliptic regularity which are needed in the continuum to define the eigenvalues are absent , the existence of solutions to discrete partial differential equations is trivial and no worries about smoothness or convergence are necessary . all we do here is to look at eigenvalues of a matrix @xmath22 defined by a finite graph @xmath9 . the discrete approach not only recovers results about the heat flow , we get results about unitary evolutions which for manifolds would require need analytic continuation : also the super trace of @xmath23 where @xmath20 is the laplace - beltrami operator is the euler characteristic . while @xmath24 solves @xmath25 , the dirac wave equation @xmath26 is solved by @xmath27 where @xmath28 is on the orthocomplement of the zero eigenspace . writing @xmath29 we have @xmath30 . the initial position and velocity of the real flow are naturally encoded in the complex wave function . that works also for a discrete time schrdinger evolution like @xmath31 @xcite if @xmath22 is scaled . + despite the fact that the dirac operator @xmath22 for a general simple graph @xmath0 is a natural object which encodes much of the geometry of the graph , it seems not have been studied in such an elementary setup . operators in @xcite have no relation to the dirac matrices @xmath22 studied here . the operator @xmath32 on has appeared in a more general setting : @xcite builds on work of dodziuk and patodi and studies the combinatorial laplacian @xmath33 acting on ech cochains . this is a bit more general : if the cover of the graph consists of unit balls , then the ech cohomology by definition agrees with the graph cohomology and the ech dirac operator agrees with @xmath22 . notice also that for the laplacian @xmath34 on scalar functions , the factorization @xmath35 is very well known since @xmath36 is the incidence matrix of the graph . the matrices @xmath37 have been used by poincar already in 1900 . + one impediment to carry over geometric results from the continuum to the discrete appears the lack of a hodge dual for a general simple graph and the absence of symmetry like poincar duality . does some geometric conditions have to be imposed on the graph like that the unit spheres are sphere - like graphs to bring over results from compact riemannian manifolds to the discrete ? the answer is no . the mckean - singer supersymmetry for eigenvalues holds in general for any finite simple graph . while it is straightforward to implement the dirac operator @xmath22 of a graph in a computer , setting up @xmath22 can be tedious when done by hand because for a complete graph of @xmath38 nodes , @xmath22 is already a @xmath39 matrix . for small networks with a dozen of nodes already , the dirac operator acts on a vector space on vector spaces having dimensions going in to thousands . therefore , even before we had the computer routines were in place which produce the dirac operator from a general graph , computer algebra software was necessary to experiment with the relatively large matrices first cobbled together by hand . having routines which encode the dirac operator for a network is useful in many ways , like for computing the cohomology groups . such code helped us also to write @xcite . homotopy deformations of graphs or nerve graphs defined by a ech cover allow to reduce the complexity of the computation of the cohomology groups . + some elementary ideas of noncommutative geometry can be explained well in this framework . let @xmath40 denote the hilbert space defined by the simplex set @xmath41 of the graph and let @xmath42 denote the banach algebra of operators on @xmath43 . it is a hilbert space itself with the inner product @xmath44 . the operator @xmath45 defines together with a subalgebra @xmath46 of @xmath42 , a * connes distance * on @xmath41 . classical geometry is when @xmath47 is the algebra of diagonal matrices . in this case , the connes metric @xmath48| \leq 1}(l , e_x - e_y)$ ] between two vertices is the geodesic shortest distance but @xmath49 extends to a metric on all simplices @xmath41 . the advantage of the set - up is that other algebras @xmath46 define other metrics on @xmath41 and more generally on * states * , elements in the unit sphere @xmath50 of @xmath51 . in the classical case , when @xmath52 , all states are * pure states * , unit vectors @xmath53 in @xmath43 and correspond to states @xmath54 in @xmath51 . for a noncommutative algebra @xmath46 , elements in @xmath50 are in general * mixed states * , to express it in quantum mechanics jargon . we do not make use of this here but note it as additional motivation to study the operator @xmath22 in graph theory . + the spectral theory of jacobi matrices and schrdinger operators show that the discrete and continuous theories are often very close but that it is not always straightforward to translate from the continuum to the discrete . the key to get the same results as in the continuum is to work with the right definitions . for translating between compact riemannian manifolds and finite simple graphs , the dirac operator can serve as a link , as we see below . while the result in this paper is purely mathematical and just restates a not a priory obvious symmetry known already in the continuum , the story can be interesting for teaching linear algebra or illustrating some ideas in physics . on the didactic side , the topic allows to underline one of the many ideas in @xcite , using linear algebra tools only . on the physics side , it illustrates classical and relativistic quantum dynamics in a simple setup . only a couple of lines in a modern computer algebra system were needed to realize the dirac operator of a general graph as a concrete finite matrix and to find the quantum evolution in equation ( [ diracsolution ] ) as well as concrete examples of arbitrary members of cohomology classes by hodge theory . the later is also here just a remark in linear algebra as indicated in an appendix . for a finite simple graph @xmath0 , the exterior bundle @xmath55 is a finite dimensional vector space of dimension @xmath56 , where @xmath6 is the cardinality of @xmath7 the set of @xmath57 simplices contained in @xmath9 . as in the continuum , we have a super commutative multiplication @xmath58 on @xmath59 but this algebra structure is not used here . the vector space @xmath60 consists of all functions on @xmath7 which are antisymmetric in all @xmath61 arguments . the bundle splits into an even @xmath62 and an odd part @xmath63 which are traditionally called the bosonic and fermionic subspaces of @xmath59 . the exterior derivative @xmath64 satisfies @xmath65 . while a concrete implementation of @xmath66 requires to give an orientation of each element in @xmath41 , all quantities we are interested in do not depend on this orientation however . it is the usual choice of a basis ambiguity as it is custom in linear algebra . given a graph @xmath0 , define the * dirac operator * @xmath1 and * laplace - beltrami * operator @xmath67 . the operators @xmath68 leaving @xmath69 invariant are called the laplace - beltrami operators @xmath70 on @xmath71-forms . * examples . * we write @xmath72 to indicate multiplicity @xmath73 . + * 1 ) * for the complete graph @xmath74 the spectrum of @xmath22 is @xmath75 . + * 2 ) * for the circle graph @xmath76 the spectrum is @xmath77 , where the notation understands that @xmath78 has multiplicity @xmath79 . the product of the nonzero eigenvalues , a measure for the complexity of the graph is @xmath80 . + * 3 ) * for the star graph @xmath81 the eigenvalues of @xmath22 are @xmath82 . the product of the nonzero eigenvalues is @xmath38 . + * 4 ) * for the linear graph @xmath83 with @xmath38 vertices and @xmath84 edges , the eigenvalues of @xmath22 are the union of @xmath85 , and @xmath86 , where @xmath87 is the @xmath88 matrix which has @xmath79 in the diagonal and @xmath89 in the side diagonal . the product of the nonzero eigenvalues is @xmath90 . + * remarks . * + * 1 ) * the scalar part @xmath34 is one of the two most commonly used laplacians on graphs @xcite . the operator @xmath20 generalizes it to all @xmath71-forms , as in the continuum . + * 2 ) * when @xmath37 is implemented as a matrix , it is called a signed * incidence matrix * of the graph . since it depends on the choice of orientation for each simplex , also the dirac operator @xmath22 depends on this choice of basis . both @xmath20 and the eigenvalues of @xmath22 do not depend on it . + * 3 ) * the laplacian for graphs has been introduced by kirkhoff in 1847 while proving the matrix - tree theorem . the computation with incidence matrices is as old as algebraic topology . poincar used the incidence matrix @xmath66 already in 1900 to compute betti numbers @xcite . + * 4 ) * while @xmath22 exchanges bosonic and fermionic spaces @xmath91 , the laplacian @xmath20 is the direct sum @xmath92 of laplacians on @xmath93 forms . + * 5 ) * the kernel of @xmath94 is called the vector space of * harmonic @xmath93 forms*. such harmonic forms represent cohomology classes . by hodge theory ( see appendix ) the dimension of the kernel is equal to the @xmath93th betti number , the dimension of the @xmath93th cohomology group @xmath95 . + * 6 ) * especially , @xmath34 is the graph laplacian @xmath96 , where @xmath97 is the diagonal matrix containing the vertex degrees and @xmath46 is the adjacency matrix of the graph . the matrix @xmath34 is one of the natural laplacians on graphs @xcite . + * 7 ) * while @xmath98 always has @xmath99 in the diagonal , the diagonal entries @xmath100 counts the number of @xmath101 simplices attached to the @xmath71 simplex @xmath102 and @xmath103 is always equal to @xmath99 . the diagonal entries @xmath104 therefore determine the number of @xmath61 dimensional simplices attached to a @xmath93 dimensional simplex @xmath102 . we will see that the dirac and laplacian matrices are useful for combinatorics when counting closed curves in @xmath41 in the same way as adjacency matrices are useful in counting paths in @xmath9 . + * 8) * the operator @xmath22 can be generalized as in the continuum to to bundles . the simplest example is to take a @xmath89-form @xmath105 and to define @xmath106 . this is possible since @xmath59 is an algebra . we have then the generalized dirac operator @xmath107 and @xmath108 . the * curvature operator * @xmath109 is no more zero in general . mckean - singer could generalizes to this more general setup as @xmath110 provides the super - symmetry . + * 9 ) * a @xmath89-form @xmath46 defines a field @xmath111 which satisfies the * maxwell equations * @xmath112 . physics asks to find the field @xmath113 , given @xmath114 . while this corresponds to the @xmath115-current in classical electro magnetism which includes charge and electric currents , it should be noted that no geometric structure is assumed for the graph . the maxwell equations hold for any finite simple graph . it defines the evolution of light on the graph . the equation @xmath116 can in a coulomb gauge where @xmath117 be written as @xmath118 where @xmath119 is the laplacian on @xmath89 forms . given a current @xmath114 , we can get @xmath46 and so the electromagnetic field @xmath113 . by solving a system of linear equations . this is possible if @xmath114 is perpendicular to the kernel of @xmath119 which by hodge theory works if @xmath9 is simply connected . linear algebra determines then determines the field @xmath113 , a function on triangles of a graph . as on a simply connected compact manifold , a simply connected graph does not feature light in vacuum since @xmath120 has only the solution @xmath121 . + * 10 ) * dirac introduce the dirac operator @xmath22 in the continuum to make quantum mechanics lorentz invariant . the quantum evolution becomes wave mechanics . already in the continuum , there is mathematically almost no difference between the wave evolution with given initial position and velocity of the wave the schrdinger evolution with a complex wave because both real and imaginary parts of @xmath122 solve the wave equation . the only difference between schrdinger and wave evolution is that the initial velocity @xmath53 of the wave must be in the orthocomplement of the zero eigenvalue . ( this has to be the case also in the continuum , if we hit a string , the initial velocity has to have average zero momentum in order not to displace the center of mass of the string . ) a similar equivalence between schrdinger and wave equation holds for a time - discretized evolution @xmath123 . + * 11 ) * an other way to implement a bundle formalism is to take a unitary group @xmath124 and replace entries @xmath89 in @xmath66 with unitary elements @xmath125 and @xmath126 with @xmath127 . now @xmath1 and @xmath128 are selfadjoint operators on @xmath129 for some @xmath130 and can be implemented as concrete @xmath131 matrices . the gauge field @xmath125 defines now curvatures , which is a @xmath124-valued function on all triangles of @xmath41 . again , the more general operator @xmath22 is symmetric and supersymmetry holds for the eigenvalues . actually , as @xmath132 can be expressed as a sum over closed paths , the eigenvalues of @xmath110 are just @xmath130 copies of the eigenvalues of @xmath22 if the multiplicative curvatures are @xmath89 ( zero curvature ) and the graph is contractible . since @xmath133 for odd functions @xmath15 and @xmath134 is independent of @xmath125 , the simplest functional on @xmath125 which involves the curvature is the wilson action @xmath135 which is zero if the curvature is zero . it is natural to ask to minimize this over the compact space of all fields @xmath46 . more general functionals are interesting like @xmath136 , the product of the nonzero eigenvalues of @xmath20 ; this is an interesting integral quantity in the flat @xmath137 case already : it is a measure for complexity of the graph and combinatorially interesting since @xmath138 is equal to the number of spanning trees in the graph @xmath9 . + let @xmath0 be a finite simple graph . if @xmath74 denotes the complete graph with @xmath38 vertices , we write @xmath139 for the set of @xmath140 subgraphs of @xmath9 . the set of all simplices @xmath141 is the super graph of @xmath9 on which the dirac operator lives . the graph @xmath9 defines @xmath41 without addition of more structure but the additional structure is useful , similarly as tangent bundles are useful for manifolds . it is useful to think of elements in @xmath41 as elementary units . if @xmath142 is the cardinality of @xmath139 then the finite sum @xmath143 is called the euler characteristic of @xmath9 . + the set @xmath144 of antisymmetric functions @xmath15 on @xmath139 is a linear space of @xmath93-forms . the exterior derivative @xmath145 defined by @xmath146 is a linear map which satisfies @xmath147 . it defines the cohomology groups @xmath148 of dimension @xmath149 for which the euler - poincar formula @xmath150 holds . the direct sum @xmath151 is the discrete analogue of the exterior bundle . + mckean and singer have noticed in the riemannian setup that the heat flow @xmath152 solving @xmath153 has constant super trace @xmath154 . the reason is that the nonzero bosonic eigenvalues can be paired bijectively with nonzero fermionic eigenvalues . it can be rephrased that @xmath22 is an isomorphism from @xmath155 , where @xmath156 is the orthogonal complement of the zero eigenspace in @xmath157 and @xmath158 the orthogonal complement of the zero eigenspace in @xmath159 . in the limit @xmath14 we get the dimensions of the harmonic forms ; in the limit @xmath160 we obtain the simplicial euler characteristic . for any complex valued continuous function @xmath15 on the real line satisfying @xmath16 we have @xmath161 especially , @xmath162 for any complex @xmath11 . since @xmath1 is a symmetric @xmath4 matrix , the kernel of @xmath22 and the kernel of @xmath128 are the same . the functional calculus defines @xmath163 for any complex valued continuous function . since @xmath22 is normal , @xmath164 , it can be diagonalized @xmath165 with a diagonal matrix @xmath166 and defines @xmath167 for any continuous function . because @xmath15 can by weierstrass be approximated by polynomials and the diagonal entries of @xmath168 are empty , we can assume that @xmath169 for an even function @xmath170 . let @xmath171 be the subspace of @xmath59 spanned by nonzero eigenvalues of @xmath20 . let @xmath172 be the subspace generated by eigenvectors to nonzero eigenvalues of @xmath20 on @xmath69 and @xmath173 be the subspace generated by eigenvectors to nonzero eigenvalues of @xmath20 on @xmath159 . since @xmath22 commutes with @xmath20 , each eigenvector @xmath15 of @xmath174 on @xmath159 has an eigenvector @xmath175 of @xmath174 on @xmath69 . since @xmath176 is invertible , there is a bijection between fermionic and bosonic eigenvalues . each nonzero eigenvalue appears the same number of times on the fermionic and bosonic part . * remarks . * + * 1 * by definition , @xmath177 agrees with the euler characteristic . it can be seen as an * analytic index * @xmath178 of the restricted dirac operator @xmath179 because @xmath180 is the space of harmonic states in @xmath157 and @xmath181 is the dimension of the fermionic harmonic space . + * 2 ) * the original mckean - singer proof works in the graph theoretical setup too as shown in the appendix . + * 3 ) * in @xcite we have for numerical purposes discretized the schroedinger flow . this shows that we can replace the flow @xmath182 by a map @xmath183 with a suitably rescaled @xmath22 which is dynamically similar to the unitary evolution and has the property that the system has * finite propagation speed*. the operator @xmath184 has the super trace @xmath185 so that also this discrete time evolution satisfies the mckean - singer formula . + * 4 ) * the heat flow proof interpolating between the identity and the projection onto harmonic forms makes the connection between euler - poincar and hodge more natural . while for @xmath13 and @xmath186 we have a lefshetz fixed point theorem , for @xmath187 it can be seen as an application of the atiyah - bott generalization of that fixed point theorem . in the discrete , atiyah - bott is very similar to brouwer - lefshetz @xcite . for @xmath13 the lefshetz fixed point theorem expresses the lefshetz number of the identity as the sum of the fixed points by poincar - hopf . in the limit @xmath14 , the lefshetz fixed point theorem sees the lefshetz number as the signed sum over all fixed points which are harmonic forms . for positive finite @xmath11 we can rephrase mckean - singer s result that the lefshetz number of the dirac bundle automorphism @xmath152 is time independent . we start with a few elementary facts about the operator @xmath22 : let @xmath188 denote the eigenvalues of @xmath22 and let @xmath189 denote the maximal degree of @xmath9 . if @xmath190 is an eigenvalue , then @xmath191 is an eigenvalue too so that @xmath192=\sum_i \lambda_i=0 $ ] . if we split the hilbert space as @xmath193 then @xmath194 $ ] , where @xmath195 is the annihilation operator and @xmath51 which maps bosonic to fermionic states and @xmath46 is the creation operator . define @xmath196 $ ] . then @xmath197 is called supersymmetry in 0 space dimensions ( see @xcite ) . if @xmath198 , then @xmath199 . apply @xmath200 again to this identity to get @xmath201 . this shows that @xmath202 is an other eigenvector . every eigenvalue @xmath190 is contained in @xmath203 $ ] so that every eigenvalue of @xmath128 is contained in the interval @xmath204 $ ] and @xmath205 = \sum_{i=1}^n \lambda_i^2/n \leq 2\deg$ ] . because all entries of @xmath22 are either @xmath126 and @xmath89 and because there are @xmath66 such entries , each row or column has maximal length @xmath206 . that the standard deviation is @xmath207 follows since the `` random variables '' @xmath208 take values in @xmath203 $ ] . also a theorem of schur @xcite confirms that @xmath209 . * remarks*. + * 1 ) * it follows that the spectrum of @xmath22 is determined by the spectrum of @xmath20 . if @xmath210 $ ] are the eigenvalues of @xmath20 , then @xmath211 are the eigenvalues of @xmath22 . it follows already from the reflection symmetry of the eigenvalues of @xmath22 that the positive eigenvalues of @xmath20 appear * in pairs*. the mckean - singer statement is stronger than that . it tells that if one member of the pair is in the bosonic sector , the other is in the fermionic one . + * 2 ) * the schur argument given in the proof is not quite irrelevant when looking at spectral statistics . since the average degree in the erdoes - rnyi probability space @xmath212 is of the order @xmath213 the standard deviation of the eigenvalues is of the order @xmath214 even so we can have eigenvalues arbitrarily close to @xmath215 . + * 3 ) * numerical computations of the spectrum of @xmath22 for large random matrices is difficult because the dirac matrices are much larger than the adjacency matrices of the graph . it would be interesting however to know more about the distribution of the eigenvalues of @xmath22 for large matrices . + * 4 * sometimes , an other laplacian @xmath216 defined as follows for graphs . let @xmath217 denote the degree of a vertex @xmath102 . define operator @xmath218 if @xmath219 and @xmath220 if @xmath221 is an edge . this operator satisfies @xmath222 , where @xmath223 is the diagonal operator for which the only nonzero entries are @xmath224 if @xmath219 . while the spectrum of @xmath34 is in @xmath204 $ ] the spectrum of @xmath87 is in @xmath225 $ ] . since @xmath34 has integer entries , it is better suited for combinatorics . in the following examples of spectra we use the notation @xmath226 indicating that the eigenvalue @xmath190 appears with multiplicity @xmath38 . + for the complete graph @xmath140 the spectrum of @xmath216 is @xmath227 while the spectrum of @xmath34 is @xmath228 . for a cycle graph @xmath76 the spectrum of @xmath216 is @xmath229 while the spectrum of @xmath34 is @xmath230 . this square graph @xmath231 shows that the estimate @xmath232 $ ] is optimal . for a star graph @xmath81 which is an example of a not a vertex regular graph , the eigenvalues of @xmath216 are @xmath233 while the spectrum of @xmath34 is @xmath234 . + the number of zero eigenvalues of @xmath22 is equal to the sum of all the betti numbers @xmath235 . this follows from the hodge theorem ( see appendix ) : the dimension of the kernel of @xmath94 is equal to @xmath236 . the * dirac complexity * of a finite graph is defined as the product of the nonzero eigenvalues of its dirac operator @xmath22 . the euler - poincar identity assures that @xmath237 . if we ignore the signs , we get a number of interest : the number of nonzero eigenvalue pairs in @xmath22 is the * sign - less euler - poincar number * @xmath238 . the sum @xmath239 is the total number of eigenvalues and by the previous proposition , the sum @xmath240 is the total number of zero eigenvalues . since the eigenvalues of @xmath22 come in pairs @xmath241 , the number of pairs is the sign - less euler - poincar number . the sign - less euler - poincar number is even if and only if the dirac complexity is positive . arrange the product of nonzero eigenvalues of @xmath22 as a product of pairs @xmath242 . if @xmath9 is a triangularization of a sphere and which has an even number of edges , then the dirac complexity is positive . we have @xmath243 and since @xmath244 and @xmath245 we have @xmath246 . we can express now the sign - less euler - poincar number in terms of the number of edges : @xmath247/2 = [ ( 2 + 2 v_1 ) - 2]/2 = 2 v_1/2 = v_1 \ ; .\ ] ] the following explains why the dodecahedron or cube have negative complexity : if @xmath9 is a connected graph without triangles which has an even number of vertices , then the dirac complexity is negative . @xmath248 . the sign - less euler - poincar number is @xmath249/2 = [ 1-b_1 + 2v_1 - ( 1+b_1)]/2 = v_1-b_1 = v_0-b_0 = v_0 - 1 $ ] . for cyclic graphs @xmath76 the complexity is @xmath80 if @xmath38 is odd and @xmath250 if @xmath38 is even . for star graphs @xmath81 , the complexity is @xmath38 is odd and @xmath90 if @xmath38 is even . for a tree , the dirac complexity is positive if and only if the number of edges is even . the dirac complexity is still @xmath251 as in the previous proof but now @xmath245 so that it is @xmath252 . we have computed the complexity for all platonic , archimedian and catalan solids in the example section . all these 31 graphs have an even number @xmath252 of edges and an even number @xmath252 of vertices . next we estimate the distance between the spectra of different graphs : we have the following variant of lidskii s theorem which i learned from @xcite : for any two selfadjoint complex @xmath253 matrices @xmath46 and @xmath97 with eigenvalues @xmath254 and @xmath255 , one has @xmath256 denote with @xmath257 the eigenvalues of the selfadjoint matrix @xmath258 and let @xmath125 be the unitary matrix diagonalizing @xmath223 so that @xmath259 . we calculate @xmath260 the claim follows now from lidskii s inequality @xmath261 ( see @xcite ) this allows to compare the spectra of laplacians @xmath34 of graphs which are close . lets define the following metric on the erdoes - rnyi space @xmath262 of graphs of order @xmath38 on the same vertex set . denote by @xmath263 the number of edges at which @xmath9 and @xmath43 differ divided by @xmath38 . define also a metric between their adjacency spectra @xmath264 as @xmath265 if the maximal degree in either @xmath266 is @xmath189 , then the adjacency spectra distance can be estimated by @xmath267 we only need to verify the result for @xmath268 , since the left hand side is the @xmath269 distance between the vectors @xmath270 and the triangle inequality gives then the result for all @xmath263 . for @xmath268 , the matrix @xmath271 differs by @xmath89 only in @xmath272 entries . the lidskii lemma implies the result . * remarks . * + * 1 ) * we take a normalized distance between graphs because this is better suited for taking graph limits . + * 2 ) * as far as we know , lidskii s theorem has not been used yet in the spectral theory of graphs . we feel that it could have more potential , especially when looking at random graph settings @xcite . + 3 ) for the laplace operator @xmath34 this estimate would be more complicated since the diagonal entries can differ by more than @xmath89 . it is more natural therefore to look at the dirac matrices for which the entries are only @xmath273 or @xmath126 . + to carry this to dirac matrices , there are two things to consider . first , the matrices depend on a choice of orientation of the simplices which requires to chose the same orientation if both graphs contain the same simplex . second , the dirac matrices have different size because @xmath22 is a @xmath4 matrix if @xmath53 is the total number of simplices in @xmath9 . also this is no impediment : take the union of all simplices which occur for @xmath9 and @xmath43 and let @xmath53 denote its cardinality . we can now write down possibly * augmented * @xmath4 matrices @xmath274 which have the same nonzero eigenvalues than the original dirac operator . indeed , the later matrices are obtained from the augmented matrices by deleting the zero rows and columns corresponding to simplices which are not present . define the * spectral distance * between @xmath9 and @xmath43 as @xmath275 where @xmath276 are the eigenvalues of the augmented dirac matrices @xmath266 , which are now both @xmath4 matrices . define the * simplex distance * @xmath277 of two graphs @xmath266 with vertex set @xmath278 as @xmath279 times the number of simplices of @xmath266 which are different in the complete graph on @xmath278 . here @xmath53 is the total number of simplices in the union when both graphs are considered subgraphs of the complete graph on @xmath278 . the * maximal simplex degree * of a simplex @xmath102 of dimension @xmath93 in a graph @xmath9 is the sum of the number of simplices of dimension @xmath61 which contain @xmath102 and the sum of the number of simplices of dimension @xmath280 which are included in @xmath102 . in other words , the maximal simplex degree is the number of nonzero entries @xmath281 in a column @xmath102 of the incidence matrix @xmath66 . we get the following perturbation result : the dirac spectra of two graphs @xmath266 with vertex set @xmath278 satisfies @xmath282 if @xmath189 is the maximal simplex degree . since we have chosen the same orientation for simplices which are present in both graphs , this assures that the matrix @xmath283 has entries of absolute value @xmath284 and are nonzero only at @xmath285 , where an incidence happens exactly at one of the graphs @xmath266 . the value @xmath286 is an upper bound for @xmath287 , which by lidskii is an upper bound for the spectral distance @xmath288 . * remarks . * + * 1 ) * while the estimate is rough in general , it can become useful when looking at graph limits or estimating spectral distances between laplace eigenvalues of different regions . if we look at graphs of fixed dimensions like classes of triangularizations of a manifold @xmath289 , then @xmath189 is related to the dimension of @xmath289 only and both sides of the inequality have a chance to behave nicely in the continuum limit . + * 2 ) * for large graphs which agree in many places , most eigenvalues of the laplacian must be close . the result is convenient to estimate the spectral distance between the complete graph @xmath74 and @xmath9 . in that case , @xmath290 and we have @xmath291 since the dirac spectrum of @xmath74 is contained is the set @xmath292 . it follows that erdoes - renyi graphs in @xmath293 for probabilities @xmath71 close to @xmath89 have a dirac spectrum concentrated near @xmath294 . if @xmath9 has @xmath73 simplices less than @xmath74 , then @xmath295 . + * examples . * + * 1 ) * let @xmath9 be the triangle and @xmath43 be the line graph with three vertices . then @xmath296 . we have @xmath297 . since the graphs differ on the triangle and one edge only , we have @xmath298 and @xmath299 . the augmented dirac matrices are @xmath300 , d_h = \left [ \begin{array}{ccccccc } 0 & 0 & 0 & -1 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & -1 & 0 \\ 0 & 0 & 0 & 1 & 0 & 1 & 0 \\ -1 & 0 & 1 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & -1 & 1 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ \end{array } \right ] \ ; .\ ] ] the spectrum of the dirac matrix of @xmath9 is @xmath301 the spectrum of the augmented dirac matrix of @xmath43 is @xmath302 . the spectral distance is @xmath303 . + * 2 ) * if @xmath9 is @xmath74 and @xmath43 is @xmath304 then @xmath305 . we have @xmath306/(2^n-1 ) = 2^{n-1}/(2^n-1 ) \sim 1/2 $ ] and @xmath307 and so @xmath308 . the dirac eigenvalues differ by @xmath309 at @xmath310 places and by @xmath311 at @xmath312 places . the spectral distance is @xmath313 which is about @xmath314 . taking away one vertex together with all the edges is quite a drastic perturbation step . it gets rid of a lot of simplices and changes the dimension of the graph . + * 3 ) * let @xmath9 be the wheel graph @xmath315 with 4 spikes . it is the simplest model for a planar region with boundary . let @xmath43 be the graph where we make a pyramid extension over one of the boundary edges . this is a homotopy deformation . the graph @xmath9 has @xmath316 simplices and the graph @xmath43 has @xmath317 simplices . the graph @xmath43 has one vertex , two edges and one triangle more than @xmath9 so that @xmath318 . the maximal degree is @xmath319 so that the right hand side of the estimate is @xmath320 . the eigenvalues of the dirac operator @xmath321 + @xmath322 $ ] + ( where , we wrote @xmath323 for typographic reasons ) of @xmath43 are @xmath324 @xmath325 @xmath326 @xmath327 @xmath328 @xmath329 . the eigenvalues of the augmented dirac operator of @xmath9 are @xmath330 @xmath331 @xmath332 @xmath333 @xmath334 @xmath335 @xmath336 . the spectral differences is @xmath337 . we were generous with the degree estimate . the new added point adds simplices with maximal degree @xmath338 so that the proof of the perturbation result could be improved to get an upper bound @xmath339 . it is still by a factor 3 larger than the spectral difference , but we get the idea : if @xmath43 is a triangularization of a large domain and we just move the boundary a bit by adding a triangle then the spectrum almost does not budge . this will allow us to study spectral differences @xmath340 of planar regions in terms of the area of the symmetric difference . adjacency matrices have always served as an algebraic bridge to study graphs . the entry @xmath341 has the interpretation as the number of paths in @xmath9 starting at a vertex @xmath342 and ending at a vertex @xmath114 . from the adjacency matrix @xmath46 , the laplacian @xmath343 is defined . similarly as for @xmath34 , we can read off geometric quantities also from the diagonal entries of the full laplacians @xmath128 on @xmath71 forms . the number of @xmath344-dimensional simplices which contain the @xmath71-dimensional simplex @xmath102 is called the * p - degree * of @xmath102 . it is denoted @xmath345 . * examples . * + * 1 ) * for @xmath346 , the degree @xmath347 is the usual degree vertex degree @xmath348 of the vertex @xmath102 . + * 2 ) * for @xmath349 the degree @xmath350 is the number of triangles which are attached to an edge @xmath102 . for @xmath351 we have @xmath352 . for @xmath346 we have @xmath353 . because @xmath354 has @xmath99 nonzero entries @xmath89 or @xmath126 in each row @xmath102 , we have @xmath355 . the reason for the special situation @xmath346 is that @xmath34 is the only laplacian which does not contain a second part @xmath356 because @xmath357 is zero on scalars : @xmath358 * remarks . * + * 1 ) * the case @xmath346 is special because @xmath34 only consists of @xmath359 while @xmath70 with @xmath351 has two parts @xmath360 . the degree formulas actually count closed paths of length @xmath79 starting at @xmath102 . for @xmath351 , paths can also lower the dimension . for example , for @xmath349 , where we start with an edge , then there are two paths which start at the edge , chose a vertex and then get back to the edge . this explains the correction @xmath99 . + * 2 ) * for @xmath349 , we can count the number of triangles attached to an edge @xmath102 with @xmath361 . + * 3 ) * a closed path interpretation of the diagonal entries will generalize the statement . paths of length @xmath79 count adjacent simplices in @xmath41 . we can read off the total number @xmath362 of @xmath71-simplices in @xmath9 from the trace of @xmath70 . this generalizes the euler handshaking result that the sum of the degrees of a finite simple graph is twice the number of edges : @xmath363 . summing up @xmath352 gives @xmath364 . we write just @xmath89 for the identity matrix . the formula @xmath365 follows from mckean - singer because @xmath366 . combinatorically it is equivalent to a statement which we can verify directly and manifests in cancellations of traces : @xmath367 adding this up gives @xmath368 . + an other consequence is @xmath369 * remarks . * + * 1 ) * each identity @xmath370 produces some `` curvatures '' on the super graph @xmath41 satisfying @xmath371 the case @xmath372 being trivial giving @xmath373 for which gauss - bonnet is the definition of the euler characteristic and where @xmath374 is the case just discussed . + while the adjacency matrix @xmath46 of a graph has @xmath375 as the number of closed paths in @xmath9 of length @xmath93 , the interpretation of @xmath376 becomes only obvious when looking at it in more generally when looking the full laplacian @xmath20 as we do here . instead of finding an interpretation where we add loops to the vertices , it is more natural to look at paths in @xmath41 : a * path * in @xmath377 as a sequence of simplices @xmath378 such that either @xmath379 is either strictly contained in @xmath380 or that @xmath380 is strictly contained in @xmath379 and a path starting in @xmath7 can additionally to @xmath7 only visit one of the neighboring spaces @xmath381 or @xmath382 along the entire trajectory . the fact that this random walk on @xmath41 can not visit three different dimension - sectors @xmath383 is a consequence of the identities @xmath384 ; a path visiting three different sectors would have cases , where @xmath66 or @xmath385 appears in a pair in the expansion of @xmath386 . algebraically it manifests itself in the fact that @xmath387 is always is a block matrix for which each block @xmath388 leaves the subspace @xmath60 of @xmath93-forms invariant . + * examples . * + * 1 ) * for a graph without triangles , a path starting at a vertex @xmath389 is a sequence @xmath390 . every path in @xmath41 of length @xmath391 corresponds to a path of length @xmath38 in @xmath9 . + * 2 ) * for a triangular graph @xmath9 , there are three closed paths of length @xmath79 starting an edge @xmath392 . the first path is @xmath393 , the second @xmath394 and the third is @xmath395 where @xmath11 is the triangle . + * 3 ) * again for the triangle , there are @xmath396 closed paths of length @xmath115 starting at a vertex : four paths crossing two edges @xmath397 and two paths crossing the same edge twice @xmath398 . there are @xmath399 paths of length @xmath115 starting at an edge . there are four paths of the form @xmath400 and four paths of the form @xmath401 and one path @xmath402 . the integer @xmath403 is the number of paths of length @xmath93 in @xmath41 starting at a simplex @xmath102 and ending at a simplex @xmath404 . the trace @xmath405 is the total number of closed paths in @xmath41 which have length @xmath406 . we expand @xmath407 . for odd @xmath93 we have expressions of the form @xmath408 or @xmath409 . for even @xmath93 , we have expressions of the form @xmath410 or @xmath411 . the second statement follows from summing over @xmath102 . as a consequence of the mckean - singer theorem we have the following corollary which is a priori not so obvious because we do not assume any symmetry for the graph . let @xmath9 be an arbitrary finite simple graph . the number of closed paths in @xmath41 of length @xmath406 starting at even dimensional simplices is equal to the number of closed paths of length @xmath406 starting at odd dimensional simplices . @xmath412 . for example , on a triangle , there are @xmath396 closed paths of length @xmath115 starting at a vertex and @xmath399 closed paths of length @xmath115 starting at an edge and @xmath399 closed paths of length @xmath115 starting at a triangle . there are @xmath413 paths starting at an even dimensional simplex ( vertex or triangle ) and @xmath414 paths starting at an odd dimensional simplex ( edge ) . finally , we want to write the curvature @xmath415 of a graph using the operator @xmath22 . for a vertex @xmath416 , denote by @xmath417 the number of @xmath57 graphs in the unit sphere @xmath418 . the curvature at a vertex @xmath102 is defined as @xmath419 it satisfies the * gauss - bonnet theorem * @xcite @xmath420 an identity which holds for any finite simple graph . the result becomes more interesting and deeper , if more gometric structure on the graph is assumed . for example , for geometric graphs , where the unit spheres are discrete spheres of fixed dimension with euler characteristic like in the continuum , then @xmath421 for odd dimensional graphs . this result @xcite relies on discrete integral geometric methods and in particular on @xcite which assures that curvature is the expectation of the index of functions . denote by @xmath422 the operator @xmath70 restricted to the unit sphere @xmath418 . it can be thought of as an analogue of a * signature * for differential operators . define the linear operator @xmath423 so that @xmath424 . the curvature @xmath415 at a vertex @xmath102 of a graph satisfies @xmath425 from @xmath426 we get @xmath427 and so @xmath428 we have now also in the discrete an operator theoretical interpretation of gauss - bonnet - chern theorem in the same way as in the continuum , where the * atiyah - singer index theorem * provides this interpretation . + if we restrict @xmath22 to the even part @xmath429 , the euler characteristic is the * analytic index * @xmath430 . the * topological index * of @xmath22 is @xmath431 . as in the continuum , the discrete gauss - bonnet theorem is an example of an index theorem . in this section we describe a general way to get dirac isospectral graphs and give example of an isospectral pairs . in the next section , among the examples , an other example is given . there is an analogue quest in the continuum for isospectral nonisometric metrics for all differential forms . as mentioned in @xcite , there are various examples known . milnor s examples with flat tori lifts to isospectrality on forms for example . two connected finite simple graphs @xmath432 which have the same number of vertices and edges and which are triangle free and isospectral with respect to @xmath34 are isospectral with respect to @xmath22 . [ lifting ] since @xmath433 are the same , hodge theory shows that @xmath119 has @xmath434 zero eigenvalues in both cases . because the graphs are connected , @xmath34 have @xmath435 zero eigenvalues . the mckean - singer theorem implies that the nonzero eigenvalues of @xmath34 match the nonzero eigenvalues of @xmath119 . this can be applied also in situations , where triangles are present and where the spectrum on @xmath436 is trivially equivalent . an example has been found in @xcite . that paper uses sunada s techniques to construct isospectral simple graphs . we show that such examples can also be isospectral with respect to the dirac operator . these examples are analogues to examples of manifolds given by gornet @xcite where isospectrality for functions and differential forms can be different . the rest of the article consists of examples . let @xmath9 be the cyclic graph with @xmath115 elements . the exterior derivative @xmath66 is a map from @xmath437 . it defines @xmath438 . the operator @xmath439 is the same than @xmath34 and @xmath440 for all @xmath11 . we have @xmath441 , l_1 = d_0 d_0^ * = \left [ \begin{array}{cccc } 2 & -1 & 0 & -1 \\ -1 & 2 & -1 & 0 \\ 0 & -1 & 2 & -1 \\ -1 & 0 & -1 & 2 \\ \end{array } \right ] \ ; .\ ] ] and so @xmath442 , l=\left [ \begin{array}{cccccccc } 2 & -1 & 0 & -1 & 0 & 0 & 0 & 0 \\ -1 & 2 & -1 & 0 & 0 & 0 & 0 & 0 \\ 0 & -1 & 2 & -1 & 0 & 0 & 0 & 0 \\ -1 & 0 & -1 & 2 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 2 & -1 & 0 & -1 \\ 0 & 0 & 0 & 0 & -1 & 2 & -1 & 0 \\ 0 & 0 & 0 & 0 & 0 & -1 & 2 & -1 \\ 0 & 0 & 0 & 0 & -1 & 0 & -1 & 2 \\ \end{array } \right ] \ ; .\ ] ] the bosonic eigenvalues are @xmath443 , the fermionic eigenvalues are @xmath443 . the example generalizes to any cyclic graph @xmath76 . the operators @xmath444 are jacobi matrices @xmath445 which fourier diagonalizes to the diagonal matrix with entries @xmath446 . let @xmath9 be the triangle . the exterior derivative @xmath447 is @xmath448 , d_1 = \left [ \begin{array}{ccc } -1 & 1 & -1 \\ \end{array } \right ] \;\ ] ] and @xmath449 , l_1=d_1 d_1^ * + d_0^ * d_0 = \left [ \begin{array}{ccc } 3 & 0 & 0 \\ 0 & 3 & 0 \\ 0 & 0 & 3 \\ \end{array } \right ] \ ; l_2 = d_1 d_1^ * = \left [ \begin{array}{c } 3 \\ \end{array } \right ] \ ; .\ ] ] the dirac and laplace operator is @xmath450 , \ ; l=\left [ \begin{array}{ccccccc } 2 & -1 & -1 & 0 & 0 & 0 & 0 \\ -1 & 2 & -1 & 0 & 0 & 0 & 0 \\ -1 & -1 & 2 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 3 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 3 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 3 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 3 \\ \end{array } \right ] \ ; .\ ] ] the later has the bosonic eigenvalues @xmath451 and the fermionic eigenvalues @xmath452 . we have @xmath453 and so @xmath454 . + for the tetrahedron @xmath455 , the exterior derivative @xmath36 is a map from @xmath456 . @xmath457 $ ] and @xmath458 $ ] . @xmath459 $ ] and @xmath460 $ ] and @xmath461 $ ] which sums up to @xmath462 . finally , @xmath463 $ ] and @xmath464 $ ] . for a complete graph @xmath465 , the eigenvalues of @xmath34 are @xmath466 . the other operators @xmath94 with @xmath467 are diagonal with entries @xmath38 . the dirac operator @xmath22 restricted to @xmath71-forms is a @xmath468 matrix , where @xmath469 is the binomial coefficient . lets just write out the matrix @xmath470 @xmath471 \ ; .\ ] ] [ cols= " < , < , < " , ] + gives the super trace @xmath472 we write @xmath473 if the eigenvalue @xmath190 appears with multiplicity @xmath93 . the bosonic eigenvalues of @xmath128 are @xmath474 the fermionic eigenvalues of @xmath20 are @xmath475 the eigenvalues @xmath78 appear only on bosonic parts matching that the betti vector @xmath476 has its support on the bosonic part . the picture colors the vertices , edges and triangles according to the ground states , the eigenvectors to the lowest eigenvalues of @xmath477 and @xmath436 . the figure shows a fullerene type graph based on an icosahedron . it has @xmath478 vertices , @xmath479 edges and @xmath480 triangles . the dirac operator @xmath22 is a @xmath481 matrix in the next picture we see the three sorted spectra of @xmath477 and @xmath436 . here is the spectrum of a two dimensional discrete torus and a discrete klein bottle : the dunce hat is an example important in homotopy theory . it is a non - contractible graph which is homotopic to a one point graph . petersen graph as mentioned in the previous section , some known isospectral graphs for the graph laplacian are also isospectral for the dirac operator . in @xcite , all isospectral connected graphs up to order @xmath482 have been computed . for order @xmath482 , there are already examples which are also isospectral for all @xmath71-forms and for the dirac operator . some of them lead to dirac isospectral graphs : + @xmath483 + @xmath484 . + the laplacian eigenvalues are @xmath485 and @xmath486 as well as @xmath487 . one of the questions which fueled research on spectral theory was whether one can hear the shape of a drum " . the discrete analogue question is whether the dirac operator can hear the shape of a discrete hexagonal region . a hexagonal region and the spectra of the laplacians @xmath488 . a convex hexagonal region and the spectra of @xmath488 as far as we know , no isospectral completely two dimensional convex graphs are known . isospectral benzenoid graphs were constructed in @xcite . they are isospectral with respect to the adjacency matrix @xmath46 . but not with respect to the laplacian @xmath489 . after triangulating the hexagons , the first pair of graphs have euler characteristic @xmath78 , the second euler characteristic @xmath89 . but after triangularization , the graphs are not isospectral with respect to @xmath46 , nor with respect to @xmath20 . isospectral domains in the plane were constructed with sundada methods by gordon , webb and wolpert @xcite . a simplified version consists of two squares and three triangles . to try out graph versions of this , we computed the spectra of a couple of graphs of this type . they are all homotopic to the figure 8 graph but are all not isospectral . this shows that discretizing sunada can not be done naivly . the next section mentions examples given in @xcite which were obtained by adapting sunada type methods to graphs . here are the picture of the dirac matrices and their spectrum of the isospectral graphs constructed in @xcite . for the 5 platonic solids , we have the following dirac spectra and complexities @xmath490 , where we write @xmath491 for @xmath190 with multiplicity @xmath38 . + lrr solid & non - negative dirac spectrum & complexity + tetrahedron & @xmath492 & @xmath493 + octahedron & @xmath494 & @xmath495 + cube & @xmath496 & @xmath497 + dodecahedron&@xmath498 & @xmath499 + icosahedron & @xmath500 & @xmath501 + the following table confirms that for two - dimensional geometric graph for which the unit sphere is a circular graph , the dirac complexity is positive . we actually came up with the statement after watching the following tables : @xmath502 \ ; , c = \left [ \begin{array}{rrr } { \rm dim}(g ) & { \rm complexity } & \chi(g ) \\ \hline 1 & -4.6449\times 10 ^ 6 & -10 \\ 2 & 3.5323\times 10^{84 } & 2 \\ 2 & 1.9246\times 10^{35 } & 2 \\ 1 & -6.6871\times 10^{15 } & -28 \\ 1 & -1.2496\times 10^{31 } & -58 \\ 1 & -7.8275\times 10^{12 } & -22 \\ 1 & -3.4164\times 10^{15 } & -22 \\ 1 & -4.0315\times 10^{37 } & -58 \\ 3 & 3.7873\times 10^{26 } & 2 \\ 3 & 1.4404\times 10^{64 } & 2 \\ 2 & 2.7042\times 10^{44 } & 2 \\ 2 & 3.4381\times 10^{18 } & 2 \\ 3 & -1.6987\times 10^{14 } & 1 \\ \end{array } \right ] \ ; .\ ] ] while the super symmetry argument is elegant , the original proof is illustrative and shows better the link to hodge theory . we follow @xcite but change notation and expand some of the steps slightly . denote by @xmath503 the eigenspace of the eigenvalue @xmath190 of @xmath20 on @xmath504 . @xmath505 for all @xmath506 . [ mckeansingerpropo ] if @xmath190 is a positive eigenvalue , we just write @xmath507 instead of @xmath508 . + ( i ) @xmath509 + proof . that the right hand side is included in the left hand side follows from @xmath510=[d^*,l]=0 $ ] . for example : if @xmath511 with @xmath512 , then @xmath513 and @xmath514 . + given @xmath515 , then @xmath516 so that the two linear spaces on the right hand side are perpendicular . assume we are given @xmath514 which is perpendicular to both subspaces . then @xmath517 and @xmath518 so that @xmath519 and @xmath520 which together with @xmath514 implies @xmath521 . + ( ii ) summing the identity @xmath522 obtained in ( i ) gives @xmath523 @xmath524+[\dim(d^ * e^{p+1})-\dim(d^ * d e^{p-1 } ) ] \geq 0 \ ; . \end{aligned}\ ] ] ( iii ) @xmath525 + from ( ii ) , we recycle @xmath526 . the claim follows now because each term is @xmath527 : @xmath528 + [ \dim(d^ * d e^p ) - \dim(d^ * e^p ) ] \leq 0 \ ; . \end{aligned}\ ] ] the two parts ( ii ) and ( iii ) imply the proposition . @xmath529 . proposition ( [ mckeansingerpropo ] ) implies @xmath530 for @xmath531 . a taylor expansion of the super trace of the heat kernel using the proposition gives @xmath532 for @xmath13 we have by definition the euler characteristic and for @xmath14 , the supertrace of the identity on harmonic forms . the proof of the hodge theorem stating that @xmath533 is isomorphic to the space @xmath534 of harmonic @xmath71-forms is clearly visible in mckean - singers proof : \a ) @xmath535 and @xmath536 . + b ) @xmath537 is equivalent to @xmath538 and @xmath539 . + c ) @xmath511 and @xmath540 are orthogonal . @xmath541 span @xmath542 . \a ) is clear from @xmath384 . + b ) if @xmath520 then @xmath543 shows that both @xmath544 and @xmath538 . the other direction is clear . + c ) @xmath545 . any @xmath71-form can be written as a sum of an exact , a coexact and a harmonic @xmath71-form . there is an orthogonal decomposition @xmath546 any @xmath170 can be written as @xmath547 where @xmath93 is harmonic . the operator @xmath548 is symmetric so that image and kernel are perpendicular . we have seen before that the image of @xmath20 splits into two orthogonal components @xmath549 and @xmath550 . the dimension of the vector space of harmonic @xmath93-forms on a simple graph is @xmath236 . every cohomology class has a unique harmonic representative . if @xmath520 , then @xmath538 and so @xmath551 . given a closed @xmath38-form @xmath170 then @xmath552 and the hodge decomposition shows @xmath553 so that @xmath170 differs from a harmonic form only by @xmath554 and so that this harmonic form is in the same cohomology class than @xmath15 . we can assign so to a cohomology class a harmonic form and this map is an isomorphism . we have mentioned that the dirac operator has appeared when computing ech cohomology of a manifold and investigating the relation between the spectrum of the graph and the spectrum of the manifold @xcite . we want to show here that ech cohomology is equivalent to graph cohomology . this is practical : while graph cohomology is easy to implement in a computer , it can be tedious to compute the operator @xmath22 and so cohomology groups . if the zero eigenvalues and so the cohomology is the only interest in @xmath22 , then one can proceed in two different but related ways : one possibilities to reduce the complexity of computing the kernel of @xmath22 is to deform the graph using homotopy steps to get a simpler graph . another possibility is to look at the dirac operator of the nerve . the following definition is equivalent to a definition of ivashchenko : a graph is called * contractible in itself * , if there is a proper subgraph @xmath43 which is contractible and which is the unit sphere of a vertex . the graph @xmath9 without the vertex @xmath102 and all connections to @xmath102 is called a contraction of @xmath9 . for a contractible graph , there is a sequence of such contraction steps which reduces @xmath9 to a one point graph . a ech cover of @xmath9 is a finite set of subgraphs @xmath555 which are contractible and for which any finite intersection of such subgraphs is either empty or contractible . a hech cover defines a new graph @xmath130 called the * nerve * of the cover . the vertices of @xmath130 are the elements of the cover and there is an edge @xmath556 if both the subgraphs @xmath557 have a nonempty intersection . * ech cohomology * is then defined as the graph cohomology of the nerve of @xmath9 . in a first step we expand each of the @xmath555 so that in each @xmath555 , the intersections @xmath560 are in a neighborhood @xmath561 of a point @xmath562 . now homotopically shrink each @xmath555 to @xmath561 . graph cohomology obviously is identical to ech cohomology if we chose the cover @xmath563 for the vertices @xmath564 which has the property that the nerve of @xmath9 is equal to @xmath9 . because the nerve @xmath130 of @xmath9 is homotopic to @xmath9 and graph cohomology of homotopic graphs is the same ( as already proven by ivashchenko @xcite ) , we are done . |